Merge branch 'master' into llvm-context

Conflicts:
	src/gallium/auxiliary/gallivm/lp_bld_tgsi_soa.c

211 files changed, 5185 insertions, 3932 deletions

diff --git a/SConstruct b/SConstruct
index 558ebf90aa..c6198041fb 100644
--- a/SConstruct
+++ b/SConstruct
@@ -3,14 +3,14 @@
 #
 # For example, invoke scons as 
 #
-#   scons debug=1 dri=0 machine=x86
+#   scons build=debug llvm=yes machine=x86
 #
 # to set configuration variables. Or you can write those options to a file
 # named config.py:
 #
 #   # config.py
-#   debug=1
-#   dri=0
+#   build='debug'
+#   llvm=True
 #   machine='x86'
 # 
 # Invoke
@@ -30,54 +30,8 @@ import common
 #######################################################################
 # Configuration options
 
-default_statetrackers = 'mesa'
-default_targets = 'graw-null'
-
-if common.default_platform in ('linux', 'freebsd', 'darwin'):
-	default_drivers = 'softpipe,galahad,failover,svga,i915,i965,trace,identity,llvmpipe'
-	default_winsys = 'xlib'
-elif common.default_platform in ('winddk',):
-	default_drivers = 'softpipe,svga,i915,i965,trace,identity'
-	default_winsys = 'all'
-elif common.default_platform in ('embedded',):
-	default_drivers = 'softpipe,llvmpipe'
-	default_winsys = 'xlib'
-else:
-	default_drivers = 'all'
-	default_winsys = 'all'
-
 opts = Variables('config.py')
 common.AddOptions(opts)
-opts.Add(ListVariable('statetrackers', 'state trackers to build', default_statetrackers,
-                     ['mesa', 'python', 'xorg', 'egl']))
-opts.Add(ListVariable('drivers', 'pipe drivers to build', default_drivers,
-                     ['softpipe', 'galahad', 'failover', 'svga', 'i915', 'i965', 'trace', 'r300', 'r600', 'identity', 'llvmpipe', 'nouveau', 'nv50', 'nvfx']))
-opts.Add(ListVariable('winsys', 'winsys drivers to build', default_winsys,
-                     ['xlib', 'vmware', 'i915', 'i965', 'gdi', 'radeon', 'r600', 'graw-xlib']))
-
-opts.Add(ListVariable('targets', 'driver targets to build', default_targets,
-		      ['dri-i915',
-		       'dri-i965',
-		       'dri-nouveau',
-		       'dri-radeong',
-		       'dri-swrast',
-		       'dri-vmwgfx',
-		       'egl-i915',
-		       'egl-i965',
-		       'egl-nouveau',
-		       'egl-radeon',
-		       'egl-swrast',
-		       'egl-vmwgfx',
-		       'graw-xlib',
-		       'graw-null',
-		       'libgl-gdi',
-		       'libgl-xlib',
-		       'xorg-i915',
-		       'xorg-i965',
-		       'xorg-nouveau',
-		       'xorg-radeon',
-		       'xorg-vmwgfx']))
-
 opts.Add(EnumVariable('MSVS_VERSION', 'MS Visual C++ version', None, allowed_values=('7.1', '8.0', '9.0')))
 
 env = Environment(
@@ -87,61 +41,26 @@ env = Environment(
 	ENV = os.environ,
 )
 
-if os.environ.has_key('CC'):
-	env['CC'] = os.environ['CC']
-if os.environ.has_key('CFLAGS'):
-	env['CCFLAGS'] += SCons.Util.CLVar(os.environ['CFLAGS'])
-if os.environ.has_key('CXX'):
-	env['CXX'] = os.environ['CXX']
-if os.environ.has_key('CXXFLAGS'):
-	env['CXXFLAGS'] += SCons.Util.CLVar(os.environ['CXXFLAGS'])
-if os.environ.has_key('LDFLAGS'):
-	env['LINKFLAGS'] += SCons.Util.CLVar(os.environ['LDFLAGS'])
+# Backwards compatability with old target configuration variable
+try:
+    targets = ARGUMENTS['targets']
+except KeyError:
+    pass
+else:
+    targets = targets.split(',')
+    print 'scons: warning: targets option is deprecated; pass the targets on their own such as'
+    print
+    print '  scons %s' % ' '.join(targets)
+    print 
+    COMMAND_LINE_TARGETS.append(targets)
 
-Help(opts.GenerateHelpText(env))
 
-# replicate options values in local variables
-debug = env['debug']
-dri = env['dri']
-machine = env['machine']
-platform = env['platform']
-
-# derived options
-x86 = machine == 'x86'
-ppc = machine == 'ppc'
-gcc = platform in ('linux', 'freebsd', 'darwin', 'embedded')
-msvc = platform in ('windows', 'winddk')
-
-Export([
-	'debug', 
-	'x86', 
-	'ppc', 
-	'dri', 
-	'platform',
-	'gcc',
-	'msvc',
-])
+Help(opts.GenerateHelpText(env))
 
 
 #######################################################################
 # Environment setup
 
-# Always build trace, rbug, identity, softpipe, and llvmpipe (where possible)
-if 'trace' not in env['drivers']:
-    env['drivers'].append('trace')
-if 'rbug' not in env['drivers']:
-    env['drivers'].append('rbug')
-if 'galahad' not in env['drivers']:
-    env['drivers'].append('galahad')
-if 'identity' not in env['drivers']:
-    env['drivers'].append('identity')
-if 'softpipe' not in env['drivers']:
-    env['drivers'].append('softpipe')
-if env['llvm'] and 'llvmpipe' not in env['drivers']:
-    env['drivers'].append('llvmpipe')
-if 'sw' not in env['drivers']:
-    env['drivers'].append('sw')
-
 # Includes
 env.Prepend(CPPPATH = [
 	'#/include',
@@ -157,7 +76,7 @@ if env['msvc']:
     env.Append(CPPPATH = ['#include/c99'])
 
 # Embedded
-if platform == 'embedded':
+if env['platform'] == 'embedded':
 	env.Append(CPPDEFINES = [
 		'_POSIX_SOURCE',
 		('_POSIX_C_SOURCE', '199309L'), 
@@ -174,7 +93,7 @@ if platform == 'embedded':
 	])
 
 # Posix
-if platform in ('posix', 'linux', 'freebsd', 'darwin'):
+if env['platform'] in ('posix', 'linux', 'freebsd', 'darwin'):
 	env.Append(CPPDEFINES = [
 		'_POSIX_SOURCE',
 		('_POSIX_C_SOURCE', '199309L'), 
@@ -184,9 +103,9 @@ if platform in ('posix', 'linux', 'freebsd', 'darwin'):
 		'PTHREADS',
 		'HAVE_POSIX_MEMALIGN',
 	])
-	if gcc:
+	if env['gcc']:
 		env.Append(CFLAGS = ['-fvisibility=hidden'])
-	if platform == 'darwin':
+	if env['platform'] == 'darwin':
 		env.Append(CPPDEFINES = ['_DARWIN_C_SOURCE'])
 	env.Append(LIBS = [
 		'm',
@@ -212,5 +131,3 @@ SConscript(
 	duplicate = 0 # http://www.scons.org/doc/0.97/HTML/scons-user/x2261.html
 )
 
-env.Default('src')
-
diff --git a/common.py b/common.py
index 13e8dc83e9..b7749c925d 100644
--- a/common.py
+++ b/common.py
@@ -8,6 +8,8 @@ import subprocess
 import sys
 import platform as _platform
 
+import SCons.Script.SConscript
+
 
 #######################################################################
 # Defaults
@@ -20,6 +22,15 @@ _platform_map = {
 default_platform = sys.platform
 default_platform = _platform_map.get(default_platform, default_platform)
 
+# Search sys.argv[] for a "platform=foo" argument since we don't have
+# an 'env' variable at this point.
+if 'platform' in SCons.Script.ARGUMENTS:
+    selected_platform = SCons.Script.ARGUMENTS['platform']
+else:
+    selected_platform = default_platform
+
+cross_compiling = selected_platform != default_platform
+
 _machine_map = {
 	'x86': 'x86',
 	'i386': 'x86',
@@ -37,38 +48,26 @@ if 'PROCESSOR_ARCHITECTURE' in os.environ:
 else:
 	default_machine = _platform.machine()
 default_machine = _machine_map.get(default_machine, 'generic')
+default_toolchain = 'default'
+
+if selected_platform == 'windows' and cross_compiling:
+    default_machine = 'x86'
+    default_toolchain = 'crossmingw'
 
 
 # find default_llvm value
 if 'LLVM' in os.environ:
     default_llvm = 'yes'
 else:
-    # Search sys.argv[] for a "platform=foo" argument since we don't have
-    # an 'env' variable at this point.
-    platform = default_platform
-    pattern = re.compile("(platform=)(.*)")
-    for arg in sys.argv:
-        m = pattern.match(arg)
-        if m:
-            platform = m.group(2)
-
     default_llvm = 'no'
     try:
-        if platform != 'windows' and subprocess.call(['llvm-config', '--version'], stdout=subprocess.PIPE) == 0:
+        if selected_platform != 'windows' and \
+           subprocess.call(['llvm-config', '--version'], stdout=subprocess.PIPE) == 0:
             default_llvm = 'yes'
     except:
         pass
 
 
-# find default_dri value
-if default_platform in ('linux', 'freebsd'):
-	default_dri = 'yes'
-elif default_platform in ('winddk', 'windows', 'wince', 'darwin'):
-	default_dri = 'no'
-else:
-	default_dri = 'no'
-
-
 #######################################################################
 # Common options
 
@@ -88,8 +87,7 @@ def AddOptions(opts):
 											 allowed_values=('generic', 'ppc', 'x86', 'x86_64')))
 	opts.Add(EnumOption('platform', 'target platform', default_platform,
 											 allowed_values=('linux', 'cell', 'windows', 'winddk', 'wince', 'darwin', 'embedded', 'cygwin', 'sunos5', 'freebsd8')))
-	opts.Add('toolchain', 'compiler toolchain', 'default')
+	opts.Add('toolchain', 'compiler toolchain', default_toolchain)
 	opts.Add(BoolOption('llvm', 'use LLVM', default_llvm))
-	opts.Add(BoolOption('dri', 'build DRI drivers', default_dri))
 	opts.Add(BoolOption('debug', 'DEPRECATED: debug build', 'yes'))
 	opts.Add(BoolOption('profile', 'DEPRECATED: profile build', 'no'))
diff --git a/configure.ac b/configure.ac
index 860ea21ca7..feb8c7fd08 100644
--- a/configure.ac
+++ b/configure.ac
@@ -493,6 +493,16 @@ AC_ARG_ENABLE([openvg],
     [enable_openvg="$enableval"],
     [enable_openvg=no])
 
+dnl smooth the transition; should be removed eventually
+if test "x$enable_openvg" = xno; then
+    case "x$with_state_trackers" in
+    x*vega*)
+        AC_MSG_WARN([vega state tracker is enabled without --enable-openvg])
+        enable_openvg=yes
+        ;;
+    esac
+fi
+
 if test "x$enable_opengl" = xno -a \
         "x$enable_gles1" = xno -a \
         "x$enable_gles2" = xno -a \
@@ -1398,6 +1408,11 @@ yes)
             PKG_CHECK_MODULES([LIBKMS_XORG], [libkms >= $LIBKMS_XORG_REQUIRED])
             HAVE_ST_XORG="yes"
             ;;
+        vega)
+            if test "x$enable_openvg" != xyes; then
+                AC_MSG_ERROR([cannot build vega state tracker without --enable-openvg])
+            fi
+            ;;
         esac
 
 	if test -n "$tracker"; then
@@ -1743,12 +1758,17 @@ echo ""
 echo "        EGL:             $enable_egl"
 if test "$enable_egl" = yes; then
     echo "        EGL platforms:   $EGL_PLATFORMS"
-    echo "        EGL drivers:     $EGL_DRIVERS_DIRS"
+
+    egl_drivers=""
+    for d in $EGL_DRIVERS_DIRS; do
+        egl_drivers="$egl_drivers egl_$d"
+    done
+
     if test "$enable_gallium" = yes -a "$HAVE_ST_EGL" = yes; then
-        echo "        EGL tracker:     yes"
-        echo "        EGL client APIs:$EGL_CLIENT_APIS"
+        echo "        EGL drivers:    ${egl_drivers} egl_gallium"
+        echo "        EGL Gallium STs:$EGL_CLIENT_APIS"
     else
-        echo "        EGL tracker:     no"
+        echo "        EGL drivers:    $egl_drivers"
     fi
 fi
 
@@ -1769,9 +1789,6 @@ if echo "$SRC_DIRS" | grep 'gallium' >/dev/null 2>&1; then
     echo "        Winsys dirs:     $GALLIUM_WINSYS_DIRS"
     echo "        Driver dirs:     $GALLIUM_DRIVERS_DIRS"
     echo "        Trackers dirs:   $GALLIUM_STATE_TRACKERS_DIRS"
-    if test "x$HAVE_ST_EGL" = xyes; then
-        echo "        EGL client APIs: $EGL_CLIENT_APIS"
-    fi
 else
     echo "        Gallium:         no"
 fi
diff --git a/docs/egl.html b/docs/egl.html
index cc6462d978..c86e543859 100644
--- a/docs/egl.html
+++ b/docs/egl.html
@@ -136,11 +136,6 @@ binaries.</p>
 specified EGL driver to be loaded.  It comes in handy when one wants to test a
 specific driver.  This variable is ignored for setuid/setgid binaries.</p>
 
-<p><code>egl_gallium</code> dynamically loads hardware drivers and client API
-modules found in <code>EGL_DRIVERS_PATH</code>.  Thus, specifying this variable
-alone is not sufficient for <code>egl_gallium</code> for an uninstalled
-build.</p>
-
 </li>
 
 <li><code>EGL_PLATFORM</code>
diff --git a/scons/gallium.py b/scons/gallium.py
index b065b7bc49..7374cf58d3 100644
--- a/scons/gallium.py
+++ b/scons/gallium.py
@@ -50,29 +50,34 @@ def symlink(target, source, env):
 
 def install(env, source, subdir):
     target_dir = os.path.join(env.Dir('#.').srcnode().abspath, env['build_dir'], subdir)
-    env.Install(target_dir, source)
+    return env.Install(target_dir, source)
 
 def install_program(env, source):
-    install(env, source, 'bin')
+    return install(env, source, 'bin')
 
 def install_shared_library(env, sources, version = ()):
+    targets = []
     install_dir = os.path.join(env.Dir('#.').srcnode().abspath, env['build_dir'])
     version = tuple(map(str, version))
     if env['SHLIBSUFFIX'] == '.dll':
         dlls = env.FindIxes(sources, 'SHLIBPREFIX', 'SHLIBSUFFIX')
-        install(env, dlls, 'bin')
+        targets += install(env, dlls, 'bin')
         libs = env.FindIxes(sources, 'LIBPREFIX', 'LIBSUFFIX')
-        install(env, libs, 'lib')
+        targets += install(env, libs, 'lib')
     else:
         for source in sources:
             target_dir =  os.path.join(install_dir, 'lib')
             target_name = '.'.join((str(source),) + version)
             last = env.InstallAs(os.path.join(target_dir, target_name), source)
+            targets += last
             while len(version):
                 version = version[:-1]
                 target_name = '.'.join((str(source),) + version)
                 action = SCons.Action.Action(symlink, "$TARGET -> $SOURCE")
                 last = env.Command(os.path.join(target_dir, target_name), last, action) 
+                targets += last
+    return targets
+
 
 def createInstallMethods(env):
     env.AddMethod(install_program, 'InstallProgram')
@@ -98,6 +103,38 @@ def num_jobs():
     return 1
 
 
+def pkg_config_modules(env, name, modules):
+    '''Simple wrapper for pkg-config.'''
+
+    env[name] = False
+
+    if env['platform'] == 'windows':
+        return
+
+    if not env.Detect('pkg-config'):
+        return
+
+    # Put -I and -L flags directly into the environment, as these don't affect
+    # the compilation of targets that do not use them
+    try:
+        env.ParseConfig('pkg-config --cflags-only-I --libs-only-L ' + ' '.join(modules))
+    except OSError:
+        return
+
+    # Other flags may affect the compilation of unrelated targets, so store
+    # them with a prefix, (e.g., XXX_CFLAGS, XXX_LIBS, etc)
+    try:
+        flags = env.ParseFlags('!pkg-config --cflags-only-other --libs-only-l --libs-only-other ' + ' '.join(modules))
+    except OSError:
+        return
+    prefix = name.upper() + '_'
+    for flag_name, flag_value in flags.iteritems():
+        env[prefix + flag_name] = flag_value
+
+    env[name] = True
+
+
+
 def generate(env):
     """Common environment generation code"""
 
@@ -110,21 +147,27 @@ def generate(env):
             env['toolchain'] = 'wcesdk'
     env.Tool(env['toolchain'])
 
-    if env['platform'] == 'embedded':
-        # Allow overriding compiler from environment
-        if os.environ.has_key('CC'):
-            env['CC'] = os.environ['CC']
-            # Update CCVERSION to match
-            pipe = SCons.Action._subproc(env, [env['CC'], '--version'],
-                                         stdin = 'devnull',
-                                         stderr = 'devnull',
-                                         stdout = subprocess.PIPE)
-            if pipe.wait() == 0:
-                line = pipe.stdout.readline()
-                match = re.search(r'[0-9]+(\.[0-9]+)+', line)
-                if match:
-                    env['CCVERSION'] = match.group(0)
-            
+    # Allow override compiler and specify additional flags from environment
+    if os.environ.has_key('CC'):
+        env['CC'] = os.environ['CC']
+        # Update CCVERSION to match
+        pipe = SCons.Action._subproc(env, [env['CC'], '--version'],
+                                     stdin = 'devnull',
+                                     stderr = 'devnull',
+                                     stdout = subprocess.PIPE)
+        if pipe.wait() == 0:
+            line = pipe.stdout.readline()
+            match = re.search(r'[0-9]+(\.[0-9]+)+', line)
+            if match:
+                env['CCVERSION'] = match.group(0)
+    if os.environ.has_key('CFLAGS'):
+        env['CCFLAGS'] += SCons.Util.CLVar(os.environ['CFLAGS'])
+    if os.environ.has_key('CXX'):
+        env['CXX'] = os.environ['CXX']
+    if os.environ.has_key('CXXFLAGS'):
+        env['CXXFLAGS'] += SCons.Util.CLVar(os.environ['CXXFLAGS'])
+    if os.environ.has_key('LDFLAGS'):
+        env['LINKFLAGS'] += SCons.Util.CLVar(os.environ['LDFLAGS'])
 
     env['gcc'] = 'gcc' in os.path.basename(env['CC']).split('-')
     env['msvc'] = env['CC'] == 'cl'
@@ -140,10 +183,16 @@ def generate(env):
     # Backwards compatability with the debug= profile= options
     if env['build'] == 'debug':
         if not env['debug']:
-            print 'scons: debug option is deprecated: use instead build=release'
+            print 'scons: warning: debug option is deprecated and will be removed eventually; use instead'
+            print
+            print ' scons build=release'
+            print
             env['build'] = 'release'
         if env['profile']:
-            print 'scons: profile option is deprecated: use instead build=profile'
+            print 'scons: warning: profile option is deprecated and will be removed eventually; use instead'
+            print
+            print ' scons build=profile'
+            print
             env['build'] = 'profile'
     if False:
         # Enforce SConscripts to use the new build variable
@@ -184,6 +233,9 @@ def generate(env):
     if env.GetOption('num_jobs') <= 1:
         env.SetOption('num_jobs', num_jobs())
 
+    env.Decider('MD5-timestamp')
+    env.SetOption('max_drift', 60)
+
     # C preprocessor options
     cppdefines = []
     if env['build'] in ('debug', 'checked'):
@@ -499,9 +551,15 @@ def generate(env):
     # Default libs
     env.Append(LIBS = [])
 
-    # Load LLVM
+    # Load tools
     if env['llvm']:
         env.Tool('llvm')
+        env.Tool('udis86')
+    
+    pkg_config_modules(env, 'x11', ['x11', 'xext'])
+    pkg_config_modules(env, 'dri', ['libdrm'])
+    pkg_config_modules(env, 'xorg', ['xorg-server'])
+    pkg_config_modules(env, 'kms', ['libkms'])
 
     # Custom builders and methods
     env.Tool('custom')
diff --git a/scons/llvm.py b/scons/llvm.py
index 39fbb910b6..1b033acb1b 100644
--- a/scons/llvm.py
+++ b/scons/llvm.py
@@ -38,6 +38,8 @@ import SCons.Util
 
 
 def generate(env):
+    env['llvm'] = False
+
     try:
         llvm_dir = os.environ['LLVM']
     except KeyError:
@@ -64,13 +66,13 @@ def generate(env):
         # XXX: There is no llvm-config on Windows, so assume a standard layout
         if llvm_dir is None:
             print 'scons: LLVM environment variable must be specified when building for windows'
-            env.Exit(1)
+            return
 
         # Try to determine the LLVM version from llvm/Config/config.h
         llvm_config = os.path.join(llvm_dir, 'include/llvm/Config/config.h')
         if not os.path.exists(llvm_config):
             print 'scons: could not find %s' % llvm_config
-            env.Exit(1)
+            return
         llvm_version_re = re.compile(r'^#define PACKAGE_VERSION "([^"]*)"')
         llvm_version = None
         for line in open(llvm_config, 'rt'):
@@ -81,7 +83,7 @@ def generate(env):
                 break
         if llvm_version is None:
             print 'scons: could not determine the LLVM version from %s' % llvm_config
-            env.Exit(1)
+            return
 
         env.Prepend(CPPPATH = [os.path.join(llvm_dir, 'include')])
         env.AppendUnique(CPPDEFINES = [
@@ -133,7 +135,7 @@ def generate(env):
     else:
         if not env.Detect('llvm-config'):
             print 'scons: llvm-config script not found' % llvm_version
-            env.Exit(1)
+            return
 
         llvm_version = env.backtick('llvm-config --version').rstrip()
         llvm_version = distutils.version.LooseVersion(llvm_version)
@@ -144,11 +146,12 @@ def generate(env):
             env.ParseConfig('llvm-config --ldflags')
         except OSError:
             print 'scons: llvm-config version %s failed' % llvm_version
-            env.Exit(1)
+            return
         else:
             env['LINK'] = env['CXX']
 
     assert llvm_version is not None
+    env['llvm'] = True
 
     print 'scons: Found LLVM version %s' % llvm_version
     env['LLVM_VERSION'] = llvm_version
diff --git a/scons/udis86.py b/scons/udis86.py
index ba71d4eb0b..bb91d3c35c 100644
--- a/scons/udis86.py
+++ b/scons/udis86.py
@@ -31,8 +31,10 @@ def generate(env):
     conf = env.Configure()
 
     if conf.CheckHeader('udis86.h'): # and conf.CheckLib('udis86'):
-        env.Append(CPPDEFINES = [('HAVE_UDIS86', '1')])
+        env['UDIS86'] = True
         env.Prepend(LIBS = ['udis86'])
+    else:
+        env['UDIS86'] = False
 
     conf.Finish()
 
diff --git a/src/SConscript b/src/SConscript
index c3e34be6f7..c42d9bff2d 100644
--- a/src/SConscript
+++ b/src/SConscript
@@ -1,19 +1,17 @@
 Import('*')
 
-if 'egl' in env['statetrackers']:
-    SConscript('mapi/vgapi/SConscript')
-    SConscript('egl/main/SConscript')
+SConscript('mapi/vgapi/SConscript')
 
-if 'mesa' in env['statetrackers']:
-    if platform == 'windows':
-        SConscript('talloc/SConscript')
+if env['platform'] == 'windows':
+    SConscript('egl/main/SConscript')
+    SConscript('talloc/SConscript')
 
-    SConscript('glsl/SConscript')
-    SConscript('mapi/glapi/SConscript')
-    SConscript('mesa/SConscript')
+SConscript('glsl/SConscript')
+SConscript('mapi/glapi/SConscript')
+SConscript('mesa/SConscript')
 
-    if platform != 'embedded':
-        SConscript('glut/glx/SConscript')
+if env['platform'] != 'embedded':
+    SConscript('glut/glx/SConscript')
 
 SConscript('gallium/SConscript')
 
diff --git a/src/egl/main/SConscript b/src/egl/main/SConscript
index 8ade85b357..f001b81600 100644
--- a/src/egl/main/SConscript
+++ b/src/egl/main/SConscript
@@ -4,49 +4,49 @@
 
 Import('*')
 
-if env['platform'] != 'winddk':
-
-	env = env.Clone()
-
-	env.Append(CPPDEFINES = [
-		'_EGL_NATIVE_PLATFORM=_EGL_PLATFORM_WINDOWS',
-		'_EGL_DRIVER_SEARCH_DIR=\\"\\"',
-		'_EGL_OS_WINDOWS',
-		'_EGL_GET_CORE_ADDRESSES',
-		'KHRONOS_DLL_EXPORTS',
-	])
-
-	env.Append(CPPPATH = [
-		'#/include',
-	])
-
-	egl_sources = [
-		'eglapi.c',
-		'eglarray.c',
-		'eglconfig.c',
-		'eglcontext.c',
-		'eglcurrent.c',
-		'egldisplay.c',
-		'egldriver.c',
-		'eglfallbacks.c',
-		'eglglobals.c',
-		'eglimage.c',
-		'egllog.c',
-		'eglmisc.c',
-		'eglmode.c',
-		'eglscreen.c',
-		'eglstring.c',
-		'eglsurface.c',
-		'eglsync.c',
-	]
-
-	egl = env.SharedLibrary(
-		target = 'libEGL',
-		source = egl_sources + ['egl.def'],
-	)
-
-	env.InstallSharedLibrary(egl, version=(1, 4, 0))
-
-	egl = [env.FindIxes(egl, 'LIBPREFIX', 'LIBSUFFIX')]
-
-	Export('egl')
+env = env.Clone()
+
+env.Append(CPPDEFINES = [
+    '_EGL_NATIVE_PLATFORM=_EGL_PLATFORM_WINDOWS',
+    '_EGL_DRIVER_SEARCH_DIR=\\"\\"',
+    '_EGL_OS_WINDOWS',
+    '_EGL_GET_CORE_ADDRESSES',
+    'KHRONOS_DLL_EXPORTS',
+])
+
+env.Append(CPPPATH = [
+    '#/include',
+])
+
+egl_sources = [
+    'eglapi.c',
+    'eglarray.c',
+    'eglconfig.c',
+    'eglcontext.c',
+    'eglcurrent.c',
+    'egldisplay.c',
+    'egldriver.c',
+    'eglfallbacks.c',
+    'eglglobals.c',
+    'eglimage.c',
+    'egllog.c',
+    'eglmisc.c',
+    'eglmode.c',
+    'eglscreen.c',
+    'eglstring.c',
+    'eglsurface.c',
+    'eglsync.c',
+]
+
+egl = env.SharedLibrary(
+    target = 'libEGL',
+    source = egl_sources + ['egl.def'],
+)
+
+installed_egl = env.InstallSharedLibrary(egl, version=(1, 4, 0))
+
+env.Alias('egl', installed_egl)
+
+egl = [env.FindIxes(egl, 'LIBPREFIX', 'LIBSUFFIX')]
+
+Export('egl')
diff --git a/src/egl/main/eglarray.h b/src/egl/main/eglarray.h
index fe92efc11e..c8309fb066 100644
--- a/src/egl/main/eglarray.h
+++ b/src/egl/main/eglarray.h
@@ -37,7 +37,7 @@ void *
 _eglFindArray(_EGLArray *array, void *elem);
 
 
-void **
+PUBLIC void **
 _eglFilterArray(_EGLArray *array, EGLint *size,
                 _EGLArrayForEach filter, void *filter_data);
 
diff --git a/src/egl/main/eglconfig.h b/src/egl/main/eglconfig.h
index 3457670bfa..2169960fd1 100644
--- a/src/egl/main/eglconfig.h
+++ b/src/egl/main/eglconfig.h
@@ -3,6 +3,7 @@
 
 
 #include <assert.h>
+#include <stddef.h>
 #include "egltypedefs.h"
 
 
diff --git a/src/egl/main/eglcurrent.c b/src/egl/main/eglcurrent.c
index c0e8c119a9..cbca9ff2c2 100644
--- a/src/egl/main/eglcurrent.c
+++ b/src/egl/main/eglcurrent.c
@@ -14,33 +14,7 @@
 static _EGLThreadInfo dummy_thread = _EGL_THREAD_INFO_INITIALIZER;
 
 
-#ifdef GLX_USE_TLS
-static __thread const _EGLThreadInfo *_egl_TSD
-   __attribute__ ((tls_model("initial-exec")));
-
-static INLINE void _eglSetTSD(const _EGLThreadInfo *t)
-{
-   _egl_TSD = t;
-}
-
-static INLINE _EGLThreadInfo *_eglGetTSD(void)
-{
-   return (_EGLThreadInfo *) _egl_TSD;
-}
-
-static INLINE void _eglFiniTSD(void)
-{
-}
-
-static INLINE EGLBoolean _eglInitTSD(void (*dtor)(_EGLThreadInfo *))
-{
-   /* TODO destroy TSD */
-   (void) dtor;
-   (void) _eglFiniTSD;
-   return EGL_TRUE;
-}
-
-#elif PTHREADS
+#if PTHREADS
 #include <pthread.h>
 
 static _EGL_DECLARE_MUTEX(_egl_TSDMutex);
@@ -48,14 +22,26 @@ static EGLBoolean _egl_TSDInitialized;
 static pthread_key_t _egl_TSD;
 static void (*_egl_FreeTSD)(_EGLThreadInfo *);
 
+#ifdef GLX_USE_TLS
+static __thread const _EGLThreadInfo *_egl_TLS
+   __attribute__ ((tls_model("initial-exec")));
+#endif
+
 static INLINE void _eglSetTSD(const _EGLThreadInfo *t)
 {
    pthread_setspecific(_egl_TSD, (const void *) t);
+#ifdef GLX_USE_TLS
+   _egl_TLS = t;
+#endif
 }
 
 static INLINE _EGLThreadInfo *_eglGetTSD(void)
 {
+#ifdef GLX_USE_TLS
+   return (_EGLThreadInfo *) _egl_TLS;
+#else
    return (_EGLThreadInfo *) pthread_getspecific(_egl_TSD);
+#endif
 }
 
 static INLINE void _eglFiniTSD(void)
diff --git a/src/egl/main/egldriver.c b/src/egl/main/egldriver.c
index 2359253ff1..ff0011c4b1 100644
--- a/src/egl/main/egldriver.c
+++ b/src/egl/main/egldriver.c
@@ -395,35 +395,62 @@ _eglPreloadForEach(const char *search_path,
 static const char *
 _eglGetSearchPath(void)
 {
-   static const char *search_path;
+   static char search_path[1024];
 
 #if defined(_EGL_OS_UNIX) || defined(_EGL_OS_WINDOWS)
-   if (!search_path) {
-      static char buffer[1024];
-      const char *p;
+   if (search_path[0] == '\0') {
+      char *buf = search_path;
+      size_t len = sizeof(search_path);
+      EGLBoolean use_env;
+      char dir_sep;
       int ret;
 
-      p = getenv("EGL_DRIVERS_PATH");
 #if defined(_EGL_OS_UNIX)
-      if (p && (geteuid() != getuid() || getegid() != getgid())) {
+      use_env = (geteuid() == getuid() && getegid() == getgid());
+      dir_sep = '/';
+#else
+      use_env = EGL_TRUE;
+      dir_sep = '\\';
+#endif
+
+      if (use_env) {
+         char *p;
+
+         /* extract the dirname from EGL_DRIVER */
+         p = getenv("EGL_DRIVER");
+         if (p && strchr(p, dir_sep)) {
+            ret = _eglsnprintf(buf, len, "%s", p);
+            if (ret > 0 && ret < len) {
+               p = strrchr(buf, dir_sep);
+               *p++ = ':';
+
+               len -= p - buf;
+               buf = p;
+            }
+         }
+
+         /* append EGL_DRIVERS_PATH */
+         p = getenv("EGL_DRIVERS_PATH");
+         if (p) {
+            ret = _eglsnprintf(buf, len, "%s:", p);
+            if (ret > 0 && ret < len) {
+               buf += ret;
+               len -= ret;
+            }
+         }
+      }
+      else {
          _eglLog(_EGL_DEBUG,
                "ignore EGL_DRIVERS_PATH for setuid/setgid binaries");
-         p = NULL;
       }
-#endif /* _EGL_OS_UNIX */
 
-      if (p) {
-         ret = _eglsnprintf(buffer, sizeof(buffer),
-               "%s:%s", p, _EGL_DRIVER_SEARCH_DIR);
-         if (ret > 0 && ret < sizeof(buffer))
-            search_path = buffer;
-      }
+      ret = _eglsnprintf(buf, len, "%s", _EGL_DRIVER_SEARCH_DIR);
+      if (ret < 0 || ret >= len)
+         search_path[0] = '\0';
+
+      _eglLog(_EGL_DEBUG, "EGL search path is %s", search_path);
    }
-   if (!search_path)
-      search_path = _EGL_DRIVER_SEARCH_DIR;
-#else
-   search_path = "";
-#endif
+#endif /* defined(_EGL_OS_UNIX) || defined(_EGL_OS_WINDOWS) */
 
    return search_path;
 }
diff --git a/src/egl/main/egllog.c b/src/egl/main/egllog.c
index 8f3bae2243..12c55f901a 100644
--- a/src/egl/main/egllog.c
+++ b/src/egl/main/egllog.c
@@ -151,6 +151,7 @@ _eglLog(EGLint level, const char *fmtStr, ...)
 {
    va_list args;
    char msg[MAXSTRING];
+   int ret;
 
    /* one-time initialization; a little race here is fine */
    if (!logging.initialized)
@@ -162,7 +163,9 @@ _eglLog(EGLint level, const char *fmtStr, ...)
 
    if (logging.logger) {
       va_start(args, fmtStr);
-      vsnprintf(msg, MAXSTRING, fmtStr, args);
+      ret = vsnprintf(msg, MAXSTRING, fmtStr, args);
+      if (ret < 0 || ret >= MAXSTRING)
+         strcpy(msg, "<message truncated>");
       va_end(args);
 
       logging.logger(level, msg);
diff --git a/src/gallium/SConscript b/src/gallium/SConscript
index 3c171552be..b5bc9ecfd9 100644
--- a/src/gallium/SConscript
+++ b/src/gallium/SConscript
@@ -1,30 +1,118 @@
-import os
+Import('env')
 
-Import('*')
+# 
+# Auxiliary modules
+#
 
 SConscript('auxiliary/SConscript')
 
-for driver in env['drivers']:
-	SConscript(os.path.join('drivers', driver, 'SConscript'))
+#
+# Drivers
+#
+
+SConscript([
+    'drivers/failover/SConscript', 
+    'drivers/galahad/SConscript',
+    'drivers/identity/SConscript', 
+    'drivers/llvmpipe/SConscript', 
+    'drivers/rbug/SConscript',
+    'drivers/softpipe/SConscript',
+    'drivers/svga/SConscript', 
+    'drivers/trace/SConscript', 
+])
+
+if not env['msvc']:
+    # These drivers do not build on MSVC compilers
+    SConscript([
+        'drivers/i915/SConscript', 
+        'drivers/i965/SConscript',
+        'drivers/r300/SConscript',
+    ])
+
+if env['dri']:
+    # These drivers depend on drm headers
+    SConscript([
+        # XXX: nouveau drivers have a tight dependency on libdrm, so to enable
+        # we need some version logic before we enable them. Also, ATM there is
+        # no nouveau target in scons
+        #'drivers/nouveau/SConscript', 
+        #'drivers/nv50/SConscript', 
+        #'drivers/nvfx/SConscript',
+        'drivers/r600/SConscript', 
+    ])
+
+#
+# State trackers
+#
 
 # Needed by some state trackers
 SConscript('winsys/sw/null/SConscript')
 
 SConscript('state_trackers/python/SConscript')
-if platform != 'embedded':
+if env['platform'] != 'embedded':
 	SConscript('state_trackers/glx/xlib/SConscript')
 	SConscript('state_trackers/dri/SConscript')
 	SConscript('state_trackers/xorg/SConscript')
 	SConscript('state_trackers/egl/SConscript')
 	SConscript('state_trackers/vega/SConscript')
 
-if platform == 'windows':
+if env['platform'] == 'windows':
 	SConscript('state_trackers/wgl/SConscript')
 
+#
+# Winsys
+# 
+
 SConscript('winsys/SConscript')
 
-SConscript('targets/SConscript')
+#
+# Targets
+#
+
+SConscript([
+    'targets/graw-null/SConscript',
+])
+
+if env['x11']:
+    SConscript([
+        'targets/graw-xlib/SConscript',
+        'targets/libgl-xlib/SConscript',
+    ])
 
-if platform != 'embedded':
+if env['platform'] == 'windows':
+    SConscript([
+        'targets/graw-gdi/SConscript',
+        'targets/libgl-gdi/SConscript',
+        #'egl-gdi/SConscript',
+    ])
+
+if env['dri']:
+    SConscript([
+        'targets/SConscript.dri',
+	    'targets/dri-i915/SConscript',
+	    'targets/dri-i965/SConscript',
+	    #'targets/dri-nouveau/SConscript',
+	    'targets/dri-r300/SConscript',
+        'targets/dri-r600/SConscript',
+	    'targets/dri-swrast/SConscript',
+	    'targets/dri-vmwgfx/SConscript',
+    ])
+
+if env['xorg']:
+    SConscript([
+        #'targets/xorg-i915/SConscript',
+        #'targets/xorg-i965/SConscript',
+        #'targets/xorg-nouveau/SConscript',
+        #'targets/xorg-radeon/SConscript',
+        'targets/xorg-vmwgfx/SConscript',
+    ])
+
+
+#
+# Unit tests & tools
+#
+
+if env['platform'] != 'embedded':
 	SConscript('tests/unit/SConscript')
 	SConscript('tests/graw/SConscript')
+
diff --git a/src/gallium/auxiliary/SConscript b/src/gallium/auxiliary/SConscript
index 94cd74424a..0e5da1374f 100644
--- a/src/gallium/auxiliary/SConscript
+++ b/src/gallium/auxiliary/SConscript
@@ -7,8 +7,6 @@ env.Append(CPPPATH = [
     'util',
 ])
 
-env.Tool('udis86')
-
 env.CodeGenerate(
     target = 'indices/u_indices_gen.c', 
     script = 'indices/u_indices_gen.py', 
@@ -202,39 +200,42 @@ source = [
 ]
 
 if env['llvm']:
+    if env['UDIS86']:
+        env.Append(CPPDEFINES = [('HAVE_UDIS86', '1')])
+
     source += [
-    'gallivm/lp_bld_arit.c',
-    'gallivm/lp_bld_assert.c',
-    'gallivm/lp_bld_bitarit.c',
-    'gallivm/lp_bld_const.c',
-    'gallivm/lp_bld_conv.c',
-    'gallivm/lp_bld_debug.c',
-    'gallivm/lp_bld_flow.c',
-    'gallivm/lp_bld_format_aos.c',
-    'gallivm/lp_bld_format_soa.c',
-    'gallivm/lp_bld_format_yuv.c',
-    'gallivm/lp_bld_gather.c',
-    'gallivm/lp_bld_init.c',
-    'gallivm/lp_bld_intr.c',
-    'gallivm/lp_bld_logic.c',
-    'gallivm/lp_bld_misc.cpp',
-    'gallivm/lp_bld_pack.c',
-    'gallivm/lp_bld_printf.c',
-    'gallivm/lp_bld_quad.c',
-    'gallivm/lp_bld_sample.c',
-    'gallivm/lp_bld_sample_aos.c',
-    'gallivm/lp_bld_sample_soa.c',
-    'gallivm/lp_bld_struct.c',
-    'gallivm/lp_bld_swizzle.c',
-    'gallivm/lp_bld_tgsi_aos.c',
-    'gallivm/lp_bld_tgsi_info.c',
-    'gallivm/lp_bld_tgsi_soa.c',
-    'gallivm/lp_bld_type.c',
-    'draw/draw_llvm.c',
-    'draw/draw_llvm_sample.c',
-    'draw/draw_llvm_translate.c',
-    'draw/draw_pt_fetch_shade_pipeline_llvm.c',
-    'draw/draw_vs_llvm.c'
+        'gallivm/lp_bld_arit.c',
+        'gallivm/lp_bld_assert.c',
+        'gallivm/lp_bld_bitarit.c',
+        'gallivm/lp_bld_const.c',
+        'gallivm/lp_bld_conv.c',
+        'gallivm/lp_bld_debug.c',
+        'gallivm/lp_bld_flow.c',
+        'gallivm/lp_bld_format_aos.c',
+        'gallivm/lp_bld_format_soa.c',
+        'gallivm/lp_bld_format_yuv.c',
+        'gallivm/lp_bld_gather.c',
+        'gallivm/lp_bld_init.c',
+        'gallivm/lp_bld_intr.c',
+        'gallivm/lp_bld_logic.c',
+        'gallivm/lp_bld_misc.cpp',
+        'gallivm/lp_bld_pack.c',
+        'gallivm/lp_bld_printf.c',
+        'gallivm/lp_bld_quad.c',
+        'gallivm/lp_bld_sample.c',
+        'gallivm/lp_bld_sample_aos.c',
+        'gallivm/lp_bld_sample_soa.c',
+        'gallivm/lp_bld_struct.c',
+        'gallivm/lp_bld_swizzle.c',
+        'gallivm/lp_bld_tgsi_aos.c',
+        'gallivm/lp_bld_tgsi_info.c',
+        'gallivm/lp_bld_tgsi_soa.c',
+        'gallivm/lp_bld_type.c',
+        'draw/draw_llvm.c',
+        'draw/draw_llvm_sample.c',
+        'draw/draw_llvm_translate.c',
+        'draw/draw_pt_fetch_shade_pipeline_llvm.c',
+        'draw/draw_vs_llvm.c'
     ]
 
 gallium = env.ConvenienceLibrary(
@@ -242,4 +243,6 @@ gallium = env.ConvenienceLibrary(
     source = source,
 )
 
+env.Alias('gallium', gallium)
+
 Export('gallium')
diff --git a/src/gallium/auxiliary/gallivm/lp_bld_tgsi_soa.c b/src/gallium/auxiliary/gallivm/lp_bld_tgsi_soa.c
index 6188c4fdd1..1de4139363 100644
--- a/src/gallium/auxiliary/gallivm/lp_bld_tgsi_soa.c
+++ b/src/gallium/auxiliary/gallivm/lp_bld_tgsi_soa.c
@@ -59,6 +59,7 @@
 #include "lp_bld_tgsi.h"
 #include "lp_bld_limits.h"
 #include "lp_bld_debug.h"
+#include "lp_bld_printf.h"
 
 
 #define FOR_EACH_CHANNEL( CHAN )\
@@ -120,9 +121,12 @@ struct lp_build_tgsi_soa_context
 {
    struct lp_build_context base;
 
-   /* Builder for integer masks and indices */
+   /* Builder for vector integer masks and indices */
    struct lp_build_context uint_bld;
 
+   /* Builder for scalar elements of shader's data type (float) */
+   struct lp_build_context elem_bld;
+
    LLVMValueRef consts_ptr;
    const LLVMValueRef *pos;
    const LLVMValueRef (*inputs)[NUM_CHANNELS];
@@ -460,7 +464,7 @@ build_gather(struct lp_build_tgsi_soa_context *bld,
       LLVMValueRef index = LLVMBuildExtractElement(bld->base.builder,
                                                    indexes, ii, "");
       LLVMValueRef scalar_ptr = LLVMBuildGEP(bld->base.builder, base_ptr,
-                                             &index, 1, "");
+                                             &index, 1, "gather_ptr");
       LLVMValueRef scalar = LLVMBuildLoad(bld->base.builder, scalar_ptr, "");
 
       res = LLVMBuildInsertElement(bld->base.builder, res, scalar, ii, "");
@@ -471,8 +475,61 @@ build_gather(struct lp_build_tgsi_soa_context *bld,
 
 
 /**
+ * Scatter/store vector.
+ */
+static void
+emit_mask_scatter(struct lp_build_tgsi_soa_context *bld,
+                  LLVMValueRef base_ptr,
+                  LLVMValueRef indexes,
+                  LLVMValueRef values,
+                  struct lp_exec_mask *mask,
+                  LLVMValueRef pred)
+{
+   struct gallivm_state *gallivm = bld->base.gallivm;
+   LLVMBuilderRef builder = bld->base.builder;
+   unsigned i;
+
+   /* Mix the predicate and execution mask */
+   if (mask->has_mask) {
+      if (pred) {
+         pred = LLVMBuildAnd(mask->bld->builder, pred, mask->exec_mask, "");
+      }
+      else {
+         pred = mask->exec_mask;
+      }
+   }
+
+   /*
+    * Loop over elements of index_vec, store scalar value.
+    */
+   for (i = 0; i < bld->base.type.length; i++) {
+      LLVMValueRef ii = lp_build_const_int32(gallivm, i);
+      LLVMValueRef index = LLVMBuildExtractElement(builder, indexes, ii, "");
+      LLVMValueRef scalar_ptr = LLVMBuildGEP(builder, base_ptr, &index, 1, "scatter_ptr");
+      LLVMValueRef val = LLVMBuildExtractElement(builder, values, ii, "scatter_val");
+      LLVMValueRef scalar_pred = pred ?
+         LLVMBuildExtractElement(builder, pred, ii, "scatter_pred") : NULL;
+
+      if (0)
+         lp_build_printf(gallivm, "scatter %d: val %f at %d %p\n",
+                         ii, val, index, scalar_ptr);
+
+      if (scalar_pred) {
+         LLVMValueRef real_val, dst_val;
+         dst_val = LLVMBuildLoad(builder, scalar_ptr, "");
+         real_val = lp_build_select(&bld->elem_bld, scalar_pred, val, dst_val);
+         LLVMBuildStore(builder, real_val, scalar_ptr);
+      }
+      else {
+         LLVMBuildStore(builder, val, scalar_ptr);
+      }
+   }
+}
+
+
+/**
  * Read the current value of the ADDR register, convert the floats to
- * ints, multiply by four and return the vector of offsets.
+ * ints, add the base index and return the vector of offsets.
  * The offsets will be used to index into the constant buffer or
  * temporary register file.
  */
@@ -753,7 +810,9 @@ emit_store(
    LLVMValueRef pred,
    LLVMValueRef value)
 {
+   struct gallivm_state *gallivm = bld->base.gallivm;
    const struct tgsi_full_dst_register *reg = &inst->Dst[index];
+   struct lp_build_context *uint_bld = &bld->uint_bld;
    LLVMValueRef indirect_index = NULL;
 
    switch( inst->Instruction.Saturate ) {
@@ -791,9 +850,40 @@ emit_store(
 
    case TGSI_FILE_TEMPORARY:
       if (reg->Register.Indirect) {
-         /* XXX not done yet */
-         debug_printf("WARNING: LLVM scatter store of temp regs"
-                      " not implemented\n");
+         LLVMBuilderRef builder = bld->base.builder;
+         LLVMValueRef chan_vec =
+            lp_build_const_int_vec(gallivm, uint_bld->type, chan_index);
+         LLVMValueRef length_vec =
+            lp_build_const_int_vec(gallivm, uint_bld->type,
+                                   bld->base.type.length);
+         LLVMValueRef index_vec;  /* indexes into the temp registers */
+         LLVMValueRef temps_array;
+         LLVMValueRef pixel_offsets;
+         LLVMTypeRef float_ptr_type;
+         int i;
+
+         /* build pixel offset vector: {0, 1, 2, 3, ...} */
+         pixel_offsets = uint_bld->undef; 
+         for (i = 0; i < bld->base.type.length; i++) {
+            LLVMValueRef ii = lp_build_const_int32(gallivm, i);
+            pixel_offsets = LLVMBuildInsertElement(builder, pixel_offsets,
+                                                   ii, ii, "");
+         }
+
+         /* index_vec = (indirect_index * 4 + chan_index) * length + offsets */
+         index_vec = lp_build_shl_imm(uint_bld, indirect_index, 2);
+         index_vec = lp_build_add(uint_bld, index_vec, chan_vec);
+         index_vec = lp_build_mul(uint_bld, index_vec, length_vec);
+         index_vec = lp_build_add(uint_bld, index_vec, pixel_offsets);
+
+         float_ptr_type =
+            LLVMPointerType(LLVMFloatTypeInContext(gallivm->context), 0);
+         temps_array = LLVMBuildBitCast(builder, bld->temps_array,
+                                        float_ptr_type, "");
+
+         /* Scatter store values into temp registers */
+         emit_mask_scatter(bld, temps_array, index_vec, value,
+                           &bld->exec_mask, pred);
       }
       else {
          LLVMValueRef temp_ptr = get_temp_ptr(bld, reg->Register.Index,
@@ -1045,6 +1135,53 @@ emit_kilp(struct lp_build_tgsi_soa_context *bld,
       lp_build_mask_check(bld->mask);
 }
 
+
+/**
+ * Emit code which will dump the value of all the temporary registers
+ * to stdout.
+ */
+static void
+emit_dump_temps(struct lp_build_tgsi_soa_context *bld)
+{
+   struct gallivm_state *gallivm = bld->base.gallivm;
+   LLVMBuilderRef builder = gallivm->builder;
+   LLVMValueRef temp_ptr;
+   LLVMValueRef i0 = lp_build_const_int32(gallivm, 0);
+   LLVMValueRef i1 = lp_build_const_int32(gallivm, 1);
+   LLVMValueRef i2 = lp_build_const_int32(gallivm, 2);
+   LLVMValueRef i3 = lp_build_const_int32(gallivm, 3);
+   int index;
+   int n = bld->info->file_max[TGSI_FILE_TEMPORARY];
+
+   for (index = 0; index < n; index++) {
+      LLVMValueRef idx = lp_build_const_int32(gallivm, index);
+      LLVMValueRef v[4][4], res;
+      int chan;
+
+      lp_build_printf(gallivm, "TEMP[%d]:\n", idx);
+
+      for (chan = 0; chan < 4; chan++) {
+         temp_ptr = get_temp_ptr(bld, index, chan);
+         res = LLVMBuildLoad(bld->base.builder, temp_ptr, "");
+         v[chan][0] = LLVMBuildExtractElement(builder, res, i0, "");
+         v[chan][1] = LLVMBuildExtractElement(builder, res, i1, "");
+         v[chan][2] = LLVMBuildExtractElement(builder, res, i2, "");
+         v[chan][3] = LLVMBuildExtractElement(builder, res, i3, "");
+      }
+
+      lp_build_printf(gallivm, "  X: %f %f %f %f\n",
+                      v[0][0], v[0][1], v[0][2], v[0][3]);
+      lp_build_printf(gallivm, "  Y: %f %f %f %f\n",
+                      v[1][0], v[1][1], v[1][2], v[1][3]);
+      lp_build_printf(gallivm, "  Z: %f %f %f %f\n",
+                      v[2][0], v[2][1], v[2][2], v[2][3]);
+      lp_build_printf(gallivm, "  W: %f %f %f %f\n",
+                      v[3][0], v[3][1], v[3][2], v[3][3]);
+   }
+}
+
+
+
 static void
 emit_declaration(
    struct lp_build_tgsi_soa_context *bld,
@@ -1064,29 +1201,31 @@ emit_declaration(
          assert(idx < LP_MAX_TGSI_TEMPS);
          if (bld->indirect_files & (1 << TGSI_FILE_TEMPORARY)) {
             LLVMValueRef array_size = lp_build_const_int32(gallivm, last * 4 + 4);
-            bld->temps_array = lp_build_array_alloca(gallivm,
-                                                     vec_type, array_size, "");
+            bld->temps_array = lp_build_array_alloca(gallivm, vec_type,
+                                                     array_size, "temparray");
          } else {
             for (i = 0; i < NUM_CHANNELS; i++)
-               bld->temps[idx][i] = lp_build_alloca(gallivm, vec_type, "");
+               bld->temps[idx][i] = lp_build_alloca(gallivm, vec_type, "temp");
          }
          break;
 
       case TGSI_FILE_OUTPUT:
          for (i = 0; i < NUM_CHANNELS; i++)
-            bld->outputs[idx][i] = lp_build_alloca(gallivm, vec_type, "");
+            bld->outputs[idx][i] = lp_build_alloca(gallivm, vec_type,
+                                                   "output");
          break;
 
       case TGSI_FILE_ADDRESS:
          assert(idx < LP_MAX_TGSI_ADDRS);
          for (i = 0; i < NUM_CHANNELS; i++)
-            bld->addr[idx][i] = lp_build_alloca(gallivm, vec_type, "");
+            bld->addr[idx][i] = lp_build_alloca(gallivm, vec_type, "addr");
          break;
 
       case TGSI_FILE_PREDICATE:
          assert(idx < LP_MAX_TGSI_PREDS);
          for (i = 0; i < NUM_CHANNELS; i++)
-            bld->preds[idx][i] = lp_build_alloca(gallivm, vec_type, "");
+            bld->preds[idx][i] = lp_build_alloca(gallivm, vec_type,
+                                                 "predicate");
          break;
 
       default:
@@ -1741,6 +1880,10 @@ emit_instruction(
       break;
 
    case TGSI_OPCODE_END:
+      if (0) {
+         /* for debugging */
+         emit_dump_temps(bld);
+      }
       *pc = -1;
       break;
 
@@ -2083,6 +2226,7 @@ lp_build_tgsi_soa(struct gallivm_state *gallivm,
    memset(&bld, 0, sizeof bld);
    lp_build_context_init(&bld.base, gallivm, type);
    lp_build_context_init(&bld.uint_bld, gallivm, lp_uint_type(type));
+   lp_build_context_init(&bld.elem_bld, gallivm, lp_elem_type(type));
    bld.mask = mask;
    bld.pos = pos;
    bld.inputs = inputs;
diff --git a/src/gallium/auxiliary/gallivm/lp_bld_type.c b/src/gallium/auxiliary/gallivm/lp_bld_type.c
index bfc123305c..ee61646766 100644
--- a/src/gallium/auxiliary/gallivm/lp_bld_type.c
+++ b/src/gallium/auxiliary/gallivm/lp_bld_type.c
@@ -189,6 +189,22 @@ lp_build_int32_vec4_type(struct gallivm_state *gallivm)
 
 
 /**
+ * Create element of vector type
+ */
+struct lp_type
+lp_elem_type(struct lp_type type)
+{
+   struct lp_type res_type;
+
+   assert(type.length > 1);
+   res_type = type;
+   res_type.length = 1;
+
+   return res_type;
+}
+
+
+/**
  * Create unsigned integer type variation of given type.
  */
 struct lp_type
diff --git a/src/gallium/auxiliary/gallivm/lp_bld_type.h b/src/gallium/auxiliary/gallivm/lp_bld_type.h
index 2edae8b3d1..e62e90d638 100644
--- a/src/gallium/auxiliary/gallivm/lp_bld_type.h
+++ b/src/gallium/auxiliary/gallivm/lp_bld_type.h
@@ -367,6 +367,10 @@ lp_unorm8_vec4_type(void)
 
 
 struct lp_type
+lp_elem_type(struct lp_type type);
+
+
+struct lp_type
 lp_uint_type(struct lp_type type);
 
 
diff --git a/src/gallium/auxiliary/tgsi/tgsi_build.c b/src/gallium/auxiliary/tgsi/tgsi_build.c
index 6dbedf15ca..16a205f206 100644
--- a/src/gallium/auxiliary/tgsi/tgsi_build.c
+++ b/src/gallium/auxiliary/tgsi/tgsi_build.c
@@ -64,25 +64,14 @@ header_bodysize_grow( struct tgsi_header *header )
 }
 
 struct tgsi_processor
-tgsi_default_processor( void )
-{
-   struct tgsi_processor processor;
-
-   processor.Processor = TGSI_PROCESSOR_FRAGMENT;
-   processor.Padding = 0;
-
-   return processor;
-}
-
-struct tgsi_processor
 tgsi_build_processor(
    unsigned type,
    struct tgsi_header *header )
 {
    struct tgsi_processor processor;
 
-   processor = tgsi_default_processor();
    processor.Processor = type;
+   processor.Padding = 0;
 
    header_headersize_grow( header );
 
@@ -93,7 +82,19 @@ tgsi_build_processor(
  * declaration
  */
 
-struct tgsi_declaration
+static void
+declaration_grow(
+   struct tgsi_declaration *declaration,
+   struct tgsi_header *header )
+{
+   assert( declaration->NrTokens < 0xFF );
+
+   declaration->NrTokens++;
+
+   header_bodysize_grow( header );
+}
+
+static struct tgsi_declaration
 tgsi_default_declaration( void )
 {
    struct tgsi_declaration declaration;
@@ -112,7 +113,7 @@ tgsi_default_declaration( void )
    return declaration;
 }
 
-struct tgsi_declaration
+static struct tgsi_declaration
 tgsi_build_declaration(
    unsigned file,
    unsigned usage_mask,
@@ -144,16 +145,85 @@ tgsi_build_declaration(
    return declaration;
 }
 
-static void
-declaration_grow(
+static struct tgsi_declaration_range
+tgsi_default_declaration_range( void )
+{
+   struct tgsi_declaration_range dr;
+
+   dr.First = 0;
+   dr.Last = 0;
+
+   return dr;
+}
+
+static struct tgsi_declaration_range
+tgsi_build_declaration_range(
+   unsigned first,
+   unsigned last,
    struct tgsi_declaration *declaration,
    struct tgsi_header *header )
 {
-   assert( declaration->NrTokens < 0xFF );
+   struct tgsi_declaration_range declaration_range;
 
-   declaration->NrTokens++;
+   assert( last >= first );
+   assert( last <= 0xFFFF );
 
-   header_bodysize_grow( header );
+   declaration_range.First = first;
+   declaration_range.Last = last;
+
+   declaration_grow( declaration, header );
+
+   return declaration_range;
+}
+
+static struct tgsi_declaration_dimension
+tgsi_build_declaration_dimension(unsigned index_2d,
+                                 struct tgsi_declaration *declaration,
+                                 struct tgsi_header *header)
+{
+   struct tgsi_declaration_dimension dd;
+
+   assert(index_2d <= 0xFFFF);
+
+   dd.Index2D = index_2d;
+   dd.Padding = 0;
+
+   declaration_grow(declaration, header);
+
+   return dd;
+}
+
+static struct tgsi_declaration_semantic
+tgsi_default_declaration_semantic( void )
+{
+   struct tgsi_declaration_semantic ds;
+
+   ds.Name = TGSI_SEMANTIC_POSITION;
+   ds.Index = 0;
+   ds.Padding = 0;
+
+   return ds;
+}
+
+static struct tgsi_declaration_semantic
+tgsi_build_declaration_semantic(
+   unsigned semantic_name,
+   unsigned semantic_index,
+   struct tgsi_declaration *declaration,
+   struct tgsi_header *header )
+{
+   struct tgsi_declaration_semantic ds;
+
+   assert( semantic_name <= TGSI_SEMANTIC_COUNT );
+   assert( semantic_index <= 0xFFFF );
+
+   ds.Name = semantic_name;
+   ds.Index = semantic_index;
+   ds.Padding = 0;
+
+   declaration_grow( declaration, header );
+
+   return ds;
 }
 
 struct tgsi_full_declaration
@@ -257,104 +327,11 @@ tgsi_build_full_declaration(
    return size;
 }
 
-struct tgsi_declaration_range
-tgsi_default_declaration_range( void )
-{
-   struct tgsi_declaration_range dr;
-
-   dr.First = 0;
-   dr.Last = 0;
-
-   return dr;
-}
-
-struct tgsi_declaration_range
-tgsi_build_declaration_range(
-   unsigned first,
-   unsigned last,
-   struct tgsi_declaration *declaration,
-   struct tgsi_header *header )
-{
-   struct tgsi_declaration_range declaration_range;
-
-   assert( last >= first );
-   assert( last <= 0xFFFF );
-
-   declaration_range = tgsi_default_declaration_range();
-   declaration_range.First = first;
-   declaration_range.Last = last;
-
-   declaration_grow( declaration, header );
-
-   return declaration_range;
-}
-
-struct tgsi_declaration_dimension
-tgsi_default_declaration_dimension(void)
-{
-   struct tgsi_declaration_dimension dd;
-
-   dd.Index2D = 0;
-   dd.Padding = 0;
-
-   return dd;
-}
-
-struct tgsi_declaration_dimension
-tgsi_build_declaration_dimension(unsigned index_2d,
-                                 struct tgsi_declaration *declaration,
-                                 struct tgsi_header *header)
-{
-   struct tgsi_declaration_dimension dd;
-
-   assert(index_2d <= 0xFFFF);
-
-   dd = tgsi_default_declaration_dimension();
-   dd.Index2D = index_2d;
-
-   declaration_grow(declaration, header);
-
-   return dd;
-}
-
-struct tgsi_declaration_semantic
-tgsi_default_declaration_semantic( void )
-{
-   struct tgsi_declaration_semantic ds;
-
-   ds.Name = TGSI_SEMANTIC_POSITION;
-   ds.Index = 0;
-   ds.Padding = 0;
-
-   return ds;
-}
-
-struct tgsi_declaration_semantic
-tgsi_build_declaration_semantic(
-   unsigned semantic_name,
-   unsigned semantic_index,
-   struct tgsi_declaration *declaration,
-   struct tgsi_header *header )
-{
-   struct tgsi_declaration_semantic ds;
-
-   assert( semantic_name <= TGSI_SEMANTIC_COUNT );
-   assert( semantic_index <= 0xFFFF );
-
-   ds = tgsi_default_declaration_semantic();
-   ds.Name = semantic_name;
-   ds.Index = semantic_index;
-
-   declaration_grow( declaration, header );
-
-   return ds;
-}
-
 /*
  * immediate
  */
 
-struct tgsi_immediate
+static struct tgsi_immediate
 tgsi_default_immediate( void )
 {
    struct tgsi_immediate immediate;
@@ -367,7 +344,7 @@ tgsi_default_immediate( void )
    return immediate;
 }
 
-struct tgsi_immediate
+static struct tgsi_immediate
 tgsi_build_immediate(
    struct tgsi_header *header )
 {
@@ -406,7 +383,7 @@ immediate_grow(
    header_bodysize_grow( header );
 }
 
-union tgsi_immediate_data
+static union tgsi_immediate_data
 tgsi_build_immediate_float32(
    float value,
    struct tgsi_immediate *immediate,
@@ -480,7 +457,7 @@ tgsi_default_instruction( void )
    return instruction;
 }
 
-struct tgsi_instruction
+static struct tgsi_instruction
 tgsi_build_instruction(unsigned opcode,
                        unsigned saturate,
                        unsigned predicate,
@@ -519,6 +496,266 @@ instruction_grow(
    header_bodysize_grow( header );
 }
 
+struct tgsi_instruction_predicate
+tgsi_default_instruction_predicate(void)
+{
+   struct tgsi_instruction_predicate instruction_predicate;
+
+   instruction_predicate.SwizzleX = TGSI_SWIZZLE_X;
+   instruction_predicate.SwizzleY = TGSI_SWIZZLE_Y;
+   instruction_predicate.SwizzleZ = TGSI_SWIZZLE_Z;
+   instruction_predicate.SwizzleW = TGSI_SWIZZLE_W;
+   instruction_predicate.Negate = 0;
+   instruction_predicate.Index = 0;
+   instruction_predicate.Padding = 0;
+
+   return instruction_predicate;
+}
+
+static struct tgsi_instruction_predicate
+tgsi_build_instruction_predicate(int index,
+                                 unsigned negate,
+                                 unsigned swizzleX,
+                                 unsigned swizzleY,
+                                 unsigned swizzleZ,
+                                 unsigned swizzleW,
+                                 struct tgsi_instruction *instruction,
+                                 struct tgsi_header *header)
+{
+   struct tgsi_instruction_predicate instruction_predicate;
+
+   instruction_predicate = tgsi_default_instruction_predicate();
+   instruction_predicate.SwizzleX = swizzleX;
+   instruction_predicate.SwizzleY = swizzleY;
+   instruction_predicate.SwizzleZ = swizzleZ;
+   instruction_predicate.SwizzleW = swizzleW;
+   instruction_predicate.Negate = negate;
+   instruction_predicate.Index = index;
+
+   instruction_grow(instruction, header);
+
+   return instruction_predicate;
+}
+
+static struct tgsi_instruction_label
+tgsi_default_instruction_label( void )
+{
+   struct tgsi_instruction_label instruction_label;
+
+   instruction_label.Label = 0;
+   instruction_label.Padding = 0;
+
+   return instruction_label;
+}
+
+static struct tgsi_instruction_label
+tgsi_build_instruction_label(
+   unsigned label,
+   struct tgsi_token  *prev_token,
+   struct tgsi_instruction *instruction,
+   struct tgsi_header *header )
+{
+   struct tgsi_instruction_label instruction_label;
+
+   instruction_label.Label = label;
+   instruction_label.Padding = 0;
+   instruction->Label = 1;
+
+   instruction_grow( instruction, header );
+
+   return instruction_label;
+}
+
+static struct tgsi_instruction_texture
+tgsi_default_instruction_texture( void )
+{
+   struct tgsi_instruction_texture instruction_texture;
+
+   instruction_texture.Texture = TGSI_TEXTURE_UNKNOWN;
+   instruction_texture.Padding = 0;
+
+   return instruction_texture;
+}
+
+static struct tgsi_instruction_texture
+tgsi_build_instruction_texture(
+   unsigned texture,
+   struct tgsi_token *prev_token,
+   struct tgsi_instruction *instruction,
+   struct tgsi_header *header )
+{
+   struct tgsi_instruction_texture instruction_texture;
+
+   instruction_texture.Texture = texture;
+   instruction_texture.Padding = 0;
+   instruction->Texture = 1;
+
+   instruction_grow( instruction, header );
+
+   return instruction_texture;
+}
+
+static struct tgsi_src_register
+tgsi_default_src_register( void )
+{
+   struct tgsi_src_register src_register;
+
+   src_register.File = TGSI_FILE_NULL;
+   src_register.SwizzleX = TGSI_SWIZZLE_X;
+   src_register.SwizzleY = TGSI_SWIZZLE_Y;
+   src_register.SwizzleZ = TGSI_SWIZZLE_Z;
+   src_register.SwizzleW = TGSI_SWIZZLE_W;
+   src_register.Negate = 0;
+   src_register.Absolute = 0;
+   src_register.Indirect = 0;
+   src_register.Dimension = 0;
+   src_register.Index = 0;
+
+   return src_register;
+}
+
+static struct tgsi_src_register
+tgsi_build_src_register(
+   unsigned file,
+   unsigned swizzle_x,
+   unsigned swizzle_y,
+   unsigned swizzle_z,
+   unsigned swizzle_w,
+   unsigned negate,
+   unsigned absolute,
+   unsigned indirect,
+   unsigned dimension,
+   int index,
+   struct tgsi_instruction *instruction,
+   struct tgsi_header *header )
+{
+   struct tgsi_src_register   src_register;
+
+   assert( file < TGSI_FILE_COUNT );
+   assert( swizzle_x <= TGSI_SWIZZLE_W );
+   assert( swizzle_y <= TGSI_SWIZZLE_W );
+   assert( swizzle_z <= TGSI_SWIZZLE_W );
+   assert( swizzle_w <= TGSI_SWIZZLE_W );
+   assert( negate <= 1 );
+   assert( index >= -0x8000 && index <= 0x7FFF );
+
+   src_register.File = file;
+   src_register.SwizzleX = swizzle_x;
+   src_register.SwizzleY = swizzle_y;
+   src_register.SwizzleZ = swizzle_z;
+   src_register.SwizzleW = swizzle_w;
+   src_register.Negate = negate;
+   src_register.Absolute = absolute;
+   src_register.Indirect = indirect;
+   src_register.Dimension = dimension;
+   src_register.Index = index;
+
+   instruction_grow( instruction, header );
+
+   return src_register;
+}
+
+static struct tgsi_dimension
+tgsi_default_dimension( void )
+{
+   struct tgsi_dimension dimension;
+
+   dimension.Indirect = 0;
+   dimension.Dimension = 0;
+   dimension.Padding = 0;
+   dimension.Index = 0;
+
+   return dimension;
+}
+
+static struct tgsi_full_src_register
+tgsi_default_full_src_register( void )
+{
+   struct tgsi_full_src_register full_src_register;
+
+   full_src_register.Register = tgsi_default_src_register();
+   full_src_register.Indirect = tgsi_default_src_register();
+   full_src_register.Dimension = tgsi_default_dimension();
+   full_src_register.DimIndirect = tgsi_default_src_register();
+
+   return full_src_register;
+}
+
+static struct tgsi_dimension
+tgsi_build_dimension(
+   unsigned indirect,
+   unsigned index,
+   struct tgsi_instruction *instruction,
+   struct tgsi_header *header )
+{
+   struct tgsi_dimension dimension;
+
+   dimension.Indirect = indirect;
+   dimension.Dimension = 0;
+   dimension.Padding = 0;
+   dimension.Index = index;
+
+   instruction_grow( instruction, header );
+
+   return dimension;
+}
+
+static struct tgsi_dst_register
+tgsi_default_dst_register( void )
+{
+   struct tgsi_dst_register dst_register;
+
+   dst_register.File = TGSI_FILE_NULL;
+   dst_register.WriteMask = TGSI_WRITEMASK_XYZW;
+   dst_register.Indirect = 0;
+   dst_register.Dimension = 0;
+   dst_register.Index = 0;
+   dst_register.Padding = 0;
+
+   return dst_register;
+}
+
+static struct tgsi_dst_register
+tgsi_build_dst_register(
+   unsigned file,
+   unsigned mask,
+   unsigned indirect,
+   unsigned dimension,
+   int index,
+   struct tgsi_instruction *instruction,
+   struct tgsi_header *header )
+{
+   struct tgsi_dst_register dst_register;
+
+   assert( file < TGSI_FILE_COUNT );
+   assert( mask <= TGSI_WRITEMASK_XYZW );
+   assert( index >= -32768 && index <= 32767 );
+
+   dst_register.File = file;
+   dst_register.WriteMask = mask;
+   dst_register.Indirect = indirect;
+   dst_register.Dimension = dimension;
+   dst_register.Index = index;
+   dst_register.Padding = 0;
+
+   instruction_grow( instruction, header );
+
+   return dst_register;
+}
+
+static struct tgsi_full_dst_register
+tgsi_default_full_dst_register( void )
+{
+   struct tgsi_full_dst_register full_dst_register;
+
+   full_dst_register.Register = tgsi_default_dst_register();
+   full_dst_register.Indirect = tgsi_default_src_register();
+   full_dst_register.Dimension = tgsi_default_dimension();
+   full_dst_register.DimIndirect = tgsi_default_src_register();
+
+   return full_dst_register;
+}
+
 struct tgsi_full_instruction
 tgsi_default_full_instruction( void )
 {
@@ -794,268 +1031,7 @@ tgsi_build_full_instruction(
    return size;
 }
 
-struct tgsi_instruction_predicate
-tgsi_default_instruction_predicate(void)
-{
-   struct tgsi_instruction_predicate instruction_predicate;
-
-   instruction_predicate.SwizzleX = TGSI_SWIZZLE_X;
-   instruction_predicate.SwizzleY = TGSI_SWIZZLE_Y;
-   instruction_predicate.SwizzleZ = TGSI_SWIZZLE_Z;
-   instruction_predicate.SwizzleW = TGSI_SWIZZLE_W;
-   instruction_predicate.Negate = 0;
-   instruction_predicate.Index = 0;
-   instruction_predicate.Padding = 0;
-
-   return instruction_predicate;
-}
-
-struct tgsi_instruction_predicate
-tgsi_build_instruction_predicate(int index,
-                                 unsigned negate,
-                                 unsigned swizzleX,
-                                 unsigned swizzleY,
-                                 unsigned swizzleZ,
-                                 unsigned swizzleW,
-                                 struct tgsi_instruction *instruction,
-                                 struct tgsi_header *header)
-{
-   struct tgsi_instruction_predicate instruction_predicate;
-
-   instruction_predicate = tgsi_default_instruction_predicate();
-   instruction_predicate.SwizzleX = swizzleX;
-   instruction_predicate.SwizzleY = swizzleY;
-   instruction_predicate.SwizzleZ = swizzleZ;
-   instruction_predicate.SwizzleW = swizzleW;
-   instruction_predicate.Negate = negate;
-   instruction_predicate.Index = index;
-
-   instruction_grow(instruction, header);
-
-   return instruction_predicate;
-}
-
-struct tgsi_instruction_label
-tgsi_default_instruction_label( void )
-{
-   struct tgsi_instruction_label instruction_label;
-
-   instruction_label.Label = 0;
-   instruction_label.Padding = 0;
-
-   return instruction_label;
-}
-
-struct tgsi_instruction_label
-tgsi_build_instruction_label(
-   unsigned label,
-   struct tgsi_token  *prev_token,
-   struct tgsi_instruction *instruction,
-   struct tgsi_header *header )
-{
-   struct tgsi_instruction_label instruction_label;
-
-   instruction_label = tgsi_default_instruction_label();
-   instruction_label.Label = label;
-   instruction->Label = 1;
-
-   instruction_grow( instruction, header );
-
-   return instruction_label;
-}
-
-struct tgsi_instruction_texture
-tgsi_default_instruction_texture( void )
-{
-   struct tgsi_instruction_texture instruction_texture;
-
-   instruction_texture.Texture = TGSI_TEXTURE_UNKNOWN;
-   instruction_texture.Padding = 0;
-
-   return instruction_texture;
-}
-
-struct tgsi_instruction_texture
-tgsi_build_instruction_texture(
-   unsigned texture,
-   struct tgsi_token *prev_token,
-   struct tgsi_instruction *instruction,
-   struct tgsi_header *header )
-{
-   struct tgsi_instruction_texture instruction_texture;
-
-   instruction_texture = tgsi_default_instruction_texture();
-   instruction_texture.Texture = texture;
-   instruction->Texture = 1;
-
-   instruction_grow( instruction, header );
-
-   return instruction_texture;
-}
-
-struct tgsi_src_register
-tgsi_default_src_register( void )
-{
-   struct tgsi_src_register src_register;
-
-   src_register.File = TGSI_FILE_NULL;
-   src_register.SwizzleX = TGSI_SWIZZLE_X;
-   src_register.SwizzleY = TGSI_SWIZZLE_Y;
-   src_register.SwizzleZ = TGSI_SWIZZLE_Z;
-   src_register.SwizzleW = TGSI_SWIZZLE_W;
-   src_register.Negate = 0;
-   src_register.Absolute = 0;
-   src_register.Indirect = 0;
-   src_register.Dimension = 0;
-   src_register.Index = 0;
-
-   return src_register;
-}
-
-struct tgsi_src_register
-tgsi_build_src_register(
-   unsigned file,
-   unsigned swizzle_x,
-   unsigned swizzle_y,
-   unsigned swizzle_z,
-   unsigned swizzle_w,
-   unsigned negate,
-   unsigned absolute,
-   unsigned indirect,
-   unsigned dimension,
-   int index,
-   struct tgsi_instruction *instruction,
-   struct tgsi_header *header )
-{
-   struct tgsi_src_register   src_register;
-
-   assert( file < TGSI_FILE_COUNT );
-   assert( swizzle_x <= TGSI_SWIZZLE_W );
-   assert( swizzle_y <= TGSI_SWIZZLE_W );
-   assert( swizzle_z <= TGSI_SWIZZLE_W );
-   assert( swizzle_w <= TGSI_SWIZZLE_W );
-   assert( negate <= 1 );
-   assert( index >= -0x8000 && index <= 0x7FFF );
-
-   src_register = tgsi_default_src_register();
-   src_register.File = file;
-   src_register.SwizzleX = swizzle_x;
-   src_register.SwizzleY = swizzle_y;
-   src_register.SwizzleZ = swizzle_z;
-   src_register.SwizzleW = swizzle_w;
-   src_register.Negate = negate;
-   src_register.Absolute = absolute;
-   src_register.Indirect = indirect;
-   src_register.Dimension = dimension;
-   src_register.Index = index;
-
-   instruction_grow( instruction, header );
-
-   return src_register;
-}
-
-struct tgsi_full_src_register
-tgsi_default_full_src_register( void )
-{
-   struct tgsi_full_src_register full_src_register;
-
-   full_src_register.Register = tgsi_default_src_register();
-   full_src_register.Indirect = tgsi_default_src_register();
-   full_src_register.Dimension = tgsi_default_dimension();
-   full_src_register.DimIndirect = tgsi_default_src_register();
-
-   return full_src_register;
-}
-
-
-struct tgsi_dimension
-tgsi_default_dimension( void )
-{
-   struct tgsi_dimension dimension;
-
-   dimension.Indirect = 0;
-   dimension.Dimension = 0;
-   dimension.Padding = 0;
-   dimension.Index = 0;
-
-   return dimension;
-}
-
-struct tgsi_dimension
-tgsi_build_dimension(
-   unsigned indirect,
-   unsigned index,
-   struct tgsi_instruction *instruction,
-   struct tgsi_header *header )
-{
-   struct tgsi_dimension dimension;
-
-   dimension = tgsi_default_dimension();
-   dimension.Indirect = indirect;
-   dimension.Index = index;
-
-   instruction_grow( instruction, header );
-
-   return dimension;
-}
-
-struct tgsi_dst_register
-tgsi_default_dst_register( void )
-{
-   struct tgsi_dst_register dst_register;
-
-   dst_register.File = TGSI_FILE_NULL;
-   dst_register.WriteMask = TGSI_WRITEMASK_XYZW;
-   dst_register.Indirect = 0;
-   dst_register.Dimension = 0;
-   dst_register.Index = 0;
-   dst_register.Padding = 0;
-
-   return dst_register;
-}
-
-struct tgsi_dst_register
-tgsi_build_dst_register(
-   unsigned file,
-   unsigned mask,
-   unsigned indirect,
-   unsigned dimension,
-   int index,
-   struct tgsi_instruction *instruction,
-   struct tgsi_header *header )
-{
-   struct tgsi_dst_register dst_register;
-
-   assert( file < TGSI_FILE_COUNT );
-   assert( mask <= TGSI_WRITEMASK_XYZW );
-   assert( index >= -32768 && index <= 32767 );
-
-   dst_register = tgsi_default_dst_register();
-   dst_register.File = file;
-   dst_register.WriteMask = mask;
-   dst_register.Index = index;
-   dst_register.Indirect = indirect;
-   dst_register.Dimension = dimension;
-
-   instruction_grow( instruction, header );
-
-   return dst_register;
-}
-
-struct tgsi_full_dst_register
-tgsi_default_full_dst_register( void )
-{
-   struct tgsi_full_dst_register full_dst_register;
-
-   full_dst_register.Register = tgsi_default_dst_register();
-   full_dst_register.Indirect = tgsi_default_src_register();
-   full_dst_register.Dimension = tgsi_default_dimension();
-   full_dst_register.DimIndirect = tgsi_default_src_register();
-
-   return full_dst_register;
-}
-
-struct tgsi_property
+static struct tgsi_property
 tgsi_default_property( void )
 {
    struct tgsi_property property;
@@ -1068,7 +1044,7 @@ tgsi_default_property( void )
    return property;
 }
 
-struct tgsi_property
+static struct tgsi_property
 tgsi_build_property(unsigned property_name,
                     struct tgsi_header *header)
 {
@@ -1107,7 +1083,7 @@ property_grow(
    header_bodysize_grow( header );
 }
 
-struct tgsi_property_data
+static struct tgsi_property_data
 tgsi_build_property_data(
    unsigned value,
    struct tgsi_property *property,
diff --git a/src/gallium/auxiliary/tgsi/tgsi_build.h b/src/gallium/auxiliary/tgsi/tgsi_build.h
index 112107a088..3f236a9c24 100644
--- a/src/gallium/auxiliary/tgsi/tgsi_build.h
+++ b/src/gallium/auxiliary/tgsi/tgsi_build.h
@@ -45,9 +45,6 @@ struct tgsi_header
 tgsi_build_header( void );
 
 struct tgsi_processor
-tgsi_default_processor( void );
-
-struct tgsi_processor
 tgsi_build_processor(
    unsigned processor,
    struct tgsi_header *header );
@@ -56,21 +53,6 @@ tgsi_build_processor(
  * declaration
  */
 
-struct tgsi_declaration
-tgsi_default_declaration( void );
-
-struct tgsi_declaration
-tgsi_build_declaration(
-   unsigned file,
-   unsigned usage_mask,
-   unsigned interpolate,
-   unsigned dimension,
-   unsigned semantic,
-   unsigned centroid,
-   unsigned invariant,
-   unsigned cylindrical_wrap,
-   struct tgsi_header *header );
-
 struct tgsi_full_declaration
 tgsi_default_full_declaration( void );
 
@@ -81,54 +63,13 @@ tgsi_build_full_declaration(
    struct tgsi_header *header,
    unsigned maxsize );
 
-struct tgsi_declaration_range
-tgsi_default_declaration_range( void );
-
-struct tgsi_declaration_range
-tgsi_build_declaration_range(
-   unsigned first,
-   unsigned last,
-   struct tgsi_declaration *declaration,
-   struct tgsi_header *header );
-
-struct tgsi_declaration_dimension
-tgsi_default_declaration_dimension(void);
-
-struct tgsi_declaration_dimension
-tgsi_build_declaration_dimension(unsigned index_2d,
-                                 struct tgsi_declaration *declaration,
-                                 struct tgsi_header *header);
-
-struct tgsi_declaration_semantic
-tgsi_default_declaration_semantic( void );
-
-struct tgsi_declaration_semantic
-tgsi_build_declaration_semantic(
-   unsigned semantic_name,
-   unsigned semantic_index,
-   struct tgsi_declaration *declaration,
-   struct tgsi_header *header );
-
 /*
  * immediate
  */
 
-struct tgsi_immediate
-tgsi_default_immediate( void );
-
-struct tgsi_immediate
-tgsi_build_immediate(
-   struct tgsi_header *header );
-
 struct tgsi_full_immediate
 tgsi_default_full_immediate( void );
 
-union tgsi_immediate_data
-tgsi_build_immediate_float32(
-   float value,
-   struct tgsi_immediate *immediate,
-   struct tgsi_header *header );
-
 unsigned
 tgsi_build_full_immediate(
    const struct tgsi_full_immediate *full_imm,
@@ -140,23 +81,9 @@ tgsi_build_full_immediate(
  * properties
  */
 
-struct tgsi_property
-tgsi_default_property( void );
-
-struct tgsi_property
-tgsi_build_property(
-   unsigned property_name,
-   struct tgsi_header *header );
-
 struct tgsi_full_property
 tgsi_default_full_property( void );
 
-struct tgsi_property_data
-tgsi_build_property_data(
-   unsigned value,
-   struct tgsi_property *property,
-   struct tgsi_header *header );
-
 unsigned
 tgsi_build_full_property(
    const struct tgsi_full_property *full_prop,
@@ -171,15 +98,6 @@ tgsi_build_full_property(
 struct tgsi_instruction
 tgsi_default_instruction( void );
 
-struct tgsi_instruction
-tgsi_build_instruction(
-   unsigned opcode,
-   unsigned saturate,
-   unsigned predicate,
-   unsigned num_dst_regs,
-   unsigned num_src_regs,
-   struct tgsi_header *header );
-
 struct tgsi_full_instruction
 tgsi_default_full_instruction( void );
 
@@ -193,85 +111,6 @@ tgsi_build_full_instruction(
 struct tgsi_instruction_predicate
 tgsi_default_instruction_predicate(void);
 
-struct tgsi_instruction_predicate
-tgsi_build_instruction_predicate(int index,
-                                 unsigned negate,
-                                 unsigned swizzleX,
-                                 unsigned swizzleY,
-                                 unsigned swizzleZ,
-                                 unsigned swizzleW,
-                                 struct tgsi_instruction *instruction,
-                                 struct tgsi_header *header);
-
-struct tgsi_instruction_label
-tgsi_default_instruction_label( void );
-
-struct tgsi_instruction_label
-tgsi_build_instruction_label(
-   unsigned label,
-   struct tgsi_token *prev_token,
-   struct tgsi_instruction *instruction,
-   struct tgsi_header *header );
-
-struct tgsi_instruction_texture
-tgsi_default_instruction_texture( void );
-
-struct tgsi_instruction_texture
-tgsi_build_instruction_texture(
-   unsigned texture,
-   struct tgsi_token *prev_token,
-   struct tgsi_instruction *instruction,
-   struct tgsi_header *header );
-
-struct tgsi_src_register
-tgsi_default_src_register( void );
-
-struct tgsi_src_register
-tgsi_build_src_register(
-   unsigned file,
-   unsigned swizzle_x,
-   unsigned swizzle_y,
-   unsigned swizzle_z,
-   unsigned swizzle_w,
-   unsigned negate,
-   unsigned absolute,
-   unsigned indirect,
-   unsigned dimension,
-   int index,
-   struct tgsi_instruction *instruction,
-   struct tgsi_header *header );
-
-struct tgsi_full_src_register
-tgsi_default_full_src_register( void );
-
-
-struct tgsi_dimension
-tgsi_default_dimension( void );
-
-struct tgsi_dimension
-tgsi_build_dimension(
-   unsigned indirect,
-   unsigned index,
-   struct tgsi_instruction *instruction,
-   struct tgsi_header *header );
-
-struct tgsi_dst_register
-tgsi_default_dst_register( void );
-
-struct tgsi_dst_register
-tgsi_build_dst_register(
-   unsigned file,
-   unsigned mask,
-   unsigned indirect,
-   unsigned dimension,
-   int index,
-   struct tgsi_instruction *instruction,
-   struct tgsi_header *header );
-
-struct tgsi_full_dst_register
-tgsi_default_full_dst_register( void );
-
-
 #if defined __cplusplus
 }
 #endif
diff --git a/src/gallium/auxiliary/tgsi/tgsi_exec.c b/src/gallium/auxiliary/tgsi/tgsi_exec.c
index 3a71540506..7892a67f04 100644
--- a/src/gallium/auxiliary/tgsi/tgsi_exec.c
+++ b/src/gallium/auxiliary/tgsi/tgsi_exec.c
@@ -429,6 +429,24 @@ micro_sne(union tgsi_exec_channel *dst,
 }
 
 static void
+micro_sfl(union tgsi_exec_channel *dst)
+{
+   dst->f[0] = 0.0f;
+   dst->f[1] = 0.0f;
+   dst->f[2] = 0.0f;
+   dst->f[3] = 0.0f;
+}
+
+static void
+micro_str(union tgsi_exec_channel *dst)
+{
+   dst->f[0] = 1.0f;
+   dst->f[1] = 1.0f;
+   dst->f[2] = 1.0f;
+   dst->f[3] = 1.0f;
+}
+
+static void
 micro_trunc(union tgsi_exec_channel *dst,
             const union tgsi_exec_channel *src)
 {
@@ -453,50 +471,12 @@ enum tgsi_exec_datatype {
 /*
  * Shorthand locations of various utility registers (_I = Index, _C = Channel)
  */
-#define TEMP_0_I           TGSI_EXEC_TEMP_00000000_I
-#define TEMP_0_C           TGSI_EXEC_TEMP_00000000_C
-#define TEMP_7F_I          TGSI_EXEC_TEMP_7FFFFFFF_I
-#define TEMP_7F_C          TGSI_EXEC_TEMP_7FFFFFFF_C
-#define TEMP_80_I          TGSI_EXEC_TEMP_80000000_I
-#define TEMP_80_C          TGSI_EXEC_TEMP_80000000_C
-#define TEMP_FF_I          TGSI_EXEC_TEMP_FFFFFFFF_I
-#define TEMP_FF_C          TGSI_EXEC_TEMP_FFFFFFFF_C
-#define TEMP_1_I           TGSI_EXEC_TEMP_ONE_I
-#define TEMP_1_C           TGSI_EXEC_TEMP_ONE_C
-#define TEMP_2_I           TGSI_EXEC_TEMP_TWO_I
-#define TEMP_2_C           TGSI_EXEC_TEMP_TWO_C
-#define TEMP_128_I         TGSI_EXEC_TEMP_128_I
-#define TEMP_128_C         TGSI_EXEC_TEMP_128_C
-#define TEMP_M128_I        TGSI_EXEC_TEMP_MINUS_128_I
-#define TEMP_M128_C        TGSI_EXEC_TEMP_MINUS_128_C
 #define TEMP_KILMASK_I     TGSI_EXEC_TEMP_KILMASK_I
 #define TEMP_KILMASK_C     TGSI_EXEC_TEMP_KILMASK_C
 #define TEMP_OUTPUT_I      TGSI_EXEC_TEMP_OUTPUT_I
 #define TEMP_OUTPUT_C      TGSI_EXEC_TEMP_OUTPUT_C
 #define TEMP_PRIMITIVE_I   TGSI_EXEC_TEMP_PRIMITIVE_I
 #define TEMP_PRIMITIVE_C   TGSI_EXEC_TEMP_PRIMITIVE_C
-#define TEMP_CC_I          TGSI_EXEC_TEMP_CC_I
-#define TEMP_CC_C          TGSI_EXEC_TEMP_CC_C
-#define TEMP_3_I           TGSI_EXEC_TEMP_THREE_I
-#define TEMP_3_C           TGSI_EXEC_TEMP_THREE_C
-#define TEMP_HALF_I        TGSI_EXEC_TEMP_HALF_I
-#define TEMP_HALF_C        TGSI_EXEC_TEMP_HALF_C
-#define TEMP_R0            TGSI_EXEC_TEMP_R0
-#define TEMP_P0            TGSI_EXEC_TEMP_P0
-
-#define IS_CHANNEL_ENABLED(INST, CHAN)\
-   ((INST).Dst[0].Register.WriteMask & (1 << (CHAN)))
-
-#define IS_CHANNEL_ENABLED2(INST, CHAN)\
-   ((INST).Dst[1].Register.WriteMask & (1 << (CHAN)))
-
-#define FOR_EACH_ENABLED_CHANNEL(INST, CHAN)\
-   for (CHAN = 0; CHAN < NUM_CHANNELS; CHAN++)\
-      if (IS_CHANNEL_ENABLED( INST, CHAN ))
-
-#define FOR_EACH_ENABLED_CHANNEL2(INST, CHAN)\
-   for (CHAN = 0; CHAN < NUM_CHANNELS; CHAN++)\
-      if (IS_CHANNEL_ENABLED2( INST, CHAN ))
 
 
 /** The execution mask depends on the conditional mask and the loop mask */
@@ -511,6 +491,14 @@ static const union tgsi_exec_channel OneVec = {
    {1.0f, 1.0f, 1.0f, 1.0f}
 };
 
+static const union tgsi_exec_channel P128Vec = {
+   {128.0f, 128.0f, 128.0f, 128.0f}
+};
+
+static const union tgsi_exec_channel M128Vec = {
+   {-128.0f, -128.0f, -128.0f, -128.0f}
+};
+
 
 /**
  * Assert that none of the float values in 'chan' are infinite or NaN.
@@ -572,8 +560,6 @@ tgsi_exec_set_constant_buffers(struct tgsi_exec_machine *mach,
 }
 
 
-
-
 /**
  * Check if there's a potential src/dst register data dependency when
  * using SOA execution.
@@ -607,18 +593,20 @@ tgsi_check_soa_dependencies(const struct tgsi_full_instruction *inst)
            inst->Dst[0].Register.File) &&
           ((inst->Src[i].Register.Index ==
             inst->Dst[0].Register.Index) ||
-	   inst->Src[i].Register.Indirect ||
-	   inst->Dst[0].Register.Indirect)) {
+           inst->Src[i].Register.Indirect ||
+           inst->Dst[0].Register.Indirect)) {
          /* loop over dest channels */
          uint channelsWritten = 0x0;
-         FOR_EACH_ENABLED_CHANNEL(*inst, chan) {
-            /* check if we're reading a channel that's been written */
-            uint swizzle = tgsi_util_get_full_src_register_swizzle(&inst->Src[i], chan);
-            if (channelsWritten & (1 << swizzle)) {
-               return TRUE;
-            }
+         for (chan = 0; chan < NUM_CHANNELS; chan++) {
+            if (inst->Dst[0].Register.WriteMask & (1 << chan)) {
+               /* check if we're reading a channel that's been written */
+               uint swizzle = tgsi_util_get_full_src_register_swizzle(&inst->Src[i], chan);
+               if (channelsWritten & (1 << swizzle)) {
+                  return TRUE;
+               }
 
-            channelsWritten |= (1 << chan);
+               channelsWritten |= (1 << chan);
+            }
          }
       }
    }
@@ -813,18 +801,18 @@ tgsi_exec_machine_create( void )
    mach->MaxGeometryShaderOutputs = TGSI_MAX_TOTAL_VERTICES;
    mach->Predicates = &mach->Temps[TGSI_EXEC_TEMP_P0];
 
-   /* Setup constants. */
+   /* Setup constants needed by the SSE2 executor. */
    for( i = 0; i < 4; i++ ) {
-      mach->Temps[TEMP_0_I].xyzw[TEMP_0_C].u[i] = 0x00000000;
-      mach->Temps[TEMP_7F_I].xyzw[TEMP_7F_C].u[i] = 0x7FFFFFFF;
-      mach->Temps[TEMP_80_I].xyzw[TEMP_80_C].u[i] = 0x80000000;
-      mach->Temps[TEMP_FF_I].xyzw[TEMP_FF_C].u[i] = 0xFFFFFFFF;
-      mach->Temps[TEMP_1_I].xyzw[TEMP_1_C].f[i] = 1.0f;
-      mach->Temps[TEMP_2_I].xyzw[TEMP_2_C].f[i] = 2.0f;
-      mach->Temps[TEMP_128_I].xyzw[TEMP_128_C].f[i] = 128.0f;
-      mach->Temps[TEMP_M128_I].xyzw[TEMP_M128_C].f[i] = -128.0f;
-      mach->Temps[TEMP_3_I].xyzw[TEMP_3_C].f[i] = 3.0f;
-      mach->Temps[TEMP_HALF_I].xyzw[TEMP_HALF_C].f[i] = 0.5f;
+      mach->Temps[TGSI_EXEC_TEMP_00000000_I].xyzw[TGSI_EXEC_TEMP_00000000_C].u[i] = 0x00000000;
+      mach->Temps[TGSI_EXEC_TEMP_7FFFFFFF_I].xyzw[TGSI_EXEC_TEMP_7FFFFFFF_C].u[i] = 0x7FFFFFFF;
+      mach->Temps[TGSI_EXEC_TEMP_80000000_I].xyzw[TGSI_EXEC_TEMP_80000000_C].u[i] = 0x80000000;
+      mach->Temps[TGSI_EXEC_TEMP_FFFFFFFF_I].xyzw[TGSI_EXEC_TEMP_FFFFFFFF_C].u[i] = 0xFFFFFFFF;    /* not used */
+      mach->Temps[TGSI_EXEC_TEMP_ONE_I].xyzw[TGSI_EXEC_TEMP_ONE_C].f[i] = 1.0f;
+      mach->Temps[TGSI_EXEC_TEMP_TWO_I].xyzw[TGSI_EXEC_TEMP_TWO_C].f[i] = 2.0f;    /* not used */
+      mach->Temps[TGSI_EXEC_TEMP_128_I].xyzw[TGSI_EXEC_TEMP_128_C].f[i] = 128.0f;
+      mach->Temps[TGSI_EXEC_TEMP_MINUS_128_I].xyzw[TGSI_EXEC_TEMP_MINUS_128_C].f[i] = -128.0f;
+      mach->Temps[TGSI_EXEC_TEMP_THREE_I].xyzw[TGSI_EXEC_TEMP_THREE_C].f[i] = 3.0f;
+      mach->Temps[TGSI_EXEC_TEMP_HALF_I].xyzw[TGSI_EXEC_TEMP_HALF_C].f[i] = 0.5f;
    }
 
 #ifdef DEBUG
@@ -886,27 +874,35 @@ micro_div(
 }
 
 static void
-micro_float_clamp(union tgsi_exec_channel *dst,
-                  const union tgsi_exec_channel *src)
+micro_rcc(union tgsi_exec_channel *dst,
+          const union tgsi_exec_channel *src)
 {
    uint i;
 
    for (i = 0; i < 4; i++) {
-      if (src->f[i] > 0.0f) {
-         if (src->f[i] > 1.884467e+019f)
+      float recip = 1.0f / src->f[i];
+
+      if (recip > 0.0f) {
+         if (recip > 1.884467e+019f) {
             dst->f[i] = 1.884467e+019f;
-         else if (src->f[i] < 5.42101e-020f)
+         }
+         else if (recip < 5.42101e-020f) {
             dst->f[i] = 5.42101e-020f;
-         else
-            dst->f[i] = src->f[i];
+         }
+         else {
+            dst->f[i] = recip;
+         }
       }
       else {
-         if (src->f[i] < -1.884467e+019f)
+         if (recip < -1.884467e+019f) {
             dst->f[i] = -1.884467e+019f;
-         else if (src->f[i] > -5.42101e-020f)
+         }
+         else if (recip > -5.42101e-020f) {
             dst->f[i] = -5.42101e-020f;
-         else
-            dst->f[i] = src->f[i];
+         }
+         else {
+            dst->f[i] = recip;
+         }
       }
    }
 }
@@ -958,60 +954,6 @@ micro_mul(union tgsi_exec_channel *dst,
    dst->f[3] = src0->f[3] * src1->f[3];
 }
 
-#if 0
-static void
-micro_imul64(
-   union tgsi_exec_channel *dst0,
-   union tgsi_exec_channel *dst1,
-   const union tgsi_exec_channel *src0,
-   const union tgsi_exec_channel *src1 )
-{
-   dst1->i[0] = src0->i[0] * src1->i[0];
-   dst1->i[1] = src0->i[1] * src1->i[1];
-   dst1->i[2] = src0->i[2] * src1->i[2];
-   dst1->i[3] = src0->i[3] * src1->i[3];
-   dst0->i[0] = 0;
-   dst0->i[1] = 0;
-   dst0->i[2] = 0;
-   dst0->i[3] = 0;
-}
-#endif
-
-#if 0
-static void
-micro_umul64(
-   union tgsi_exec_channel *dst0,
-   union tgsi_exec_channel *dst1,
-   const union tgsi_exec_channel *src0,
-   const union tgsi_exec_channel *src1 )
-{
-   dst1->u[0] = src0->u[0] * src1->u[0];
-   dst1->u[1] = src0->u[1] * src1->u[1];
-   dst1->u[2] = src0->u[2] * src1->u[2];
-   dst1->u[3] = src0->u[3] * src1->u[3];
-   dst0->u[0] = 0;
-   dst0->u[1] = 0;
-   dst0->u[2] = 0;
-   dst0->u[3] = 0;
-}
-#endif
-
-
-#if 0
-static void
-micro_movc(
-   union tgsi_exec_channel *dst,
-   const union tgsi_exec_channel *src0,
-   const union tgsi_exec_channel *src1,
-   const union tgsi_exec_channel *src2 )
-{
-   dst->u[0] = src0->u[0] ? src1->u[0] : src2->u[0];
-   dst->u[1] = src0->u[1] ? src1->u[1] : src2->u[1];
-   dst->u[2] = src0->u[2] ? src1->u[2] : src2->u[2];
-   dst->u[3] = src0->u[3] ? src1->u[3] : src2->u[3];
-}
-#endif
-
 static void
 micro_neg(
    union tgsi_exec_channel *dst,
@@ -1607,9 +1549,6 @@ store_dest(struct tgsi_exec_machine *mach,
 #define FETCH(VAL,INDEX,CHAN)\
     fetch_source(mach, VAL, &inst->Src[INDEX], CHAN, TGSI_EXEC_DATA_FLOAT)
 
-#define STORE(VAL,INDEX,CHAN)\
-   store_dest(mach, VAL, &inst->Dst[INDEX], inst, CHAN, TGSI_EXEC_DATA_FLOAT)
-
 
 /**
  * Execute ARB-style KIL which is predicated by a src register.
@@ -1753,7 +1692,7 @@ exec_tex(struct tgsi_exec_machine *mach,
    union tgsi_exec_channel r[4];
    const union tgsi_exec_channel *lod = &ZeroVec;
    enum tgsi_sampler_control control;
-   uint chan_index;
+   uint chan;
 
    if (modifier != TEX_MODIFIER_NONE) {
       FETCH(&r[3], 0, CHAN_W);
@@ -1825,8 +1764,10 @@ exec_tex(struct tgsi_exec_machine *mach,
       assert(0);
    }
 
-   FOR_EACH_ENABLED_CHANNEL(*inst, chan_index) {
-      STORE(&r[chan_index], 0, chan_index);
+   for (chan = 0; chan < NUM_CHANNELS; chan++) {
+      if (inst->Dst[0].Register.WriteMask & (1 << chan)) {
+         store_dest(mach, &r[chan], &inst->Dst[0], inst, chan, TGSI_EXEC_DATA_FLOAT);
+      }
    }
 }
 
@@ -1836,7 +1777,7 @@ exec_txd(struct tgsi_exec_machine *mach,
 {
    const uint unit = inst->Src[3].Register.Index;
    union tgsi_exec_channel r[4];
-   uint chan_index;
+   uint chan;
 
    /*
     * XXX: This is fake TXD -- the derivatives are not taken into account, yet.
@@ -1886,8 +1827,10 @@ exec_txd(struct tgsi_exec_machine *mach,
       assert(0);
    }
 
-   FOR_EACH_ENABLED_CHANNEL(*inst, chan_index) {
-      STORE(&r[chan_index], 0, chan_index);
+   for (chan = 0; chan < NUM_CHANNELS; chan++) {
+      if (inst->Dst[0].Register.WriteMask & (1 << chan)) {
+         store_dest(mach, &r[chan], &inst->Dst[0], inst, chan, TGSI_EXEC_DATA_FLOAT);
+      }
    }
 }
 
@@ -2021,6 +1964,26 @@ exec_declaration(struct tgsi_exec_machine *mach,
    }
 }
 
+typedef void (* micro_op)(union tgsi_exec_channel *dst);
+
+static void
+exec_vector(struct tgsi_exec_machine *mach,
+            const struct tgsi_full_instruction *inst,
+            micro_op op,
+            enum tgsi_exec_datatype dst_datatype)
+{
+   unsigned int chan;
+
+   for (chan = 0; chan < NUM_CHANNELS; chan++) {
+      if (inst->Dst[0].Register.WriteMask & (1 << chan)) {
+         union tgsi_exec_channel dst;
+
+         op(&dst);
+         store_dest(mach, &dst, &inst->Dst[0], inst, chan, dst_datatype);
+      }
+   }
+}
+
 typedef void (* micro_unary_op)(union tgsi_exec_channel *dst,
                                 const union tgsi_exec_channel *src);
 
@@ -2074,6 +2037,27 @@ typedef void (* micro_binary_op)(union tgsi_exec_channel *dst,
                                  const union tgsi_exec_channel *src1);
 
 static void
+exec_scalar_binary(struct tgsi_exec_machine *mach,
+                   const struct tgsi_full_instruction *inst,
+                   micro_binary_op op,
+                   enum tgsi_exec_datatype dst_datatype,
+                   enum tgsi_exec_datatype src_datatype)
+{
+   unsigned int chan;
+   union tgsi_exec_channel src[2];
+   union tgsi_exec_channel dst;
+
+   fetch_source(mach, &src[0], &inst->Src[0], CHAN_X, src_datatype);
+   fetch_source(mach, &src[1], &inst->Src[1], CHAN_Y, src_datatype);
+   op(&dst, &src[0], &src[1]);
+   for (chan = 0; chan < NUM_CHANNELS; chan++) {
+      if (inst->Dst[0].Register.WriteMask & (1 << chan)) {
+         store_dest(mach, &dst, &inst->Dst[0], inst, chan, dst_datatype);
+      }
+   }
+}
+
+static void
 exec_vector_binary(struct tgsi_exec_machine *mach,
                    const struct tgsi_full_instruction *inst,
                    micro_binary_op op,
@@ -2320,6 +2304,289 @@ exec_nrm3(struct tgsi_exec_machine *mach,
 }
 
 static void
+exec_scs(struct tgsi_exec_machine *mach,
+         const struct tgsi_full_instruction *inst)
+{
+   if (inst->Dst[0].Register.WriteMask & TGSI_WRITEMASK_XY) {
+      union tgsi_exec_channel arg;
+      union tgsi_exec_channel result;
+
+      fetch_source(mach, &arg, &inst->Src[0], CHAN_X, TGSI_EXEC_DATA_FLOAT);
+
+      if (inst->Dst[0].Register.WriteMask & TGSI_WRITEMASK_X) {
+         micro_cos(&result, &arg);
+         store_dest(mach, &result, &inst->Dst[0], inst, CHAN_X, TGSI_EXEC_DATA_FLOAT);
+      }
+      if (inst->Dst[0].Register.WriteMask & TGSI_WRITEMASK_Y) {
+         micro_sin(&result, &arg);
+         store_dest(mach, &result, &inst->Dst[0], inst, CHAN_Y, TGSI_EXEC_DATA_FLOAT);
+      }
+   }
+   if (inst->Dst[0].Register.WriteMask & TGSI_WRITEMASK_Z) {
+      store_dest(mach, &ZeroVec, &inst->Dst[0], inst, CHAN_Z, TGSI_EXEC_DATA_FLOAT);
+   }
+   if (inst->Dst[0].Register.WriteMask & TGSI_WRITEMASK_W) {
+      store_dest(mach, &OneVec, &inst->Dst[0], inst, CHAN_W, TGSI_EXEC_DATA_FLOAT);
+   }
+}
+
+static void
+exec_x2d(struct tgsi_exec_machine *mach,
+         const struct tgsi_full_instruction *inst)
+{
+   union tgsi_exec_channel r[4];
+   union tgsi_exec_channel d[2];
+
+   fetch_source(mach, &r[0], &inst->Src[1], CHAN_X, TGSI_EXEC_DATA_FLOAT);
+   fetch_source(mach, &r[1], &inst->Src[1], CHAN_Y, TGSI_EXEC_DATA_FLOAT);
+   if (inst->Dst[0].Register.WriteMask & TGSI_WRITEMASK_XZ) {
+      fetch_source(mach, &r[2], &inst->Src[2], CHAN_X, TGSI_EXEC_DATA_FLOAT);
+      micro_mul(&r[2], &r[2], &r[0]);
+      fetch_source(mach, &r[3], &inst->Src[2], CHAN_Y, TGSI_EXEC_DATA_FLOAT);
+      micro_mul(&r[3], &r[3], &r[1]);
+      micro_add(&r[2], &r[2], &r[3]);
+      fetch_source(mach, &r[3], &inst->Src[0], CHAN_X, TGSI_EXEC_DATA_FLOAT);
+      micro_add(&d[0], &r[2], &r[3]);
+   }
+   if (inst->Dst[0].Register.WriteMask & TGSI_WRITEMASK_YW) {
+      fetch_source(mach, &r[2], &inst->Src[2], CHAN_Z, TGSI_EXEC_DATA_FLOAT);
+      micro_mul(&r[2], &r[2], &r[0]);
+      fetch_source(mach, &r[3], &inst->Src[2], CHAN_W, TGSI_EXEC_DATA_FLOAT);
+      micro_mul(&r[3], &r[3], &r[1]);
+      micro_add(&r[2], &r[2], &r[3]);
+      fetch_source(mach, &r[3], &inst->Src[0], CHAN_Y, TGSI_EXEC_DATA_FLOAT);
+      micro_add(&d[1], &r[2], &r[3]);
+   }
+   if (inst->Dst[0].Register.WriteMask & TGSI_WRITEMASK_X) {
+      store_dest(mach, &d[0], &inst->Dst[0], inst, CHAN_X, TGSI_EXEC_DATA_FLOAT);
+   }
+   if (inst->Dst[0].Register.WriteMask & TGSI_WRITEMASK_Y) {
+      store_dest(mach, &d[1], &inst->Dst[0], inst, CHAN_Y, TGSI_EXEC_DATA_FLOAT);
+   }
+   if (inst->Dst[0].Register.WriteMask & TGSI_WRITEMASK_Z) {
+      store_dest(mach, &d[0], &inst->Dst[0], inst, CHAN_Z, TGSI_EXEC_DATA_FLOAT);
+   }
+   if (inst->Dst[0].Register.WriteMask & TGSI_WRITEMASK_W) {
+      store_dest(mach, &d[1], &inst->Dst[0], inst, CHAN_W, TGSI_EXEC_DATA_FLOAT);
+   }
+}
+
+static void
+exec_rfl(struct tgsi_exec_machine *mach,
+         const struct tgsi_full_instruction *inst)
+{
+   union tgsi_exec_channel r[9];
+
+   if (inst->Dst[0].Register.WriteMask & TGSI_WRITEMASK_XYZ) {
+      /* r0 = dp3(src0, src0) */
+      fetch_source(mach, &r[2], &inst->Src[0], CHAN_X, TGSI_EXEC_DATA_FLOAT);
+      micro_mul(&r[0], &r[2], &r[2]);
+      fetch_source(mach, &r[4], &inst->Src[0], CHAN_Y, TGSI_EXEC_DATA_FLOAT);
+      micro_mul(&r[8], &r[4], &r[4]);
+      micro_add(&r[0], &r[0], &r[8]);
+      fetch_source(mach, &r[6], &inst->Src[0], CHAN_Z, TGSI_EXEC_DATA_FLOAT);
+      micro_mul(&r[8], &r[6], &r[6]);
+      micro_add(&r[0], &r[0], &r[8]);
+
+      /* r1 = dp3(src0, src1) */
+      fetch_source(mach, &r[3], &inst->Src[1], CHAN_X, TGSI_EXEC_DATA_FLOAT);
+      micro_mul(&r[1], &r[2], &r[3]);
+      fetch_source(mach, &r[5], &inst->Src[1], CHAN_Y, TGSI_EXEC_DATA_FLOAT);
+      micro_mul(&r[8], &r[4], &r[5]);
+      micro_add(&r[1], &r[1], &r[8]);
+      fetch_source(mach, &r[7], &inst->Src[1], CHAN_Z, TGSI_EXEC_DATA_FLOAT);
+      micro_mul(&r[8], &r[6], &r[7]);
+      micro_add(&r[1], &r[1], &r[8]);
+
+      /* r1 = 2 * r1 / r0 */
+      micro_add(&r[1], &r[1], &r[1]);
+      micro_div(&r[1], &r[1], &r[0]);
+
+      if (inst->Dst[0].Register.WriteMask & TGSI_WRITEMASK_X) {
+         micro_mul(&r[2], &r[2], &r[1]);
+         micro_sub(&r[2], &r[2], &r[3]);
+         store_dest(mach, &r[2], &inst->Dst[0], inst, CHAN_X, TGSI_EXEC_DATA_FLOAT);
+      }
+      if (inst->Dst[0].Register.WriteMask & TGSI_WRITEMASK_Y) {
+         micro_mul(&r[4], &r[4], &r[1]);
+         micro_sub(&r[4], &r[4], &r[5]);
+         store_dest(mach, &r[4], &inst->Dst[0], inst, CHAN_Y, TGSI_EXEC_DATA_FLOAT);
+      }
+      if (inst->Dst[0].Register.WriteMask & TGSI_WRITEMASK_Z) {
+         micro_mul(&r[6], &r[6], &r[1]);
+         micro_sub(&r[6], &r[6], &r[7]);
+         store_dest(mach, &r[6], &inst->Dst[0], inst, CHAN_Z, TGSI_EXEC_DATA_FLOAT);
+      }
+   }
+   if (inst->Dst[0].Register.WriteMask & TGSI_WRITEMASK_W) {
+      store_dest(mach, &OneVec, &inst->Dst[0], inst, CHAN_W, TGSI_EXEC_DATA_FLOAT);
+   }
+}
+
+static void
+exec_xpd(struct tgsi_exec_machine *mach,
+         const struct tgsi_full_instruction *inst)
+{
+   union tgsi_exec_channel r[6];
+   union tgsi_exec_channel d[3];
+
+   fetch_source(mach, &r[0], &inst->Src[0], CHAN_Y, TGSI_EXEC_DATA_FLOAT);
+   fetch_source(mach, &r[1], &inst->Src[1], CHAN_Z, TGSI_EXEC_DATA_FLOAT);
+
+   micro_mul(&r[2], &r[0], &r[1]);
+
+   fetch_source(mach, &r[3], &inst->Src[0], CHAN_Z, TGSI_EXEC_DATA_FLOAT);
+   fetch_source(mach, &r[4], &inst->Src[1], CHAN_Y, TGSI_EXEC_DATA_FLOAT);
+
+   micro_mul(&r[5], &r[3], &r[4] );
+   micro_sub(&d[CHAN_X], &r[2], &r[5]);
+
+   fetch_source(mach, &r[2], &inst->Src[1], CHAN_X, TGSI_EXEC_DATA_FLOAT);
+
+   micro_mul(&r[3], &r[3], &r[2]);
+
+   fetch_source(mach, &r[5], &inst->Src[0], CHAN_X, TGSI_EXEC_DATA_FLOAT);
+
+   micro_mul(&r[1], &r[1], &r[5]);
+   micro_sub(&d[CHAN_Y], &r[3], &r[1]);
+
+   micro_mul(&r[5], &r[5], &r[4]);
+   micro_mul(&r[0], &r[0], &r[2]);
+   micro_sub(&d[CHAN_Z], &r[5], &r[0]);
+
+   if (inst->Dst[0].Register.WriteMask & TGSI_WRITEMASK_X) {
+      store_dest(mach, &d[CHAN_X], &inst->Dst[0], inst, CHAN_X, TGSI_EXEC_DATA_FLOAT);
+   }
+   if (inst->Dst[0].Register.WriteMask & TGSI_WRITEMASK_Y) {
+      store_dest(mach, &d[CHAN_Y], &inst->Dst[0], inst, CHAN_Y, TGSI_EXEC_DATA_FLOAT);
+   }
+   if (inst->Dst[0].Register.WriteMask & TGSI_WRITEMASK_Z) {
+      store_dest(mach, &d[CHAN_Z], &inst->Dst[0], inst, CHAN_Z, TGSI_EXEC_DATA_FLOAT);
+   }
+   if (inst->Dst[0].Register.WriteMask & TGSI_WRITEMASK_W) {
+      store_dest(mach, &OneVec, &inst->Dst[0], inst, CHAN_W, TGSI_EXEC_DATA_FLOAT);
+   }
+}
+
+static void
+exec_dst(struct tgsi_exec_machine *mach,
+         const struct tgsi_full_instruction *inst)
+{
+   union tgsi_exec_channel r[2];
+   union tgsi_exec_channel d[4];
+
+   if (inst->Dst[0].Register.WriteMask & TGSI_WRITEMASK_Y) {
+      fetch_source(mach, &r[0], &inst->Src[0], CHAN_Y, TGSI_EXEC_DATA_FLOAT);
+      fetch_source(mach, &r[1], &inst->Src[1], CHAN_Y, TGSI_EXEC_DATA_FLOAT);
+      micro_mul(&d[CHAN_Y], &r[0], &r[1]);
+   }
+   if (inst->Dst[0].Register.WriteMask & TGSI_WRITEMASK_Z) {
+      fetch_source(mach, &d[CHAN_Z], &inst->Src[0], CHAN_Z, TGSI_EXEC_DATA_FLOAT);
+   }
+   if (inst->Dst[0].Register.WriteMask & TGSI_WRITEMASK_W) {
+      fetch_source(mach, &d[CHAN_W], &inst->Src[1], CHAN_W, TGSI_EXEC_DATA_FLOAT);
+   }
+
+   if (inst->Dst[0].Register.WriteMask & TGSI_WRITEMASK_X) {
+      store_dest(mach, &OneVec, &inst->Dst[0], inst, CHAN_X, TGSI_EXEC_DATA_FLOAT);
+   }
+   if (inst->Dst[0].Register.WriteMask & TGSI_WRITEMASK_Y) {
+      store_dest(mach, &d[CHAN_Y], &inst->Dst[0], inst, CHAN_Y, TGSI_EXEC_DATA_FLOAT);
+   }
+   if (inst->Dst[0].Register.WriteMask & TGSI_WRITEMASK_Z) {
+      store_dest(mach, &d[CHAN_Z], &inst->Dst[0], inst, CHAN_Z, TGSI_EXEC_DATA_FLOAT);
+   }
+   if (inst->Dst[0].Register.WriteMask & TGSI_WRITEMASK_W) {
+      store_dest(mach, &d[CHAN_W], &inst->Dst[0], inst, CHAN_W, TGSI_EXEC_DATA_FLOAT);
+   }
+}
+
+static void
+exec_log(struct tgsi_exec_machine *mach,
+         const struct tgsi_full_instruction *inst)
+{
+   union tgsi_exec_channel r[3];
+
+   fetch_source(mach, &r[0], &inst->Src[0], CHAN_X, TGSI_EXEC_DATA_FLOAT);
+   micro_abs(&r[2], &r[0]);  /* r2 = abs(r0) */
+   micro_lg2(&r[1], &r[2]);  /* r1 = lg2(r2) */
+   micro_flr(&r[0], &r[1]);  /* r0 = floor(r1) */
+   if (inst->Dst[0].Register.WriteMask & TGSI_WRITEMASK_X) {
+      store_dest(mach, &r[0], &inst->Dst[0], inst, CHAN_X, TGSI_EXEC_DATA_FLOAT);
+   }
+   if (inst->Dst[0].Register.WriteMask & TGSI_WRITEMASK_Y) {
+      micro_exp2(&r[0], &r[0]);       /* r0 = 2 ^ r0 */
+      micro_div(&r[0], &r[2], &r[0]); /* r0 = r2 / r0 */
+      store_dest(mach, &r[0], &inst->Dst[0], inst, CHAN_Y, TGSI_EXEC_DATA_FLOAT);
+   }
+   if (inst->Dst[0].Register.WriteMask & TGSI_WRITEMASK_Z) {
+      store_dest(mach, &r[1], &inst->Dst[0], inst, CHAN_Z, TGSI_EXEC_DATA_FLOAT);
+   }
+   if (inst->Dst[0].Register.WriteMask & TGSI_WRITEMASK_W) {
+      store_dest(mach, &OneVec, &inst->Dst[0], inst, CHAN_W, TGSI_EXEC_DATA_FLOAT);
+   }
+}
+
+static void
+exec_exp(struct tgsi_exec_machine *mach,
+         const struct tgsi_full_instruction *inst)
+{
+   union tgsi_exec_channel r[3];
+
+   fetch_source(mach, &r[0], &inst->Src[0], CHAN_X, TGSI_EXEC_DATA_FLOAT);
+   micro_flr(&r[1], &r[0]);  /* r1 = floor(r0) */
+   if (inst->Dst[0].Register.WriteMask & TGSI_WRITEMASK_X) {
+      micro_exp2(&r[2], &r[1]);       /* r2 = 2 ^ r1 */
+      store_dest(mach, &r[2], &inst->Dst[0], inst, CHAN_X, TGSI_EXEC_DATA_FLOAT);
+   }
+   if (inst->Dst[0].Register.WriteMask & TGSI_WRITEMASK_Y) {
+      micro_sub(&r[2], &r[0], &r[1]); /* r2 = r0 - r1 */
+      store_dest(mach, &r[2], &inst->Dst[0], inst, CHAN_Y, TGSI_EXEC_DATA_FLOAT);
+   }
+   if (inst->Dst[0].Register.WriteMask & TGSI_WRITEMASK_Z) {
+      micro_exp2(&r[2], &r[0]);       /* r2 = 2 ^ r0 */
+      store_dest(mach, &r[2], &inst->Dst[0], inst, CHAN_Z, TGSI_EXEC_DATA_FLOAT);
+   }
+   if (inst->Dst[0].Register.WriteMask & TGSI_WRITEMASK_W) {
+      store_dest(mach, &OneVec, &inst->Dst[0], inst, CHAN_W, TGSI_EXEC_DATA_FLOAT);
+   }
+}
+
+static void
+exec_lit(struct tgsi_exec_machine *mach,
+         const struct tgsi_full_instruction *inst)
+{
+   union tgsi_exec_channel r[3];
+   union tgsi_exec_channel d[3];
+
+   if (inst->Dst[0].Register.WriteMask & TGSI_WRITEMASK_X) {
+      store_dest(mach, &OneVec, &inst->Dst[0], inst, CHAN_X, TGSI_EXEC_DATA_FLOAT);
+   }
+   if (inst->Dst[0].Register.WriteMask & TGSI_WRITEMASK_YZ) {
+      fetch_source(mach, &r[0], &inst->Src[0], CHAN_X, TGSI_EXEC_DATA_FLOAT);
+      if (inst->Dst[0].Register.WriteMask & TGSI_WRITEMASK_Y) {
+         micro_max(&d[CHAN_Y], &r[0], &ZeroVec);
+         store_dest(mach, &d[CHAN_Y], &inst->Dst[0], inst, CHAN_Y, TGSI_EXEC_DATA_FLOAT);
+      }
+
+      if (inst->Dst[0].Register.WriteMask & TGSI_WRITEMASK_Z) {
+         fetch_source(mach, &r[1], &inst->Src[0], CHAN_Y, TGSI_EXEC_DATA_FLOAT);
+         micro_max(&r[1], &r[1], &ZeroVec);
+
+         fetch_source(mach, &r[2], &inst->Src[0], CHAN_W, TGSI_EXEC_DATA_FLOAT);
+         micro_min(&r[2], &r[2], &P128Vec);
+         micro_max(&r[2], &r[2], &M128Vec);
+         micro_pow(&r[1], &r[1], &r[2]);
+         micro_lt(&d[CHAN_Z], &ZeroVec, &r[0], &r[1], &ZeroVec);
+         store_dest(mach, &d[CHAN_Z], &inst->Dst[0], inst, CHAN_Z, TGSI_EXEC_DATA_FLOAT);
+      }
+   }
+   if (inst->Dst[0].Register.WriteMask & TGSI_WRITEMASK_W) {
+      store_dest(mach, &OneVec, &inst->Dst[0], inst, CHAN_W, TGSI_EXEC_DATA_FLOAT);
+   }
+}
+
+static void
 exec_break(struct tgsi_exec_machine *mach)
 {
    if (mach->BreakType == TGSI_EXEC_BREAK_INSIDE_LOOP) {
@@ -2702,9 +2969,7 @@ exec_instruction(
    const struct tgsi_full_instruction *inst,
    int *pc )
 {
-   uint chan_index;
    union tgsi_exec_channel r[10];
-   union tgsi_exec_channel d[8];
 
    (*pc)++;
 
@@ -2718,36 +2983,7 @@ exec_instruction(
       break;
 
    case TGSI_OPCODE_LIT:
-      if (IS_CHANNEL_ENABLED( *inst, CHAN_Y ) || IS_CHANNEL_ENABLED( *inst, CHAN_Z )) {
-         FETCH( &r[0], 0, CHAN_X );
-         if (IS_CHANNEL_ENABLED( *inst, CHAN_Y )) {
-            micro_max(&d[CHAN_Y], &r[0], &mach->Temps[TEMP_0_I].xyzw[TEMP_0_C]);
-         }
-
-         if (IS_CHANNEL_ENABLED( *inst, CHAN_Z )) {
-            FETCH( &r[1], 0, CHAN_Y );
-            micro_max( &r[1], &r[1], &mach->Temps[TEMP_0_I].xyzw[TEMP_0_C] );
-
-            FETCH( &r[2], 0, CHAN_W );
-            micro_min( &r[2], &r[2], &mach->Temps[TEMP_128_I].xyzw[TEMP_128_C] );
-            micro_max( &r[2], &r[2], &mach->Temps[TEMP_M128_I].xyzw[TEMP_M128_C] );
-            micro_pow( &r[1], &r[1], &r[2] );
-            micro_lt(&d[CHAN_Z], &mach->Temps[TEMP_0_I].xyzw[TEMP_0_C], &r[0], &r[1], &mach->Temps[TEMP_0_I].xyzw[TEMP_0_C]);
-         }
-
-         if (IS_CHANNEL_ENABLED(*inst, CHAN_Y)) {
-            STORE(&d[CHAN_Y], 0, CHAN_Y);
-         }
-         if (IS_CHANNEL_ENABLED(*inst, CHAN_Z)) {
-            STORE(&d[CHAN_Z], 0, CHAN_Z);
-         }
-      }
-      if (IS_CHANNEL_ENABLED( *inst, CHAN_X )) {
-         STORE( &mach->Temps[TEMP_1_I].xyzw[TEMP_1_C], 0, CHAN_X );
-      }
-      if (IS_CHANNEL_ENABLED( *inst, CHAN_W )) {
-         STORE( &mach->Temps[TEMP_1_I].xyzw[TEMP_1_C], 0, CHAN_W );
-      }
+      exec_lit(mach, inst);
       break;
 
    case TGSI_OPCODE_RCP:
@@ -2759,44 +2995,11 @@ exec_instruction(
       break;
 
    case TGSI_OPCODE_EXP:
-      FETCH( &r[0], 0, CHAN_X );
-      micro_flr( &r[1], &r[0] );  /* r1 = floor(r0) */
-      if (IS_CHANNEL_ENABLED( *inst, CHAN_X )) {
-         micro_exp2( &r[2], &r[1] );       /* r2 = 2 ^ r1 */
-         STORE( &r[2], 0, CHAN_X );        /* store r2 */
-      }
-      if (IS_CHANNEL_ENABLED( *inst, CHAN_Y )) {
-         micro_sub( &r[2], &r[0], &r[1] ); /* r2 = r0 - r1 */
-         STORE( &r[2], 0, CHAN_Y );        /* store r2 */
-      }
-      if (IS_CHANNEL_ENABLED( *inst, CHAN_Z )) {
-         micro_exp2( &r[2], &r[0] );       /* r2 = 2 ^ r0 */
-         STORE( &r[2], 0, CHAN_Z );        /* store r2 */
-      }
-      if (IS_CHANNEL_ENABLED( *inst, CHAN_W )) {
-         STORE( &mach->Temps[TEMP_1_I].xyzw[TEMP_1_C], 0, CHAN_W );
-      }
+      exec_exp(mach, inst);
       break;
 
    case TGSI_OPCODE_LOG:
-      FETCH( &r[0], 0, CHAN_X );
-      micro_abs( &r[2], &r[0] );  /* r2 = abs(r0) */
-      micro_lg2( &r[1], &r[2] );  /* r1 = lg2(r2) */
-      micro_flr( &r[0], &r[1] );  /* r0 = floor(r1) */
-      if (IS_CHANNEL_ENABLED( *inst, CHAN_X )) {
-         STORE( &r[0], 0, CHAN_X );
-      }
-      if (IS_CHANNEL_ENABLED( *inst, CHAN_Y )) {
-         micro_exp2( &r[0], &r[0] );       /* r0 = 2 ^ r0 */
-         micro_div( &r[0], &r[2], &r[0] ); /* r0 = r2 / r0 */
-         STORE( &r[0], 0, CHAN_Y );
-      }
-      if (IS_CHANNEL_ENABLED( *inst, CHAN_Z )) {
-         STORE( &r[1], 0, CHAN_Z );
-      }
-      if (IS_CHANNEL_ENABLED( *inst, CHAN_W )) {
-         STORE( &mach->Temps[TEMP_1_I].xyzw[TEMP_1_C], 0, CHAN_W );
-      }
+      exec_log(mach, inst);
       break;
 
    case TGSI_OPCODE_MUL:
@@ -2816,30 +3019,7 @@ exec_instruction(
       break;
 
    case TGSI_OPCODE_DST:
-      if (IS_CHANNEL_ENABLED( *inst, CHAN_Y )) {
-         FETCH( &r[0], 0, CHAN_Y );
-         FETCH( &r[1], 1, CHAN_Y);
-         micro_mul(&d[CHAN_Y], &r[0], &r[1]);
-      }
-      if (IS_CHANNEL_ENABLED( *inst, CHAN_Z )) {
-         FETCH(&d[CHAN_Z], 0, CHAN_Z);
-      }
-      if (IS_CHANNEL_ENABLED( *inst, CHAN_W )) {
-         FETCH(&d[CHAN_W], 1, CHAN_W);
-      }
-
-      if (IS_CHANNEL_ENABLED(*inst, CHAN_X)) {
-         STORE(&mach->Temps[TEMP_1_I].xyzw[TEMP_1_C], 0, CHAN_X);
-      }
-      if (IS_CHANNEL_ENABLED(*inst, CHAN_Y)) {
-         STORE(&d[CHAN_Y], 0, CHAN_Y);
-      }
-      if (IS_CHANNEL_ENABLED(*inst, CHAN_Z)) {
-         STORE(&d[CHAN_Z], 0, CHAN_Z);
-      }
-      if (IS_CHANNEL_ENABLED(*inst, CHAN_W)) {
-         STORE(&d[CHAN_W], 0, CHAN_W);
-      }
+      exec_dst(mach, inst);
       break;
 
    case TGSI_OPCODE_MIN:
@@ -2903,53 +3083,11 @@ exec_instruction(
       break;
 
    case TGSI_OPCODE_POW:
-      FETCH(&r[0], 0, CHAN_X);
-      FETCH(&r[1], 1, CHAN_X);
-
-      micro_pow( &r[0], &r[0], &r[1] );
-
-      FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) {
-         STORE( &r[0], 0, chan_index );
-      }
+      exec_scalar_binary(mach, inst, micro_pow, TGSI_EXEC_DATA_FLOAT, TGSI_EXEC_DATA_FLOAT);
       break;
 
    case TGSI_OPCODE_XPD:
-      FETCH(&r[0], 0, CHAN_Y);
-      FETCH(&r[1], 1, CHAN_Z);
-
-      micro_mul( &r[2], &r[0], &r[1] );
-
-      FETCH(&r[3], 0, CHAN_Z);
-      FETCH(&r[4], 1, CHAN_Y);
-
-      micro_mul( &r[5], &r[3], &r[4] );
-      micro_sub(&d[CHAN_X], &r[2], &r[5]);
-
-      FETCH(&r[2], 1, CHAN_X);
-
-      micro_mul( &r[3], &r[3], &r[2] );
-
-      FETCH(&r[5], 0, CHAN_X);
-
-      micro_mul( &r[1], &r[1], &r[5] );
-      micro_sub(&d[CHAN_Y], &r[3], &r[1]);
-
-      micro_mul( &r[5], &r[5], &r[4] );
-      micro_mul( &r[0], &r[0], &r[2] );
-      micro_sub(&d[CHAN_Z], &r[5], &r[0]);
-
-      if (IS_CHANNEL_ENABLED(*inst, CHAN_X)) {
-         STORE(&d[CHAN_X], 0, CHAN_X);
-      }
-      if (IS_CHANNEL_ENABLED(*inst, CHAN_Y)) {
-         STORE(&d[CHAN_Y], 0, CHAN_Y);
-      }
-      if (IS_CHANNEL_ENABLED(*inst, CHAN_Z)) {
-         STORE(&d[CHAN_Z], 0, CHAN_Z);
-      }
-      if (IS_CHANNEL_ENABLED( *inst, CHAN_W )) {
-         STORE( &mach->Temps[TEMP_1_I].xyzw[TEMP_1_C], 0, CHAN_W );
-      }
+      exec_xpd(mach, inst);
       break;
 
    case TGSI_OPCODE_ABS:
@@ -2957,12 +3095,7 @@ exec_instruction(
       break;
 
    case TGSI_OPCODE_RCC:
-      FETCH(&r[0], 0, CHAN_X);
-      micro_div(&r[0], &mach->Temps[TEMP_1_I].xyzw[TEMP_1_C], &r[0]);
-      micro_float_clamp(&r[0], &r[0]);
-      FOR_EACH_ENABLED_CHANNEL(*inst, chan_index) {
-         STORE(&r[0], 0, chan_index);
-      }
+      exec_scalar_unary(mach, inst, micro_rcc, TGSI_EXEC_DATA_FLOAT, TGSI_EXEC_DATA_FLOAT);
       break;
 
    case TGSI_OPCODE_DPH:
@@ -3006,52 +3139,7 @@ exec_instruction(
       break;
 
    case TGSI_OPCODE_RFL:
-      if (IS_CHANNEL_ENABLED(*inst, CHAN_X) ||
-          IS_CHANNEL_ENABLED(*inst, CHAN_Y) ||
-          IS_CHANNEL_ENABLED(*inst, CHAN_Z)) {
-         /* r0 = dp3(src0, src0) */
-         FETCH(&r[2], 0, CHAN_X);
-         micro_mul(&r[0], &r[2], &r[2]);
-         FETCH(&r[4], 0, CHAN_Y);
-         micro_mul(&r[8], &r[4], &r[4]);
-         micro_add(&r[0], &r[0], &r[8]);
-         FETCH(&r[6], 0, CHAN_Z);
-         micro_mul(&r[8], &r[6], &r[6]);
-         micro_add(&r[0], &r[0], &r[8]);
-
-         /* r1 = dp3(src0, src1) */
-         FETCH(&r[3], 1, CHAN_X);
-         micro_mul(&r[1], &r[2], &r[3]);
-         FETCH(&r[5], 1, CHAN_Y);
-         micro_mul(&r[8], &r[4], &r[5]);
-         micro_add(&r[1], &r[1], &r[8]);
-         FETCH(&r[7], 1, CHAN_Z);
-         micro_mul(&r[8], &r[6], &r[7]);
-         micro_add(&r[1], &r[1], &r[8]);
-
-         /* r1 = 2 * r1 / r0 */
-         micro_add(&r[1], &r[1], &r[1]);
-         micro_div(&r[1], &r[1], &r[0]);
-
-         if (IS_CHANNEL_ENABLED(*inst, CHAN_X)) {
-            micro_mul(&r[2], &r[2], &r[1]);
-            micro_sub(&r[2], &r[2], &r[3]);
-            STORE(&r[2], 0, CHAN_X);
-         }
-         if (IS_CHANNEL_ENABLED(*inst, CHAN_Y)) {
-            micro_mul(&r[4], &r[4], &r[1]);
-            micro_sub(&r[4], &r[4], &r[5]);
-            STORE(&r[4], 0, CHAN_Y);
-         }
-         if (IS_CHANNEL_ENABLED(*inst, CHAN_Z)) {
-            micro_mul(&r[6], &r[6], &r[1]);
-            micro_sub(&r[6], &r[6], &r[7]);
-            STORE(&r[6], 0, CHAN_Z);
-         }
-      }
-      if (IS_CHANNEL_ENABLED(*inst, CHAN_W)) {
-         STORE(&mach->Temps[TEMP_1_I].xyzw[TEMP_1_C], 0, CHAN_W);
-      }
+      exec_rfl(mach, inst);
       break;
 
    case TGSI_OPCODE_SEQ:
@@ -3059,9 +3147,7 @@ exec_instruction(
       break;
 
    case TGSI_OPCODE_SFL:
-      FOR_EACH_ENABLED_CHANNEL(*inst, chan_index) {
-         STORE(&mach->Temps[TEMP_0_I].xyzw[TEMP_0_C], 0, chan_index);
-      }
+      exec_vector(mach, inst, micro_sfl, TGSI_EXEC_DATA_FLOAT);
       break;
 
    case TGSI_OPCODE_SGT:
@@ -3081,9 +3167,7 @@ exec_instruction(
       break;
 
    case TGSI_OPCODE_STR:
-      FOR_EACH_ENABLED_CHANNEL(*inst, chan_index) {
-         STORE(&mach->Temps[TEMP_1_I].xyzw[TEMP_1_C], 0, chan_index);
-      }
+      exec_vector(mach, inst, micro_str, TGSI_EXEC_DATA_FLOAT);
       break;
 
    case TGSI_OPCODE_TEX:
@@ -3140,42 +3224,7 @@ exec_instruction(
       break;
 
    case TGSI_OPCODE_X2D:
-      FETCH(&r[0], 1, CHAN_X);
-      FETCH(&r[1], 1, CHAN_Y);
-      if (IS_CHANNEL_ENABLED(*inst, CHAN_X) ||
-          IS_CHANNEL_ENABLED(*inst, CHAN_Z)) {
-         FETCH(&r[2], 2, CHAN_X);
-         micro_mul(&r[2], &r[2], &r[0]);
-         FETCH(&r[3], 2, CHAN_Y);
-         micro_mul(&r[3], &r[3], &r[1]);
-         micro_add(&r[2], &r[2], &r[3]);
-         FETCH(&r[3], 0, CHAN_X);
-         micro_add(&d[CHAN_X], &r[2], &r[3]);
-         
-      }
-      if (IS_CHANNEL_ENABLED(*inst, CHAN_Y) ||
-          IS_CHANNEL_ENABLED(*inst, CHAN_W)) {
-         FETCH(&r[2], 2, CHAN_Z);
-         micro_mul(&r[2], &r[2], &r[0]);
-         FETCH(&r[3], 2, CHAN_W);
-         micro_mul(&r[3], &r[3], &r[1]);
-         micro_add(&r[2], &r[2], &r[3]);
-         FETCH(&r[3], 0, CHAN_Y);
-         micro_add(&d[CHAN_Y], &r[2], &r[3]);
-         
-      }
-      if (IS_CHANNEL_ENABLED(*inst, CHAN_X)) {
-         STORE(&d[CHAN_X], 0, CHAN_X);
-      }
-      if (IS_CHANNEL_ENABLED(*inst, CHAN_Y)) {
-         STORE(&d[CHAN_Y], 0, CHAN_Y);
-      }
-      if (IS_CHANNEL_ENABLED(*inst, CHAN_Z)) {
-         STORE(&d[CHAN_X], 0, CHAN_Z);
-      }
-      if (IS_CHANNEL_ENABLED(*inst, CHAN_W)) {
-         STORE(&d[CHAN_Y], 0, CHAN_W);
-      }
+      exec_x2d(mach, inst);
       break;
 
    case TGSI_OPCODE_ARA:
@@ -3283,23 +3332,7 @@ exec_instruction(
       break;
 
    case TGSI_OPCODE_SCS:
-      if( IS_CHANNEL_ENABLED( *inst, CHAN_X ) || IS_CHANNEL_ENABLED( *inst, CHAN_Y ) ) {
-         FETCH( &r[0], 0, CHAN_X );
-         if (IS_CHANNEL_ENABLED(*inst, CHAN_X)) {
-            micro_cos(&r[1], &r[0]);
-            STORE(&r[1], 0, CHAN_X);
-         }
-         if (IS_CHANNEL_ENABLED(*inst, CHAN_Y)) {
-            micro_sin(&r[1], &r[0]);
-            STORE(&r[1], 0, CHAN_Y);
-         }
-      }
-      if( IS_CHANNEL_ENABLED( *inst, CHAN_Z ) ) {
-         STORE( &mach->Temps[TEMP_0_I].xyzw[TEMP_0_C], 0, CHAN_Z );
-      }
-      if( IS_CHANNEL_ENABLED( *inst, CHAN_W ) ) {
-         STORE( &mach->Temps[TEMP_1_I].xyzw[TEMP_1_C], 0, CHAN_W );
-      }
+      exec_scs(mach, inst);
       break;
 
    case TGSI_OPCODE_NRM:
@@ -3684,14 +3717,6 @@ tgsi_exec_machine_run( struct tgsi_exec_machine *mach )
       mach->Primitives[0] = 0;
    }
 
-   for (i = 0; i < QUAD_SIZE; i++) {
-      mach->Temps[TEMP_CC_I].xyzw[TEMP_CC_C].u[i] =
-         (TGSI_EXEC_CC_EQ << TGSI_EXEC_CC_X_SHIFT) |
-         (TGSI_EXEC_CC_EQ << TGSI_EXEC_CC_Y_SHIFT) |
-         (TGSI_EXEC_CC_EQ << TGSI_EXEC_CC_Z_SHIFT) |
-         (TGSI_EXEC_CC_EQ << TGSI_EXEC_CC_W_SHIFT);
-   }
-
    /* execute declarations (interpolants) */
    for (i = 0; i < mach->NumDeclarations; i++) {
       exec_declaration( mach, mach->Declarations+i );
diff --git a/src/gallium/auxiliary/tgsi/tgsi_exec.h b/src/gallium/auxiliary/tgsi/tgsi_exec.h
index 9d62c1d7e7..bd5492e949 100644
--- a/src/gallium/auxiliary/tgsi/tgsi_exec.h
+++ b/src/gallium/auxiliary/tgsi/tgsi_exec.h
@@ -131,34 +131,15 @@ struct tgsi_sampler
 #define TGSI_EXEC_TEMP_PRIMITIVE_I  (TGSI_EXEC_NUM_TEMPS + 2)
 #define TGSI_EXEC_TEMP_PRIMITIVE_C  2
 
-/* NVIDIA condition code (CC) vector
- */
-#define TGSI_EXEC_CC_GT       0x01
-#define TGSI_EXEC_CC_EQ       0x02
-#define TGSI_EXEC_CC_LT       0x04
-#define TGSI_EXEC_CC_UN       0x08
-
-#define TGSI_EXEC_CC_X_MASK   0x000000ff
-#define TGSI_EXEC_CC_X_SHIFT  0
-#define TGSI_EXEC_CC_Y_MASK   0x0000ff00
-#define TGSI_EXEC_CC_Y_SHIFT  8
-#define TGSI_EXEC_CC_Z_MASK   0x00ff0000
-#define TGSI_EXEC_CC_Z_SHIFT  16
-#define TGSI_EXEC_CC_W_MASK   0xff000000
-#define TGSI_EXEC_CC_W_SHIFT  24
-
-#define TGSI_EXEC_TEMP_CC_I         (TGSI_EXEC_NUM_TEMPS + 2)
-#define TGSI_EXEC_TEMP_CC_C         3
-
-#define TGSI_EXEC_TEMP_THREE_I      (TGSI_EXEC_NUM_TEMPS + 3)
-#define TGSI_EXEC_TEMP_THREE_C      0
+#define TGSI_EXEC_TEMP_THREE_I      (TGSI_EXEC_NUM_TEMPS + 2)
+#define TGSI_EXEC_TEMP_THREE_C      3
 
 #define TGSI_EXEC_TEMP_HALF_I       (TGSI_EXEC_NUM_TEMPS + 3)
-#define TGSI_EXEC_TEMP_HALF_C       1
+#define TGSI_EXEC_TEMP_HALF_C       0
 
 /* execution mask, each value is either 0 or ~0 */
 #define TGSI_EXEC_MASK_I            (TGSI_EXEC_NUM_TEMPS + 3)
-#define TGSI_EXEC_MASK_C            2
+#define TGSI_EXEC_MASK_C            1
 
 /* 4 register buffer for various purposes */
 #define TGSI_EXEC_TEMP_R0           (TGSI_EXEC_NUM_TEMPS + 4)
diff --git a/src/gallium/drivers/failover/SConscript b/src/gallium/drivers/failover/SConscript
index f8e9b1b491..9347431f4e 100644
--- a/src/gallium/drivers/failover/SConscript
+++ b/src/gallium/drivers/failover/SConscript
@@ -10,4 +10,6 @@ failover = env.ConvenienceLibrary(
 		'fo_context.c',
 	])
 
+env.Alias('failover', failover)
+
 Export('failover')
diff --git a/src/gallium/drivers/galahad/SConscript b/src/gallium/drivers/galahad/SConscript
index b398a3f061..1343dfd454 100644
--- a/src/gallium/drivers/galahad/SConscript
+++ b/src/gallium/drivers/galahad/SConscript
@@ -10,4 +10,6 @@ galahad = env.ConvenienceLibrary(
 		'glhd_screen.c',
 	])
 
+env.Alias('galahad', galahad)
+
 Export('galahad')
diff --git a/src/gallium/drivers/i915/SConscript b/src/gallium/drivers/i915/SConscript
index d4bf6fef13..8f5deed64a 100644
--- a/src/gallium/drivers/i915/SConscript
+++ b/src/gallium/drivers/i915/SConscript
@@ -2,10 +2,6 @@ Import('*')
 
 env = env.Clone()
 
-if msvc:
-	print 'warning: not building i915g'
-	Return()
-
 i915 = env.ConvenienceLibrary(
 	target = 'i915',
 	source = [
@@ -34,4 +30,6 @@ i915 = env.ConvenienceLibrary(
 		'i915_resource_texture.c',
 	])
 
+env.Alias('i915', i915)
+
 Export('i915')
diff --git a/src/gallium/drivers/i965/SConscript b/src/gallium/drivers/i965/SConscript
index 119f914a16..019af682f6 100644
--- a/src/gallium/drivers/i965/SConscript
+++ b/src/gallium/drivers/i965/SConscript
@@ -2,10 +2,6 @@ Import('*')
 
 env = env.Clone()
 
-if msvc:
-	print 'warning: not building i965g'
-	Return();
-
 i965 = env.ConvenienceLibrary(
 	target = 'i965',
 	source = [
diff --git a/src/gallium/drivers/llvmpipe/README b/src/gallium/drivers/llvmpipe/README
index ec30d4d708..e9374cc6ef 100644
--- a/src/gallium/drivers/llvmpipe/README
+++ b/src/gallium/drivers/llvmpipe/README
@@ -1,53 +1,6 @@
 LLVMPIPE -- a fork of softpipe that employs LLVM for code generation.
 
 
-Status
-======
-
-Done so far is:
-
- - the whole fragment pipeline is code generated in a single function
- 
-   - input interpolation
-   
-   - depth testing
- 
-   - texture sampling
-     - 1D/2D/3D/cube maps supported
-     - all texture wrap modes supported
-     - all texture filtering modes supported
-     - perhaps not all texture formats yet supported
-   
-   - fragment shader TGSI translation
-     - same level of support as the TGSI SSE2 exec machine, with the exception
-       we don't fallback to TGSI interpretation when an unsupported opcode is
-       found, but just ignore it
-     - done in SoA layout
-     - input interpolation also code generated
- 
-   - alpha testing
- 
-   - blend (including logic ops)
-     - both in SoA and AoS layouts, but only the former used for now
- 
- - code is generic
-   - intermediates can be vectors of floats, ubytes, fixed point, etc, and of
-     any width and length
-   - not all operations are implemented for these types yet though
-
-Most mesa/progs/demos/* work. 
-
-To do (probably by this order):
-
- - code generate stipple and stencil testing
-
- - translate TGSI control flow instructions, and all other remaining opcodes
- 
- - integrate with the draw module for VS code generation
-
- - code generate the triangle setup and rasterization
-
-
 Requirements
 ============
 
@@ -98,7 +51,7 @@ Building
 
 To build everything on Linux invoke scons as:
 
-  scons debug=yes statetrackers=mesa drivers=llvmpipe winsys=xlib dri=false
+  scons build=debug libgl-xlib
 
 Alternatively, you can build it with GNU make, if you prefer, by invoking it as
 
@@ -108,19 +61,16 @@ but the rest of these instructions assume that scons is used.
 
 For windows is everything the except except the winsys:
 
-  scons debug=yes statetrackers=mesa drivers=llvmpipe winsys=gdi dri=false
+  scons build=debug libgl-gdi
 
 Using
 =====
 
-On Linux, building will create a drop-in alternative for libGL.so. To use it
-set the environment variables:
-
-  export LD_LIBRARY_PATH=$PWD/build/linux-x86_64-debug/lib:$LD_LIBRARY_PATH
+On Linux, building will create a drop-in alternative for libGL.so into
 
-or
+  build/foo/gallium/targets/libgl-xlib/libGL.so
 
-  export LD_LIBRARY_PATH=$PWD/build/linux-x86-debug/lib:$LD_LIBRARY_PATH
+To use it set the LD_LIBRARY_PATH environment variable accordingly.
 
 For performance evaluation pass debug=no to scons, and use the corresponding
 lib directory without the "-debug" suffix.
@@ -136,7 +86,7 @@ Profiling
 
 To profile llvmpipe you should pass the options
 
-  scons debug=no profile=yes <same-as-before>
+  scons build=profile <same-as-before>
 
 This will ensure that frame pointers are used both in C and JIT functions, and
 that no tail call optimizations are done by gcc.
@@ -200,5 +150,4 @@ Development Notes
   interfaces very closely, and appear to be complete enough for code
   generation. See 
   http://npcontemplation.blogspot.com/2008/06/secret-of-llvm-c-bindings.html
-  for a stand-alone example.
-  See the llvm-c/Core.h file for reference.
+  for a stand-alone example.  See the llvm-c/Core.h file for reference.
diff --git a/src/gallium/drivers/llvmpipe/SConscript b/src/gallium/drivers/llvmpipe/SConscript
index 49950153a4..26b258b956 100644
--- a/src/gallium/drivers/llvmpipe/SConscript
+++ b/src/gallium/drivers/llvmpipe/SConscript
@@ -9,8 +9,6 @@ if not env['llvm']:
 
 env = env.Clone()
 
-env.Tool('udis86')
-
 env.Append(CPPPATH = ['.'])
 
 env.CodeGenerate(
@@ -78,6 +76,8 @@ llvmpipe = env.ConvenienceLibrary(
         lp_tile_soa_os,
 	])
 
+env.Alias('llvmpipe', llvmpipe)
+
 
 if env['platform'] != 'embedded':
     env = env.Clone()
@@ -92,7 +92,7 @@ if env['platform'] != 'embedded':
 	'sincos',
     ]
 
-    if not msvc:
+    if not env['msvc']:
         tests.append('round')
 
     for test in tests:
diff --git a/src/gallium/drivers/llvmpipe/lp_rast.c b/src/gallium/drivers/llvmpipe/lp_rast.c
index d358a98394..decf3bd449 100644
--- a/src/gallium/drivers/llvmpipe/lp_rast.c
+++ b/src/gallium/drivers/llvmpipe/lp_rast.c
@@ -485,8 +485,11 @@ static void
 lp_rast_end_query(struct lp_rasterizer_task *task,
                   const union lp_rast_cmd_arg arg)
 {
-   task->query->count[task->thread_index] += task->vis_counter;
-   task->query = NULL;
+   assert(task->query);
+   if (task->query) {
+      task->query->count[task->thread_index] += task->vis_counter;
+      task->query = NULL;
+   }
 }
 
 
diff --git a/src/gallium/drivers/llvmpipe/lp_setup_line.c b/src/gallium/drivers/llvmpipe/lp_setup_line.c
index 827413bb33..29c231714e 100644
--- a/src/gallium/drivers/llvmpipe/lp_setup_line.c
+++ b/src/gallium/drivers/llvmpipe/lp_setup_line.c
@@ -569,7 +569,10 @@ try_setup_line( struct lp_setup_context *setup,
       return TRUE;
    }
 
-   u_rect_find_intersection(&setup->draw_region, &bbox);
+   /* Can safely discard negative regions:
+    */
+   bbox.x0 = MAX2(bbox.x0, 0);
+   bbox.y0 = MAX2(bbox.y0, 0);
 
    line = lp_setup_alloc_triangle(scene,
                                   key->num_inputs,
@@ -680,24 +683,26 @@ try_setup_line( struct lp_setup_context *setup,
     * these planes elsewhere.
     */
    if (nr_planes == 8) {
+      const struct u_rect *scissor = &setup->scissor;
+
       plane[4].dcdx = -1;
       plane[4].dcdy = 0;
-      plane[4].c = 1-bbox.x0;
+      plane[4].c = 1-scissor->x0;
       plane[4].eo = 1;
 
       plane[5].dcdx = 1;
       plane[5].dcdy = 0;
-      plane[5].c = bbox.x1+1;
+      plane[5].c = scissor->x1+1;
       plane[5].eo = 0;
 
       plane[6].dcdx = 0;
       plane[6].dcdy = 1;
-      plane[6].c = 1-bbox.y0;
+      plane[6].c = 1-scissor->y0;
       plane[6].eo = 1;
 
       plane[7].dcdx = 0;
       plane[7].dcdy = -1;
-      plane[7].c = bbox.y1+1;
+      plane[7].c = scissor->y1+1;
       plane[7].eo = 0;
    }
 
diff --git a/src/gallium/drivers/llvmpipe/lp_setup_tri.c b/src/gallium/drivers/llvmpipe/lp_setup_tri.c
index 4ab0b72a57..bfb6bf277b 100644
--- a/src/gallium/drivers/llvmpipe/lp_setup_tri.c
+++ b/src/gallium/drivers/llvmpipe/lp_setup_tri.c
@@ -295,7 +295,12 @@ do_triangle_ccw(struct lp_setup_context *setup,
       return TRUE;
    }
 
-   u_rect_find_intersection(&setup->draw_region, &bbox);
+   /* Can safely discard negative regions, but need to keep hold of
+    * information about when the triangle extends past screen
+    * boundaries.  See trimmed_box in lp_setup_bin_triangle().
+    */
+   bbox.x0 = MAX2(bbox.x0, 0);
+   bbox.y0 = MAX2(bbox.y0, 0);
 
    tri = lp_setup_alloc_triangle(scene,
                                  key->num_inputs,
@@ -501,24 +506,26 @@ do_triangle_ccw(struct lp_setup_context *setup,
     * these planes elsewhere.
     */
    if (nr_planes == 7) {
+      const struct u_rect *scissor = &setup->scissor;
+
       plane[3].dcdx = -1;
       plane[3].dcdy = 0;
-      plane[3].c = 1-bbox.x0;
+      plane[3].c = 1-scissor->x0;
       plane[3].eo = 1;
 
       plane[4].dcdx = 1;
       plane[4].dcdy = 0;
-      plane[4].c = bbox.x1+1;
+      plane[4].c = scissor->x1+1;
       plane[4].eo = 0;
 
       plane[5].dcdx = 0;
       plane[5].dcdy = 1;
-      plane[5].c = 1-bbox.y0;
+      plane[5].c = 1-scissor->y0;
       plane[5].eo = 1;
 
       plane[6].dcdx = 0;
       plane[6].dcdy = -1;
-      plane[6].c = bbox.y1+1;
+      plane[6].c = scissor->y1+1;
       plane[6].eo = 0;
    }
 
@@ -559,6 +566,7 @@ lp_setup_bin_triangle( struct lp_setup_context *setup,
                        int nr_planes )
 {
    struct lp_scene *scene = setup->scene;
+   struct u_rect trimmed_box = *bbox;   
    int i;
 
    /* What is the largest power-of-two boundary this triangle crosses:
@@ -572,6 +580,13 @@ lp_setup_bin_triangle( struct lp_setup_context *setup,
    int sz = floor_pot((bbox->x1 - (bbox->x0 & ~3)) |
 		      (bbox->y1 - (bbox->y0 & ~3)));
 
+   /* Now apply scissor, etc to the bounding box.  Could do this
+    * earlier, but it confuses the logic for tri-16 and would force
+    * the rasterizer to also respect scissor, etc, just for the rare
+    * cases where a small triangle extends beyond the scissor.
+    */
+   u_rect_find_intersection(&setup->draw_region, &trimmed_box);
+
    /* Determine which tile(s) intersect the triangle's bounding box
     */
    if (dx < TILE_SIZE)
@@ -626,15 +641,16 @@ lp_setup_bin_triangle( struct lp_setup_context *setup,
       struct lp_rast_plane *plane = GET_PLANES(tri);
       int c[MAX_PLANES];
       int ei[MAX_PLANES];
+
       int eo[MAX_PLANES];
       int xstep[MAX_PLANES];
       int ystep[MAX_PLANES];
       int x, y;
 
-      int ix0 = bbox->x0 / TILE_SIZE;
-      int iy0 = bbox->y0 / TILE_SIZE;
-      int ix1 = bbox->x1 / TILE_SIZE;
-      int iy1 = bbox->y1 / TILE_SIZE;
+      int ix0 = trimmed_box.x0 / TILE_SIZE;
+      int iy0 = trimmed_box.y0 / TILE_SIZE;
+      int ix1 = trimmed_box.x1 / TILE_SIZE;
+      int iy1 = trimmed_box.y1 / TILE_SIZE;
       
       for (i = 0; i < nr_planes; i++) {
          c[i] = (plane[i].c + 
@@ -799,6 +815,16 @@ static void triangle_both( struct lp_setup_context *setup,
 {
    float area = calc_area(v0, v1, v2);
 
+   if (0) {
+      assert(!util_is_inf_or_nan(v0[0][0]));
+      assert(!util_is_inf_or_nan(v0[0][1]));
+      assert(!util_is_inf_or_nan(v1[0][0]));
+      assert(!util_is_inf_or_nan(v1[0][1]));
+      assert(!util_is_inf_or_nan(v2[0][0]));
+      assert(!util_is_inf_or_nan(v2[0][1]));
+      assert(!util_is_inf_or_nan(area));
+   }
+
    if (area > 0.0f) 
       retry_triangle_ccw( setup, v0, v1, v2, setup->ccw_is_frontface );
    else if (area < 0.0f)
diff --git a/src/gallium/drivers/llvmpipe/lp_state_fs.c b/src/gallium/drivers/llvmpipe/lp_state_fs.c
index 9c39845980..2c4943a69f 100644
--- a/src/gallium/drivers/llvmpipe/lp_state_fs.c
+++ b/src/gallium/drivers/llvmpipe/lp_state_fs.c
@@ -1074,11 +1074,11 @@ llvmpipe_bind_fs_state(struct pipe_context *pipe, void *fs)
 
    draw_flush(llvmpipe->draw);
 
+   llvmpipe->fs = (struct lp_fragment_shader *) fs;
+
    draw_bind_fragment_shader(llvmpipe->draw,
                              (llvmpipe->fs ? llvmpipe->fs->draw_data : NULL));
 
-   llvmpipe->fs = fs;
-
    llvmpipe->dirty |= LP_NEW_FS;
 }
 
diff --git a/src/gallium/drivers/nv50/nv50_shader_state.c b/src/gallium/drivers/nv50/nv50_shader_state.c
index 6c41e8f456..306aa81d98 100644
--- a/src/gallium/drivers/nv50/nv50_shader_state.c
+++ b/src/gallium/drivers/nv50/nv50_shader_state.c
@@ -39,6 +39,9 @@ nv50_transfer_constbuf(struct nv50_context *nv50,
    uint32_t *map;
    unsigned count, start;
 
+   if (buf == NULL)
+      return;
+
    map = pipe_buffer_map(pipe, buf, PIPE_TRANSFER_READ, &transfer);
    if (!map)
       return;
diff --git a/src/gallium/drivers/r300/SConscript b/src/gallium/drivers/r300/SConscript
index bf023daaa5..b49db93799 100644
--- a/src/gallium/drivers/r300/SConscript
+++ b/src/gallium/drivers/r300/SConscript
@@ -39,5 +39,7 @@ r300 = env.ConvenienceLibrary(
         'r300_transfer.c',
     ] + r300compiler) + r300compiler
 
+env.Alias('r300', r300)
+
 Export('r300')
 
diff --git a/src/gallium/drivers/r300/r300_state.c b/src/gallium/drivers/r300/r300_state.c
index f2479a994c..f513f87ca4 100644
--- a/src/gallium/drivers/r300/r300_state.c
+++ b/src/gallium/drivers/r300/r300_state.c
@@ -1789,7 +1789,7 @@ static void r300_set_constant_buffer(struct pipe_context *pipe,
 {
     struct r300_context* r300 = r300_context(pipe);
     struct r300_constant_buffer *cbuf;
-    uint32_t *mapped = r300_buffer(buf)->user_buffer;
+    uint32_t *mapped;
 
     switch (shader) {
         case PIPE_SHADER_VERTEX:
diff --git a/src/gallium/drivers/r600/SConscript b/src/gallium/drivers/r600/SConscript
index bf0ad8571b..3fc1fa94c2 100644
--- a/src/gallium/drivers/r600/SConscript
+++ b/src/gallium/drivers/r600/SConscript
@@ -25,10 +25,14 @@ r600 = env.ConvenienceLibrary(
         'r600_resource.c',
         'r600_shader.c',
         'r600_state.c',
+        'r600_state_common.c',
         'r600_texture.c',
+        'r600_translate.c',
         'r700_asm.c',
         'evergreen_state.c',
         'eg_asm.c',
     ])
 
+env.Alias('r600', r600)
+
 Export('r600')
diff --git a/src/gallium/drivers/r600/evergreen_state.c b/src/gallium/drivers/r600/evergreen_state.c
index 7222348506..4725b5e75e 100644
--- a/src/gallium/drivers/r600/evergreen_state.c
+++ b/src/gallium/drivers/r600/evergreen_state.c
@@ -281,14 +281,21 @@ static void *evergreen_create_rs_state(struct pipe_context *ctx,
 	tmp = (unsigned)(state->point_size * 8.0);
 	r600_pipe_state_add_reg(rstate, R_028A00_PA_SU_POINT_SIZE, S_028A00_HEIGHT(tmp) | S_028A00_WIDTH(tmp), 0xFFFFFFFF, NULL);
 	r600_pipe_state_add_reg(rstate, R_028A04_PA_SU_POINT_MINMAX, 0x80000000, 0xFFFFFFFF, NULL);
-	r600_pipe_state_add_reg(rstate, R_028A08_PA_SU_LINE_CNTL, 0x00000008, 0xFFFFFFFF, NULL);
+
+	tmp = (unsigned)state->line_width * 8;
+	r600_pipe_state_add_reg(rstate, R_028A08_PA_SU_LINE_CNTL, S_028A08_WIDTH(tmp), 0xFFFFFFFF, NULL);
+
 	r600_pipe_state_add_reg(rstate, R_028C00_PA_SC_LINE_CNTL, 0x00000400, 0xFFFFFFFF, NULL);
 	r600_pipe_state_add_reg(rstate, R_028C0C_PA_CL_GB_VERT_CLIP_ADJ, 0x3F800000, 0xFFFFFFFF, NULL);
 	r600_pipe_state_add_reg(rstate, R_028C10_PA_CL_GB_VERT_DISC_ADJ, 0x3F800000, 0xFFFFFFFF, NULL);
 	r600_pipe_state_add_reg(rstate, R_028C14_PA_CL_GB_HORZ_CLIP_ADJ, 0x3F800000, 0xFFFFFFFF, NULL);
 	r600_pipe_state_add_reg(rstate, R_028C18_PA_CL_GB_HORZ_DISC_ADJ, 0x3F800000, 0xFFFFFFFF, NULL);
 	r600_pipe_state_add_reg(rstate, R_028B7C_PA_SU_POLY_OFFSET_CLAMP, 0x0, 0xFFFFFFFF, NULL);
-	r600_pipe_state_add_reg(rstate, R_028C08_PA_SU_VTX_CNTL, 0x00000005, 0xFFFFFFFF, NULL);
+
+	r600_pipe_state_add_reg(rstate, R_028C08_PA_SU_VTX_CNTL,
+				S_028C08_PIX_CENTER_HALF(state->gl_rasterization_rules),
+				0xFFFFFFFF, NULL);
+
 	r600_pipe_state_add_reg(rstate, R_02820C_PA_SC_CLIPRECT_RULE, clip_rule, 0xFFFFFFFF, NULL);
 	return rstate;
 }
@@ -826,6 +833,13 @@ static void evergreen_set_constant_buffer(struct pipe_context *ctx, uint shader,
 	struct r600_pipe_context *rctx = (struct r600_pipe_context *)ctx;
 	struct r600_resource *rbuffer = (struct r600_resource*)buffer;
 
+	/* Note that the state tracker can unbind constant buffers by
+	 * passing NULL here.
+	 */
+	if (buffer == NULL) {
+		return;
+	}
+
 	switch (shader) {
 	case PIPE_SHADER_VERTEX:
 		rctx->vs_const_buffer.nregs = 0;
diff --git a/src/gallium/drivers/r600/evergreend.h b/src/gallium/drivers/r600/evergreend.h
index 8e96f9355e..a337916c09 100644
--- a/src/gallium/drivers/r600/evergreend.h
+++ b/src/gallium/drivers/r600/evergreend.h
@@ -1636,6 +1636,9 @@
 #define R_028980_ALU_CONST_CACHE_VS_0                0x00028980
 #define R_028A04_PA_SU_POINT_MINMAX                  0x00028A04
 #define R_028A08_PA_SU_LINE_CNTL                     0x00028A08
+#define   S_028A08_WIDTH(x)                            (((x) & 0xFFFF) << 0)
+#define   G_028A08_WIDTH(x)                            (((x) >> 0) & 0xFFFF)
+#define   C_028A08_WIDTH                               0xFFFF0000
 #define R_028A10_VGT_OUTPUT_PATH_CNTL                0x00028A10
 #define R_028A14_VGT_HOS_CNTL                        0x00028A14
 #define R_028A18_VGT_HOS_MAX_TESS_LEVEL              0x00028A18
@@ -1687,6 +1690,9 @@
 #define R_028C00_PA_SC_LINE_CNTL                     0x00028C00
 #define R_028C04_PA_SC_AA_CONFIG                     0x00028C04
 #define R_028C08_PA_SU_VTX_CNTL                      0x00028C08
+#define   S_028C08_PIX_CENTER_HALF(x)                  (((x) & 0x1) << 0)
+#define   G_028C08_PIX_CENTER_HALF(x)                  (((x) >> 0) & 0x1)
+#define   C_028C08_PIX_CENTER_HALF                     0xFFFFFFFE
 #define R_028C0C_PA_CL_GB_VERT_CLIP_ADJ              0x00028C0C
 #define R_028C10_PA_CL_GB_VERT_DISC_ADJ              0x00028C10
 #define R_028C14_PA_CL_GB_HORZ_CLIP_ADJ              0x00028C14
diff --git a/src/gallium/drivers/r600/r600.h b/src/gallium/drivers/r600/r600.h
index 62d983269f..5ec607bbd8 100644
--- a/src/gallium/drivers/r600/r600.h
+++ b/src/gallium/drivers/r600/r600.h
@@ -112,7 +112,8 @@ struct r600_tiling_info *r600_get_tiling_info(struct radeon *radeon);
 /* r600_bo.c */
 struct r600_bo;
 struct r600_bo *r600_bo(struct radeon *radeon,
-				  unsigned size, unsigned alignment, unsigned usage);
+                        unsigned size, unsigned alignment,
+                        unsigned binding, unsigned usage);
 struct r600_bo *r600_bo_handle(struct radeon *radeon,
 			       unsigned handle, unsigned *array_mode);
 void *r600_bo_map(struct radeon *radeon, struct r600_bo *bo, unsigned usage, void *ctx);
diff --git a/src/gallium/drivers/r600/r600_buffer.c b/src/gallium/drivers/r600/r600_buffer.c
index 455aa2e81f..ed97b6e69a 100644
--- a/src/gallium/drivers/r600/r600_buffer.c
+++ b/src/gallium/drivers/r600/r600_buffer.c
@@ -38,32 +38,6 @@
 
 extern struct u_resource_vtbl r600_buffer_vtbl;
 
-u32 r600_domain_from_usage(unsigned usage)
-{
-	u32 domain = RADEON_GEM_DOMAIN_GTT;
-
-	if (usage & PIPE_BIND_RENDER_TARGET) {
-		domain |= RADEON_GEM_DOMAIN_VRAM;
-	}
-	if (usage & PIPE_BIND_DEPTH_STENCIL) {
-		domain |= RADEON_GEM_DOMAIN_VRAM;
-	}
-	if (usage & PIPE_BIND_SAMPLER_VIEW) {
-		domain |= RADEON_GEM_DOMAIN_VRAM;
-	}
-	/* also need BIND_BLIT_SOURCE/DESTINATION ? */
-	if (usage & PIPE_BIND_VERTEX_BUFFER) {
-		domain |= RADEON_GEM_DOMAIN_GTT;
-	}
-	if (usage & PIPE_BIND_INDEX_BUFFER) {
-		domain |= RADEON_GEM_DOMAIN_GTT;
-	}
-	if (usage & PIPE_BIND_CONSTANT_BUFFER) {
-		domain |= RADEON_GEM_DOMAIN_VRAM;
-	}
-
-	return domain;
-}
 
 struct pipe_resource *r600_buffer_create(struct pipe_screen *screen,
 					 const struct pipe_resource *templ)
@@ -85,8 +59,7 @@ struct pipe_resource *r600_buffer_create(struct pipe_screen *screen,
 	rbuffer->r.base.b.screen = screen;
 	rbuffer->r.base.vtbl = &r600_buffer_vtbl;
 	rbuffer->r.size = rbuffer->r.base.b.width0;
-	rbuffer->r.domain = r600_domain_from_usage(rbuffer->r.base.b.bind);
-	bo = r600_bo((struct radeon*)screen->winsys, rbuffer->r.base.b.width0, alignment, rbuffer->r.base.b.bind);
+	bo = r600_bo((struct radeon*)screen->winsys, rbuffer->r.base.b.width0, alignment, rbuffer->r.base.b.bind, rbuffer->r.base.b.usage);
 	if (bo == NULL) {
 		FREE(rbuffer);
 		return NULL;
@@ -156,8 +129,9 @@ static void *r600_buffer_transfer_map(struct pipe_context *pipe,
 				r600_bo_reference((struct radeon*)pipe->winsys, &rbuffer->r.bo, NULL);
 				rbuffer->num_ranges = 0;
 				rbuffer->r.bo = r600_bo((struct radeon*)pipe->winsys,
-							     rbuffer->r.base.b.width0, 0,
-							     rbuffer->r.base.b.bind);
+                                                        rbuffer->r.base.b.width0, 0,
+                                                        rbuffer->r.base.b.bind,
+                                                        rbuffer->r.base.b.usage);
 				break;
 			}
 		}
diff --git a/src/gallium/drivers/r600/r600_pipe.c b/src/gallium/drivers/r600/r600_pipe.c
index 3564539843..128c998b93 100644
--- a/src/gallium/drivers/r600/r600_pipe.c
+++ b/src/gallium/drivers/r600/r600_pipe.c
@@ -77,13 +77,16 @@ static void r600_destroy_context(struct pipe_context *context)
 {
 	struct r600_pipe_context *rctx = (struct r600_pipe_context *)context;
 
+	rctx->context.delete_depth_stencil_alpha_state(&rctx->context, rctx->custom_dsa_flush);
+
 	r600_context_fini(&rctx->ctx);
+
+	util_blitter_destroy(rctx->blitter);
+
 	for (int i = 0; i < R600_PIPE_NSTATES; i++) {
 		free(rctx->states[i]);
 	}
 
-	util_blitter_destroy(rctx->blitter);
-
 	u_upload_destroy(rctx->upload_vb);
 	u_upload_destroy(rctx->upload_ib);
 
@@ -273,6 +276,7 @@ static int r600_get_param(struct pipe_screen* pscreen, enum pipe_cap param)
 	/* Unsupported features (boolean caps). */
 	case PIPE_CAP_TIMER_QUERY:
 	case PIPE_CAP_STREAM_OUTPUT:
+	case PIPE_CAP_PRIMITIVE_RESTART:
 	case PIPE_CAP_INDEP_BLEND_FUNC: /* FIXME allow this */
 		return 0;
 
@@ -428,6 +432,9 @@ static void r600_destroy_screen(struct pipe_screen* pscreen)
 
 	if (rscreen == NULL)
 		return;
+
+	radeon_decref(rscreen->radeon);
+
 	FREE(rscreen);
 }
 
diff --git a/src/gallium/drivers/r600/r600_resource.h b/src/gallium/drivers/r600/r600_resource.h
index d152285815..7a2d1f4412 100644
--- a/src/gallium/drivers/r600/r600_resource.h
+++ b/src/gallium/drivers/r600/r600_resource.h
@@ -35,7 +35,7 @@ struct r600_transfer {
 	/* Buffer transfer. */
 	struct pipe_transfer		*buffer_transfer;
 	unsigned			offset;
-	struct pipe_resource		*linear_texture;
+	struct pipe_resource		*staging_texture;
 };
 
 /* This gets further specialized into either buffer or texture
@@ -45,8 +45,6 @@ struct r600_transfer {
 struct r600_resource {
 	struct u_resource		base;
 	struct r600_bo			*bo;
-	u32				domain;
-	u32				flink;
 	u32				size;
 };
 
@@ -68,9 +66,6 @@ struct r600_resource_texture {
 
 void r600_init_screen_resource_functions(struct pipe_screen *screen);
 
-/* r600_buffer */
-u32 r600_domain_from_usage(unsigned usage);
-
 /* r600_texture */
 struct pipe_resource *r600_texture_create(struct pipe_screen *screen,
 					const struct pipe_resource *templ);
diff --git a/src/gallium/drivers/r600/r600_shader.c b/src/gallium/drivers/r600/r600_shader.c
index 4106587398..1a0b35d9bf 100644
--- a/src/gallium/drivers/r600/r600_shader.c
+++ b/src/gallium/drivers/r600/r600_shader.c
@@ -218,7 +218,7 @@ static int r600_pipe_shader(struct pipe_context *ctx, struct r600_pipe_shader *s
 
 	/* copy new shader */
 	if (shader->bo == NULL) {
-		shader->bo = r600_bo(rctx->radeon, rshader->bc.ndw * 4, 4096, 0);
+		shader->bo = r600_bo(rctx->radeon, rshader->bc.ndw * 4, 4096, 0, 0);
 		if (shader->bo == NULL) {
 			return -ENOMEM;
 		}
diff --git a/src/gallium/drivers/r600/r600_state.c b/src/gallium/drivers/r600/r600_state.c
index b3e0d49321..54cc79b154 100644
--- a/src/gallium/drivers/r600/r600_state.c
+++ b/src/gallium/drivers/r600/r600_state.c
@@ -469,12 +469,17 @@ static void *r600_create_rs_state(struct pipe_context *ctx,
 	r600_pipe_state_add_reg(rstate, R_028A00_PA_SU_POINT_SIZE, S_028A00_HEIGHT(tmp) | S_028A00_WIDTH(tmp), 0xFFFFFFFF, NULL);
 	r600_pipe_state_add_reg(rstate, R_028A04_PA_SU_POINT_MINMAX, 0x80000000, 0xFFFFFFFF, NULL);
 
-	tmp = (unsigned)(state->line_width * 8.0);
+	tmp = (unsigned)state->line_width * 8;
 	r600_pipe_state_add_reg(rstate, R_028A08_PA_SU_LINE_CNTL, S_028A08_WIDTH(tmp), 0xFFFFFFFF, NULL);
 
 	r600_pipe_state_add_reg(rstate, R_028A0C_PA_SC_LINE_STIPPLE, 0x00000005, 0xFFFFFFFF, NULL);
 	r600_pipe_state_add_reg(rstate, R_028A48_PA_SC_MPASS_PS_CNTL, 0x00000000, 0xFFFFFFFF, NULL);
 	r600_pipe_state_add_reg(rstate, R_028C00_PA_SC_LINE_CNTL, 0x00000400, 0xFFFFFFFF, NULL);
+        
+	r600_pipe_state_add_reg(rstate, R_028C08_PA_SU_VTX_CNTL,
+				S_028C08_PIX_CENTER_HALF(state->gl_rasterization_rules),
+				0xFFFFFFFF, NULL);
+
 	r600_pipe_state_add_reg(rstate, R_028C0C_PA_CL_GB_VERT_CLIP_ADJ, 0x3F800000, 0xFFFFFFFF, NULL);
 	r600_pipe_state_add_reg(rstate, R_028C10_PA_CL_GB_VERT_DISC_ADJ, 0x3F800000, 0xFFFFFFFF, NULL);
 	r600_pipe_state_add_reg(rstate, R_028C14_PA_CL_GB_HORZ_CLIP_ADJ, 0x3F800000, 0xFFFFFFFF, NULL);
@@ -1018,6 +1023,13 @@ static void r600_set_constant_buffer(struct pipe_context *ctx, uint shader, uint
 	struct r600_pipe_context *rctx = (struct r600_pipe_context *)ctx;
 	struct r600_resource *rbuffer = (struct r600_resource*)buffer;
 
+	/* Note that the state tracker can unbind constant buffers by
+	 * passing NULL here.
+	 */
+	if (buffer == NULL) {
+		return;
+	}
+
 	switch (shader) {
 	case PIPE_SHADER_VERTEX:
 		rctx->vs_const_buffer.nregs = 0;
diff --git a/src/gallium/drivers/r600/r600_texture.c b/src/gallium/drivers/r600/r600_texture.c
index 4ebd5b754b..a63990fbff 100644
--- a/src/gallium/drivers/r600/r600_texture.c
+++ b/src/gallium/drivers/r600/r600_texture.c
@@ -31,6 +31,7 @@
 #include <util/u_inlines.h>
 #include <util/u_memory.h>
 #include "state_tracker/drm_driver.h"
+#include "pipebuffer/pb_buffer.h"
 #include "r600_pipe.h"
 #include "r600_resource.h"
 #include "r600_state_inlines.h"
@@ -39,8 +40,8 @@
 
 extern struct u_resource_vtbl r600_texture_vtbl;
 
-/* Copy from a tiled texture to a detiled one. */
-static void r600_copy_from_tiled_texture(struct pipe_context *ctx, struct r600_transfer *rtransfer)
+/* Copy from a full GPU texture to a transfer's staging one. */
+static void r600_copy_to_staging_texture(struct pipe_context *ctx, struct r600_transfer *rtransfer)
 {
 	struct pipe_transfer *transfer = (struct pipe_transfer*)rtransfer;
 	struct pipe_resource *texture = transfer->resource;
@@ -48,15 +49,15 @@ static void r600_copy_from_tiled_texture(struct pipe_context *ctx, struct r600_t
 
 	subdst.face = 0;
 	subdst.level = 0;
-	ctx->resource_copy_region(ctx, rtransfer->linear_texture,
+	ctx->resource_copy_region(ctx, rtransfer->staging_texture,
 				subdst, 0, 0, 0, texture, transfer->sr,
 				transfer->box.x, transfer->box.y, transfer->box.z,
 				transfer->box.width, transfer->box.height);
 }
 
 
-/* Copy from a detiled texture to a tiled one. */
-static void r600_copy_into_tiled_texture(struct pipe_context *ctx, struct r600_transfer *rtransfer)
+/* Copy from a transfer's staging texture to a full GPU one. */
+static void r600_copy_from_staging_texture(struct pipe_context *ctx, struct r600_transfer *rtransfer)
 {
 	struct pipe_transfer *transfer = (struct pipe_transfer*)rtransfer;
 	struct pipe_resource *texture = transfer->resource;
@@ -66,7 +67,7 @@ static void r600_copy_into_tiled_texture(struct pipe_context *ctx, struct r600_t
 	subsrc.level = 0;
 	ctx->resource_copy_region(ctx, texture, transfer->sr,
 				  transfer->box.x, transfer->box.y, transfer->box.z,
-				  rtransfer->linear_texture, subsrc,
+				  rtransfer->staging_texture, subsrc,
 				  0, 0, 0,
 				  transfer->box.width, transfer->box.height);
 
@@ -283,7 +284,6 @@ r600_texture_create_object(struct pipe_screen *screen,
 	pipe_reference_init(&resource->base.b.reference, 1);
 	resource->base.b.screen = screen;
 	resource->bo = bo;
-	resource->domain = r600_domain_from_usage(resource->base.b.bind);
 	rtex->pitch_override = pitch_in_bytes_override;
 
 	if (array_mode)
@@ -293,7 +293,7 @@ r600_texture_create_object(struct pipe_screen *screen,
 	resource->size = rtex->size;
 
 	if (!resource->bo) {
-		resource->bo = r600_bo(radeon, rtex->size, 4096, 0);
+		resource->bo = r600_bo(radeon, rtex->size, 4096, base->bind, base->usage);
 		if (!resource->bo) {
 			FREE(rtex);
 			return NULL;
@@ -306,8 +306,14 @@ struct pipe_resource *r600_texture_create(struct pipe_screen *screen,
 						const struct pipe_resource *templ)
 {
 	unsigned array_mode = 0;
+        static int force_tiling = -1;
 
-	if (debug_get_bool_option("R600_FORCE_TILING", FALSE)) {
+        /* Would like some magic "get_bool_option_once" routine.
+	 */
+	if (force_tiling == -1)
+                force_tiling = debug_get_bool_option("R600_FORCE_TILING", FALSE);
+
+	if (force_tiling) {
 		if (!(templ->flags & R600_RESOURCE_FLAG_TRANSFER) &&
 		    !(templ->bind & PIPE_BIND_SCANOUT)) {
 			array_mode = V_038000_ARRAY_2D_TILED_THIN1;
@@ -434,10 +440,20 @@ int r600_texture_depth_flush(struct pipe_context *ctx,
 	}
 
 out:
+	/* XXX: only do this if the depth texture has actually changed:
+	 */
 	r600_blit_uncompress_depth_ptr(ctx, rtex);
 	return 0;
 }
 
+/* Needs adjustment for pixelformat:
+ */
+static INLINE unsigned u_box_volume( const struct pipe_box *box )
+{
+        return box->width * box->depth * box->height;
+};
+
+
 struct pipe_transfer* r600_texture_get_transfer(struct pipe_context *ctx,
 						struct pipe_resource *texture,
 						struct pipe_subresource sr,
@@ -448,6 +464,35 @@ struct pipe_transfer* r600_texture_get_transfer(struct pipe_context *ctx,
 	struct pipe_resource resource;
 	struct r600_transfer *trans;
 	int r;
+	boolean use_staging_texture = FALSE;
+	boolean discard = FALSE;
+
+	if (!(usage & PIPE_TRANSFER_READ) && (usage & PIPE_TRANSFER_DISCARD))
+		discard = TRUE;
+
+	/* We cannot map a tiled texture directly because the data is
+	 * in a different order, therefore we do detiling using a blit.
+	 *
+	 * Also, use a temporary in GTT memory for read transfers, as
+	 * the CPU is much happier reading out of cached system memory
+	 * than uncached VRAM.
+	 */
+	if (rtex->tiled)
+		use_staging_texture = TRUE;
+
+        if (usage & PIPE_TRANSFER_READ &&
+            u_box_volume(box) > 1024)
+                use_staging_texture = TRUE;
+
+        /* XXX: Use a staging texture for uploads if the underlying BO
+         * is busy.  No interface for checking that currently? so do
+         * it eagerly whenever the transfer doesn't require a readback
+         * and might block.
+         */
+        if ((usage & PIPE_TRANSFER_WRITE) &&
+            discard &&
+            !(usage & (PIPE_TRANSFER_DONTBLOCK | PIPE_TRANSFER_UNSYNCHRONIZED)))
+                use_staging_texture = TRUE;
 
 	trans = CALLOC_STRUCT(r600_transfer);
 	if (trans == NULL)
@@ -457,6 +502,10 @@ struct pipe_transfer* r600_texture_get_transfer(struct pipe_context *ctx,
 	trans->transfer.usage = usage;
 	trans->transfer.box = *box;
 	if (rtex->depth) {
+                /* XXX: only readback the rectangle which is being mapped?
+                 */
+                /* XXX: when discard is true, no need to read back from depth texture
+                 */
 		r = r600_texture_depth_flush(ctx, texture);
 		if (r < 0) {
 			R600_ERR("failed to create temporary texture to hold untiled copy\n");
@@ -464,7 +513,7 @@ struct pipe_transfer* r600_texture_get_transfer(struct pipe_context *ctx,
 			FREE(trans);
 			return NULL;
 		}
-	} else if (rtex->tiled) {
+	} else if (use_staging_texture) {
 		resource.target = PIPE_TEXTURE_2D;
 		resource.format = texture->format;
 		resource.width0 = box->width;
@@ -472,7 +521,7 @@ struct pipe_transfer* r600_texture_get_transfer(struct pipe_context *ctx,
 		resource.depth0 = 1;
 		resource.last_level = 0;
 		resource.nr_samples = 0;
-		resource.usage = PIPE_USAGE_DYNAMIC;
+		resource.usage = PIPE_USAGE_STAGING;
 		resource.bind = 0;
 		resource.flags = R600_RESOURCE_FLAG_TRANSFER;
 		/* For texture reading, the temporary (detiled) texture is used as
@@ -486,8 +535,8 @@ struct pipe_transfer* r600_texture_get_transfer(struct pipe_context *ctx,
 			resource.bind |= PIPE_BIND_SAMPLER_VIEW;
 		}
 		/* Create the temporary texture. */
-		trans->linear_texture = ctx->screen->resource_create(ctx->screen, &resource);
-		if (trans->linear_texture == NULL) {
+		trans->staging_texture = ctx->screen->resource_create(ctx->screen, &resource);
+		if (trans->staging_texture == NULL) {
 			R600_ERR("failed to create temporary texture to hold untiled copy\n");
 			pipe_resource_reference(&trans->transfer.resource, NULL);
 			FREE(trans);
@@ -495,11 +544,9 @@ struct pipe_transfer* r600_texture_get_transfer(struct pipe_context *ctx,
 		}
 
 		trans->transfer.stride =
-		  ((struct r600_resource_texture *)trans->linear_texture)->pitch_in_bytes[0];
-		if (usage & PIPE_TRANSFER_READ) {
-			/* We cannot map a tiled texture directly because the data is
-			 * in a different order, therefore we do detiling using a blit. */
-			r600_copy_from_tiled_texture(ctx, trans);
+                        ((struct r600_resource_texture *)trans->staging_texture)->pitch_in_bytes[0];
+		if (!discard) {
+			r600_copy_to_staging_texture(ctx, trans);
 			/* Always referenced in the blit. */
 			ctx->flush(ctx, 0, NULL);
 		}
@@ -516,11 +563,11 @@ void r600_texture_transfer_destroy(struct pipe_context *ctx,
 	struct r600_transfer *rtransfer = (struct r600_transfer*)transfer;
 	struct r600_resource_texture *rtex = (struct r600_resource_texture*)transfer->resource;
 
-	if (rtransfer->linear_texture) {
+	if (rtransfer->staging_texture) {
 		if (transfer->usage & PIPE_TRANSFER_WRITE) {
-			r600_copy_into_tiled_texture(ctx, rtransfer);
+			r600_copy_from_staging_texture(ctx, rtransfer);
 		}
-		pipe_resource_reference(&rtransfer->linear_texture, NULL);
+		pipe_resource_reference(&rtransfer->staging_texture, NULL);
 	}
 	if (rtex->flushed_depth_texture) {
 		pipe_resource_reference((struct pipe_resource **)&rtex->flushed_depth_texture, NULL);
@@ -537,10 +584,11 @@ void* r600_texture_transfer_map(struct pipe_context *ctx,
 	enum pipe_format format = transfer->resource->format;
 	struct radeon *radeon = (struct radeon *)ctx->screen->winsys;
 	unsigned offset = 0;
+	unsigned usage = 0;
 	char *map;
 
-	if (rtransfer->linear_texture) {
-		bo = ((struct r600_resource *)rtransfer->linear_texture)->bo;
+	if (rtransfer->staging_texture) {
+		bo = ((struct r600_resource *)rtransfer->staging_texture)->bo;
 	} else {
 		struct r600_resource_texture *rtex = (struct r600_resource_texture*)transfer->resource;
 
@@ -553,7 +601,30 @@ void* r600_texture_transfer_map(struct pipe_context *ctx,
 			transfer->box.y / util_format_get_blockheight(format) * transfer->stride +
 			transfer->box.x / util_format_get_blockwidth(format) * util_format_get_blocksize(format);
 	}
-	map = r600_bo_map(radeon, bo, 0, ctx);
+
+	if (transfer->usage & PIPE_TRANSFER_WRITE) {
+		usage |= PB_USAGE_CPU_WRITE;
+
+		if (transfer->usage & PIPE_TRANSFER_DISCARD) {
+		}
+
+		if (transfer->usage & PIPE_TRANSFER_FLUSH_EXPLICIT) {
+		}
+	}
+
+	if (transfer->usage & PIPE_TRANSFER_READ) {
+		usage |= PB_USAGE_CPU_READ;
+	}
+
+	if (transfer->usage & PIPE_TRANSFER_DONTBLOCK) {
+		usage |= PB_USAGE_DONTBLOCK;
+	}
+
+	if (transfer->usage & PIPE_TRANSFER_UNSYNCHRONIZED) {
+		usage |= PB_USAGE_UNSYNCHRONIZED;
+	}
+
+	map = r600_bo_map(radeon, bo, usage, ctx);
 	if (!map) {
 		return NULL;
 	}
@@ -568,8 +639,8 @@ void r600_texture_transfer_unmap(struct pipe_context *ctx,
 	struct radeon *radeon = (struct radeon *)ctx->screen->winsys;
 	struct r600_bo *bo;
 
-	if (rtransfer->linear_texture) {
-		bo = ((struct r600_resource *)rtransfer->linear_texture)->bo;
+	if (rtransfer->staging_texture) {
+		bo = ((struct r600_resource *)rtransfer->staging_texture)->bo;
 	} else {
 		struct r600_resource_texture *rtex = (struct r600_resource_texture*)transfer->resource;
 
diff --git a/src/gallium/drivers/r600/r600d.h b/src/gallium/drivers/r600/r600d.h
index a3cb5b8600..ae19bfb828 100644
--- a/src/gallium/drivers/r600/r600d.h
+++ b/src/gallium/drivers/r600/r600d.h
@@ -2100,6 +2100,10 @@
 #define   G_028C00_LAST_PIXEL(x)                       (((x) >> 10) & 0x1)
 #define   C_028C00_LAST_PIXEL                          0xFFFFFBFF
 #define R_028C04_PA_SC_AA_CONFIG                     0x028C04
+#define R_028C08_PA_SU_VTX_CNTL                      0x028C08
+#define   S_028C08_PIX_CENTER_HALF(x)                  (((x) & 0x1) << 0)
+#define   G_028C08_PIX_CENTER_HALF(x)                  (((x) >> 0) & 0x1)
+#define   C_028C08_PIX_CENTER_HALF                     0xFFFFFFFE
 #define R_028C1C_PA_SC_AA_SAMPLE_LOCS_MCTX           0x028C1C
 #define R_028C48_PA_SC_AA_MASK                       0x028C48
 #define R_028810_PA_CL_CLIP_CNTL                     0x028810
diff --git a/src/gallium/drivers/rbug/README b/src/gallium/drivers/rbug/README
index b6d3a5cf35..9d7bd4e769 100644
--- a/src/gallium/drivers/rbug/README
+++ b/src/gallium/drivers/rbug/README
@@ -7,24 +7,10 @@ This directory contains a Gallium3D remote debugger pipe driver.
 It provides remote debugging functionality.
 
 
-= Build Instructions =
-
-To build, invoke scons on the top dir as
- 
- scons dri=no statetrackers=mesa winsys=xlib
-
-
 = Usage =
 
-To use do
-
-   export LD_LIBRARY_PATH=$PWD/build/linux-x86-debug/lib
-
-ensure the right libGL.so is being picked by doing
-
-   ldd progs/trivial/tri 
+Do
 
-   export XMESA_TRACE=y
    GALLIUM_RBUG=true progs/trivial/tri
 
 which should open gallium remote debugging session. While the program is running
diff --git a/src/gallium/drivers/rbug/SConscript b/src/gallium/drivers/rbug/SConscript
index 3da6ac104a..169c2718dc 100644
--- a/src/gallium/drivers/rbug/SConscript
+++ b/src/gallium/drivers/rbug/SConscript
@@ -11,4 +11,6 @@ rbug = env.ConvenienceLibrary(
 		'rbug_screen.c',
 	])
 
+env.Alias('rbug', rbug)
+
 Export('rbug')
diff --git a/src/gallium/drivers/softpipe/SConscript b/src/gallium/drivers/softpipe/SConscript
index d5f4d28aef..ea10e8a9f9 100644
--- a/src/gallium/drivers/softpipe/SConscript
+++ b/src/gallium/drivers/softpipe/SConscript
@@ -37,4 +37,6 @@ softpipe = env.ConvenienceLibrary(
 		'sp_tile_cache.c',
 	])
 
+env.Alias('softpipe', softpipe)
+
 Export('softpipe')
diff --git a/src/gallium/drivers/sw/SConscript b/src/gallium/drivers/sw/SConscript
index e9ebf751dd..40d01db2f6 100644
--- a/src/gallium/drivers/sw/SConscript
+++ b/src/gallium/drivers/sw/SConscript
@@ -20,7 +20,6 @@ if True:
 
 if env['llvm']:
     env.Append(CPPDEFINES = 'GALLIUM_LLVMPIPE')
-    env.Tool('udis86')
     env.Prepend(LIBS = [llvmpipe])
     extra.append(llvmpipe)
 
diff --git a/src/gallium/drivers/trace/README b/src/gallium/drivers/trace/README
index cdcd8d2b4b..c210cba032 100644
--- a/src/gallium/drivers/trace/README
+++ b/src/gallium/drivers/trace/README
@@ -7,23 +7,8 @@ This directory contains a Gallium3D trace debugger pipe driver.
 It can traces all incoming calls.
 
 
-= Build Instructions =
-
-To build, invoke scons on the top dir as
- 
- scons dri=no statetrackers=mesa winsys=xlib
-
-
 = Usage =
 
-To use do
-
- export LD_LIBRARY_PATH=$PWD/build/linux-x86-debug/lib
-
-ensure the right libGL.so is being picked by doing
-
- ldd progs/trivial/tri 
-
 == Tracing ==
 
 For tracing then do
@@ -40,6 +25,7 @@ For remote debugging see:
 
   src/gallium/drivers/rbug/README
 
+
 = Integrating =
 
 You can integrate the trace pipe driver either inside the state tracker or the 
@@ -60,5 +46,5 @@ are automatically wrapped by trace_screen.
 
 
 --
-Jose Fonseca <jrfonseca@tungstengraphics.com>
+Jose Fonseca <jfonseca@vmware.com>
 Jakob Bornecrantz <jakob@vmware.com>
diff --git a/src/gallium/drivers/trace/SConscript b/src/gallium/drivers/trace/SConscript
index 06b0c4863a..1384fe33d7 100644
--- a/src/gallium/drivers/trace/SConscript
+++ b/src/gallium/drivers/trace/SConscript
@@ -12,4 +12,6 @@ trace = env.ConvenienceLibrary(
         'tr_texture.c',
     ])
 
+env.Alias('trace', trace)
+
 Export('trace')
diff --git a/src/gallium/state_trackers/d3d1x/dxgi/src/dxgi_native.cpp b/src/gallium/state_trackers/d3d1x/dxgi/src/dxgi_native.cpp
index e1c34611d1..c246fc5ef7 100644
--- a/src/gallium/state_trackers/d3d1x/dxgi/src/dxgi_native.cpp
+++ b/src/gallium/state_trackers/d3d1x/dxgi/src/dxgi_native.cpp
@@ -1101,6 +1101,7 @@ struct GalliumDXGISwapChain : public GalliumDXGIObject<IDXGISwapChain, GalliumDX
 		struct pipe_resource* dst;
 		struct pipe_resource* src;
 		struct pipe_surface* dst_surface;
+		enum native_attachment att;
 
 		void* present_cookie;
 		hr = parent->backend->BeginPresent(desc.OutputWindow, &present_cookie, &cur_window, &rect, &rgndata, &preserve_aspect_ratio);
@@ -1221,16 +1222,9 @@ struct GalliumDXGISwapChain : public GalliumDXGIObject<IDXGISwapChain, GalliumDX
 
 		pipe->flush(pipe, PIPE_FLUSH_RENDER_CACHE | PIPE_FLUSH_FRAME, 0);
 
-		if(db)
-		{
-			if(!surface->swap_buffers(surface))
-				return DXGI_ERROR_DEVICE_REMOVED;
-		}
-		else
-		{
-			if(!surface->flush_frontbuffer(surface))
-				return DXGI_ERROR_DEVICE_REMOVED;
-		}
+		att = (db) ? NATIVE_ATTACHMENT_BACK_LEFT : NATIVE_ATTACHMENT_FRONT_LEFT;
+		if(!surface->present(surface, att, FALSE, 0))
+			return DXGI_ERROR_DEVICE_REMOVED;
 
 end_present:
 		parent->backend->EndPresent(desc.OutputWindow, present_cookie);
diff --git a/src/gallium/state_trackers/egl/SConscript b/src/gallium/state_trackers/egl/SConscript
index efcce25e31..8d34e47df0 100644
--- a/src/gallium/state_trackers/egl/SConscript
+++ b/src/gallium/state_trackers/egl/SConscript
@@ -3,34 +3,35 @@
 
 Import('*')
 
-if 'egl' in env['statetrackers']:
+if env['platform'] != 'windows':
+    Return()
 
-    env = env.Clone()
+env = env.Clone()
 
-    env.Append(CPPPATH = [
-	'#/src/egl/main',
-	'#/src/gallium/winsys/sw',
-	'.',
-    ])
-    env.Append(CPPDEFINES = [
-	'HAVE_GDI_BACKEND',
-    ])
+env.Append(CPPPATH = [
+    '#/src/egl/main',
+    '#/src/gallium/winsys/sw',
+    '.',
+])
+env.Append(CPPDEFINES = [
+    'HAVE_GDI_BACKEND',
+])
 
-    common_sources = [
-        'common/egl_g3d.c',
-        'common/egl_g3d_api.c',
-        'common/egl_g3d_image.c',
-        'common/egl_g3d_st.c',
-        'common/egl_g3d_sync.c',
-        'common/native_helper.c',
-    ]
+common_sources = [
+    'common/egl_g3d.c',
+    'common/egl_g3d_api.c',
+    'common/egl_g3d_image.c',
+    'common/egl_g3d_st.c',
+    'common/egl_g3d_sync.c',
+    'common/native_helper.c',
+]
 
-    gdi_sources = common_sources + [
-        'gdi/native_gdi.c',
-    ]
+gdi_sources = common_sources + [
+    'gdi/native_gdi.c',
+]
 
-    st_egl_gdi = env.ConvenienceLibrary(
-	target = 'st_egl_gdi',
-	source = gdi_sources,
-    )
-    Export('st_egl_gdi')
+st_egl_gdi = env.ConvenienceLibrary(
+    target = 'st_egl_gdi',
+    source = gdi_sources,
+)
+Export('st_egl_gdi')
diff --git a/src/gallium/state_trackers/egl/common/egl_g3d.c b/src/gallium/state_trackers/egl/common/egl_g3d.c
index 30ddcd5bc1..a3750ac56f 100644
--- a/src/gallium/state_trackers/egl/common/egl_g3d.c
+++ b/src/gallium/state_trackers/egl/common/egl_g3d.c
@@ -156,7 +156,8 @@ egl_g3d_add_screens(_EGLDriver *drv, _EGLDisplay *dpy)
  */
 static EGLBoolean
 init_config_attributes(_EGLConfig *conf, const struct native_config *nconf,
-                       EGLint api_mask, enum pipe_format depth_stencil_format)
+                       EGLint api_mask, enum pipe_format depth_stencil_format,
+                       EGLBoolean preserve_buffer, EGLint max_swap_interval)
 {
    uint rgba[4], depth_stencil[2], buffer_size;
    EGLint surface_type;
@@ -238,6 +239,11 @@ init_config_attributes(_EGLConfig *conf, const struct native_config *nconf,
       conf->TransparentBlueValue = nconf->transparent_rgb_values[2];
    }
 
+   conf->MinSwapInterval = 0;
+   conf->MaxSwapInterval = max_swap_interval;
+   if (preserve_buffer)
+      conf->SurfaceType |= EGL_SWAP_BEHAVIOR_PRESERVED_BIT;
+
    return _eglValidateConfig(conf, EGL_FALSE);
 }
 
@@ -247,7 +253,8 @@ init_config_attributes(_EGLConfig *conf, const struct native_config *nconf,
 static EGLBoolean
 egl_g3d_init_config(_EGLDriver *drv, _EGLDisplay *dpy,
                     _EGLConfig *conf, const struct native_config *nconf,
-                    enum pipe_format depth_stencil_format)
+                    enum pipe_format depth_stencil_format,
+                    int preserve_buffer, int max_swap_interval)
 {
    struct egl_g3d_config *gconf = egl_g3d_config(conf);
    EGLint buffer_mask, api_mask;
@@ -288,7 +295,8 @@ egl_g3d_init_config(_EGLDriver *drv, _EGLDisplay *dpy,
    }
 
    valid = init_config_attributes(&gconf->base,
-         nconf, api_mask, depth_stencil_format);
+         nconf, api_mask, depth_stencil_format,
+         preserve_buffer, max_swap_interval);
    if (!valid) {
       _eglLog(_EGL_DEBUG, "skip invalid config 0x%x", nconf->native_visual_id);
       return EGL_FALSE;
@@ -349,6 +357,7 @@ egl_g3d_add_configs(_EGLDriver *drv, _EGLDisplay *dpy, EGLint id)
    const struct native_config **native_configs;
    enum pipe_format depth_stencil_formats[8];
    int num_formats, num_configs, i, j;
+   int preserve_buffer, max_swap_interval;
 
    native_configs = gdpy->native->get_configs(gdpy->native, &num_configs);
    if (!num_configs) {
@@ -357,6 +366,11 @@ egl_g3d_add_configs(_EGLDriver *drv, _EGLDisplay *dpy, EGLint id)
       return id;
    }
 
+   preserve_buffer =
+      gdpy->native->get_param(gdpy->native, NATIVE_PARAM_PRESERVE_BUFFER);
+   max_swap_interval =
+      gdpy->native->get_param(gdpy->native, NATIVE_PARAM_MAX_SWAP_INTERVAL);
+
    num_formats = egl_g3d_fill_depth_stencil_formats(dpy,
          depth_stencil_formats);
 
@@ -368,7 +382,8 @@ egl_g3d_add_configs(_EGLDriver *drv, _EGLDisplay *dpy, EGLint id)
          if (gconf) {
             _eglInitConfig(&gconf->base, dpy, id);
             if (!egl_g3d_init_config(drv, dpy, &gconf->base,
-                     native_configs[i], depth_stencil_formats[j])) {
+                     native_configs[i], depth_stencil_formats[j],
+                     preserve_buffer, max_swap_interval)) {
                FREE(gconf);
                break;
             }
@@ -538,7 +553,8 @@ egl_g3d_initialize(_EGLDriver *drv, _EGLDisplay *dpy,
 
    if (dpy->Platform == _EGL_PLATFORM_DRM) {
       dpy->Extensions.MESA_drm_display = EGL_TRUE;
-      dpy->Extensions.MESA_drm_image = EGL_TRUE;
+      if (gdpy->native->buffer)
+         dpy->Extensions.MESA_drm_image = EGL_TRUE;
    }
 
    if (egl_g3d_add_configs(drv, dpy, 1) == 1) {
diff --git a/src/gallium/state_trackers/egl/common/egl_g3d_api.c b/src/gallium/state_trackers/egl/common/egl_g3d_api.c
index c10245bb06..fd7dc8f814 100644
--- a/src/gallium/state_trackers/egl/common/egl_g3d_api.c
+++ b/src/gallium/state_trackers/egl/common/egl_g3d_api.c
@@ -97,6 +97,70 @@ egl_g3d_choose_st(_EGLDriver *drv, _EGLContext *ctx,
    return stapi;
 }
 
+static int
+egl_g3d_compare_config(const _EGLConfig *conf1, const _EGLConfig *conf2,
+                       void *priv_data)
+{
+   const _EGLConfig *criteria = (const _EGLConfig *) priv_data;
+
+   /* EGL_NATIVE_VISUAL_TYPE ignored? */
+   return _eglCompareConfigs(conf1, conf2, criteria, EGL_TRUE);
+}
+
+static EGLBoolean
+egl_g3d_match_config(const _EGLConfig *conf, const _EGLConfig *criteria)
+{
+   if (!_eglMatchConfig(conf, criteria))
+      return EGL_FALSE;
+
+   if (criteria->MatchNativePixmap != EGL_NONE &&
+       criteria->MatchNativePixmap != EGL_DONT_CARE) {
+      struct egl_g3d_display *gdpy = egl_g3d_display(conf->Display);
+      struct egl_g3d_config *gconf = egl_g3d_config(conf);
+      EGLNativePixmapType pix =
+         (EGLNativePixmapType) criteria->MatchNativePixmap;
+
+      if (!gdpy->native->is_pixmap_supported(gdpy->native, pix, gconf->native))
+         return EGL_FALSE;
+   }
+
+   return EGL_TRUE;
+}
+
+static EGLBoolean
+egl_g3d_choose_config(_EGLDriver *drv, _EGLDisplay *dpy, const EGLint *attribs,
+                      EGLConfig *configs, EGLint size, EGLint *num_configs)
+{
+   _EGLConfig **tmp_configs, criteria;
+   EGLint tmp_size, i;
+
+   if (!num_configs)
+      return _eglError(EGL_BAD_PARAMETER, "eglChooseConfigs");
+
+   if (!_eglParseConfigAttribList(&criteria, dpy, attribs))
+      return _eglError(EGL_BAD_ATTRIBUTE, "eglChooseConfig");
+
+   tmp_configs = (_EGLConfig **) _eglFilterArray(dpy->Configs, &tmp_size,
+         (_EGLArrayForEach) egl_g3d_match_config, (void *) &criteria);
+   if (!tmp_configs)
+      return _eglError(EGL_BAD_ALLOC, "eglChooseConfig(out of memory)");
+
+   /* perform sorting of configs */
+   if (tmp_configs && tmp_size) {
+      _eglSortConfigs((const _EGLConfig **) tmp_configs, tmp_size,
+            egl_g3d_compare_config, (void *) &criteria);
+      size = MIN2(tmp_size, size);
+      for (i = 0; i < size; i++)
+         configs[i] = _eglGetConfigHandle(tmp_configs[i]);
+   }
+
+   free(tmp_configs);
+
+   *num_configs = size;
+
+   return EGL_TRUE;
+}
+
 static _EGLContext *
 egl_g3d_create_context(_EGLDriver *drv, _EGLDisplay *dpy, _EGLConfig *conf,
                        _EGLContext *share, const EGLint *attribs)
@@ -539,7 +603,10 @@ egl_g3d_swap_buffers(_EGLDriver *drv, _EGLDisplay *dpy, _EGLSurface *surf)
             PIPE_FLUSH_RENDER_CACHE | PIPE_FLUSH_FRAME, NULL);
    }
 
-   return gsurf->native->swap_buffers(gsurf->native);
+   return gsurf->native->present(gsurf->native,
+         NATIVE_ATTACHMENT_BACK_LEFT,
+         gsurf->base.SwapBehavior == EGL_BUFFER_PRESERVED,
+         gsurf->base.SwapInterval);
 }
 
 /**
@@ -607,8 +674,7 @@ egl_g3d_copy_buffers(_EGLDriver *drv, _EGLDisplay *dpy, _EGLSurface *surf,
       if (psrc) {
          gdpy->pipe->resource_copy_region(gdpy->pipe, ptex, subdst, 0, 0, 0,
                gsurf->render_texture, subsrc, 0, 0, 0, ptex->width0, ptex->height0);
-
-         nsurf->flush_frontbuffer(nsurf);
+         nsurf->present(nsurf, NATIVE_ATTACHMENT_FRONT_LEFT, FALSE, 0);
       }
 
       pipe_resource_reference(&ptex, NULL);
@@ -838,6 +904,8 @@ egl_g3d_init_driver_api(_EGLDriver *drv)
 {
    _eglInitDriverFallbacks(drv);
 
+   drv->API.ChooseConfig = egl_g3d_choose_config;
+
    drv->API.CreateContext = egl_g3d_create_context;
    drv->API.DestroyContext = egl_g3d_destroy_context;
    drv->API.CreateWindowSurface = egl_g3d_create_window_surface;
diff --git a/src/gallium/state_trackers/egl/common/egl_g3d_image.c b/src/gallium/state_trackers/egl/common/egl_g3d_image.c
index be9c88e5e4..6a1f8cc697 100644
--- a/src/gallium/state_trackers/egl/common/egl_g3d_image.c
+++ b/src/gallium/state_trackers/egl/common/egl_g3d_image.c
@@ -38,7 +38,7 @@
 #include "egl_g3d_api.h"
 #include "egl_g3d_image.h"
 
-/* move this to native display? */
+/* for struct winsys_handle */
 #include "state_tracker/drm_driver.h"
 
 /**
@@ -137,13 +137,11 @@ egl_g3d_reference_drm_buffer(_EGLDisplay *dpy, EGLint name,
                              _EGLImage *img, const EGLint *attribs)
 {
    struct egl_g3d_display *gdpy = egl_g3d_display(dpy);
-   struct pipe_screen *screen = gdpy->native->screen;
    struct pipe_resource templ;
    struct winsys_handle wsh;
    _EGLImageAttribs attrs;
    EGLint format;
 
-   /* winsys_handle is in theory platform-specific */
    if (dpy->Platform != _EGL_PLATFORM_DRM)
       return NULL;
 
@@ -178,9 +176,10 @@ egl_g3d_reference_drm_buffer(_EGLDisplay *dpy, EGLint name,
 
    memset(&wsh, 0, sizeof(wsh));
    wsh.handle = (unsigned) name;
-   wsh.stride = attrs.DRMBufferStrideMESA;
+   wsh.stride =
+      attrs.DRMBufferStrideMESA * util_format_get_blocksize(templ.format);
 
-   return screen->resource_from_handle(screen, &templ, &wsh);
+   return gdpy->native->buffer->import_buffer(gdpy->native, &templ, &wsh);
 }
 
 #endif /* EGL_MESA_drm_image */
@@ -302,10 +301,8 @@ egl_g3d_export_drm_image(_EGLDriver *drv, _EGLDisplay *dpy, _EGLImage *img,
 {
    struct egl_g3d_display *gdpy = egl_g3d_display(dpy);
    struct egl_g3d_image *gimg = egl_g3d_image(img);
-   struct pipe_screen *screen = gdpy->native->screen;
    struct winsys_handle wsh;
 
-   /* winsys_handle is in theory platform-specific */
    if (dpy->Platform != _EGL_PLATFORM_DRM)
       return EGL_FALSE;
 
@@ -313,9 +310,9 @@ egl_g3d_export_drm_image(_EGLDriver *drv, _EGLDisplay *dpy, _EGLImage *img,
    if (name) {
       memset(&handle, 0, sizeof(handle));
       wsh.type = DRM_API_HANDLE_TYPE_SHARED;
-      if (!screen->resource_get_handle(screen, gimg->texture, &wsh)) {
+      if (!gdpy->native->buffer->export_buffer(gdpy->native,
+                                               gimg->texture, &wsh))
          return EGL_FALSE;
-      }
 
       *name = wsh.handle;
    }
@@ -324,7 +321,8 @@ egl_g3d_export_drm_image(_EGLDriver *drv, _EGLDisplay *dpy, _EGLImage *img,
    if (handle || stride) {
       memset(&wsh, 0, sizeof(wsh));
       wsh.type = DRM_API_HANDLE_TYPE_KMS;
-      if (!screen->resource_get_handle(screen, gimg->texture, &wsh))
+      if (!gdpy->native->buffer->export_buffer(gdpy->native,
+                                               gimg->texture, &wsh))
          return EGL_FALSE;
 
       if (handle)
diff --git a/src/gallium/state_trackers/egl/common/egl_g3d_st.c b/src/gallium/state_trackers/egl/common/egl_g3d_st.c
index 0affe632cf..25e2999590 100644
--- a/src/gallium/state_trackers/egl/common/egl_g3d_st.c
+++ b/src/gallium/state_trackers/egl/common/egl_g3d_st.c
@@ -192,7 +192,8 @@ egl_g3d_st_framebuffer_flush_front(struct st_framebuffer_iface *stfbi,
    _EGLSurface *surf = (_EGLSurface *) stfbi->st_manager_private;
    struct egl_g3d_surface *gsurf = egl_g3d_surface(surf);
 
-   return gsurf->native->flush_frontbuffer(gsurf->native);
+   return gsurf->native->present(gsurf->native,
+         NATIVE_ATTACHMENT_FRONT_LEFT, FALSE, 0);
 }
 
 static boolean 
diff --git a/src/gallium/state_trackers/egl/common/native.h b/src/gallium/state_trackers/egl/common/native.h
index 3c3f57e267..3886ca2056 100644
--- a/src/gallium/state_trackers/egl/common/native.h
+++ b/src/gallium/state_trackers/egl/common/native.h
@@ -34,6 +34,11 @@
 #include "pipe/p_state.h"
 #include "state_tracker/sw_winsys.h"
 
+#ifdef __cplusplus
+extern "C" {
+#endif
+
+#include "native_buffer.h"
 #include "native_modeset.h"
 
 /**
@@ -54,7 +59,17 @@ enum native_param_type {
     * Return TRUE if window/pixmap surfaces use the buffers of the native
     * types.
     */
-   NATIVE_PARAM_USE_NATIVE_BUFFER
+   NATIVE_PARAM_USE_NATIVE_BUFFER,
+
+   /**
+    * Return TRUE if native_surface::present can preserve the buffer.
+    */
+   NATIVE_PARAM_PRESERVE_BUFFER,
+
+   /**
+    * Return the maximum supported swap interval.
+    */
+   NATIVE_PARAM_MAX_SWAP_INTERVAL
 };
 
 struct native_surface {
@@ -66,17 +81,12 @@ struct native_surface {
    void (*destroy)(struct native_surface *nsurf);
 
    /**
-    * Swap the front and back buffers so that the back buffer is visible.  It
-    * is no-op if the surface is single-buffered.  The contents of the back
-    * buffer after swapping may or may not be preserved.
-    */
-   boolean (*swap_buffers)(struct native_surface *nsurf);
-
-   /**
-    * Make the front buffer visible.  In some native displays, changes to the
-    * front buffer might not be visible immediately and require manual flush.
+    * Present the given buffer to the native engine.
     */
-   boolean (*flush_frontbuffer)(struct native_surface *nsurf);
+   boolean (*present)(struct native_surface *nsurf,
+                      enum native_attachment natt,
+                      boolean preserve,
+                      uint swap_interval);
 
    /**
     * Validate the buffers of the surface.  textures, if not NULL, points to an
@@ -181,6 +191,7 @@ struct native_display {
                                                    EGLNativePixmapType pix,
                                                    const struct native_config *nconf);
 
+   const struct native_display_buffer *buffer;
    const struct native_display_modeset *modeset;
 };
 
@@ -232,4 +243,8 @@ native_get_drm_platform(void);
 const struct native_platform *
 native_get_fbdev_platform(void);
 
+#ifdef __cplusplus
+}
+#endif
+
 #endif /* _NATIVE_H_ */
diff --git a/src/gallium/state_trackers/egl/common/native_buffer.h b/src/gallium/state_trackers/egl/common/native_buffer.h
new file mode 100644
index 0000000000..5c29ab9741
--- /dev/null
+++ b/src/gallium/state_trackers/egl/common/native_buffer.h
@@ -0,0 +1,59 @@
+/*
+ * Mesa 3-D graphics library
+ * Version:  7.9
+ *
+ * Copyright (C) 2010 LunarG Inc.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included
+ * in all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+ * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
+ * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER
+ * DEALINGS IN THE SOFTWARE.
+ *
+ * Authors:
+ *    Chia-I Wu <olv@lunarg.com>
+ */
+
+#ifndef _NATIVE_BUFFER_H_
+#define _NATIVE_BUFFER_H_
+
+#include "pipe/p_compiler.h"
+
+struct native_display;
+struct pipe_resource;
+
+/**
+ * Buffer interface of the native display.  It allows native buffers to be
+ * imported and exported.
+ *
+ * Just like a native window or a native pixmap, a native buffer is another
+ * native type.  Its definition depends on the native display.
+ *
+ * For DRM platform, the type of a native buffer is struct winsys_handle.
+ */
+struct native_display_buffer {
+   struct pipe_resource *(*import_buffer)(struct native_display *ndpy,
+                                          const struct pipe_resource *templ,
+                                          void *buf);
+
+   /**
+    * The resource must be creatred with PIPE_BIND_SHARED.
+    */
+   boolean (*export_buffer)(struct native_display *ndpy,
+                            struct pipe_resource *res,
+                            void *buf);
+};
+
+#endif /* _NATIVE_BUFFER_H_ */
diff --git a/src/gallium/state_trackers/egl/drm/modeset.c b/src/gallium/state_trackers/egl/drm/modeset.c
index 5ed22f7b9d..0cc06caa2a 100644
--- a/src/gallium/state_trackers/egl/drm/modeset.c
+++ b/src/gallium/state_trackers/egl/drm/modeset.c
@@ -167,6 +167,32 @@ drm_surface_swap_buffers(struct native_surface *nsurf)
    return TRUE;
 }
 
+static boolean
+drm_surface_present(struct native_surface *nsurf,
+                    enum native_attachment natt,
+                    boolean preserve,
+                    uint swap_interval)
+{
+   boolean ret;
+
+   if (preserve || swap_interval)
+      return FALSE;
+
+   switch (natt) {
+   case NATIVE_ATTACHMENT_FRONT_LEFT:
+      ret = drm_surface_flush_frontbuffer(nsurf);
+      break;
+   case NATIVE_ATTACHMENT_BACK_LEFT:
+      ret = drm_surface_swap_buffers(nsurf);
+      break;
+   default:
+      ret = FALSE;
+      break;
+   }
+
+   return ret;
+}
+
 static void
 drm_surface_wait(struct native_surface *nsurf)
 {
@@ -225,8 +251,7 @@ drm_display_create_surface(struct native_display *ndpy,
    resource_surface_set_size(drmsurf->rsurf, drmsurf->width, drmsurf->height);
 
    drmsurf->base.destroy = drm_surface_destroy;
-   drmsurf->base.swap_buffers = drm_surface_swap_buffers;
-   drmsurf->base.flush_frontbuffer = drm_surface_flush_frontbuffer;
+   drmsurf->base.present = drm_surface_present;
    drmsurf->base.validate = drm_surface_validate;
    drmsurf->base.wait = drm_surface_wait;
 
diff --git a/src/gallium/state_trackers/egl/drm/native_drm.c b/src/gallium/state_trackers/egl/drm/native_drm.c
index f6dc558437..3759c2a26d 100644
--- a/src/gallium/state_trackers/egl/drm/native_drm.c
+++ b/src/gallium/state_trackers/egl/drm/native_drm.c
@@ -103,6 +103,9 @@ drm_display_get_param(struct native_display *ndpy,
    int val;
 
    switch (param) {
+   case NATIVE_PARAM_USE_NATIVE_BUFFER:
+   case NATIVE_PARAM_PRESERVE_BUFFER:
+   case NATIVE_PARAM_MAX_SWAP_INTERVAL:
    default:
       val = 0;
       break;
@@ -182,6 +185,29 @@ drm_display_init_screen(struct native_display *ndpy)
    return TRUE;
 }
 
+static struct pipe_resource *
+drm_display_import_buffer(struct native_display *ndpy,
+                          const struct pipe_resource *templ,
+                          void *buf)
+{
+   return ndpy->screen->resource_from_handle(ndpy->screen,
+         templ, (struct winsys_handle *) buf);
+}
+
+static boolean
+drm_display_export_buffer(struct native_display *ndpy,
+                          struct pipe_resource *res,
+                          void *buf)
+{
+   return ndpy->screen->resource_get_handle(ndpy->screen,
+         res, (struct winsys_handle *) buf);
+}
+
+static struct native_display_buffer drm_display_buffer = {
+   drm_display_import_buffer,
+   drm_display_export_buffer
+};
+
 static struct native_display *
 drm_create_display(int fd, struct native_event_handler *event_handler,
                    void *user_data)
@@ -205,6 +231,7 @@ drm_create_display(int fd, struct native_event_handler *event_handler,
    drmdpy->base.get_param = drm_display_get_param;
    drmdpy->base.get_configs = drm_display_get_configs;
 
+   drmdpy->base.buffer = &drm_display_buffer;
    drm_display_init_modeset(&drmdpy->base);
 
    return &drmdpy->base;
diff --git a/src/gallium/state_trackers/egl/fbdev/native_fbdev.c b/src/gallium/state_trackers/egl/fbdev/native_fbdev.c
index e459402076..1b5ea8bf9d 100644
--- a/src/gallium/state_trackers/egl/fbdev/native_fbdev.c
+++ b/src/gallium/state_trackers/egl/fbdev/native_fbdev.c
@@ -137,6 +137,32 @@ fbdev_surface_swap_buffers(struct native_surface *nsurf)
    return ret;
 }
 
+static boolean
+fbdev_surface_present(struct native_surface *nsurf,
+                      enum native_attachment natt,
+                      boolean preserve,
+                      uint swap_interval)
+{
+   boolean ret;
+
+   if (preserve || swap_interval)
+      return FALSE;
+
+   switch (natt) {
+   case NATIVE_ATTACHMENT_FRONT_LEFT:
+      ret = fbdev_surface_flush_frontbuffer(nsurf);
+      break;
+   case NATIVE_ATTACHMENT_BACK_LEFT:
+      ret = fbdev_surface_swap_buffers(nsurf);
+      break;
+   default:
+      ret = FALSE;
+      break;
+   }
+
+   return ret;
+}
+
 static void
 fbdev_surface_wait(struct native_surface *nsurf)
 {
@@ -181,8 +207,7 @@ fbdev_display_create_scanout_surface(struct native_display *ndpy,
    resource_surface_set_size(fbsurf->rsurf, fbsurf->width, fbsurf->height);
 
    fbsurf->base.destroy = fbdev_surface_destroy;
-   fbsurf->base.swap_buffers = fbdev_surface_swap_buffers;
-   fbsurf->base.flush_frontbuffer = fbdev_surface_flush_frontbuffer;
+   fbsurf->base.present = fbdev_surface_present;
    fbsurf->base.validate = fbdev_surface_validate;
    fbsurf->base.wait = fbdev_surface_wait;
 
@@ -279,6 +304,9 @@ fbdev_display_get_param(struct native_display *ndpy,
    int val;
 
    switch (param) {
+   case NATIVE_PARAM_USE_NATIVE_BUFFER:
+   case NATIVE_PARAM_PRESERVE_BUFFER:
+   case NATIVE_PARAM_MAX_SWAP_INTERVAL:
    default:
       val = 0;
       break;
diff --git a/src/gallium/state_trackers/egl/gdi/native_gdi.c b/src/gallium/state_trackers/egl/gdi/native_gdi.c
index 91701e5b7d..d259e6edc8 100644
--- a/src/gallium/state_trackers/egl/gdi/native_gdi.c
+++ b/src/gallium/state_trackers/egl/gdi/native_gdi.c
@@ -160,6 +160,32 @@ gdi_surface_swap_buffers(struct native_surface *nsurf)
 }
 
 static boolean
+gdi_surface_present(struct native_surface *nsurf,
+                    enum native_attachment natt,
+                    boolean preserve,
+                    uint swap_interval)
+{
+   boolean ret;
+
+   if (preserve || swap_interval)
+      return FALSE;
+
+   switch (natt) {
+   case NATIVE_ATTACHMENT_FRONT_LEFT:
+      ret = gdi_surface_flush_frontbuffer(nsurf);
+      break;
+   case NATIVE_ATTACHMENT_BACK_LEFT:
+      ret = gdi_surface_swap_buffers(nsurf);
+      break;
+   default:
+      ret = FALSE;
+      break;
+   }
+
+   return ret;
+}
+
+static boolean
 gdi_surface_validate(struct native_surface *nsurf, uint attachment_mask,
                         unsigned int *seq_num, struct pipe_resource **textures,
                         int *width, int *height)
@@ -231,8 +257,7 @@ gdi_display_create_window_surface(struct native_display *ndpy,
    gdi_surface_update_geometry(&gsurf->base);
 
    gsurf->base.destroy = gdi_surface_destroy;
-   gsurf->base.swap_buffers = gdi_surface_swap_buffers;
-   gsurf->base.flush_frontbuffer = gdi_surface_flush_frontbuffer;
+   gsurf->base.present = gdi_surface_present;
    gsurf->base.validate = gdi_surface_validate;
    gsurf->base.wait = gdi_surface_wait;
 
@@ -321,6 +346,8 @@ gdi_display_get_param(struct native_display *ndpy,
       /* private buffers are allocated */
       val = FALSE;
       break;
+   case NATIVE_PARAM_PRESERVE_BUFFER:
+   case NATIVE_PARAM_MAX_SWAP_INTERVAL:
    default:
       val = 0;
       break;
diff --git a/src/gallium/state_trackers/egl/x11/native_dri2.c b/src/gallium/state_trackers/egl/x11/native_dri2.c
index 1169e273c3..331a7de432 100644
--- a/src/gallium/state_trackers/egl/x11/native_dri2.c
+++ b/src/gallium/state_trackers/egl/x11/native_dri2.c
@@ -338,6 +338,32 @@ dri2_surface_swap_buffers(struct native_surface *nsurf)
 }
 
 static boolean
+dri2_surface_present(struct native_surface *nsurf,
+                     enum native_attachment natt,
+                     boolean preserve,
+                     uint swap_interval)
+{
+   boolean ret;
+
+   if (swap_interval)
+      return FALSE;
+
+   switch (natt) {
+   case NATIVE_ATTACHMENT_FRONT_LEFT:
+      ret = dri2_surface_flush_frontbuffer(nsurf);
+      break;
+   case NATIVE_ATTACHMENT_BACK_LEFT:
+      ret = dri2_surface_swap_buffers(nsurf);
+      break;
+   default:
+      ret = FALSE;
+      break;
+   }
+
+   return ret;
+}
+
+static boolean
 dri2_surface_validate(struct native_surface *nsurf, uint attachment_mask,
                       unsigned int *seq_num, struct pipe_resource **textures,
                       int *width, int *height)
@@ -430,8 +456,7 @@ dri2_display_create_surface(struct native_display *ndpy,
    dri2surf->color_format = dri2conf->base.color_format;
 
    dri2surf->base.destroy = dri2_surface_destroy;
-   dri2surf->base.swap_buffers = dri2_surface_swap_buffers;
-   dri2surf->base.flush_frontbuffer = dri2_surface_flush_frontbuffer;
+   dri2surf->base.present = dri2_surface_present;
    dri2surf->base.validate = dri2_surface_validate;
    dri2surf->base.wait = dri2_surface_wait;
 
@@ -630,9 +655,14 @@ dri2_display_get_param(struct native_display *ndpy,
 
    switch (param) {
    case NATIVE_PARAM_USE_NATIVE_BUFFER:
-      /* DRI2GetBuffers use the native buffers */
+      /* DRI2GetBuffers uses the native buffers */
+      val = TRUE;
+      break;
+   case NATIVE_PARAM_PRESERVE_BUFFER:
+      /* DRI2CopyRegion is used */
       val = TRUE;
       break;
+   case NATIVE_PARAM_MAX_SWAP_INTERVAL:
    default:
       val = 0;
       break;
diff --git a/src/gallium/state_trackers/egl/x11/native_ximage.c b/src/gallium/state_trackers/egl/x11/native_ximage.c
index 4b32f6e36e..84811fb6e1 100644
--- a/src/gallium/state_trackers/egl/x11/native_ximage.c
+++ b/src/gallium/state_trackers/egl/x11/native_ximage.c
@@ -175,6 +175,32 @@ ximage_surface_swap_buffers(struct native_surface *nsurf)
 }
 
 static boolean
+ximage_surface_present(struct native_surface *nsurf,
+                       enum native_attachment natt,
+                       boolean preserve,
+                       uint swap_interval)
+{
+   boolean ret;
+
+   if (preserve || swap_interval)
+      return FALSE;
+
+   switch (natt) {
+   case NATIVE_ATTACHMENT_FRONT_LEFT:
+      ret = ximage_surface_flush_frontbuffer(nsurf);
+      break;
+   case NATIVE_ATTACHMENT_BACK_LEFT:
+      ret = ximage_surface_swap_buffers(nsurf);
+      break;
+   default:
+      ret = FALSE;
+      break;
+   }
+
+   return ret;
+}
+
+static boolean
 ximage_surface_validate(struct native_surface *nsurf, uint attachment_mask,
                         unsigned int *seq_num, struct pipe_resource **textures,
                         int *width, int *height)
@@ -257,8 +283,7 @@ ximage_display_create_surface(struct native_display *ndpy,
    xsurf->xdraw.drawable = xsurf->drawable;
 
    xsurf->base.destroy = ximage_surface_destroy;
-   xsurf->base.swap_buffers = ximage_surface_swap_buffers;
-   xsurf->base.flush_frontbuffer = ximage_surface_flush_frontbuffer;
+   xsurf->base.present = ximage_surface_present;
    xsurf->base.validate = ximage_surface_validate;
    xsurf->base.wait = ximage_surface_wait;
 
@@ -416,6 +441,8 @@ ximage_display_get_param(struct native_display *ndpy,
       /* private buffers are allocated */
       val = FALSE;
       break;
+   case NATIVE_PARAM_PRESERVE_BUFFER:
+   case NATIVE_PARAM_MAX_SWAP_INTERVAL:
    default:
       val = 0;
       break;
diff --git a/src/gallium/state_trackers/glx/xlib/SConscript b/src/gallium/state_trackers/glx/xlib/SConscript
index 9df351a276..775000eacc 100644
--- a/src/gallium/state_trackers/glx/xlib/SConscript
+++ b/src/gallium/state_trackers/glx/xlib/SConscript
@@ -3,25 +3,25 @@
 
 Import('*')
 
-if env['platform'] == 'linux' \
-        and 'mesa' in env['statetrackers']:
+if not env['x11']:
+    Return()
 
-    env = env.Clone()
+env = env.Clone()
 
-    env.Append(CPPPATH = [
-        '#/src/mapi',
-        '#/src/mesa',
-        '#/src/mesa/main',
-    ])
+env.Append(CPPPATH = [
+    '#/src/mapi',
+    '#/src/mesa',
+    '#/src/mesa/main',
+])
 
-    st_xlib = env.ConvenienceLibrary(
-	target = 'st_xlib',
-	source = [
-		'glx_api.c',
-		'glx_getproc.c',
-		'glx_usefont.c',
-		'xm_api.c',
-		'xm_st.c',
-		]
-    )
-    Export('st_xlib')
+st_xlib = env.ConvenienceLibrary(
+target = 'st_xlib',
+source = [
+    'glx_api.c',
+    'glx_getproc.c',
+    'glx_usefont.c',
+    'xm_api.c',
+    'xm_st.c',
+    ]
+)
+Export('st_xlib')
diff --git a/src/gallium/state_trackers/python/SConscript b/src/gallium/state_trackers/python/SConscript
index aadeaa0a35..b8371865a6 100644
--- a/src/gallium/state_trackers/python/SConscript
+++ b/src/gallium/state_trackers/python/SConscript
@@ -3,57 +3,64 @@ import os.path
 
 Import('*')
 
-if 'python' in env['statetrackers']:
-
-    env = env.Clone()
-    
-    env.Tool('python')
-    
-    env.Tool('swig')
-    env.Append(SWIGPATH = ['#src/gallium/include', '#src/gallium/include/pipe'])
-    env.Append(SWIGFLAGS = ['-python', '-keyword'])
-    
-    env.Append(CPPPATH = '.')
-
-    if env['platform'] == 'windows':
-        env.Append(LIBS = [
-            'opengl32',
-            'gdi32',
-            'user32',
-            'kernel32',
-            'ws2_32',
-        ])
-    else:
-        env.Append(CPPDEFINES = ['GCC_HASCLASSVISIBILITY'])
-        env.Append(LIBS = [
-        	'GL',
-            'X11',
-        ])
-
-    sources = [
-        'gallium.i',
-        'st_device.c',
-        'st_sample.c',
-        'st_hardpipe_winsys.c',
-        'st_softpipe_winsys.c',
-    ]
-
-    env.Prepend(LIBS = [
-        ws_null,
-        trace,
-        gallium,
+if env['toolchain'] == 'crossmingw':
+    # Cross-compilation not supported
+    Return()
+
+if not env.Detect(['swig']):
+    Return()
+
+env = env.Clone()
+
+env.Tool('python')
+
+env.Tool('swig')
+env.Append(SWIGPATH = ['#src/gallium/include', '#src/gallium/include/pipe'])
+env.Append(SWIGFLAGS = ['-python', '-keyword'])
+
+env.Append(CPPPATH = '.')
+
+if env['platform'] == 'windows':
+    env.Append(LIBS = [
+        'opengl32',
+        'gdi32',
+        'user32',
+        'kernel32',
+        'ws2_32',
+    ])
+else:
+    env.Append(CPPDEFINES = ['GCC_HASCLASSVISIBILITY'])
+    env.Append(LIBS = [
+        'GL',
+        'X11',
     ])
 
-    if env['llvm']:
-        env.Append(CPPDEFINES = ['HAVE_LLVMPIPE'])
-        env.Prepend(LIBS = [llvmpipe])
-    if True:
-        env.Append(CPPDEFINES = ['HAVE_SOFTPIPE'])
-        env.Prepend(LIBS = [softpipe])
+sources = [
+    'gallium.i',
+    'st_device.c',
+    'st_sample.c',
+    'st_hardpipe_winsys.c',
+    'st_softpipe_winsys.c',
+]
+
+env.Prepend(LIBS = [
+    ws_null,
+    trace,
+    gallium,
+])
+
+if env['llvm']:
+    env.Append(CPPDEFINES = ['HAVE_LLVMPIPE'])
+    env.Prepend(LIBS = [llvmpipe])
+if True:
+    env.Append(CPPDEFINES = ['HAVE_SOFTPIPE'])
+    env.Prepend(LIBS = [softpipe])
+
+env['no_import_lib'] = 1
 
-    env['no_import_lib'] = 1
+pyst = env.SharedLibrary(
+    target = '_gallium',
+    source = sources,
+)
 
-    env.SharedLibrary(
-        target = '_gallium',
-        source = sources,
-    )
+env.Alias('python', pyst)
diff --git a/src/gallium/state_trackers/vega/SConscript b/src/gallium/state_trackers/vega/SConscript
index 548053eb64..4c6e18f42b 100644
--- a/src/gallium/state_trackers/vega/SConscript
+++ b/src/gallium/state_trackers/vega/SConscript
@@ -3,49 +3,48 @@
 
 Import('*')
 
-if 'egl' in env['statetrackers']:
+env = env.Clone()
 
-    env = env.Clone()
+env.Append(CPPPATH = [
+    '#/src/mapi',
+])
 
-    env.Append(CPPPATH = [
-	'#/src/mapi',
-    ])
+vega_sources = [
+   'api.c',
+   'api_context.c',
+   'api_filters.c',
+   'api_images.c',
+   'api_masks.c',
+   'api_misc.c',
+   'api_paint.c',
+   'api_params.c',
+   'api_path.c',
+   'api_text.c',
+   'api_transform.c',
+   'vgu.c',
+   'vg_context.c',
+   'vg_manager.c',
+   'vg_state.c',
+   'vg_translate.c',
+   'polygon.c',
+   'bezier.c',
+   'path.c',
+   'paint.c',
+   'arc.c',
+   'image.c',
+   'renderer.c',
+   'stroker.c',
+   'mask.c',
+   'shader.c',
+   'shaders_cache.c',
+]
 
-    vega_sources = [
-       'api.c',
-       'api_context.c',
-       'api_filters.c',
-       'api_images.c',
-       'api_masks.c',
-       'api_misc.c',
-       'api_paint.c',
-       'api_params.c',
-       'api_path.c',
-       'api_text.c',
-       'api_transform.c',
-       'vgu.c',
-       'vg_context.c',
-       'vg_manager.c',
-       'vg_state.c',
-       'vg_translate.c',
-       'polygon.c',
-       'bezier.c',
-       'path.c',
-       'paint.c',
-       'arc.c',
-       'image.c',
-       'renderer.c',
-       'stroker.c',
-       'mask.c',
-       'shader.c',
-       'shaders_cache.c',
-    ]
+# vgapi_header must be generated first
+env.Depends(vega_sources, vgapi_header)
 
-    # vgapi_header must be generated first
-    env.Depends(vega_sources, vgapi_header)
+st_vega = env.ConvenienceLibrary(
+    target = 'st_vega',
+    source = vega_sources,
+)
 
-    st_vega = env.ConvenienceLibrary(
-	target = 'st_vega',
-	source = vega_sources,
-    )
-    Export('st_vega')
+Export('st_vega')
diff --git a/src/gallium/state_trackers/vega/api_filters.c b/src/gallium/state_trackers/vega/api_filters.c
index 8ace985336..4bd5d7e4bc 100644
--- a/src/gallium/state_trackers/vega/api_filters.c
+++ b/src/gallium/state_trackers/vega/api_filters.c
@@ -37,10 +37,8 @@
 #include "pipe/p_state.h"
 #include "util/u_inlines.h"
 #include "pipe/p_screen.h"
-#include "pipe/p_shader_tokens.h"
 
 #include "util/u_format.h"
-#include "util/u_memory.h"
 #include "util/u_sampler.h"
 #include "util/u_string.h"
 
diff --git a/src/gallium/state_trackers/vega/api_images.c b/src/gallium/state_trackers/vega/api_images.c
index c36b3d2f3c..7054d9bb7e 100644
--- a/src/gallium/state_trackers/vega/api_images.c
+++ b/src/gallium/state_trackers/vega/api_images.c
@@ -36,9 +36,7 @@
 #include "pipe/p_context.h"
 #include "pipe/p_screen.h"
 #include "util/u_inlines.h"
-#include "util/u_blit.h"
 #include "util/u_tile.h"
-#include "util/u_memory.h"
 
 static INLINE VGboolean supported_image_format(VGImageFormat format)
 {
diff --git a/src/gallium/state_trackers/vega/api_masks.c b/src/gallium/state_trackers/vega/api_masks.c
index 94c1ff5375..189390ec2d 100644
--- a/src/gallium/state_trackers/vega/api_masks.c
+++ b/src/gallium/state_trackers/vega/api_masks.c
@@ -27,7 +27,6 @@
 #include "VG/openvg.h"
 
 #include "mask.h"
-#include "renderer.h"
 #include "api.h"
 
 #include "vg_context.h"
@@ -36,7 +35,6 @@
 
 #include "util/u_pack_color.h"
 #include "util/u_draw_quad.h"
-#include "util/u_memory.h"
 
 #define DISABLE_1_1_MASKING 1
 
diff --git a/src/gallium/state_trackers/vega/api_paint.c b/src/gallium/state_trackers/vega/api_paint.c
index d88341b04f..1411806455 100644
--- a/src/gallium/state_trackers/vega/api_paint.c
+++ b/src/gallium/state_trackers/vega/api_paint.c
@@ -28,7 +28,6 @@
 
 #include "vg_context.h"
 #include "paint.h"
-#include "image.h"
 #include "api.h"
 
 VGPaint vegaCreatePaint(void)
diff --git a/src/gallium/state_trackers/vega/api_path.c b/src/gallium/state_trackers/vega/api_path.c
index 6c9a2e8d65..f76adddb58 100644
--- a/src/gallium/state_trackers/vega/api_path.c
+++ b/src/gallium/state_trackers/vega/api_path.c
@@ -28,13 +28,9 @@
 
 #include "vg_context.h"
 #include "path.h"
-#include "polygon.h"
-#include "paint.h"
 #include "api.h"
 
 #include "pipe/p_context.h"
-#include "util/u_inlines.h"
-#include "util/u_draw_quad.h"
 
 VGPath vegaCreatePath(VGint pathFormat,
                       VGPathDatatype datatype,
diff --git a/src/gallium/state_trackers/vega/api_text.c b/src/gallium/state_trackers/vega/api_text.c
index b35f3be90a..2a62da0a1d 100644
--- a/src/gallium/state_trackers/vega/api_text.c
+++ b/src/gallium/state_trackers/vega/api_text.c
@@ -27,7 +27,6 @@
 #include "VG/openvg.h"
 
 #include "vg_context.h"
-#include "api.h"
 
 #include "util/u_memory.h"
 
diff --git a/src/gallium/state_trackers/vega/image.c b/src/gallium/state_trackers/vega/image.c
index c12dc71b86..28bbe420a2 100644
--- a/src/gallium/state_trackers/vega/image.c
+++ b/src/gallium/state_trackers/vega/image.c
@@ -32,13 +32,11 @@
 #include "renderer.h"
 #include "util_array.h"
 #include "api_consts.h"
-#include "shaders_cache.h"
 #include "shader.h"
 
 #include "pipe/p_context.h"
 #include "pipe/p_screen.h"
 #include "util/u_inlines.h"
-#include "util/u_blit.h"
 #include "util/u_format.h"
 #include "util/u_tile.h"
 #include "util/u_memory.h"
diff --git a/src/gallium/state_trackers/vega/paint.c b/src/gallium/state_trackers/vega/paint.c
index 2c0eb6b23d..c0c8cddabe 100644
--- a/src/gallium/state_trackers/vega/paint.c
+++ b/src/gallium/state_trackers/vega/paint.c
@@ -26,7 +26,6 @@
 
 #include "paint.h"
 
-#include "shaders_cache.h"
 #include "matrix.h"
 #include "image.h"
 #include "st_inlines.h"
@@ -34,7 +33,6 @@
 #include "pipe/p_compiler.h"
 #include "util/u_inlines.h"
 
-#include "util/u_format.h"
 #include "util/u_memory.h"
 #include "util/u_math.h"
 #include "util/u_sampler.h"
diff --git a/src/gallium/state_trackers/vega/polygon.c b/src/gallium/state_trackers/vega/polygon.c
index bc94170eb9..ca8831064c 100644
--- a/src/gallium/state_trackers/vega/polygon.c
+++ b/src/gallium/state_trackers/vega/polygon.c
@@ -29,7 +29,6 @@
 #include "matrix.h" /*for floatsEqual*/
 #include "vg_context.h"
 #include "vg_state.h"
-#include "paint.h"
 #include "renderer.h"
 #include "util_array.h"
 #include "VG/openvg.h"
diff --git a/src/gallium/state_trackers/vega/renderer.c b/src/gallium/state_trackers/vega/renderer.c
index 8c023044c4..c6f5f7a05b 100644
--- a/src/gallium/state_trackers/vega/renderer.c
+++ b/src/gallium/state_trackers/vega/renderer.c
@@ -35,12 +35,9 @@
 #include "pipe/p_shader_tokens.h"
 
 #include "util/u_draw_quad.h"
-#include "util/u_format.h"
 #include "util/u_simple_shaders.h"
 #include "util/u_memory.h"
-#include "util/u_rect.h"
 #include "util/u_sampler.h"
-#include "util/u_surface.h"
 
 #include "cso_cache/cso_context.h"
 
diff --git a/src/gallium/state_trackers/vega/shaders_cache.c b/src/gallium/state_trackers/vega/shaders_cache.c
index 53e6bfcf16..e002a7ed42 100644
--- a/src/gallium/state_trackers/vega/shaders_cache.c
+++ b/src/gallium/state_trackers/vega/shaders_cache.c
@@ -30,8 +30,6 @@
 
 #include "pipe/p_context.h"
 #include "pipe/p_defines.h"
-#include "util/u_inlines.h"
-#include "pipe/p_screen.h"
 #include "pipe/p_shader_tokens.h"
 
 #include "tgsi/tgsi_build.h"
diff --git a/src/gallium/state_trackers/vega/vg_context.c b/src/gallium/state_trackers/vega/vg_context.c
index 5cb2590602..b1f98fa477 100644
--- a/src/gallium/state_trackers/vega/vg_context.c
+++ b/src/gallium/state_trackers/vega/vg_context.c
@@ -37,7 +37,6 @@
 
 #include "pipe/p_context.h"
 #include "util/u_inlines.h"
-#include "pipe/p_shader_tokens.h"
 
 #include "cso_cache/cso_context.h"
 
diff --git a/src/gallium/state_trackers/xorg/SConscript b/src/gallium/state_trackers/xorg/SConscript
index 0b598dab6e..bc7821f8a8 100644
--- a/src/gallium/state_trackers/xorg/SConscript
+++ b/src/gallium/state_trackers/xorg/SConscript
@@ -3,34 +3,38 @@
 
 Import('*')
 
-if 'xorg' in env['statetrackers']:
+if not env['x11'] or not env['dri'] or not env['xorg']:
+    Return()
 
-    env = env.Clone()
+env = env.Clone()
 
-    env.Append(CPPPATH = [
-	'#/src/mesa',
-    ])
+env.Append(CPPPATH = [
+    '#/src/mesa',
+])
 
-    env.ParseConfig('pkg-config --cflags --libs libdrm xorg-server')
+env.ParseConfig('pkg-config --cflags --libs libdrm xorg-server')
 
-    conf = env.Configure()
+if env['kms']:
+    env.Append(CPPDEFINES = ['HAVE_LIBKMS'])
 
-    if conf.CheckHeader('X11/extensions/dpmsconst.h'):
-        env.Append(CPPDEFINES = [('HAVE_XEXTPROTO_71', '1')])
+conf = env.Configure()
 
-    conf.Finish()
+if conf.CheckHeader('X11/extensions/dpmsconst.h'):
+    env.Append(CPPDEFINES = [('HAVE_XEXTPROTO_71', '1')])
 
-    st_xorg = env.ConvenienceLibrary(
-	target = 'st_xorg',
-	source = [ 'xorg_composite.c',
-		'xorg_crtc.c',
-		'xorg_dri2.c',
-		'xorg_driver.c',
-		'xorg_exa.c',
-		'xorg_exa_tgsi.c',
-		'xorg_output.c',
-		'xorg_renderer.c',
-		'xorg_xv.c',
-		]
-    )
-    Export('st_xorg')
+conf.Finish()
+
+st_xorg = env.ConvenienceLibrary(
+target = 'st_xorg',
+source = [ 'xorg_composite.c',
+    'xorg_crtc.c',
+    'xorg_dri2.c',
+    'xorg_driver.c',
+    'xorg_exa.c',
+    'xorg_exa_tgsi.c',
+    'xorg_output.c',
+    'xorg_renderer.c',
+    'xorg_xv.c',
+    ]
+)
+Export('st_xorg')
diff --git a/src/gallium/state_trackers/xorg/xorg_driver.c b/src/gallium/state_trackers/xorg/xorg_driver.c
index 1ec772df17..1ee79ae177 100644
--- a/src/gallium/state_trackers/xorg/xorg_driver.c
+++ b/src/gallium/state_trackers/xorg/xorg_driver.c
@@ -447,7 +447,7 @@ drv_pre_init(ScrnInfoPtr pScrn, int flags)
 
     if (!drv_init_resource_management(pScrn)) {
 	xf86DrvMsg(pScrn->scrnIndex, X_ERROR, "Could not init "
-					       "Gallium3D or libKMS.");
+					       "Gallium3D or libKMS.\n");
 	return FALSE;
     }
 
diff --git a/src/gallium/targets/SConscript b/src/gallium/targets/SConscript
deleted file mode 100644
index e447d09361..0000000000
--- a/src/gallium/targets/SConscript
+++ /dev/null
@@ -1,41 +0,0 @@
-import os
-Import('*')
-
-# Compatibility with old build scripts:
-#
-if 'mesa' in env['statetrackers']:
-    if 'xlib' in env['winsys'] and 'libgl-xlib' not in env['targets']:
-        env['targets'].append('libgl-xlib')
-    if 'gdi' in env['winsys'] and 'libgl-gdi' not in env['targets']:
-        env['targets'].append('libgl-gdi')
-
-if not 'graw-xlib' in env['targets'] and not 'graw-null' in env['targets'] and not env['msvc']:
-        # XXX: disable until MSVC can link correctly
-        SConscript('graw-null/SConscript')
-
-
-if env['dri']:
-	SConscript([
-		'SConscript.dri'
-	])
-
-if 'xorg' in env['statetrackers']:
-	if 'vmware' in env['winsys']:
-		SConscript([
-			'xorg-vmwgfx/SConscript',
-		])
-
-if 'egl' in env['statetrackers']:
-	SConscript([
-		'egl-gdi/SConscript',
-	])
-
-# Ideally all non-target directories would produce convenience
-# libraries, and the actual shared libraries and other installables
-# would be finally assembled in the targets subtree:
-#
-for target in env['targets']:
-    SConscript(os.path.join(target, 'SConscript'))
-
-
-
diff --git a/src/gallium/targets/SConscript.dri b/src/gallium/targets/SConscript.dri
index bc8d179e3d..092223b369 100644
--- a/src/gallium/targets/SConscript.dri
+++ b/src/gallium/targets/SConscript.dri
@@ -3,6 +3,9 @@
 
 Import('*')
 
+if not env['dri']:
+    Return()
+
 drienv = env.Clone()
 
 drienv.Replace(CPPPATH = [
@@ -77,32 +80,3 @@ Export([
 	'COMMON_DRI_SW_OBJECTS',
 	'COMMON_DRI_DRM_OBJECTS',
 ])
-
-SConscript([
-	'dri-swrast/SConscript',
-])
-
-if 'vmware' in env['winsys']:
-	SConscript([
-		'dri-vmwgfx/SConscript',
-	])
-
-if 'i915' in env['winsys']:
-	SConscript([
-		'dri-i915/SConscript',
-	])
-
-if 'i965' in env['winsys']:
-	SConscript([
-		'dri-i965/SConscript',
-	])
-
-if 'radeon' in env['winsys']:
-	SConscript([
-		'dri-radeong/SConscript',
-	])
-
-if 'r600' in env['winsys']:
-	SConscript([
-		'dri-r600/SConscript',
-	])
diff --git a/src/gallium/targets/dri-i915/SConscript b/src/gallium/targets/dri-i915/SConscript
index 172f92d6b8..ab60013830 100644
--- a/src/gallium/targets/dri-i915/SConscript
+++ b/src/gallium/targets/dri-i915/SConscript
@@ -1,9 +1,5 @@
 Import('*')
 
-if not 'i915' in env['drivers']:
-    print 'warning: i915 pipe driver not built skipping i915_dri.so'
-    Return()
-
 env = drienv.Clone()
 
 env.ParseConfig('pkg-config --cflags --libs libdrm_intel')
@@ -24,8 +20,10 @@ env.Prepend(LIBS = [
     COMMON_DRI_DRM_OBJECTS
 ])
 
-env.LoadableModule(
+module = env.LoadableModule(
     target = 'i915_dri.so',
     source = 'target.c',
     SHLIBPREFIX = '',
 )
+
+env.Alias('dri-i915', module)
\ No newline at end of file
diff --git a/src/gallium/targets/dri-i965/SConscript b/src/gallium/targets/dri-i965/SConscript
index 684e3488f7..669f70d6b8 100644
--- a/src/gallium/targets/dri-i965/SConscript
+++ b/src/gallium/targets/dri-i965/SConscript
@@ -1,9 +1,5 @@
 Import('*')
 
-if not 'i965' in env['drivers']:
-    print 'warning: i965 pipe driver not built skipping i965_dri.so'
-    Return()
-
 env = drienv.Clone()
 
 env.ParseConfig('pkg-config --cflags --libs libdrm_intel')
@@ -27,8 +23,10 @@ env.Prepend(LIBS = [
     COMMON_DRI_DRM_OBJECTS
 ])
 
-env.LoadableModule(
+module = env.LoadableModule(
     target = 'i965_dri.so',
     source = 'target.c',
     SHLIBPREFIX = '',
 )
+
+env.Alias('dri-i965', module)
\ No newline at end of file
diff --git a/src/gallium/targets/dri-r300/SConscript b/src/gallium/targets/dri-r300/SConscript
index 33a458f2e6..005b4bbf7f 100644
--- a/src/gallium/targets/dri-r300/SConscript
+++ b/src/gallium/targets/dri-r300/SConscript
@@ -1,9 +1,5 @@
 Import('*')
 
-if not 'r300' in env['drivers']:
-    print 'warning: r300 pipe driver not built skipping r300_dri.so'
-    Return()
-
 env = drienv.Clone()
 
 env.ParseConfig('pkg-config --cflags --libs libdrm_radeon')
@@ -23,8 +19,10 @@ env.Prepend(LIBS = [
     COMMON_DRI_DRM_OBJECTS
 ])
 
-env.SharedLibrary(
+module = env.SharedLibrary(
     target ='r300_dri.so',
     source = 'target.c',
     SHLIBPREFIX = '',
 )
+
+env.Alias('dri-r300', module)
\ No newline at end of file
diff --git a/src/gallium/targets/dri-r600/SConscript b/src/gallium/targets/dri-r600/SConscript
index 64d6d2a7f6..aa771db2d1 100644
--- a/src/gallium/targets/dri-r600/SConscript
+++ b/src/gallium/targets/dri-r600/SConscript
@@ -1,9 +1,5 @@
 Import('*')
 
-if not 'r600' in env['drivers']:
-    print 'warning: r600 pipe driver not built skipping r600_dri.so'
-    Return()
-
 env = drienv.Clone()
 
 env.ParseConfig('pkg-config --cflags --libs libdrm_radeon')
@@ -22,8 +18,10 @@ env.Prepend(LIBS = [
     COMMON_DRI_DRM_OBJECTS
 ])
 
-env.SharedLibrary(
+module = env.SharedLibrary(
     target ='r600_dri.so',
     source = 'target.c',
     SHLIBPREFIX = '',
 )
+
+env.Alias('dri-r600', module)
diff --git a/src/gallium/targets/dri-swrast/SConscript b/src/gallium/targets/dri-swrast/SConscript
index d814347119..b67483800e 100644
--- a/src/gallium/targets/dri-swrast/SConscript
+++ b/src/gallium/targets/dri-swrast/SConscript
@@ -27,15 +27,16 @@ if True:
 
 if env['llvm']:
     env.Append(CPPDEFINES = 'GALLIUM_LLVMPIPE')
-    env.Tool('udis86')
     env.Prepend(LIBS = [llvmpipe])
 
 swrastg_sources = [
     'swrast_drm_api.c'
 ]
 
-env.LoadableModule(
+module = env.LoadableModule(
     target ='swrastg_dri.so',
     source = swrastg_sources,
     SHLIBPREFIX = '',
 )
+
+env.Alias('dri-swrast', module)
diff --git a/src/gallium/targets/dri-vmwgfx/SConscript b/src/gallium/targets/dri-vmwgfx/SConscript
index 7afabc7429..7888e4f2c8 100644
--- a/src/gallium/targets/dri-vmwgfx/SConscript
+++ b/src/gallium/targets/dri-vmwgfx/SConscript
@@ -1,9 +1,5 @@
 Import('*')
 
-if not 'svga' in env['drivers']:
-    print 'warning: svga pipe driver not built skipping vmwgfx_dri.so'
-    Return()
-
 env = drienv.Clone()
 
 env.Append(CPPDEFINES = ['GALLIUM_RBUG', 'GALLIUM_TRACE'])
@@ -20,8 +16,10 @@ env.Prepend(LIBS = [
     COMMON_DRI_DRM_OBJECTS
 ])
 
-env.LoadableModule(
+module = env.LoadableModule(
     target = 'vmwgfx_dri.so',
     source = 'target.c',
     SHLIBPREFIX = '',
 )
+
+env.Alias('dri-vmwgfx', module)
\ No newline at end of file
diff --git a/src/gallium/targets/egl-gdi/SConscript b/src/gallium/targets/egl-gdi/SConscript
index 8f8b28ef67..d52eeb70fd 100644
--- a/src/gallium/targets/egl-gdi/SConscript
+++ b/src/gallium/targets/egl-gdi/SConscript
@@ -3,45 +3,53 @@
 
 Import('*')
 
-if env['platform'] == 'windows':
-
-    env = env.Clone()
-
-    env.Append(CPPPATH = [
-            '#/src/gallium/state_trackers/egl',
-            '#/src/gallium/state_trackers/vega',
-            '#/src/egl/main',
-            '#/src/mesa',
-    ])
-
-    env.Append(CPPDEFINES = [
-            'FEATURE_VG=1',
-            'GALLIUM_SOFTPIPE',
-            'GALLIUM_RBUG',
-            'GALLIUM_TRACE',
-    ])
-
-    env.Append(LIBS = [
-        'gdi32',
-        'user32',
-        'kernel32',
-        'ws2_32',
-    ])
-
-    env['no_import_lib'] = 1
-
-    drivers = [softpipe]
-    if env['llvm']:
-        env.Append(CPPDEFINES = 'GALLIUM_LLVMPIPE')
-        drivers += [llvmpipe]
-    drivers += [identity, trace, rbug]
-
-    apis = [vgapi, st_vega]
-
-    egl_gallium = env.SharedLibrary(
-        target ='egl_gallium',
-        source = 'egl-static.c',
-        LIBS = st_egl_gdi + ws_gdi + drivers + apis + gallium + egl + env['LIBS'],
-    )
-
-    env.InstallSharedLibrary(egl_gallium)
+env = env.Clone()
+
+env.Append(CPPPATH = [
+        '#/src/gallium/state_trackers/egl',
+        '#/src/gallium/state_trackers/vega',
+        '#/src/egl/main',
+        '#/src/mesa',
+])
+
+env.Append(CPPDEFINES = [
+        'FEATURE_VG=1',
+        'GALLIUM_SOFTPIPE',
+        'GALLIUM_RBUG',
+        'GALLIUM_TRACE',
+])
+
+env.Append(LIBS = [
+    'gdi32',
+    'user32',
+    'kernel32',
+    'ws2_32',
+])
+
+env.Prepend(LIBS = [
+    st_egl_gdi,
+    ws_gdi,
+    identity,
+    trace,
+    rbug,
+    softpipe,
+    vgapi,
+    st_vega,
+    gallium,
+    egl,
+])
+
+if env['llvm']:
+    env.Append(CPPDEFINES = 'GALLIUM_LLVMPIPE')
+    env.Prepend(LIBS = [llvmpipe])
+
+egl_gallium = env.SharedLibrary(
+    target ='egl_gallium',
+    source = 'egl-static.c',
+)
+
+env['no_import_lib'] = 1
+
+egl_gdi = env.InstallSharedLibrary(egl_gallium)
+
+env.Alias('egl-gdi', egl_gdi)
diff --git a/src/gallium/targets/egl/Makefile b/src/gallium/targets/egl/Makefile
index 57979c4e9d..63e9352144 100644
--- a/src/gallium/targets/egl/Makefile
+++ b/src/gallium/targets/egl/Makefile
@@ -54,9 +54,8 @@ egl_LIBS += $(TOP)/src/gallium/winsys/sw/fbdev/libfbdev.a
 endif
 
 # EGL_RENDERABLE_TYPE is a compile time attribute
-egl_CPPFLAGS += $(API_DEFINES)
 ifneq ($(filter $(GL_LIB), $(EGL_CLIENT_APIS)),)
-egl_CPPFLAGS += -DFEATURE_GL=1
+egl_CPPFLAGS += $(API_DEFINES)
 endif
 ifneq ($(filter $(GLESv1_CM_LIB), $(EGL_CLIENT_APIS)),)
 egl_CPPFLAGS += -DFEATURE_ES1=1
diff --git a/src/gallium/targets/graw-gdi/SConscript b/src/gallium/targets/graw-gdi/SConscript
new file mode 100644
index 0000000000..8ee8915ace
--- /dev/null
+++ b/src/gallium/targets/graw-gdi/SConscript
@@ -0,0 +1,41 @@
+#######################################################################
+# SConscript for graw-gdi
+
+Import('*')
+
+env = env.Clone()
+
+env.Append(CPPPATH = [
+    '#src/gallium/winsys/sw',
+])
+
+env.Prepend(LIBS = [
+    gallium,
+    'gdi32',
+    identity,
+    rbug,
+    trace,
+    'user32',
+    'ws2_32',
+])
+
+sources = [
+    'graw_gdi.c',
+    graw_util,
+]
+
+env.Append(CPPDEFINES = 'GALLIUM_SOFTPIPE')
+env.Prepend(LIBS = [softpipe])
+
+graw = env.SharedLibrary(
+    target = 'graw',
+    source = sources,
+    LIBS = ws_gdi + env['LIBS'],
+)
+
+if env['platform'] == 'windows':
+    graw = env.FindIxes(graw, 'LIBPREFIX', 'LIBSUFFIX')
+else:
+    graw = env.FindIxes(graw, 'SHLIBPREFIX', 'SHLIBSUFFIX')
+
+env.Alias('graw-gdi', graw)
diff --git a/src/gallium/targets/graw-gdi/graw_gdi.c b/src/gallium/targets/graw-gdi/graw_gdi.c
new file mode 100644
index 0000000000..52a4e3cb23
--- /dev/null
+++ b/src/gallium/targets/graw-gdi/graw_gdi.c
@@ -0,0 +1,158 @@
+/**************************************************************************
+ *
+ * Copyright 2010 VMware, Inc.
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDERS, AUTHORS AND/OR ITS SUPPLIERS BE LIABLE FOR ANY CLAIM,
+ * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR
+ * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE
+ * USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ *
+ *
+ **************************************************************************/
+
+#include "gdi/gdi_sw_winsys.h"
+#include "pipe/p_screen.h"
+#include "softpipe/sp_public.h"
+#include "state_tracker/graw.h"
+#include "sw/sw_public.h"
+#include "target-helpers/wrap_screen.h"
+#include <windows.h>
+
+
+static LRESULT CALLBACK
+window_proc(HWND hWnd,
+            UINT uMsg,
+            WPARAM wParam,
+            LPARAM lParam)
+{
+   switch (uMsg) {
+   case WM_DESTROY:
+      PostQuitMessage(0);
+      break;
+
+   default:
+      return DefWindowProc(hWnd, uMsg, wParam, lParam);
+   }
+
+   return 0;
+}
+
+static struct {
+   void (* draw)(void);
+} graw;
+
+struct pipe_screen *
+graw_create_window_and_screen(int x,
+                              int y,
+                              unsigned width,
+                              unsigned height,
+                              enum pipe_format format,
+                              void **handle)
+{
+   struct sw_winsys *winsys = NULL;
+   struct pipe_screen *screen = NULL;
+   WNDCLASSEX wc = {sizeof(wc)};
+   UINT style = WS_VISIBLE | WS_TILEDWINDOW;
+   RECT rect;
+   HWND hWnd = NULL;
+   HDC hDC = NULL;
+
+   if (format != PIPE_FORMAT_R8G8B8A8_UNORM)
+      goto fail;
+
+   winsys = gdi_create_sw_winsys();
+   if (winsys == NULL)
+      goto fail;
+
+   screen = softpipe_create_screen(winsys);
+   if (screen == NULL)
+      goto fail;
+
+   wc.style = CS_OWNDC | CS_HREDRAW | CS_VREDRAW;
+   wc.lpfnWndProc = window_proc;
+   wc.lpszClassName = "graw-gdi";
+   wc.hInstance = GetModuleHandle(NULL);
+   wc.hIcon = LoadIcon(NULL, IDI_APPLICATION);
+   wc.hCursor = LoadCursor(NULL, IDC_ARROW);
+   wc.hbrBackground = (HBRUSH)(COLOR_BTNFACE + 1);
+   RegisterClassEx(&wc);
+
+   SetRect(&rect, 0, 0, width, height);
+   AdjustWindowRectEx(&rect, style, FALSE, 0);
+
+   hWnd = CreateWindowEx(0,
+                         wc.lpszClassName,
+                         wc.lpszClassName,
+                         style,
+                         x,
+                         y,
+                         rect.right - rect.left,
+                         rect.bottom - rect.top,
+                         NULL,
+                         NULL,
+                         wc.hInstance,
+                         0);
+   if (hWnd == NULL)
+      goto fail;
+
+   hDC = GetDC(hWnd);
+   if (hDC == NULL)
+      goto fail;
+
+   *handle = (void *)hDC;
+
+   return gallium_wrap_screen(screen);
+
+fail:
+   if (hWnd)
+      DestroyWindow(hWnd);
+
+   if (screen)
+      screen->destroy(screen);
+
+   return NULL;
+}
+
+void 
+graw_set_display_func(void (* draw)(void))
+{
+   graw.draw = draw;
+}
+
+void
+graw_main_loop(void)
+{
+   for (;;) {
+      MSG msg;
+
+      while (PeekMessage(&msg, NULL, 0, 0, PM_REMOVE)) {
+         if (msg.message == WM_QUIT) {
+            return;
+         }
+         TranslateMessage(&msg);
+         DispatchMessage(&msg);
+      }
+
+      if (graw.draw) {
+         graw.draw();
+      }
+
+      Sleep(0);
+   }
+}
diff --git a/src/gallium/targets/graw-null/SConscript b/src/gallium/targets/graw-null/SConscript
index 3416989d8e..ebac1728f0 100644
--- a/src/gallium/targets/graw-null/SConscript
+++ b/src/gallium/targets/graw-null/SConscript
@@ -5,54 +5,28 @@ Import('*')
 
 env = env.Clone()
 
-env.Prepend(LIBS = [
-    ws_null,
-    trace,
-    rbug,
-    identity,
-#    gallium,
-])
-
-env.Append(CPPPATH = [
-    '#src/gallium/drivers',
-])
+graw_util = env.SharedObject(
+    source = ['graw_util.c'],
+)
 
-if env['platform'] == 'windows':
-    # For trace
-    env.Append(LIBS = [
-        'ws2_32',
-    ])
+env = env.Clone()
 
 sources = [
     'graw_null.c',
-    '../graw-xlib/graw_util.c',
+    graw_util,
 ]
 
-if True:
-    env.Append(CPPDEFINES = 'GALLIUM_SOFTPIPE')
-    env.Prepend(LIBS = [softpipe])
-
-if env['llvm']:
-    env.Append(CPPDEFINES = 'GALLIUM_LLVMPIPE')
-    env.Tool('udis86')
-    env.Prepend(LIBS = [llvmpipe])
-
-# Need this for trace, identity drivers referenced by
-# gallium_wrap_screen().
-#
 env.Prepend(LIBS = [gallium])
 
 # TODO: write a wrapper function http://www.scons.org/wiki/WrapperFunctions
 graw = env.SharedLibrary(
-    target ='graw',
+    target = 'graw',
     source = sources,
 )
 
-env.InstallSharedLibrary(graw, version=(1, 0))
-
 if env['platform'] == 'windows':
     graw = env.FindIxes(graw, 'LIBPREFIX', 'LIBSUFFIX')
 else:
     graw = env.FindIxes(graw, 'SHLIBPREFIX', 'SHLIBSUFFIX')
 
-Export('graw')
+Export('graw_util', 'graw')
diff --git a/src/gallium/targets/graw-null/graw_null.c b/src/gallium/targets/graw-null/graw_null.c
index 5939a5acd3..f1fe3872c9 100644
--- a/src/gallium/targets/graw-null/graw_null.c
+++ b/src/gallium/targets/graw-null/graw_null.c
@@ -1,31 +1,5 @@
-#include "pipe/p_compiler.h"
-#include "util/u_debug.h"
-#include "util/u_memory.h"
-#include "target-helpers/wrap_screen.h"
-#include "sw/null/null_sw_winsys.h"
-#include "os/os_time.h"
 #include "state_tracker/graw.h"
 
-#ifdef GALLIUM_SOFTPIPE
-#include "softpipe/sp_public.h"
-#endif
-
-#ifdef GALLIUM_LLVMPIPE
-#include "llvmpipe/lp_public.h"
-#endif
-
-/* Haven't figured out a decent way to build the helper code yet -
- * #include it here temporarily.
- */
-#include "sw/sw_public.h"
-#include "sw/sw.c"
-
-#include <stdio.h>
-
-
-static struct {
-   void (*draw)(void);
-} graw;
 
 
 
@@ -37,45 +11,7 @@ graw_create_window_and_screen( int x,
                                enum pipe_format format,
                                void **handle)
 {
-   const char *default_driver;
-   const char *driver;
-   struct pipe_screen *screen = NULL;
-   struct sw_winsys *winsys = NULL;
-   static int dummy;
-
-
-   /* Create the underlying winsys, which performs presents to Xlib
-    * drawables:
-    */
-   winsys = null_sw_create();
-   if (winsys == NULL)
-      return NULL;
-
-#if defined(GALLIUM_LLVMPIPE)
-   default_driver = "llvmpipe";
-#elif defined(GALLIUM_SOFTPIPE)
-   default_driver = "softpipe";
-#else
-   default_driver = "";
-#endif
-
-   driver = debug_get_option("GALLIUM_DRIVER", default_driver);
-
-#if defined(GALLIUM_LLVMPIPE)
-   if (screen == NULL && strcmp(driver, "llvmpipe") == 0)
-      screen = llvmpipe_create_screen( winsys );
-#endif
-
-#if defined(GALLIUM_SOFTPIPE)
-   if (screen == NULL)
-      screen = softpipe_create_screen( winsys );
-#endif
-
-   *handle = &dummy;
-
-   /* Inject any wrapping layers we want to here:
-    */
-   return gallium_wrap_screen( screen );
+   return NULL;
 }
 
 
@@ -83,13 +19,10 @@ graw_create_window_and_screen( int x,
 void 
 graw_set_display_func( void (*draw)( void ) )
 {
-   graw.draw = draw;
 }
 
 
 void
 graw_main_loop( void )
 {
-   graw.draw();
-   os_time_sleep(100000);
 }
diff --git a/src/gallium/targets/graw-xlib/graw_util.c b/src/gallium/targets/graw-null/graw_util.c
index fc7c9ae6f9..531757f145 100644
--- a/src/gallium/targets/graw-xlib/graw_util.c
+++ b/src/gallium/targets/graw-null/graw_util.c
@@ -9,8 +9,9 @@
 
 /* Helper functions.  These are the same for all graw implementations.
  */
-void *graw_parse_geometry_shader(struct pipe_context *pipe,
-                                 const char *text)
+PUBLIC void *
+graw_parse_geometry_shader(struct pipe_context *pipe,
+                           const char *text)
 {
    struct tgsi_token tokens[1024];
    struct pipe_shader_state state;
@@ -22,8 +23,9 @@ void *graw_parse_geometry_shader(struct pipe_context *pipe,
    return pipe->create_gs_state(pipe, &state);
 }
 
-void *graw_parse_vertex_shader(struct pipe_context *pipe,
-                               const char *text)
+PUBLIC void *
+graw_parse_vertex_shader(struct pipe_context *pipe,
+                         const char *text)
 {
    struct tgsi_token tokens[1024];
    struct pipe_shader_state state;
@@ -35,8 +37,9 @@ void *graw_parse_vertex_shader(struct pipe_context *pipe,
    return pipe->create_vs_state(pipe, &state);
 }
 
-void *graw_parse_fragment_shader(struct pipe_context *pipe,
-                                 const char *text)
+PUBLIC void *
+graw_parse_fragment_shader(struct pipe_context *pipe,
+                           const char *text)
 {
    struct tgsi_token tokens[1024];
    struct pipe_shader_state state;
diff --git a/src/gallium/targets/graw-xlib/SConscript b/src/gallium/targets/graw-xlib/SConscript
index 21fce948f4..e50eb8a03d 100644
--- a/src/gallium/targets/graw-xlib/SConscript
+++ b/src/gallium/targets/graw-xlib/SConscript
@@ -3,56 +3,41 @@
 
 Import('*')
 
-if env['platform'] != 'linux':
-    Return()
-
 env = env.Clone()
 
-env.Tool('x11')
-
 env.Prepend(LIBS = [
     ws_xlib,
     trace,
     rbug,
     identity,
-#    gallium,
+    gallium,
 ])
 
+env.Prepend(LIBS = env['X11_LIBS'])
+
 env.Append(CPPPATH = [
     '#src/gallium/drivers',
     '#src/gallium/include/state_tracker',
 ])
 
-
 sources = [
     'graw_xlib.c',
-    'graw_util.c',
+    graw_util
 ]
 
-env.Tool('x11')
-
 if True:
     env.Append(CPPDEFINES = 'GALLIUM_SOFTPIPE')
     env.Prepend(LIBS = [softpipe])
 
 if env['llvm']:
     env.Append(CPPDEFINES = 'GALLIUM_LLVMPIPE')
-    env.Tool('udis86')
     env.Prepend(LIBS = [llvmpipe])
 
-# Need this for trace, identity drivers referenced by
-# gallium_wrap_screen().
-#
-env.Prepend(LIBS = [gallium])
-
-# TODO: write a wrapper function http://www.scons.org/wiki/WrapperFunctions
 graw = env.SharedLibrary(
     target ='graw',
     source = sources,
 )
 
-env.InstallSharedLibrary(graw, version=(1, 0))
-
-graw = env.FindIxes(graw, 'SHLIBPREFIX', 'SHLIBSUFFIX')
+graw = env.InstallSharedLibrary(graw, version=(1, 0))
 
-Export('graw')
+env.Alias('graw-xlib', graw)
diff --git a/src/gallium/targets/libgl-gdi/SConscript b/src/gallium/targets/libgl-gdi/SConscript
index 12fe403f62..339238756a 100644
--- a/src/gallium/targets/libgl-gdi/SConscript
+++ b/src/gallium/targets/libgl-gdi/SConscript
@@ -3,45 +3,46 @@
 
 Import('*')
 
-if env['platform'] == 'windows':
-
-    env = env.Clone()
-
-    env.Append(CPPPATH = [
-        '#src/gallium/state_trackers/wgl',
-        '#src/gallium/winsys/sw',
-    ])
-
-    env.Append(LIBS = [
-        'gdi32',
-        'user32',
-        'kernel32',
-        'ws2_32',
-        talloc,
-    ])
-
-    sources = []
-    drivers = []
-
-    if True:
-        sources = ['gdi_softpipe_winsys.c']
-        drivers = [softpipe]
-
-    if env['llvm']:
-        sources = ['gdi_llvmpipe_winsys.c']
-        drivers = [llvmpipe]
-
-    if env['gcc']:
-        sources += ['#src/gallium/state_trackers/wgl/opengl32.mingw.def']
-    else:
-        sources += ['#src/gallium/state_trackers/wgl/opengl32.def']
-        
-    drivers += [trace, rbug]
-
-    env['no_import_lib'] = 1
-
-    env.SharedLibrary(
-        target ='opengl32',
-        source = sources,
-        LIBS = wgl + ws_gdi + glapi + mesa + drivers + gallium + glsl + env['LIBS'],
-    )
+env = env.Clone()
+
+env.Append(CPPPATH = [
+    '#src/gallium/state_trackers/wgl',
+    '#src/gallium/winsys/sw',
+])
+
+env.Append(LIBS = [
+    'gdi32',
+    'user32',
+    'kernel32',
+    'ws2_32',
+    talloc,
+])
+
+sources = []
+drivers = []
+
+if True:
+    sources = ['gdi_softpipe_winsys.c']
+    drivers = [softpipe]
+
+if env['llvm']:
+    sources = ['gdi_llvmpipe_winsys.c']
+    drivers = [llvmpipe]
+
+if env['gcc']:
+    sources += ['#src/gallium/state_trackers/wgl/opengl32.mingw.def']
+else:
+    sources += ['#src/gallium/state_trackers/wgl/opengl32.def']
+    
+drivers += [trace, rbug]
+
+env['no_import_lib'] = 1
+
+opengl32 = env.SharedLibrary(
+    target ='opengl32',
+    source = sources,
+    LIBS = wgl + ws_gdi + glapi + mesa + drivers + gallium + glsl + env['LIBS'],
+)
+
+env.Alias('opengl32', opengl32)
+env.Alias('libgl-gdi', opengl32)
diff --git a/src/gallium/targets/libgl-xlib/SConscript b/src/gallium/targets/libgl-xlib/SConscript
index 27b562e1d5..582760eac9 100644
--- a/src/gallium/targets/libgl-xlib/SConscript
+++ b/src/gallium/targets/libgl-xlib/SConscript
@@ -3,17 +3,6 @@
 
 Import('*')
 
-if env['platform'] != 'linux':
-    Return()
-
-if 'mesa' not in env['statetrackers']:
-    print 'warning: Mesa state tracker disabled: skipping build of xlib libGL.so'
-    Return()
-
-if env['dri']:
-    print 'warning: DRI enabled: skipping build of xlib libGL.so'
-    Return()
-
 env = env.Clone()
 
 env.Append(CPPPATH = [
@@ -25,6 +14,8 @@ env.Append(CPPPATH = [
 
 env.Append(CPPDEFINES = ['USE_XSHM'])
 
+env.Prepend(LIBS = env['X11_LIBS'])
+
 env.Prepend(LIBS = [
     st_xlib,
     ws_xlib,
@@ -42,8 +33,6 @@ sources = [
     'xlib.c',
 ]
 
-env.Tool('x11')
-
 if True:
     env.Append(CPPDEFINES = 'GALLIUM_SOFTPIPE')
     env.Prepend(LIBS = [softpipe])
@@ -54,10 +43,10 @@ if True:
 
 if env['llvm']:
     env.Append(CPPDEFINES = 'GALLIUM_LLVMPIPE')
-    env.Tool('udis86')
     env.Prepend(LIBS = [llvmpipe])
     
-if 'cell' in env['drivers']:
+if False:
+    # TODO: Detect Cell SDK
     env.Append(CPPDEFINES = 'GALLIUM_CELL')
     env.Prepend(LIBS = [cell])
 
@@ -67,6 +56,8 @@ libgl = env.SharedLibrary(
     source = sources,
 )
 
-if not env['dri']:
-    # Only install this libGL.so if DRI not enabled
-    env.InstallSharedLibrary(libgl, version=(1, 5))
+if True:
+    # XXX: Only install this libGL.so if DRI not enabled
+    libgl = env.InstallSharedLibrary(libgl, version=(1, 5))
+
+env.Alias('libgl-xlib', libgl)
diff --git a/src/gallium/targets/xorg-vmwgfx/SConscript b/src/gallium/targets/xorg-vmwgfx/SConscript
index 43b2c74f9c..f9172aae68 100644
--- a/src/gallium/targets/xorg-vmwgfx/SConscript
+++ b/src/gallium/targets/xorg-vmwgfx/SConscript
@@ -2,62 +2,62 @@ import os.path
 
 Import('*')
 
-if not 'svga' in env['drivers']:
-    print 'warning: svga pipe driver not built skipping vmwgfx_drv.so'
-    Return()
-
-if env['platform'] == 'linux':
-
-	env = env.Clone()
-
-	env.ParseConfig('pkg-config --cflags --libs libdrm xorg-server')
-
-	env.Prepend(CPPPATH = [
-		'#/include',
-		'#/src/gallium',
-		'#/src/mesa',
-		'#/src/gallium/drivers/svga',
-		'#/src/gallium/drivers/svga/include',
-	])
-
-	env.Append(CPPDEFINES = [
-	])
-
-	if env['gcc']:
-		env.Append(CPPDEFINES = [
-			'HAVE_STDINT_H',
-			'HAVE_SYS_TYPES_H',
-		])
-
-	env.Append(CFLAGS = [
-		'-std=gnu99',
-		'-D_FILE_OFFSET_BITS=64',
-	])
-
-	env.Prepend(LIBPATH = [
-	])
-
-	env.Prepend(LIBS = [
-		trace,
-		rbug,
-		st_xorg,
-		svgadrm,
-		svga,
-                gallium,
-	])
-
-	sources = [
-		'vmw_ioctl.c',
-		'vmw_ctrl.c',
-		'vmw_screen.c',
-		'vmw_video.c',
-		'vmw_xorg.c',
-	]
-
-	# TODO: write a wrapper function http://www.scons.org/wiki/WrapperFunctions
-	env.LoadableModule(
-		target ='vmwgfx_drv.so',
-		source = sources,
-		LIBS = env['LIBS'],
-		SHLIBPREFIX = '',
-	)
+env = env.Clone()
+
+env.ParseConfig('pkg-config --cflags --libs libdrm xorg-server')
+
+if env['kms']:
+    env.ParseConfig('pkg-config --cflags --libs libkms')
+
+env.Prepend(CPPPATH = [
+    '#/include',
+    '#/src/gallium',
+    '#/src/mesa',
+    '#/src/gallium/drivers/svga',
+    '#/src/gallium/drivers/svga/include',
+])
+
+env.Append(CPPDEFINES = [
+])
+
+if env['gcc']:
+    env.Append(CPPDEFINES = [
+        'HAVE_STDINT_H',
+        'HAVE_SYS_TYPES_H',
+    ])
+
+env.Append(CFLAGS = [
+    '-std=gnu99',
+    '-D_FILE_OFFSET_BITS=64',
+])
+
+env.Prepend(LIBPATH = [
+])
+
+env.Prepend(LIBS = [
+    trace,
+    rbug,
+    st_xorg,
+    svgadrm,
+    svga,
+            gallium,
+])
+
+sources = [
+    'vmw_ioctl.c',
+    'vmw_ctrl.c',
+    'vmw_screen.c',
+    'vmw_target.c',
+    'vmw_video.c',
+    'vmw_xorg.c',
+]
+
+# TODO: write a wrapper function http://www.scons.org/wiki/WrapperFunctions
+module = env.LoadableModule(
+    target ='vmwgfx_drv.so',
+    source = sources,
+    LIBS = env['LIBS'],
+    SHLIBPREFIX = '',
+)
+
+env.Alias('xorg-vmwgfx', module)
diff --git a/src/gallium/tests/graw/SConscript b/src/gallium/tests/graw/SConscript
index ffde61965b..3341b88498 100644
--- a/src/gallium/tests/graw/SConscript
+++ b/src/gallium/tests/graw/SConscript
@@ -1,20 +1,17 @@
 Import('*')
 
-try:
-    graw
-except NameError:
-    print 'warning: graw library not avaiable: skipping build of graw test'
-    Return()
-
 env = env.Clone()
 
+env.Prepend(LIBS = [gallium])
+
 env.Prepend(LIBPATH = [graw.dir])
-env.Prepend(LIBS = ['graw'] + gallium)
+env.Prepend(LIBS = ['graw'])
+    
 
-if platform in ('freebsd8', 'sunos5'):
+if env['platform'] in ('freebsd8', 'sunos5'):
     env.Append(LIBS = ['m'])
 
-if platform == 'freebsd8':
+if env['platform'] == 'freebsd8':
     env.Append(LIBS = ['pthread'])
 
 progs = [
@@ -29,9 +26,10 @@ progs = [
     'tri-gs',
 ]
 
-for prog in progs:
-    env.Program(
-        target = prog,
-        source = prog + '.c',
+for name in progs:
+    program = env.Program(
+        target = name,
+        source = name + '.c',
     )
-
+    #env.Depends(program, graw)
+    env.Alias('graw-progs', program)
diff --git a/src/gallium/tests/unit/SConscript b/src/gallium/tests/unit/SConscript
index 359759e22b..655e8a9b41 100644
--- a/src/gallium/tests/unit/SConscript
+++ b/src/gallium/tests/unit/SConscript
@@ -4,10 +4,10 @@ env = env.Clone()
 
 env.Prepend(LIBS = [gallium])
 
-if platform in ('freebsd8', 'sunos5'):
+if env['platform'] in ('freebsd8', 'sunos5'):
     env.Append(LIBS = ['m'])
 
-if platform == 'freebsd8':
+if env['platform'] == 'freebsd8':
     env.Append(LIBS = ['pthread'])
 
 progs = [
diff --git a/src/gallium/winsys/SConscript b/src/gallium/winsys/SConscript
index 65b12287df..d74f8a2e98 100644
--- a/src/gallium/winsys/SConscript
+++ b/src/gallium/winsys/SConscript
@@ -2,45 +2,38 @@ Import('*')
 
 
 SConscript([
-	'sw/wrapper/SConscript',
+    'sw/wrapper/SConscript',
 ])
 
-if 'xlib' in env['winsys']:
-	SConscript([
-		'sw/xlib/SConscript',
-	])
+SConscript([
+    'sw/xlib/SConscript',
+])
 
-if 'gdi' in env['winsys']:
-	SConscript([
-		'sw/gdi/SConscript',
-	])
+SConscript([
+    'sw/gdi/SConscript',
+])
 
 if env['dri']:
-	SConscript([
-		'sw/dri/SConscript',
-	])
-
-	if 'vmware' in env['winsys']:
-		SConscript([
-			'svga/drm/SConscript',
-		])
-
-	if 'i915' in env['winsys']:
-		SConscript([
-			'i915/drm/SConscript',
-		])
-
-	if 'i965' in env['winsys']:
-		SConscript([
-			'i965/drm/SConscript',
-		])
-
-	if 'radeon' in env['winsys']:
-		SConscript([
-			'radeon/drm/SConscript',
-		])
-
-	if 'r600' in env['winsys']:
-		SConscript([
-			'r600/drm/SConscript',
-		])
+    SConscript([
+        'sw/dri/SConscript',
+    ])
+
+    SConscript([
+        'svga/drm/SConscript',
+    ])
+
+    SConscript([
+        'i915/drm/SConscript',
+    ])
+
+    SConscript([
+        'i965/drm/SConscript',
+    ])
+
+    SConscript([
+        'radeon/drm/SConscript',
+    ])
+
+    SConscript([
+        'r600/drm/SConscript',
+    ])
diff --git a/src/gallium/winsys/r600/drm/r600_bo.c b/src/gallium/winsys/r600/drm/r600_bo.c
index 7d54ff18fc..9b9aec5f64 100644
--- a/src/gallium/winsys/r600/drm/r600_bo.c
+++ b/src/gallium/winsys/r600/drm/r600_bo.c
@@ -29,23 +29,37 @@
 #include "radeon_drm.h"
 #include "r600_priv.h"
 #include "r600d.h"
+#include "radeon_drm.h"
 
 struct r600_bo *r600_bo(struct radeon *radeon,
-				  unsigned size, unsigned alignment, unsigned usage)
+			unsigned size, unsigned alignment,
+			unsigned binding, unsigned usage)
 {
 	struct r600_bo *ws_bo = calloc(1, sizeof(struct r600_bo));
 	struct pb_desc desc;
 	struct pb_manager *man;
 
 	desc.alignment = alignment;
-	desc.usage = usage;
+	desc.usage = (PB_USAGE_CPU_READ_WRITE | PB_USAGE_GPU_READ_WRITE);
 	ws_bo->size = size;
 
-	if (usage & (PIPE_BIND_CONSTANT_BUFFER | PIPE_BIND_VERTEX_BUFFER | PIPE_BIND_INDEX_BUFFER))
+	if (binding & (PIPE_BIND_CONSTANT_BUFFER | PIPE_BIND_VERTEX_BUFFER | PIPE_BIND_INDEX_BUFFER))
 		man = radeon->cman;
 	else
 		man = radeon->kman;
 
+	/* Staging resources particpate in transfers and blits only
+	 * and are used for uploads and downloads from regular
+	 * resources.  We generate them internally for some transfers.
+	 */
+	if (usage == PIPE_USAGE_STAGING)
+                ws_bo->domains = RADEON_GEM_DOMAIN_CPU | RADEON_GEM_DOMAIN_GTT;
+        else
+                ws_bo->domains = (RADEON_GEM_DOMAIN_CPU |
+                                  RADEON_GEM_DOMAIN_GTT |
+                                  RADEON_GEM_DOMAIN_VRAM);
+
+
 	ws_bo->pb = man->create_buffer(man, size, &desc);
 	if (ws_bo->pb == NULL) {
 		free(ws_bo);
@@ -69,6 +83,10 @@ struct r600_bo *r600_bo_handle(struct radeon *radeon,
 	}
 	bo = radeon_bo_pb_get_bo(ws_bo->pb);
 	ws_bo->size = bo->size;
+	ws_bo->domains = (RADEON_GEM_DOMAIN_CPU |
+			  RADEON_GEM_DOMAIN_GTT |
+			  RADEON_GEM_DOMAIN_VRAM);
+
 	pipe_reference_init(&ws_bo->reference, 1);
 
 	radeon_bo_get_tiling_flags(radeon, bo, &ws_bo->tiling_flags,
diff --git a/src/gallium/winsys/r600/drm/r600_hw_context.c b/src/gallium/winsys/r600/drm/r600_hw_context.c
index 2521ff9647..bcbe71e126 100644
--- a/src/gallium/winsys/r600/drm/r600_hw_context.c
+++ b/src/gallium/winsys/r600/drm/r600_hw_context.c
@@ -44,7 +44,7 @@
 int r600_context_init_fence(struct r600_context *ctx)
 {
 	ctx->fence = 1;
-	ctx->fence_bo = r600_bo(ctx->radeon, 4096, 0, 0);
+	ctx->fence_bo = r600_bo(ctx->radeon, 4096, 0, 0, 0);
 	if (ctx->fence_bo == NULL) {
 		return -ENOMEM;
 	}
@@ -384,6 +384,7 @@ static const struct r600_reg r600_context_reg_list[] = {
 	{PKT3_SET_CONTEXT_REG, R600_CONTEXT_REG_OFFSET, R_028A0C_PA_SC_LINE_STIPPLE, 0, 0, 0},
 	{PKT3_SET_CONTEXT_REG, R600_CONTEXT_REG_OFFSET, R_028A48_PA_SC_MPASS_PS_CNTL, 0, 0, 0},
 	{PKT3_SET_CONTEXT_REG, R600_CONTEXT_REG_OFFSET, R_028C00_PA_SC_LINE_CNTL, 0, 0, 0},
+	{PKT3_SET_CONTEXT_REG, R600_CONTEXT_REG_OFFSET, R_028C08_PA_SU_VTX_CNTL, 0, 0, 0},
 	{PKT3_SET_CONTEXT_REG, R600_CONTEXT_REG_OFFSET, R_028C0C_PA_CL_GB_VERT_CLIP_ADJ, 0, 0, 0},
 	{PKT3_SET_CONTEXT_REG, R600_CONTEXT_REG_OFFSET, R_028C10_PA_CL_GB_VERT_DISC_ADJ, 0, 0, 0},
 	{PKT3_SET_CONTEXT_REG, R600_CONTEXT_REG_OFFSET, R_028C14_PA_CL_GB_HORZ_CLIP_ADJ, 0, 0, 0},
@@ -611,7 +612,9 @@ void r600_context_fini(struct r600_context *ctx)
 		}
 		free(ctx->range[i].blocks);
 	}
+	free(ctx->blocks);
 	free(ctx->reloc);
+	free(ctx->bo);
 	free(ctx->pm4);
 	if (ctx->fence_bo) {
 		r600_bo_reference(ctx->radeon, &ctx->fence_bo, NULL);
@@ -785,8 +788,8 @@ void r600_context_bo_reloc(struct r600_context *ctx, u32 *pm4, struct r600_bo *r
 	bo->reloc = &ctx->reloc[ctx->creloc];
 	bo->reloc_id = ctx->creloc * sizeof(struct r600_reloc) / 4;
 	ctx->reloc[ctx->creloc].handle = bo->handle;
-	ctx->reloc[ctx->creloc].read_domain = RADEON_GEM_DOMAIN_GTT | RADEON_GEM_DOMAIN_VRAM;
-	ctx->reloc[ctx->creloc].write_domain = RADEON_GEM_DOMAIN_GTT | RADEON_GEM_DOMAIN_VRAM;
+	ctx->reloc[ctx->creloc].read_domain = rbo->domains & (RADEON_GEM_DOMAIN_GTT | RADEON_GEM_DOMAIN_VRAM);
+	ctx->reloc[ctx->creloc].write_domain = rbo->domains & (RADEON_GEM_DOMAIN_GTT | RADEON_GEM_DOMAIN_VRAM);
 	ctx->reloc[ctx->creloc].flags = 0;
 	radeon_bo_reference(ctx->radeon, &ctx->bo[ctx->creloc], bo);
 	ctx->creloc++;
@@ -1304,7 +1307,12 @@ struct r600_query *r600_context_query_create(struct r600_context *ctx, unsigned
 	query->type = query_type;
 	query->buffer_size = 4096;
 
-	query->buffer = r600_bo(ctx->radeon, query->buffer_size, 1, 0);
+	/* As of GL4, query buffers are normally read by the CPU after
+	 * being written by the gpu, hence staging is probably a good
+	 * usage pattern.
+	 */
+	query->buffer = r600_bo(ctx->radeon, query->buffer_size, 1, 0,
+				PIPE_USAGE_STAGING);
 	if (!query->buffer) {
 		free(query);
 		return NULL;
diff --git a/src/gallium/winsys/r600/drm/r600_priv.h b/src/gallium/winsys/r600/drm/r600_priv.h
index b5bd7bd92c..97c582397a 100644
--- a/src/gallium/winsys/r600/drm/r600_priv.h
+++ b/src/gallium/winsys/r600/drm/r600_priv.h
@@ -62,7 +62,7 @@ struct radeon_bo {
 	unsigned			handle;
 	unsigned			size;
 	unsigned			alignment;
-	unsigned			map_count;
+	int				map_count;
 	void				*data;
 	struct list_head		fencedlist;
 	unsigned			fence;
@@ -79,6 +79,7 @@ struct r600_bo {
 	unsigned			size;
 	unsigned			tiling_flags;
 	unsigned                        kernel_pitch;
+	unsigned			domains;
 };
 
 
@@ -90,7 +91,7 @@ struct radeon *radeon_decref(struct radeon *radeon);
 
 /* radeon_bo.c */
 struct radeon_bo *radeon_bo(struct radeon *radeon, unsigned handle,
-			    unsigned size, unsigned alignment, void *ptr);
+			    unsigned size, unsigned alignment);
 void radeon_bo_reference(struct radeon *radeon, struct radeon_bo **dst,
 			 struct radeon_bo *src);
 int radeon_bo_wait(struct radeon *radeon, struct radeon_bo *bo);
diff --git a/src/gallium/winsys/r600/drm/r600d.h b/src/gallium/winsys/r600/drm/r600d.h
index d91f7737af..5ca7456e90 100644
--- a/src/gallium/winsys/r600/drm/r600d.h
+++ b/src/gallium/winsys/r600/drm/r600d.h
@@ -795,6 +795,7 @@
 #define R_028A48_PA_SC_MPASS_PS_CNTL                 0x028A48
 #define R_028C00_PA_SC_LINE_CNTL                     0x028C00
 #define R_028C04_PA_SC_AA_CONFIG                     0x028C04
+#define R_028C08_PA_SU_VTX_CNTL                      0x028C08
 #define R_028C1C_PA_SC_AA_SAMPLE_LOCS_MCTX           0x028C1C
 #define R_028C48_PA_SC_AA_MASK                       0x028C48
 #define R_028810_PA_CL_CLIP_CNTL                     0x028810
diff --git a/src/gallium/winsys/r600/drm/radeon_bo.c b/src/gallium/winsys/r600/drm/radeon_bo.c
index 2d08686fc4..dcee00de86 100644
--- a/src/gallium/winsys/r600/drm/radeon_bo.c
+++ b/src/gallium/winsys/r600/drm/radeon_bo.c
@@ -69,7 +69,7 @@ static void radeon_bo_fixed_unmap(struct radeon *radeon, struct radeon_bo *bo)
 }
 
 struct radeon_bo *radeon_bo(struct radeon *radeon, unsigned handle,
-			unsigned size, unsigned alignment, void *ptr)
+			unsigned size, unsigned alignment)
 {
 	struct radeon_bo *bo;
 	int r;
@@ -121,9 +121,6 @@ struct radeon_bo *radeon_bo(struct radeon *radeon, unsigned handle,
 		radeon_bo_reference(radeon, &bo, NULL);
 		return bo;
 	}
-	if (ptr) {
-		memcpy(bo->data, ptr, size);
-	}
 	return bo;
 }
 
diff --git a/src/gallium/winsys/r600/drm/radeon_bo_pb.c b/src/gallium/winsys/r600/drm/radeon_bo_pb.c
index a3452027f2..312552f075 100644
--- a/src/gallium/winsys/r600/drm/radeon_bo_pb.c
+++ b/src/gallium/winsys/r600/drm/radeon_bo_pb.c
@@ -63,11 +63,13 @@ static void radeon_bo_pb_destroy(struct pb_buffer *_buf)
 {
 	struct radeon_bo_pb *buf = radeon_bo_pb(_buf);
 
-	LIST_DEL(&buf->maplist);
-
-	if (buf->bo->data != NULL) {
+	/* If this buffer is on the list of buffers to unmap,
+	 * do the unmapping now.
+	 */
+	if (!LIST_IS_EMPTY(&buf->maplist))
 		radeon_bo_unmap(buf->mgr->radeon, buf->bo);
-	}
+
+	LIST_DEL(&buf->maplist);
 	radeon_bo_reference(buf->mgr->radeon, &buf->bo, NULL);
 	FREE(buf);
 }
@@ -80,7 +82,7 @@ radeon_bo_pb_map_internal(struct pb_buffer *_buf,
 	struct pipe_context *pctx = ctx;
 
 	if (flags & PB_USAGE_UNSYNCHRONIZED) {
-		if (!buf->bo->data && radeon_bo_map(buf->mgr->radeon, buf->bo)) {
+		if (radeon_bo_map(buf->mgr->radeon, buf->bo)) {
 			return NULL;
 		}
 		LIST_DELINIT(&buf->maplist);
@@ -106,18 +108,12 @@ radeon_bo_pb_map_internal(struct pb_buffer *_buf,
 		goto out;
 	}
 
-	if (buf->bo->data != NULL) {
-		if (radeon_bo_wait(buf->mgr->radeon, buf->bo)) {
-			return NULL;
-		}
-	} else {
-		if (radeon_bo_map(buf->mgr->radeon, buf->bo)) {
-			return NULL;
-		}
-		if (radeon_bo_wait(buf->mgr->radeon, buf->bo)) {
-			radeon_bo_unmap(buf->mgr->radeon, buf->bo);
-			return NULL;
-		}
+	if (radeon_bo_map(buf->mgr->radeon, buf->bo)) {
+		return NULL;
+	}
+	if (radeon_bo_wait(buf->mgr->radeon, buf->bo)) {
+		radeon_bo_unmap(buf->mgr->radeon, buf->bo);
+		return NULL;
 	}
 out:
 	LIST_DELINIT(&buf->maplist);
@@ -172,7 +168,7 @@ radeon_bo_pb_create_buffer_from_handle(struct pb_manager *_mgr,
 	struct radeon_bo_pb *bo;
 	struct radeon_bo *hw_bo;
 
-	hw_bo = radeon_bo(radeon, handle, 0, 0, NULL);
+	hw_bo = radeon_bo(radeon, handle, 0, 0);
 	if (hw_bo == NULL)
 		return NULL;
 
@@ -217,8 +213,7 @@ radeon_bo_pb_create_buffer(struct pb_manager *_mgr,
 
 	LIST_INITHEAD(&bo->maplist);
 
-	bo->bo = radeon_bo(radeon, 0, size,
-			   desc->alignment, NULL);
+	bo->bo = radeon_bo(radeon, 0, size, desc->alignment);
 	if (bo->bo == NULL)
 		goto error2;
 	return &bo->b;
diff --git a/src/glsl/builtin_function.cpp b/src/glsl/builtin_function.cpp
index f0206431fb..942f2335eb 100644
--- a/src/glsl/builtin_function.cpp
+++ b/src/glsl/builtin_function.cpp
@@ -265,14 +265,14 @@ static const char builtin_atan[] =
    "((function atan\n"
    "   (signature float\n"
    "     (parameters\n"
-   "       (declare (in) float x))\n"
+   "       (declare (in) float y_over_x))\n"
    "     ((return (call asin ((expression float *\n"
-   "			   (var_ref x)\n"
+   "			   (var_ref y_over_x)\n"
    "			   (expression float rsq\n"
    "			    (expression float +\n"
    "			     (expression float *\n"
-   "			      (var_ref x)\n"
-   "			      (var_ref x))\n"
+   "			      (var_ref y_over_x)\n"
+   "			      (var_ref y_over_x))\n"
    "			     (constant float (1.0))))))))))\n"
    "\n"
    "   (signature vec2\n"
@@ -318,30 +318,15 @@ static const char builtin_atan[] =
    "    )\n"
    "    (\n"
    "      (declare () float r)\n"
-   "      (declare ( ) float abs_retval)\n"
-   "      (assign (constant bool (1)) (x) (var_ref abs_retval)  (call abs ((var_ref x) ))\n"
-   ")\n"
-   "      (if (expression bool > (var_ref abs_retval) (constant float (0.000100)) ) (\n"
-   "        (declare ( ) float atan_retval)\n"
-   "        (assign (constant bool (1)) (x) (var_ref atan_retval)  (call atan ((expression float / (var_ref y) (var_ref x) ) ))\n"
-   ")\n"
-   "        (assign (constant bool (1)) (x) (var_ref r)  (var_ref atan_retval) )\n"
+   "      (if (expression bool > (expression float abs (var_ref x)) (constant float (0.000100))) (\n"
+   "        (assign (constant bool (1)) (x) (var_ref r) (call atan ((expression float / (var_ref y) (var_ref x)))))\n"
    "        (if (expression bool < (var_ref x) (constant float (0.000000)) ) (\n"
-   "          (if (expression bool >= (var_ref y) (constant float (0.000000)) ) (\n"
-   "            (declare ( ) float assignment_tmp)\n"
-   "            (assign (constant bool (1)) (x) (var_ref assignment_tmp)  (expression float + (var_ref r) (constant float (3.141593)) ) )\n"
-   "            (assign (constant bool (1)) (x) (var_ref r)  (var_ref assignment_tmp) )\n"
-   "          )\n"
-   "          (\n"
-   "            (declare ( ) float assignment_tmp)\n"
-   "            (assign (constant bool (1)) (x) (var_ref assignment_tmp)  (expression float - (var_ref r) (constant float (3.141593)) ) )\n"
-   "            (assign (constant bool (1)) (x) (var_ref r)  (var_ref assignment_tmp) )\n"
-   "          ))\n"
-   "\n"
+   "          (if (expression bool >= (var_ref y) (constant float (0.000000)) )\n"
+   "              ((assign (constant bool (1)) (x) (var_ref r) (expression float + (var_ref r) (constant float (3.141593)))))\n"
+   "              ((assign (constant bool (1)) (x) (var_ref r) (expression float - (var_ref r) (constant float (3.141593))))))\n"
    "        )\n"
    "        (\n"
    "        ))\n"
-   "\n"
    "      )\n"
    "      (\n"
    "        (declare () float sgn)\n"
diff --git a/src/glsl/builtins/ir/atan b/src/glsl/builtins/ir/atan
index 6dc99d74d3..3f97e0d46f 100644
--- a/src/glsl/builtins/ir/atan
+++ b/src/glsl/builtins/ir/atan
@@ -1,14 +1,14 @@
 ((function atan
    (signature float
      (parameters
-       (declare (in) float x))
+       (declare (in) float y_over_x))
      ((return (call asin ((expression float *
-			   (var_ref x)
+			   (var_ref y_over_x)
 			   (expression float rsq
 			    (expression float +
 			     (expression float *
-			      (var_ref x)
-			      (var_ref x))
+			      (var_ref y_over_x)
+			      (var_ref y_over_x))
 			     (constant float (1.0))))))))))
 
    (signature vec2
@@ -54,30 +54,15 @@
     )
     (
       (declare () float r)
-      (declare ( ) float abs_retval)
-      (assign (constant bool (1)) (x) (var_ref abs_retval)  (call abs ((var_ref x) ))
-)
-      (if (expression bool > (var_ref abs_retval) (constant float (0.000100)) ) (
-        (declare ( ) float atan_retval)
-        (assign (constant bool (1)) (x) (var_ref atan_retval)  (call atan ((expression float / (var_ref y) (var_ref x) ) ))
-)
-        (assign (constant bool (1)) (x) (var_ref r)  (var_ref atan_retval) )
+      (if (expression bool > (expression float abs (var_ref x)) (constant float (0.000100))) (
+        (assign (constant bool (1)) (x) (var_ref r) (call atan ((expression float / (var_ref y) (var_ref x)))))
         (if (expression bool < (var_ref x) (constant float (0.000000)) ) (
-          (if (expression bool >= (var_ref y) (constant float (0.000000)) ) (
-            (declare ( ) float assignment_tmp)
-            (assign (constant bool (1)) (x) (var_ref assignment_tmp)  (expression float + (var_ref r) (constant float (3.141593)) ) )
-            (assign (constant bool (1)) (x) (var_ref r)  (var_ref assignment_tmp) )
-          )
-          (
-            (declare ( ) float assignment_tmp)
-            (assign (constant bool (1)) (x) (var_ref assignment_tmp)  (expression float - (var_ref r) (constant float (3.141593)) ) )
-            (assign (constant bool (1)) (x) (var_ref r)  (var_ref assignment_tmp) )
-          ))
-
+          (if (expression bool >= (var_ref y) (constant float (0.000000)) )
+              ((assign (constant bool (1)) (x) (var_ref r) (expression float + (var_ref r) (constant float (3.141593)))))
+              ((assign (constant bool (1)) (x) (var_ref r) (expression float - (var_ref r) (constant float (3.141593))))))
         )
         (
         ))
-
       )
       (
         (declare () float sgn)
diff --git a/src/glsl/ir.h b/src/glsl/ir.h
index 06198e4f3f..80e0f67d6d 100644
--- a/src/glsl/ir.h
+++ b/src/glsl/ir.h
@@ -37,10 +37,6 @@ extern "C" {
 #include "ir_visitor.h"
 #include "ir_hierarchical_visitor.h"
 
-#ifndef ARRAY_SIZE
-#define ARRAY_SIZE(x) (sizeof(x) / sizeof(x[0]))
-#endif
-
 /**
  * \defgroup IR Intermediate representation nodes
  *
diff --git a/src/glsl/ir_reader.cpp b/src/glsl/ir_reader.cpp
index a9cbf8ea94..7a22a945ec 100644
--- a/src/glsl/ir_reader.cpp
+++ b/src/glsl/ir_reader.cpp
@@ -64,9 +64,12 @@ static ir_texture *read_texture(_mesa_glsl_parse_state *, s_list *);
 
 static ir_dereference *read_dereference(_mesa_glsl_parse_state *,
 				        s_expression *);
-static ir_dereference *read_var_ref(_mesa_glsl_parse_state *, s_list *);
-static ir_dereference *read_array_ref(_mesa_glsl_parse_state *, s_list *);
-static ir_dereference *read_record_ref(_mesa_glsl_parse_state *, s_list *);
+static ir_dereference_variable *
+read_var_ref(_mesa_glsl_parse_state *, s_list *);
+static ir_dereference_array *
+read_array_ref(_mesa_glsl_parse_state *, s_list *);
+static ir_dereference_record *
+read_record_ref(_mesa_glsl_parse_state *, s_list *);
 
 void
 _mesa_glsl_read_ir(_mesa_glsl_parse_state *state, exec_list *instructions,
@@ -260,7 +263,7 @@ read_function_sig(_mesa_glsl_parse_state *st, ir_function *f, s_list *list,
       return;
 
    s_list *paramlist = SX_AS_LIST(type_expr->next);
-   s_list *body_list = SX_AS_LIST(paramlist->next);
+   s_list *body_list = SX_AS_LIST(type_expr->next->next);
    if (paramlist == NULL || body_list == NULL) {
       ir_read_error(st, list, "Expected (signature <type> (parameters ...) "
 			      "(<instruction> ...))");
@@ -674,7 +677,7 @@ read_call(_mesa_glsl_parse_state *st, s_list *list)
    }
 
    s_symbol *name = SX_AS_SYMBOL(list->subexpressions.head->next);
-   s_list *params = SX_AS_LIST(name->next);
+   s_list *params = SX_AS_LIST(list->subexpressions.head->next->next);
    if (name == NULL || params == NULL) {
       ir_read_error(st, list, "expected (call <name> (<param> ...))");
       return NULL;
@@ -796,12 +799,6 @@ read_swizzle(_mesa_glsl_parse_state *st, s_list *list)
    }
 
    s_expression *sub = (s_expression*) swiz->next;
-   if (sub == NULL) {
-      ir_read_error(st, list, "expected rvalue: (swizzle %s <rvalue>)",
-		    swiz->value());
-      return NULL;
-   }
-
    ir_rvalue *rvalue = read_rvalue(st, sub);
    if (rvalue == NULL)
       return NULL;
@@ -930,7 +927,7 @@ read_dereference(_mesa_glsl_parse_state *st, s_expression *expr)
    return NULL;
 }
 
-static ir_dereference *
+static ir_dereference_variable *
 read_var_ref(_mesa_glsl_parse_state *st, s_list *list)
 {
    void *ctx = st;
@@ -953,7 +950,7 @@ read_var_ref(_mesa_glsl_parse_state *st, s_list *list)
    return new(ctx) ir_dereference_variable(var);
 }
 
-static ir_dereference *
+static ir_dereference_array *
 read_array_ref(_mesa_glsl_parse_state *st, s_list *list)
 {
    void *ctx = st;
@@ -974,7 +971,7 @@ read_array_ref(_mesa_glsl_parse_state *st, s_list *list)
    return new(ctx) ir_dereference_array(subject, idx);
 }
 
-static ir_dereference *
+static ir_dereference_record *
 read_record_ref(_mesa_glsl_parse_state *st, s_list *list)
 {
    void *ctx = st;
@@ -1053,8 +1050,8 @@ read_texture(_mesa_glsl_parse_state *st, s_list *list)
       return NULL;
    }
    s_int *offset_x = SX_AS_INT(offset_list->subexpressions.head);
-   s_int *offset_y = SX_AS_INT(offset_x->next);
-   s_int *offset_z = SX_AS_INT(offset_y->next);
+   s_int *offset_y = SX_AS_INT(offset_list->subexpressions.head->next);
+   s_int *offset_z = SX_AS_INT(offset_list->subexpressions.head->next->next);
    if (offset_x == NULL || offset_y == NULL || offset_z == NULL) {
       ir_read_error(st, offset_list, "expected (<int> <int> <int>)");
       return NULL;
diff --git a/src/glsl/s_expression.cpp b/src/glsl/s_expression.cpp
index 4458c48d6b..4c8829fea9 100644
--- a/src/glsl/s_expression.cpp
+++ b/src/glsl/s_expression.cpp
@@ -133,7 +133,8 @@ void s_list::print()
    foreach_iter(exec_list_iterator, it, this->subexpressions) {
       s_expression *expr = (s_expression*) it.get();
       expr->print();
-      printf(" ");
+      if (!expr->next->is_tail_sentinel())
+	 printf(" ");
    }
    printf(")");
 }
diff --git a/src/glut/glx/SConscript b/src/glut/glx/SConscript
index 29d2cc0149..9363b5ca5c 100644
--- a/src/glut/glx/SConscript
+++ b/src/glut/glx/SConscript
@@ -78,7 +78,6 @@ if env['platform'] == 'windows':
         'glut.def',
     ]
 else:
-    env.Tool('x11')
     env.PrependUnique(LIBS = [
         'GLU',
         'GL',
diff --git a/src/mapi/glapi/SConscript b/src/mapi/glapi/SConscript
index 40db237fcb..77a5f9bd2e 100644
--- a/src/mapi/glapi/SConscript
+++ b/src/mapi/glapi/SConscript
@@ -6,77 +6,78 @@ Import('*')
 
 if env['platform'] != 'winddk':
 
-	env = env.Clone()
-	
-	env.Append(CPPDEFINES = [
-		'MAPI_GLAPI_CURRENT',
-	])
+    env = env.Clone()
+    
+    env.Append(CPPDEFINES = [
+        'MAPI_GLAPI_CURRENT',
+    ])
 
-	if env['platform'] == 'windows':
-		env.Append(CPPDEFINES = [
-			'_GDI32_', # prevent gl* being declared __declspec(dllimport) in MS headers
-			'BUILD_GL32', # declare gl* as __declspec(dllexport) in Mesa headers
-			'WIN32_THREADS', # use Win32 thread API
-		])
+    if env['platform'] == 'windows':
+        env.Append(CPPDEFINES = [
+            '_GDI32_', # prevent gl* being declared __declspec(dllimport) in MS headers
+            'BUILD_GL32', # declare gl* as __declspec(dllexport) in Mesa headers
+            'WIN32_THREADS', # use Win32 thread API
+        ])
 
-	env.Append(CPPPATH = [
-		'#/src/mapi',
-		'#/src/mesa',
-	])
-	
-	glapi_sources = [
-		'glapi_dispatch.c',
-		'glapi_entrypoint.c',
-		'glapi_getproc.c',
-		'glapi_nop.c',
-		'glthread.c',
-	]
-	
-	mapi_sources = [
-		'u_current.c',
-		'u_execmem.c',
-		'u_thread.c',
-	]
-	for s in mapi_sources:
-		o = env.SharedObject(s[:-2], '../mapi/' + s)
-		glapi_sources.append(o)
+    env.Append(CPPPATH = [
+        '#/src/mapi',
+        '#/src/mesa',
+    ])
+    
+    glapi_sources = [
+        'glapi_dispatch.c',
+        'glapi_entrypoint.c',
+        'glapi_getproc.c',
+        'glapi_nop.c',
+        'glthread.c',
+    ]
+    
+    mapi_sources = [
+        'u_current.c',
+        'u_execmem.c',
+        'u_thread.c',
+    ]
+    for s in mapi_sources:
+        o = env.SharedObject(s[:-2], '../mapi/' + s)
+        glapi_sources.append(o)
 
-	#
-	# Assembly sources
-	#
-	if gcc and env['machine'] == 'x86':
-		env.Append(CPPDEFINES = [
-			'USE_X86_ASM', 
-			'USE_MMX_ASM',
-			'USE_3DNOW_ASM',
-			'USE_SSE_ASM',
-		])
-		glapi_sources += [
-			'glapi_x86.S',
-		]
-	elif gcc and env['machine'] == 'x86_64':
-		env.Append(CPPDEFINES = [
-			'USE_X86_64_ASM', 
-		])
-		glapi_sources += [
-			'glapi_x86-64.S'
-		]
-	elif gcc and env['machine'] == 'ppc':
-		env.Append(CPPDEFINES = [
-			'USE_PPC_ASM', 
-			'USE_VMX_ASM', 
-		])
-		glapi_sources += [
-		]
-	elif gcc and env['machine'] == 'sparc':
-		glapi_sources += [
-			'glapi_sparc.S'
-		]
-	else:
-		pass
-	
-	glapi = env.ConvenienceLibrary(
-		target = 'glapi',
-		source = glapi_sources,
-	)
-	Export('glapi')
+    #
+    # Assembly sources
+    #
+    if env['gcc'] and env['platform'] != 'windows':
+        if env['machine'] == 'x86':
+            env.Append(CPPDEFINES = [
+                'USE_X86_ASM', 
+                'USE_MMX_ASM',
+                'USE_3DNOW_ASM',
+                'USE_SSE_ASM',
+            ])
+            glapi_sources += [
+                'glapi_x86.S',
+            ]
+        elif env['machine'] == 'x86_64':
+            env.Append(CPPDEFINES = [
+                'USE_X86_64_ASM', 
+            ])
+            glapi_sources += [
+                'glapi_x86-64.S'
+            ]
+        elif env['machine'] == 'ppc':
+            env.Append(CPPDEFINES = [
+                'USE_PPC_ASM', 
+                'USE_VMX_ASM', 
+            ])
+            glapi_sources += [
+            ]
+        elif env['machine'] == 'sparc':
+            glapi_sources += [
+                'glapi_sparc.S'
+            ]
+        else:
+            pass
+    
+    glapi = env.ConvenienceLibrary(
+        target = 'glapi',
+        source = glapi_sources,
+    )
+    Export('glapi')
diff --git a/src/mapi/vgapi/.gitignore b/src/mapi/vgapi/.gitignore
new file mode 100644
index 0000000000..5becb99386
--- /dev/null
+++ b/src/mapi/vgapi/.gitignore
@@ -0,0 +1 @@
+vgapi_tmp.h
diff --git a/src/mesa/SConscript b/src/mesa/SConscript
index e54737aed1..5fe5d39fb1 100644
--- a/src/mesa/SConscript
+++ b/src/mesa/SConscript
@@ -4,322 +4,324 @@
 
 Import('*')
 
-if env['platform'] != 'winddk':
+env = env.Clone()
 
-	env = env.Clone()
-	
-	env.Append(CPPPATH = [
-		'#/src/mapi',
-		'#/src/glsl',
-		'#/src/mesa',
-	])
-	
-	if env['platform'] == 'windows':
-		env.Append(CPPDEFINES = [
-			'_GDI32_', # prevent gl* being declared __declspec(dllimport) in MS headers
-			'BUILD_GL32', # declare gl* as __declspec(dllexport) in Mesa headers 
-			'WIN32_THREADS', # use Win32 thread API
-		])	 
-        env.Prepend(CPPPATH = ['#src/talloc'])
+env.Append(CPPPATH = [
+    '#/src/mapi',
+    '#/src/glsl',
+    '#/src/mesa',
+])
 
-	#
-	# Source files
-	#
-	
-	main_sources = [
-		'main/api_arrayelt.c',
-		'main/api_exec.c',
-		'main/api_loopback.c',
-		'main/api_noop.c',
-		'main/api_validate.c',
-		'main/accum.c',
-		'main/arbprogram.c',
-		'main/atifragshader.c',
-		'main/attrib.c',
-		'main/arrayobj.c',
-		'main/blend.c',
-		'main/bufferobj.c',
-		'main/buffers.c',
-		'main/clear.c',
-		'main/clip.c',
-		'main/colortab.c',
-		'main/condrender.c',
-		'main/context.c',
-		'main/convolve.c',
-		'main/cpuinfo.c',
-		'main/debug.c',
-		'main/depth.c',
-		'main/depthstencil.c',
-		'main/dlist.c',
-		'main/dlopen.c',
-		'main/drawpix.c',
-		'main/drawtex.c',
-		'main/enable.c',
-		'main/enums.c',
-		'main/eval.c',
-		'main/execmem.c',
-		'main/extensions.c',
-		'main/fbobject.c',
-		'main/feedback.c',
-		'main/ffvertex_prog.c',
-		'main/fog.c',
-		'main/formats.c',
-		'main/framebuffer.c',
-		'main/get.c',
-		'main/getstring.c',
-		'main/hash.c',
-		'main/hint.c',
-		'main/histogram.c',
-		'main/image.c',
-		'main/imports.c',
-		'main/light.c',
-		'main/lines.c',
-		'main/matrix.c',
-		'main/mipmap.c',
-		'main/mm.c',
-		'main/multisample.c',
-		'main/nvprogram.c',
-		'main/pack.c',
-		'main/pixel.c',
-		'main/pixelstore.c',
-		'main/pixeltransfer.c',
-		'main/points.c',
-		'main/polygon.c',
-		'main/querymatrix.c',
-		'main/queryobj.c',
-		'main/rastpos.c',
-		'main/readpix.c',
-		'main/remap.c',
-		'main/renderbuffer.c',
-		'main/scissor.c',
-		'main/shaderapi.c',
-		'main/shaderobj.c',
-		'main/shared.c',
-		'main/state.c',
-		'main/stencil.c',
-		'main/syncobj.c',
-		'main/texcompress.c',
-		'main/texcompress_s3tc.c',
-		'main/texcompress_fxt1.c',
-		'main/texenv.c',
-		'main/texenvprogram.c',
-		'main/texfetch.c',
-		'main/texformat.c',
-		'main/texgen.c',
-		'main/texgetimage.c',
-		'main/teximage.c',
-		'main/texobj.c',
-		'main/texpal.c',
-		'main/texparam.c',
-		'main/texrender.c',
-		'main/texstate.c',
-		'main/texstore.c',
-		'main/transformfeedback.c',
-		'main/uniforms.c',
-		'main/varray.c',
-		'main/version.c',
-		'main/viewport.c',
-		'main/vtxfmt.c',
-	]
-	
-	math_sources = [
-		'math/m_debug_clip.c',
-		'math/m_debug_norm.c',
-		'math/m_debug_xform.c',
-		'math/m_eval.c',
-		'math/m_matrix.c',
-		'math/m_translate.c',
-		'math/m_vector.c',
-		'math/m_xform.c',
-	]
-	
-	vbo_sources = [
-		'vbo/vbo_context.c',
-		'vbo/vbo_exec.c',
-		'vbo/vbo_exec_api.c',
-		'vbo/vbo_exec_array.c',
-		'vbo/vbo_exec_draw.c',
-		'vbo/vbo_exec_eval.c',
-		'vbo/vbo_rebase.c',
-		'vbo/vbo_split.c',
-		'vbo/vbo_split_copy.c',
-		'vbo/vbo_split_inplace.c',
-		'vbo/vbo_save.c',
-		'vbo/vbo_save_api.c',
-		'vbo/vbo_save_draw.c',
-		'vbo/vbo_save_loopback.c',
-	]
-	
-	vf_sources = [
-		'vf/vf.c',
-		'vf/vf_generic.c',
-		'vf/vf_sse.c',
-	]
-	
-	statetracker_sources = [
-		'state_tracker/st_atom.c',
-		'state_tracker/st_atom_blend.c',
-		'state_tracker/st_atom_clip.c',
-		'state_tracker/st_atom_constbuf.c',
-		'state_tracker/st_atom_depth.c',
-		'state_tracker/st_atom_framebuffer.c',
-		'state_tracker/st_atom_msaa.c',
-		'state_tracker/st_atom_pixeltransfer.c',
-		'state_tracker/st_atom_sampler.c',
-		'state_tracker/st_atom_scissor.c',
-		'state_tracker/st_atom_shader.c',
-		'state_tracker/st_atom_rasterizer.c',
-		'state_tracker/st_atom_stipple.c',
-		'state_tracker/st_atom_texture.c',
-		'state_tracker/st_atom_viewport.c',
-		'state_tracker/st_cb_accum.c',
-		'state_tracker/st_cb_bitmap.c',
-		'state_tracker/st_cb_blit.c',
-		'state_tracker/st_cb_bufferobjects.c',
-		'state_tracker/st_cb_clear.c',
-		'state_tracker/st_cb_condrender.c',
-		'state_tracker/st_cb_flush.c',
-		'state_tracker/st_cb_drawpixels.c',
-		'state_tracker/st_cb_drawtex.c',
-		'state_tracker/st_cb_eglimage.c',
-		'state_tracker/st_cb_fbo.c',
-		'state_tracker/st_cb_feedback.c',
-		'state_tracker/st_cb_program.c',
-		'state_tracker/st_cb_queryobj.c',
-		'state_tracker/st_cb_rasterpos.c',
-		'state_tracker/st_cb_readpixels.c',
-		'state_tracker/st_cb_strings.c',
-		'state_tracker/st_cb_texture.c',
-		'state_tracker/st_cb_viewport.c',
-		'state_tracker/st_cb_xformfb.c',
-		'state_tracker/st_context.c',
-		'state_tracker/st_debug.c',
-		'state_tracker/st_draw.c',
-		'state_tracker/st_draw_feedback.c',
-		'state_tracker/st_extensions.c',
-		'state_tracker/st_format.c',
-		'state_tracker/st_gen_mipmap.c',
-		'state_tracker/st_manager.c',
-		'state_tracker/st_mesa_to_tgsi.c',
-		'state_tracker/st_program.c',
-		'state_tracker/st_texture.c',
-	]
-	
-	program_sources = [
-		'program/arbprogparse.c',
-		'program/hash_table.c',
-        'program/ir_to_mesa.cpp',
-		'program/lex.yy.c',
-		'program/nvfragparse.c',
-		'program/nvvertparse.c',
-		'program/program.c',
-		'program/program_parse.tab.c',
-		'program/program_parse_extra.c',
-		'program/prog_cache.c',
-		'program/prog_execute.c',
-		'program/prog_instruction.c',
-		'program/prog_noise.c',
-		'program/prog_optimize.c',
-		'program/prog_parameter.c',
-		'program/prog_parameter_layout.c',
-		'program/prog_print.c',
-		'program/prog_statevars.c',
-		'program/prog_uniform.c',
-		'program/programopt.c',
-		'program/sampler.cpp',
-		'program/symbol_table.c',
-	]
-	
-	mesa_sources = (
-		main_sources +
-		math_sources +
-		program_sources +
-		vbo_sources +
-		vf_sources +
-		statetracker_sources
-	)
+if env['platform'] == 'windows':
+    env.Append(CPPDEFINES = [
+        '_GDI32_', # prevent gl* being declared __declspec(dllimport) in MS headers
+        'BUILD_GL32', # declare gl* as __declspec(dllexport) in Mesa headers
+        'WIN32_THREADS', # use Win32 thread API
+    ])
+    env.Prepend(CPPPATH = ['#src/talloc'])
 
-	#
-	# Assembly sources
-	#
-	if gcc and env['machine'] == 'x86':
-		env.Append(CPPDEFINES = [
-			'USE_X86_ASM', 
-			'USE_MMX_ASM',
-			'USE_3DNOW_ASM',
-			'USE_SSE_ASM',
-		])
-		mesa_sources += [
-			'x86/common_x86.c',
-			'x86/x86_xform.c',
-			'x86/3dnow.c',
-			'x86/sse.c',
-			'x86/common_x86_asm.S',
-			'x86/x86_xform2.S',
-			'x86/x86_xform3.S',
-			'x86/x86_xform4.S',
-			'x86/x86_cliptest.S',
-			'x86/mmx_blend.S',
-			'x86/3dnow_xform1.S',
-			'x86/3dnow_xform2.S',
-			'x86/3dnow_xform3.S',
-			'x86/3dnow_xform4.S',
-			'x86/3dnow_normal.S',
-			'x86/sse_xform1.S',
-			'x86/sse_xform2.S',
-			'x86/sse_xform3.S',
-			'x86/sse_xform4.S',
-			'x86/sse_normal.S',
-			'x86/read_rgba_span_x86.S',
-		]
-	elif gcc and env['machine'] == 'x86_64':
-		env.Append(CPPDEFINES = [
-			'USE_X86_64_ASM', 
-		])
-		mesa_sources += [
-			'x86-64/x86-64.c',
-			'x86-64/xform4.S',
-		]
-	elif gcc and env['machine'] == 'ppc':
-		env.Append(CPPDEFINES = [
-			'USE_PPC_ASM', 
-			'USE_VMX_ASM', 
-		])
-		mesa_sources += [
-			'ppc/common_ppc.c',
-		]
-	elif gcc and env['machine'] == 'sparc':
-		mesa_sources += [
-			'sparc/sparc.c',
-			'sparc/clip.S',
-			'sparc/norm.S',
-			'sparc/xform.S',
-		]
-	else:
-		pass
-	
-	# Generate matypes.h
-	if gcc and env['machine'] in ('x86', 'x86_64'):
-		# See http://www.scons.org/wiki/UsingCodeGenerators
-		gen_matypes = env.Program(
-			target = 'gen_matypes',
-			source = 'x86/gen_matypes.c',
-		)
-		matypes = env.Command(
-			'matypes.h',
-			gen_matypes,
-			gen_matypes[0].abspath + ' > $TARGET',
-		)
-		# Add the dir containing the generated header (somewhere inside  the
-		# build dir) to the include path  
-		env.Append(CPPPATH = [matypes[0].dir])
+#
+# Source files
+#
 
-	#
-	# Libraries
-	# 
+main_sources = [
+    'main/api_arrayelt.c',
+    'main/api_exec.c',
+    'main/api_loopback.c',
+    'main/api_noop.c',
+    'main/api_validate.c',
+    'main/accum.c',
+    'main/arbprogram.c',
+    'main/atifragshader.c',
+    'main/attrib.c',
+    'main/arrayobj.c',
+    'main/blend.c',
+    'main/bufferobj.c',
+    'main/buffers.c',
+    'main/clear.c',
+    'main/clip.c',
+    'main/colortab.c',
+    'main/condrender.c',
+    'main/context.c',
+    'main/convolve.c',
+    'main/cpuinfo.c',
+    'main/debug.c',
+    'main/depth.c',
+    'main/depthstencil.c',
+    'main/dlist.c',
+    'main/dlopen.c',
+    'main/drawpix.c',
+    'main/drawtex.c',
+    'main/enable.c',
+    'main/enums.c',
+    'main/eval.c',
+    'main/execmem.c',
+    'main/extensions.c',
+    'main/fbobject.c',
+    'main/feedback.c',
+    'main/ffvertex_prog.c',
+    'main/fog.c',
+    'main/formats.c',
+    'main/framebuffer.c',
+    'main/get.c',
+    'main/getstring.c',
+    'main/hash.c',
+    'main/hint.c',
+    'main/histogram.c',
+    'main/image.c',
+    'main/imports.c',
+    'main/light.c',
+    'main/lines.c',
+    'main/matrix.c',
+    'main/mipmap.c',
+    'main/mm.c',
+    'main/multisample.c',
+    'main/nvprogram.c',
+    'main/pack.c',
+    'main/pixel.c',
+    'main/pixelstore.c',
+    'main/pixeltransfer.c',
+    'main/points.c',
+    'main/polygon.c',
+    'main/querymatrix.c',
+    'main/queryobj.c',
+    'main/rastpos.c',
+    'main/readpix.c',
+    'main/remap.c',
+    'main/renderbuffer.c',
+    'main/scissor.c',
+    'main/shaderapi.c',
+    'main/shaderobj.c',
+    'main/shared.c',
+    'main/state.c',
+    'main/stencil.c',
+    'main/syncobj.c',
+    'main/texcompress.c',
+    'main/texcompress_s3tc.c',
+    'main/texcompress_fxt1.c',
+    'main/texenv.c',
+    'main/texenvprogram.c',
+    'main/texfetch.c',
+    'main/texformat.c',
+    'main/texgen.c',
+    'main/texgetimage.c',
+    'main/teximage.c',
+    'main/texobj.c',
+    'main/texpal.c',
+    'main/texparam.c',
+    'main/texrender.c',
+    'main/texstate.c',
+    'main/texstore.c',
+    'main/transformfeedback.c',
+    'main/uniforms.c',
+    'main/varray.c',
+    'main/version.c',
+    'main/viewport.c',
+    'main/vtxfmt.c',
+]
 
-	mesa = env.ConvenienceLibrary(
-		target = 'mesa',
-		source = mesa_sources,
-	)
-	Export('mesa')
+math_sources = [
+    'math/m_debug_clip.c',
+    'math/m_debug_norm.c',
+    'math/m_debug_xform.c',
+    'math/m_eval.c',
+    'math/m_matrix.c',
+    'math/m_translate.c',
+    'math/m_vector.c',
+    'math/m_xform.c',
+]
+
+vbo_sources = [
+    'vbo/vbo_context.c',
+    'vbo/vbo_exec.c',
+    'vbo/vbo_exec_api.c',
+    'vbo/vbo_exec_array.c',
+    'vbo/vbo_exec_draw.c',
+    'vbo/vbo_exec_eval.c',
+    'vbo/vbo_rebase.c',
+    'vbo/vbo_split.c',
+    'vbo/vbo_split_copy.c',
+    'vbo/vbo_split_inplace.c',
+    'vbo/vbo_save.c',
+    'vbo/vbo_save_api.c',
+    'vbo/vbo_save_draw.c',
+    'vbo/vbo_save_loopback.c',
+]
+
+vf_sources = [
+    'vf/vf.c',
+    'vf/vf_generic.c',
+    'vf/vf_sse.c',
+]
+
+statetracker_sources = [
+    'state_tracker/st_atom.c',
+    'state_tracker/st_atom_blend.c',
+    'state_tracker/st_atom_clip.c',
+    'state_tracker/st_atom_constbuf.c',
+    'state_tracker/st_atom_depth.c',
+    'state_tracker/st_atom_framebuffer.c',
+    'state_tracker/st_atom_msaa.c',
+    'state_tracker/st_atom_pixeltransfer.c',
+    'state_tracker/st_atom_sampler.c',
+    'state_tracker/st_atom_scissor.c',
+    'state_tracker/st_atom_shader.c',
+    'state_tracker/st_atom_rasterizer.c',
+    'state_tracker/st_atom_stipple.c',
+    'state_tracker/st_atom_texture.c',
+    'state_tracker/st_atom_viewport.c',
+    'state_tracker/st_cb_accum.c',
+    'state_tracker/st_cb_bitmap.c',
+    'state_tracker/st_cb_blit.c',
+    'state_tracker/st_cb_bufferobjects.c',
+    'state_tracker/st_cb_clear.c',
+    'state_tracker/st_cb_condrender.c',
+    'state_tracker/st_cb_flush.c',
+    'state_tracker/st_cb_drawpixels.c',
+    'state_tracker/st_cb_drawtex.c',
+    'state_tracker/st_cb_eglimage.c',
+    'state_tracker/st_cb_fbo.c',
+    'state_tracker/st_cb_feedback.c',
+    'state_tracker/st_cb_program.c',
+    'state_tracker/st_cb_queryobj.c',
+    'state_tracker/st_cb_rasterpos.c',
+    'state_tracker/st_cb_readpixels.c',
+    'state_tracker/st_cb_strings.c',
+    'state_tracker/st_cb_texture.c',
+    'state_tracker/st_cb_viewport.c',
+    'state_tracker/st_cb_xformfb.c',
+    'state_tracker/st_context.c',
+    'state_tracker/st_debug.c',
+    'state_tracker/st_draw.c',
+    'state_tracker/st_draw_feedback.c',
+    'state_tracker/st_extensions.c',
+    'state_tracker/st_format.c',
+    'state_tracker/st_gen_mipmap.c',
+    'state_tracker/st_manager.c',
+    'state_tracker/st_mesa_to_tgsi.c',
+    'state_tracker/st_program.c',
+    'state_tracker/st_texture.c',
+]
+
+program_sources = [
+    'program/arbprogparse.c',
+    'program/hash_table.c',
+    'program/ir_to_mesa.cpp',
+    'program/lex.yy.c',
+    'program/nvfragparse.c',
+    'program/nvvertparse.c',
+    'program/program.c',
+    'program/program_parse.tab.c',
+    'program/program_parse_extra.c',
+    'program/prog_cache.c',
+    'program/prog_execute.c',
+    'program/prog_instruction.c',
+    'program/prog_noise.c',
+    'program/prog_optimize.c',
+    'program/prog_parameter.c',
+    'program/prog_parameter_layout.c',
+    'program/prog_print.c',
+    'program/prog_statevars.c',
+    'program/prog_uniform.c',
+    'program/programopt.c',
+    'program/sampler.cpp',
+    'program/symbol_table.c',
+]
+
+mesa_sources = (
+    main_sources +
+    math_sources +
+    program_sources +
+    vbo_sources +
+    vf_sources +
+    statetracker_sources
+)
+
+#
+# Assembly sources
+#
+if env['gcc'] and env['platform'] != 'windows':
+    if env['machine'] == 'x86':
+        env.Append(CPPDEFINES = [
+            'USE_X86_ASM',
+            'USE_MMX_ASM',
+            'USE_3DNOW_ASM',
+            'USE_SSE_ASM',
+        ])
+        mesa_sources += [
+            'x86/common_x86.c',
+            'x86/x86_xform.c',
+            'x86/3dnow.c',
+            'x86/sse.c',
+            'x86/common_x86_asm.S',
+            'x86/x86_xform2.S',
+            'x86/x86_xform3.S',
+            'x86/x86_xform4.S',
+            'x86/x86_cliptest.S',
+            'x86/mmx_blend.S',
+            'x86/3dnow_xform1.S',
+            'x86/3dnow_xform2.S',
+            'x86/3dnow_xform3.S',
+            'x86/3dnow_xform4.S',
+            'x86/3dnow_normal.S',
+            'x86/sse_xform1.S',
+            'x86/sse_xform2.S',
+            'x86/sse_xform3.S',
+            'x86/sse_xform4.S',
+            'x86/sse_normal.S',
+            'x86/read_rgba_span_x86.S',
+        ]
+    elif env['machine'] == 'x86_64':
+        env.Append(CPPDEFINES = [
+            'USE_X86_64_ASM',
+        ])
+        mesa_sources += [
+            'x86-64/x86-64.c',
+            'x86-64/xform4.S',
+        ]
+    elif env['machine'] == 'ppc':
+        env.Append(CPPDEFINES = [
+            'USE_PPC_ASM',
+            'USE_VMX_ASM',
+        ])
+        mesa_sources += [
+            'ppc/common_ppc.c',
+        ]
+    elif env['machine'] == 'sparc':
+        mesa_sources += [
+            'sparc/sparc.c',
+            'sparc/clip.S',
+            'sparc/norm.S',
+            'sparc/xform.S',
+        ]
+    else:
+        pass
+
+    # Generate matypes.h
+    if env['machine'] in ('x86', 'x86_64'):
+        # See http://www.scons.org/wiki/UsingCodeGenerators
+        gen_matypes = env.Program(
+            target = 'gen_matypes',
+            source = 'x86/gen_matypes.c',
+        )
+        matypes = env.Command(
+            'matypes.h',
+            gen_matypes,
+            gen_matypes[0].abspath + ' > $TARGET',
+        )
+        # Add the dir containing the generated header (somewhere inside  the
+        # build dir) to the include path
+        env.Append(CPPPATH = [matypes[0].dir])
+
+#
+# Libraries
+#
+
+mesa = env.ConvenienceLibrary(
+    target = 'mesa',
+    source = mesa_sources,
+)
+
+env.Alias('mesa', mesa)
+
+Export('mesa')
diff --git a/src/mesa/drivers/common/meta.c b/src/mesa/drivers/common/meta.c
index 1bfd76a665..6433c89d81 100644
--- a/src/mesa/drivers/common/meta.c
+++ b/src/mesa/drivers/common/meta.c
@@ -52,6 +52,7 @@
 #include "main/readpix.h"
 #include "main/scissor.h"
 #include "main/shaderapi.h"
+#include "main/shaderobj.h"
 #include "main/state.h"
 #include "main/stencil.h"
 #include "main/texobj.h"
@@ -103,6 +104,8 @@ struct save_state
 
    /** META_ALPHA_TEST */
    GLboolean AlphaEnabled;
+   GLenum AlphaFunc;
+   GLclampf AlphaRef;
 
    /** META_BLEND */
    GLbitfield BlendEnabled;
@@ -143,10 +146,10 @@ struct save_state
    struct gl_vertex_program *VertexProgram;
    GLboolean FragmentProgramEnabled;
    struct gl_fragment_program *FragmentProgram;
-   GLuint VertexShader;
-   GLuint GeometryShader;
-   GLuint FragmentShader;
-   GLuint ActiveShader;
+   struct gl_shader_program *VertexShader;
+   struct gl_shader_program *GeometryShader;
+   struct gl_shader_program *FragmentShader;
+   struct gl_shader_program *ActiveShader;
 
    /** META_STENCIL_TEST */
    struct gl_stencil_attrib Stencil;
@@ -327,6 +330,8 @@ _mesa_meta_begin(struct gl_context *ctx, GLbitfield state)
 
    if (state & META_ALPHA_TEST) {
       save->AlphaEnabled = ctx->Color.AlphaEnabled;
+      save->AlphaFunc = ctx->Color.AlphaFunc;
+      save->AlphaRef = ctx->Color.AlphaRef;
       if (ctx->Color.AlphaEnabled)
          _mesa_set_enable(ctx, GL_ALPHA_TEST, GL_FALSE);
    }
@@ -436,14 +441,14 @@ _mesa_meta_begin(struct gl_context *ctx, GLbitfield state)
       }
 
       if (ctx->Extensions.ARB_shader_objects) {
-         save->VertexShader = ctx->Shader.CurrentVertexProgram ?
-            ctx->Shader.CurrentVertexProgram->Name : 0;
-         save->GeometryShader = ctx->Shader.CurrentGeometryProgram ?
-            ctx->Shader.CurrentGeometryProgram->Name : 0;
-         save->FragmentShader = ctx->Shader.CurrentFragmentProgram ?
-            ctx->Shader.CurrentFragmentProgram->Name : 0;
-         save->ActiveShader = ctx->Shader.ActiveProgram ?
-            ctx->Shader.ActiveProgram->Name : 0;
+	 _mesa_reference_shader_program(ctx, &save->VertexShader,
+					ctx->Shader.CurrentVertexProgram);
+	 _mesa_reference_shader_program(ctx, &save->GeometryShader,
+					ctx->Shader.CurrentGeometryProgram);
+	 _mesa_reference_shader_program(ctx, &save->FragmentShader,
+					ctx->Shader.CurrentFragmentProgram);
+	 _mesa_reference_shader_program(ctx, &save->ActiveShader,
+					ctx->Shader.CurrentFragmentProgram);
 
          _mesa_UseProgramObjectARB(0);
       }
@@ -575,6 +580,7 @@ _mesa_meta_end(struct gl_context *ctx)
    if (state & META_ALPHA_TEST) {
       if (ctx->Color.AlphaEnabled != save->AlphaEnabled)
          _mesa_set_enable(ctx, GL_ALPHA_TEST, save->AlphaEnabled);
+      _mesa_AlphaFunc(save->AlphaFunc, save->AlphaRef);
    }
 
    if (state & META_BLEND) {
@@ -675,16 +681,18 @@ _mesa_meta_end(struct gl_context *ctx)
       }
 
       if (ctx->Extensions.ARB_vertex_shader)
-	 _mesa_UseShaderProgramEXT(GL_VERTEX_SHADER, save->VertexShader);
+	 _mesa_use_shader_program(ctx, GL_VERTEX_SHADER, save->VertexShader);
 
       if (ctx->Extensions.ARB_geometry_shader4)
-	 _mesa_UseShaderProgramEXT(GL_GEOMETRY_SHADER_ARB,
-				   save->GeometryShader);
+	 _mesa_use_shader_program(ctx, GL_GEOMETRY_SHADER_ARB,
+				  save->GeometryShader);
 
       if (ctx->Extensions.ARB_fragment_shader)
-	 _mesa_UseShaderProgramEXT(GL_FRAGMENT_SHADER, save->FragmentShader);
+	 _mesa_use_shader_program(ctx, GL_FRAGMENT_SHADER,
+				  save->FragmentShader);
 
-      _mesa_ActiveProgramEXT(save->ActiveShader);
+      _mesa_reference_shader_program(ctx, &ctx->Shader.ActiveProgram,
+				     save->ActiveShader);
    }
 
    if (state & META_STENCIL_TEST) {
@@ -1989,6 +1997,7 @@ _mesa_meta_Bitmap(struct gl_context *ctx,
    struct temp_texture *tex = get_bitmap_temp_texture(ctx);
    const GLenum texIntFormat = GL_ALPHA;
    const struct gl_pixelstore_attrib unpackSave = *unpack;
+   GLubyte fg, bg;
    struct vertex {
       GLfloat x, y, z, s, t, r, g, b, a;
    };
@@ -2090,21 +2099,26 @@ _mesa_meta_Bitmap(struct gl_context *ctx,
       _mesa_BufferSubDataARB(GL_ARRAY_BUFFER_ARB, 0, sizeof(verts), verts);
    }
 
+   /* choose different foreground/background alpha values */
+   CLAMPED_FLOAT_TO_UBYTE(fg, ctx->Current.RasterColor[ACOMP]);
+   bg = (fg > 127 ? 0 : 255);
+
    bitmap1 = _mesa_map_pbo_source(ctx, &unpackSave, bitmap1);
    if (!bitmap1) {
       _mesa_meta_end(ctx);
       return;
    }
 
-   bitmap8 = (GLubyte *) calloc(1, width * height);
+   bitmap8 = (GLubyte *) malloc(width * height);
    if (bitmap8) {
+      memset(bitmap8, bg, width * height);
       _mesa_expand_bitmap(width, height, &unpackSave, bitmap1,
-                          bitmap8, width, 0xff);
+                          bitmap8, width, fg);
 
       _mesa_set_enable(ctx, tex->Target, GL_TRUE);
 
       _mesa_set_enable(ctx, GL_ALPHA_TEST, GL_TRUE);
-      _mesa_AlphaFunc(GL_GREATER, 0.0);
+      _mesa_AlphaFunc(GL_NOTEQUAL, UBYTE_TO_FLOAT(bg));
 
       setup_drawpix_texture(ctx, tex, newTex, texIntFormat, width, height,
                             GL_ALPHA, GL_UNSIGNED_BYTE, bitmap8);
diff --git a/src/mesa/drivers/dri/common/utils.c b/src/mesa/drivers/dri/common/utils.c
index c195c4fd8f..42be77fd7c 100644
--- a/src/mesa/drivers/dri/common/utils.c
+++ b/src/mesa/drivers/dri/common/utils.c
@@ -738,12 +738,18 @@ static const struct { unsigned int attrib, offset; } attribMap[] = {
 
 #define ARRAY_SIZE(a) (sizeof (a) / sizeof ((a)[0]))
 
+
+/**
+ * Return the value of a configuration attribute.  The attribute is
+ * indicated by the index.
+ */
 static int
 driGetConfigAttribIndex(const __DRIconfig *config,
 			unsigned int index, unsigned int *value)
 {
     switch (attribMap[index].attrib) {
     case __DRI_ATTRIB_RENDER_TYPE:
+        /* no support for color index mode */
 	*value = __DRI_ATTRIB_RGBA_BIT;
 	break;
     case __DRI_ATTRIB_CONFIG_CAVEAT:
@@ -755,13 +761,16 @@ driGetConfigAttribIndex(const __DRIconfig *config,
 	    *value = 0;
 	break;
     case __DRI_ATTRIB_SWAP_METHOD:
+        /* XXX no return value??? */
 	break;
 
     case __DRI_ATTRIB_FLOAT_MODE:
+        /* this field is not int-sized */
         *value = config->modes.floatMode;
         break;
 
     default:
+        /* any other int-sized field */
 	*value = *(unsigned int *)
 	    ((char *) &config->modes + attribMap[index].offset);
 	
@@ -771,6 +780,13 @@ driGetConfigAttribIndex(const __DRIconfig *config,
     return GL_TRUE;
 }
 
+
+/**
+ * Get the value of a configuration attribute.
+ * \param attrib  the attribute (one of the _DRI_ATTRIB_x tokens)
+ * \param value  returns the attribute's value
+ * \return 1 for success, 0 for failure
+ */
 int
 driGetConfigAttrib(const __DRIconfig *config,
 		   unsigned int attrib, unsigned int *value)
@@ -784,6 +800,14 @@ driGetConfigAttrib(const __DRIconfig *config,
     return GL_FALSE;
 }
 
+
+/**
+ * Get a configuration attribute name and value, given an index.
+ * \param index  which field of the __DRIconfig to query
+ * \param attrib  returns the attribute name (one of the _DRI_ATTRIB_x tokens)
+ * \param value  returns the attribute's value
+ * \return 1 for success, 0 for failure
+ */
 int
 driIndexConfigAttrib(const __DRIconfig *config, int index,
 		     unsigned int *attrib, unsigned int *value)
diff --git a/src/mesa/drivers/dri/i915/intel_structs.h b/src/mesa/drivers/dri/i915/intel_structs.h
deleted file mode 100644
index 522e3bd92c..0000000000
--- a/src/mesa/drivers/dri/i915/intel_structs.h
+++ /dev/null
@@ -1,132 +0,0 @@
-#ifndef INTEL_STRUCTS_H
-#define INTEL_STRUCTS_H
-
-struct br0 {
-   GLuint length:8;
-   GLuint pad0:3;
-   GLuint dst_tiled:1;
-   GLuint pad1:8;
-   GLuint write_rgb:1;
-   GLuint write_alpha:1;
-   GLuint opcode:7;
-   GLuint client:3;
-};
-
-   
-struct br13 {
-   GLint dest_pitch:16;
-   GLuint rop:8;
-   GLuint color_depth:2;
-   GLuint pad1:3;
-   GLuint mono_source_transparency:1;
-   GLuint clipping_enable:1;
-   GLuint pad0:1;
-};
-
-
-
-/* This is an attempt to move some of the 2D interaction in this
- * driver to using structs for packets rather than a bunch of #defines
- * and dwords.
- */
-struct xy_color_blit {
-   struct br0 br0;
-   struct br13 br13;
-
-   struct {
-      GLuint dest_x1:16;
-      GLuint dest_y1:16;
-   } dw2;
-
-   struct {
-      GLuint dest_x2:16;
-      GLuint dest_y2:16;
-   } dw3;
-   
-   GLuint dest_base_addr;
-   GLuint color;
-};
-
-struct xy_src_copy_blit {
-   struct br0 br0;
-   struct br13 br13;
-
-   struct {
-      GLuint dest_x1:16;
-      GLuint dest_y1:16;
-   } dw2;
-
-   struct {
-      GLuint dest_x2:16;
-      GLuint dest_y2:16;
-   } dw3;
-   
-   GLuint dest_base_addr;
-
-   struct {
-      GLuint src_x1:16;
-      GLuint src_y1:16;
-   } dw5;
-
-   struct {
-      GLint src_pitch:16;
-      GLuint pad:16;
-   } dw6;
-   
-   GLuint src_base_addr;
-};
-
-struct xy_setup_blit {
-   struct br0 br0;
-   struct br13 br13;
-
-   struct {
-      GLuint clip_x1:16;
-      GLuint clip_y1:16;
-   } dw2;
-
-   struct {
-      GLuint clip_x2:16;
-      GLuint clip_y2:16;
-   } dw3;
-      
-   GLuint dest_base_addr;
-   GLuint background_color;
-   GLuint foreground_color;
-   GLuint pattern_base_addr;
-};
-
-
-struct xy_text_immediate_blit {
-   struct {
-      GLuint length:8;
-      GLuint pad2:3;
-      GLuint dst_tiled:1;
-      GLuint pad1:4;
-      GLuint byte_packed:1;
-      GLuint pad0:5;
-      GLuint opcode:7;
-      GLuint client:3;
-   } dw0;
-
-   struct {
-      GLuint dest_x1:16;
-      GLuint dest_y1:16;
-   } dw1;
-
-   struct {
-      GLuint dest_x2:16;
-      GLuint dest_y2:16;
-   } dw2;   
-
-   /* Src bitmap data follows as inline dwords.
-    */
-};
-
-
-#define CLIENT_2D 0x2
-#define OPCODE_XY_SETUP_BLT 0x1
-#define OPCODE_XY_COLOR_BLT 0x50
-#define OPCODE_XY_TEXT_IMMEDIATE_BLT 0x31
-
-#endif
diff --git a/src/mesa/drivers/dri/i965/brw_cc.c b/src/mesa/drivers/dri/i965/brw_cc.c
index 00418760da..a8369b07c3 100644
--- a/src/mesa/drivers/dri/i965/brw_cc.c
+++ b/src/mesa/drivers/dri/i965/brw_cc.c
@@ -204,7 +204,7 @@ static void upload_cc_unit(struct brw_context *brw)
       cc.cc2.depth_write_enable = ctx->Depth.Mask;
    }
 
-   if (intel->stats_wm || (INTEL_DEBUG & DEBUG_STATS))
+   if (intel->stats_wm || unlikely(INTEL_DEBUG & DEBUG_STATS))
       cc.cc5.statistics_enable = 1;
 
    /* CACHE_NEW_CC_VP */
diff --git a/src/mesa/drivers/dri/i965/brw_clip.c b/src/mesa/drivers/dri/i965/brw_clip.c
index 15e60bf3ce..1be165cc9a 100644
--- a/src/mesa/drivers/dri/i965/brw_clip.c
+++ b/src/mesa/drivers/dri/i965/brw_clip.c
@@ -133,13 +133,13 @@ static void compile_clip_prog( struct brw_context *brw,
     */
    program = brw_get_program(&c.func, &program_size);
 
-    if (INTEL_DEBUG & DEBUG_CLIP) {
+   if (unlikely(INTEL_DEBUG & DEBUG_CLIP)) {
       printf("clip:\n");
       for (i = 0; i < program_size / sizeof(struct brw_instruction); i++)
 	 brw_disasm(stdout, &((struct brw_instruction *)program)[i],
 		    intel->gen);
       printf("\n");
-    }
+   }
 
    /* Upload
     */
diff --git a/src/mesa/drivers/dri/i965/brw_clip_state.c b/src/mesa/drivers/dri/i965/brw_clip_state.c
index 885167da90..60fd5fa7d9 100644
--- a/src/mesa/drivers/dri/i965/brw_clip_state.c
+++ b/src/mesa/drivers/dri/i965/brw_clip_state.c
@@ -114,10 +114,10 @@ clip_unit_create_from_key(struct brw_context *brw,
       clip.thread4.max_threads = 1 - 1;
    }
 
-   if (INTEL_DEBUG & DEBUG_SINGLE_THREAD)
+   if (unlikely(INTEL_DEBUG & DEBUG_SINGLE_THREAD))
       clip.thread4.max_threads = 0;
 
-   if (INTEL_DEBUG & DEBUG_STATS)
+   if (unlikely(INTEL_DEBUG & DEBUG_STATS))
       clip.thread4.stats_enable = 1;
 
    clip.clip5.userclip_enable_flags = 0x7f;
diff --git a/src/mesa/drivers/dri/i965/brw_draw.c b/src/mesa/drivers/dri/i965/brw_draw.c
index 04bc8cb2db..a1f403ca4e 100644
--- a/src/mesa/drivers/dri/i965/brw_draw.c
+++ b/src/mesa/drivers/dri/i965/brw_draw.c
@@ -42,7 +42,7 @@
 
 #include "intel_batchbuffer.h"
 
-#define FILE_DEBUG_FLAG DEBUG_BATCH
+#define FILE_DEBUG_FLAG DEBUG_PRIMS
 
 static GLuint prim_to_hw_prim[GL_POLYGON+1] = {
    _3DPRIM_POINTLIST,
@@ -83,8 +83,7 @@ static GLuint brw_set_prim(struct brw_context *brw,
    struct gl_context *ctx = &brw->intel.ctx;
    GLenum mode = prim->mode;
 
-   if (INTEL_DEBUG & DEBUG_PRIMS)
-      printf("PRIM: %s\n", _mesa_lookup_enum_by_nr(prim->mode));
+   DBG("PRIM: %s\n", _mesa_lookup_enum_by_nr(prim->mode));
 
    /* Slight optimization to avoid the GS program when not needed:
     */
@@ -133,9 +132,8 @@ static void brw_emit_prim(struct brw_context *brw,
    struct brw_3d_primitive prim_packet;
    struct intel_context *intel = &brw->intel;
 
-   if (INTEL_DEBUG & DEBUG_PRIMS)
-      printf("PRIM: %s %d %d\n", _mesa_lookup_enum_by_nr(prim->mode), 
-		   prim->start, prim->count);
+   DBG("PRIM: %s %d %d\n", _mesa_lookup_enum_by_nr(prim->mode),
+       prim->start, prim->count);
 
    prim_packet.header.opcode = CMD_3D_PRIM;
    prim_packet.header.length = sizeof(prim_packet)/4 - 2;
diff --git a/src/mesa/drivers/dri/i965/brw_draw_upload.c b/src/mesa/drivers/dri/i965/brw_draw_upload.c
index c4654360d4..2cefe614dd 100644
--- a/src/mesa/drivers/dri/i965/brw_draw_upload.c
+++ b/src/mesa/drivers/dri/i965/brw_draw_upload.c
@@ -168,7 +168,7 @@ static GLuint byte_types_scale[5] = {
 static GLuint get_surface_type( GLenum type, GLuint size,
                                 GLenum format, GLboolean normalized )
 {
-   if (INTEL_DEBUG & DEBUG_VERTS)
+   if (unlikely(INTEL_DEBUG & DEBUG_VERTS))
       printf("type %s size %d normalized %d\n", 
 		   _mesa_lookup_enum_by_nr(type), size, normalized);
 
diff --git a/src/mesa/drivers/dri/i965/brw_fs.cpp b/src/mesa/drivers/dri/i965/brw_fs.cpp
index 2ed59d3f5d..3caddae4a1 100644
--- a/src/mesa/drivers/dri/i965/brw_fs.cpp
+++ b/src/mesa/drivers/dri/i965/brw_fs.cpp
@@ -3129,7 +3129,7 @@ fs_visitor::generate_code()
    const char *last_annotation_string = NULL;
    ir_instruction *last_annotation_ir = NULL;
 
-   if (INTEL_DEBUG & DEBUG_WM) {
+   if (unlikely(INTEL_DEBUG & DEBUG_WM)) {
       printf("Native code for fragment shader %d:\n",
 	     ctx->Shader.CurrentFragmentProgram->Name);
    }
@@ -3141,7 +3141,7 @@ fs_visitor::generate_code()
       fs_inst *inst = (fs_inst *)iter.get();
       struct brw_reg src[3], dst;
 
-      if (INTEL_DEBUG & DEBUG_WM) {
+      if (unlikely(INTEL_DEBUG & DEBUG_WM)) {
 	 if (last_annotation_ir != inst->ir) {
 	    last_annotation_ir = inst->ir;
 	    if (last_annotation_ir) {
@@ -3335,7 +3335,7 @@ fs_visitor::generate_code()
 	 this->fail = true;
       }
 
-      if (INTEL_DEBUG & DEBUG_WM) {
+      if (unlikely(INTEL_DEBUG & DEBUG_WM)) {
 	 for (unsigned int i = last_native_inst; i < p->nr_insn; i++) {
 	    if (0) {
 	       printf("0x%08x 0x%08x 0x%08x 0x%08x ",
@@ -3376,7 +3376,7 @@ brw_wm_fs_emit(struct brw_context *brw, struct brw_wm_compile *c)
     */
    c->dispatch_width = 8;
 
-   if (INTEL_DEBUG & DEBUG_WM) {
+   if (unlikely(INTEL_DEBUG & DEBUG_WM)) {
       printf("GLSL IR for native fragment shader %d:\n", prog->Name);
       _mesa_print_ir(shader->ir, NULL);
       printf("\n");
diff --git a/src/mesa/drivers/dri/i965/brw_gs.c b/src/mesa/drivers/dri/i965/brw_gs.c
index cfcc8ea4d6..b0c76f4094 100644
--- a/src/mesa/drivers/dri/i965/brw_gs.c
+++ b/src/mesa/drivers/dri/i965/brw_gs.c
@@ -127,8 +127,8 @@ static void compile_gs_prog( struct brw_context *brw,
     */
    program = brw_get_program(&c.func, &program_size);
 
-    if (INTEL_DEBUG & DEBUG_GS) {
-       int i;
+   if (unlikely(INTEL_DEBUG & DEBUG_GS)) {
+      int i;
 
       printf("gs:\n");
       for (i = 0; i < program_size / sizeof(struct brw_instruction); i++)
diff --git a/src/mesa/drivers/dri/i965/brw_gs_state.c b/src/mesa/drivers/dri/i965/brw_gs_state.c
index 63562ebcfc..69a5f7a666 100644
--- a/src/mesa/drivers/dri/i965/brw_gs_state.c
+++ b/src/mesa/drivers/dri/i965/brw_gs_state.c
@@ -101,7 +101,7 @@ gs_unit_create_from_key(struct brw_context *brw, struct brw_gs_unit_key *key)
    if (intel->gen == 5)
       gs.thread4.rendering_enable = 1;
 
-   if (INTEL_DEBUG & DEBUG_STATS)
+   if (unlikely(INTEL_DEBUG & DEBUG_STATS))
       gs.thread4.stats_enable = 1;
 
    bo = brw_upload_cache(&brw->cache, BRW_GS_UNIT,
diff --git a/src/mesa/drivers/dri/i965/brw_misc_state.c b/src/mesa/drivers/dri/i965/brw_misc_state.c
index 24041e57b0..1d350bc041 100644
--- a/src/mesa/drivers/dri/i965/brw_misc_state.c
+++ b/src/mesa/drivers/dri/i965/brw_misc_state.c
@@ -555,7 +555,7 @@ static void upload_invarient_state( struct brw_context *brw )
       memset(&vfs, 0, sizeof(vfs));
 
       vfs.opcode = brw->CMD_VF_STATISTICS;
-      if (INTEL_DEBUG & DEBUG_STATS)
+      if (unlikely(INTEL_DEBUG & DEBUG_STATS))
 	 vfs.statistics_enable = 1; 
 
       BRW_BATCH_STRUCT(brw, &vfs);
diff --git a/src/mesa/drivers/dri/i965/brw_sf.c b/src/mesa/drivers/dri/i965/brw_sf.c
index 7dbd70daae..6da155b1a9 100644
--- a/src/mesa/drivers/dri/i965/brw_sf.c
+++ b/src/mesa/drivers/dri/i965/brw_sf.c
@@ -108,7 +108,7 @@ static void compile_sf_prog( struct brw_context *brw,
     */
    program = brw_get_program(&c.func, &program_size);
 
-   if (INTEL_DEBUG & DEBUG_SF) {
+   if (unlikely(INTEL_DEBUG & DEBUG_SF)) {
       printf("sf:\n");
       for (i = 0; i < program_size / sizeof(struct brw_instruction); i++)
 	 brw_disasm(stdout, &((struct brw_instruction *)program)[i],
diff --git a/src/mesa/drivers/dri/i965/brw_sf_state.c b/src/mesa/drivers/dri/i965/brw_sf_state.c
index 6ad9e1b48a..bd3a21ed9e 100644
--- a/src/mesa/drivers/dri/i965/brw_sf_state.c
+++ b/src/mesa/drivers/dri/i965/brw_sf_state.c
@@ -210,10 +210,10 @@ sf_unit_create_from_key(struct brw_context *brw, struct brw_sf_unit_key *key,
 
    sf.thread4.max_threads = MIN2(chipset_max_threads, key->nr_urb_entries) - 1;
 
-   if (INTEL_DEBUG & DEBUG_SINGLE_THREAD)
+   if (unlikely(INTEL_DEBUG & DEBUG_SINGLE_THREAD))
       sf.thread4.max_threads = 0;
 
-   if (INTEL_DEBUG & DEBUG_STATS)
+   if (unlikely(INTEL_DEBUG & DEBUG_STATS))
       sf.thread4.stats_enable = 1;
 
    /* CACHE_NEW_SF_VP */
diff --git a/src/mesa/drivers/dri/i965/brw_state_cache.c b/src/mesa/drivers/dri/i965/brw_state_cache.c
index b31d84953a..58ff528d44 100644
--- a/src/mesa/drivers/dri/i965/brw_state_cache.c
+++ b/src/mesa/drivers/dri/i965/brw_state_cache.c
@@ -61,6 +61,7 @@
 #include "intel_batchbuffer.h"
 #include "brw_wm.h"
 
+#define FILE_DEBUG_FLAG DEBUG_STATE
 
 static GLuint
 hash_key(struct brw_cache_item *item)
@@ -265,10 +266,9 @@ brw_upload_cache_with_auxdata(struct brw_cache *cache,
       *(void **)aux_return = (void *)((char *)item->key + item->key_size);
    }
 
-   if (INTEL_DEBUG & DEBUG_STATE)
-      printf("upload %s: %d bytes to cache id %d\n",
-		   cache->name[cache_id],
-		   data_size, cache_id);
+   DBG("upload %s: %d bytes to cache id %d\n",
+       cache->name[cache_id],
+       data_size, cache_id);
 
    /* Copy data to the buffer */
    drm_intel_bo_subdata(bo, 0, data_size, data);
@@ -407,8 +407,7 @@ brw_clear_cache(struct brw_context *brw, struct brw_cache *cache)
    struct brw_cache_item *c, *next;
    GLuint i;
 
-   if (INTEL_DEBUG & DEBUG_STATE)
-      printf("%s\n", __FUNCTION__);
+   DBG("%s\n", __FUNCTION__);
 
    for (i = 0; i < cache->size; i++) {
       for (c = cache->items[i]; c; c = next) {
@@ -434,8 +433,7 @@ brw_clear_cache(struct brw_context *brw, struct brw_cache *cache)
 void
 brw_state_cache_check_size(struct brw_context *brw)
 {
-   if (INTEL_DEBUG & DEBUG_STATE)
-      printf("%s (n_items=%d)\n", __FUNCTION__, brw->cache.n_items);
+   DBG("%s (n_items=%d)\n", __FUNCTION__, brw->cache.n_items);
 
    /* un-tuned guess.  Each object is generally a page, so 1000 of them is 4 MB of
     * state cache.
@@ -450,8 +448,7 @@ brw_destroy_cache(struct brw_context *brw, struct brw_cache *cache)
 {
    GLuint i;
 
-   if (INTEL_DEBUG & DEBUG_STATE)
-      printf("%s\n", __FUNCTION__);
+   DBG("%s\n", __FUNCTION__);
 
    brw_clear_cache(brw, cache);
    for (i = 0; i < BRW_MAX_CACHE; i++) {
diff --git a/src/mesa/drivers/dri/i965/brw_state_upload.c b/src/mesa/drivers/dri/i965/brw_state_upload.c
index f3b6a90f61..338f3876b3 100644
--- a/src/mesa/drivers/dri/i965/brw_state_upload.c
+++ b/src/mesa/drivers/dri/i965/brw_state_upload.c
@@ -435,7 +435,7 @@ void brw_upload_state(struct brw_context *brw)
 
    brw_clear_validated_bos(brw);
 
-   if (INTEL_DEBUG) {
+   if (unlikely(INTEL_DEBUG)) {
       /* Debug version which enforces various sanity checks on the
        * state flags which are generated and checked to help ensure
        * state atoms are ordered correctly in the list.
@@ -487,7 +487,7 @@ void brw_upload_state(struct brw_context *brw)
       }
    }
 
-   if (INTEL_DEBUG & DEBUG_STATE) {
+   if (unlikely(INTEL_DEBUG & DEBUG_STATE)) {
       brw_update_dirty_count(mesa_bits, state->mesa);
       brw_update_dirty_count(brw_bits, state->brw);
       brw_update_dirty_count(cache_bits, state->cache);
diff --git a/src/mesa/drivers/dri/i965/brw_urb.c b/src/mesa/drivers/dri/i965/brw_urb.c
index 0f597184b4..dfc1551aca 100644
--- a/src/mesa/drivers/dri/i965/brw_urb.c
+++ b/src/mesa/drivers/dri/i965/brw_urb.c
@@ -190,12 +190,12 @@ static void recalculate_urb_fence( struct brw_context *brw )
 	    exit(1);
 	 }
 	 
-	 if (INTEL_DEBUG & (DEBUG_URB|DEBUG_FALLBACKS))
+	 if (unlikely(INTEL_DEBUG & (DEBUG_URB|DEBUG_FALLBACKS)))
 	    printf("URB CONSTRAINED\n");
       }
 
 done:
-      if (INTEL_DEBUG & DEBUG_URB)
+      if (unlikely(INTEL_DEBUG & DEBUG_URB))
 	 printf("URB fence: %d ..VS.. %d ..GS.. %d ..CLP.. %d ..SF.. %d ..CS.. %d\n",
 		      brw->urb.vs_start,
 		      brw->urb.gs_start,
diff --git a/src/mesa/drivers/dri/i965/brw_vs_emit.c b/src/mesa/drivers/dri/i965/brw_vs_emit.c
index e560cdac1f..aa8048c9fc 100644
--- a/src/mesa/drivers/dri/i965/brw_vs_emit.c
+++ b/src/mesa/drivers/dri/i965/brw_vs_emit.c
@@ -394,7 +394,7 @@ static void brw_vs_alloc_regs( struct brw_vs_compile *c )
 
    c->prog_data.total_grf = reg;
 
-   if (INTEL_DEBUG & DEBUG_VS) {
+   if (unlikely(INTEL_DEBUG & DEBUG_VS)) {
       printf("%s NumAddrRegs %d\n", __FUNCTION__, c->vp->program.Base.NumAddressRegs);
       printf("%s NumTemps %d\n", __FUNCTION__, c->vp->program.Base.NumTemporaries);
       printf("%s reg = %d\n", __FUNCTION__, reg);
@@ -1678,7 +1678,7 @@ void brw_vs_emit(struct brw_vs_compile *c )
    GLuint index;
    GLuint file;
 
-   if (INTEL_DEBUG & DEBUG_VS) {
+   if (unlikely(INTEL_DEBUG & DEBUG_VS)) {
       printf("vs-mesa:\n");
       _mesa_fprint_program_opt(stdout, &c->vp->program.Base, PROG_PRINT_DEBUG,
 			       GL_TRUE);
@@ -2048,7 +2048,7 @@ void brw_vs_emit(struct brw_vs_compile *c )
 
    brw_optimize(p);
 
-   if (INTEL_DEBUG & DEBUG_VS) {
+   if (unlikely(INTEL_DEBUG & DEBUG_VS)) {
       int i;
 
       printf("vs-native:\n");
diff --git a/src/mesa/drivers/dri/i965/brw_vs_state.c b/src/mesa/drivers/dri/i965/brw_vs_state.c
index ebae94269f..be92313861 100644
--- a/src/mesa/drivers/dri/i965/brw_vs_state.c
+++ b/src/mesa/drivers/dri/i965/brw_vs_state.c
@@ -154,7 +154,7 @@ vs_unit_create_from_key(struct brw_context *brw, struct brw_vs_unit_key *key)
     */
    vs.vs5.sampler_count = 0;
 
-   if (INTEL_DEBUG & DEBUG_STATS)
+   if (unlikely(INTEL_DEBUG & DEBUG_STATS))
       vs.thread4.stats_enable = 1;
 
    /* Vertex program always enabled:
diff --git a/src/mesa/drivers/dri/i965/brw_wm.c b/src/mesa/drivers/dri/i965/brw_wm.c
index 7f3ba5f058..a6d2a2377f 100644
--- a/src/mesa/drivers/dri/i965/brw_wm.c
+++ b/src/mesa/drivers/dri/i965/brw_wm.c
@@ -198,7 +198,7 @@ static void do_wm_prog( struct brw_context *brw,
       c->prog_data.total_scratch = 0;
    }
 
-   if (INTEL_DEBUG & DEBUG_WM)
+   if (unlikely(INTEL_DEBUG & DEBUG_WM))
       fprintf(stderr, "\n");
 
    /* get the program
diff --git a/src/mesa/drivers/dri/i965/brw_wm_emit.c b/src/mesa/drivers/dri/i965/brw_wm_emit.c
index d06c49fd5b..272476219a 100644
--- a/src/mesa/drivers/dri/i965/brw_wm_emit.c
+++ b/src/mesa/drivers/dri/i965/brw_wm_emit.c
@@ -1920,7 +1920,7 @@ void brw_wm_emit( struct brw_wm_compile *c )
 	brw_remove_grf_to_mrf_moves(p);
    }
 
-   if (INTEL_DEBUG & DEBUG_WM) {
+   if (unlikely(INTEL_DEBUG & DEBUG_WM)) {
       int i;
 
      printf("wm-native:\n");
diff --git a/src/mesa/drivers/dri/i965/brw_wm_fp.c b/src/mesa/drivers/dri/i965/brw_wm_fp.c
index 15a238cda6..2cae698880 100644
--- a/src/mesa/drivers/dri/i965/brw_wm_fp.c
+++ b/src/mesa/drivers/dri/i965/brw_wm_fp.c
@@ -663,7 +663,7 @@ static void precalc_tex( struct brw_wm_compile *c,
 			 const struct prog_instruction *inst )
 {
    struct prog_src_register coord;
-   struct prog_dst_register tmpcoord;
+   struct prog_dst_register tmpcoord = { 0 };
    const GLuint unit = c->fp->program.Base.SamplerUnits[inst->TexSrcUnit];
 
    assert(unit < BRW_MAX_TEX_UNIT);
@@ -963,7 +963,7 @@ static void emit_render_target_writes( struct brw_wm_compile *c )
    struct prog_src_register outcolor;
    GLuint i;
 
-   struct prog_instruction *inst, *last_inst;
+   struct prog_instruction *inst, *last_inst = NULL;
 
    /* The inst->Aux field is used for FB write target and the EOT marker */
 
@@ -1058,7 +1058,7 @@ void brw_wm_pass_fp( struct brw_wm_compile *c )
    struct brw_fragment_program *fp = c->fp;
    GLuint insn;
 
-   if (INTEL_DEBUG & DEBUG_WM) {
+   if (unlikely(INTEL_DEBUG & DEBUG_WM)) {
       printf("pre-fp:\n");
       _mesa_fprint_program_opt(stdout, &fp->program.Base, PROG_PRINT_DEBUG,
 			       GL_TRUE);
@@ -1174,7 +1174,7 @@ void brw_wm_pass_fp( struct brw_wm_compile *c )
       }
    }
 
-   if (INTEL_DEBUG & DEBUG_WM) {
+   if (unlikely(INTEL_DEBUG & DEBUG_WM)) {
       printf("pass_fp:\n");
       print_insns( c->prog_instructions, c->nr_fp_insns );
       printf("\n");
diff --git a/src/mesa/drivers/dri/i965/brw_wm_glsl.c b/src/mesa/drivers/dri/i965/brw_wm_glsl.c
index d325f85ce0..7fe8ab1f33 100644
--- a/src/mesa/drivers/dri/i965/brw_wm_glsl.c
+++ b/src/mesa/drivers/dri/i965/brw_wm_glsl.c
@@ -19,7 +19,7 @@ GLboolean brw_wm_is_glsl(const struct gl_fragment_program *fp)
 {
     int i;
 
-    if (INTEL_DEBUG & DEBUG_GLSL_FORCE)
+    if (unlikely(INTEL_DEBUG & DEBUG_GLSL_FORCE))
        return GL_TRUE;
 
     for (i = 0; i < fp->Base.NumInstructions; i++) {
@@ -1002,7 +1002,7 @@ static void brw_wm_emit_glsl(struct brw_context *brw, struct brw_wm_compile *c)
     }
     post_wm_emit(c);
 
-    if (INTEL_DEBUG & DEBUG_WM) {
+    if (unlikely(INTEL_DEBUG & DEBUG_WM)) {
       printf("wm-native:\n");
       for (i = 0; i < p->nr_insn; i++)
 	 brw_disasm(stdout, &p->store[i], intel->gen);
@@ -1016,7 +1016,7 @@ static void brw_wm_emit_glsl(struct brw_context *brw, struct brw_wm_compile *c)
  */
 void brw_wm_glsl_emit(struct brw_context *brw, struct brw_wm_compile *c)
 {
-    if (INTEL_DEBUG & DEBUG_WM) {
+    if (unlikely(INTEL_DEBUG & DEBUG_WM)) {
         printf("brw_wm_glsl_emit:\n");
     }
 
@@ -1026,7 +1026,7 @@ void brw_wm_glsl_emit(struct brw_context *brw, struct brw_wm_compile *c)
     /* actual code generation */
     brw_wm_emit_glsl(brw, c);
 
-    if (INTEL_DEBUG & DEBUG_WM) {
+    if (unlikely(INTEL_DEBUG & DEBUG_WM)) {
         brw_wm_print_program(c, "brw_wm_glsl_emit done");
     }
 
diff --git a/src/mesa/drivers/dri/i965/brw_wm_pass0.c b/src/mesa/drivers/dri/i965/brw_wm_pass0.c
index d6aa9f957a..83152526b3 100644
--- a/src/mesa/drivers/dri/i965/brw_wm_pass0.c
+++ b/src/mesa/drivers/dri/i965/brw_wm_pass0.c
@@ -440,7 +440,7 @@ void brw_wm_pass0( struct brw_wm_compile *c )
       }
    }
  
-   if (INTEL_DEBUG & DEBUG_WM) {
+   if (unlikely(INTEL_DEBUG & DEBUG_WM)) {
       brw_wm_print_program(c, "pass0");
    }
 }
diff --git a/src/mesa/drivers/dri/i965/brw_wm_pass1.c b/src/mesa/drivers/dri/i965/brw_wm_pass1.c
index 962515a99e..3a2874b6dd 100644
--- a/src/mesa/drivers/dri/i965/brw_wm_pass1.c
+++ b/src/mesa/drivers/dri/i965/brw_wm_pass1.c
@@ -291,7 +291,7 @@ void brw_wm_pass1( struct brw_wm_compile *c )
       track_arg(c, inst, 2, read2);
    }
 
-   if (INTEL_DEBUG & DEBUG_WM) {
+   if (unlikely(INTEL_DEBUG & DEBUG_WM)) {
       brw_wm_print_program(c, "pass1");
    }
 }
diff --git a/src/mesa/drivers/dri/i965/brw_wm_pass2.c b/src/mesa/drivers/dri/i965/brw_wm_pass2.c
index 54acb3038b..44e3953814 100644
--- a/src/mesa/drivers/dri/i965/brw_wm_pass2.c
+++ b/src/mesa/drivers/dri/i965/brw_wm_pass2.c
@@ -331,13 +331,13 @@ void brw_wm_pass2( struct brw_wm_compile *c )
       }
    }
 
-   if (INTEL_DEBUG & DEBUG_WM) {
+   if (unlikely(INTEL_DEBUG & DEBUG_WM)) {
       brw_wm_print_program(c, "pass2");
    }
 
    c->state = PASS2_DONE;
 
-   if (INTEL_DEBUG & DEBUG_WM) {
+   if (unlikely(INTEL_DEBUG & DEBUG_WM)) {
        brw_wm_print_program(c, "pass2/done");
    }
 }
diff --git a/src/mesa/drivers/dri/i965/brw_wm_state.c b/src/mesa/drivers/dri/i965/brw_wm_state.c
index 9a27b93710..76de7b7b6f 100644
--- a/src/mesa/drivers/dri/i965/brw_wm_state.c
+++ b/src/mesa/drivers/dri/i965/brw_wm_state.c
@@ -249,7 +249,7 @@ wm_unit_create_from_key(struct brw_context *brw, struct brw_wm_unit_key *key,
 
    wm.wm5.line_stipple = key->line_stipple;
 
-   if (INTEL_DEBUG & DEBUG_STATS || key->stats_wm)
+   if (unlikely(INTEL_DEBUG & DEBUG_STATS) || key->stats_wm)
       wm.wm4.stats_enable = 1;
 
    bo = brw_upload_cache(&brw->cache, BRW_WM_UNIT,
diff --git a/src/mesa/drivers/dri/i965/gen6_wm_state.c b/src/mesa/drivers/dri/i965/gen6_wm_state.c
index 36d4ab93ba..ea5418bacf 100644
--- a/src/mesa/drivers/dri/i965/gen6_wm_state.c
+++ b/src/mesa/drivers/dri/i965/gen6_wm_state.c
@@ -73,7 +73,7 @@ prepare_wm_constants(struct brw_context *brw)
 const struct brw_tracked_state gen6_wm_constants = {
    .dirty = {
       .mesa  = _NEW_PROGRAM_CONSTANTS,
-      .brw   = 0,
+      .brw   = BRW_NEW_FRAGMENT_PROGRAM,
       .cache = 0,
    },
    .prepare = prepare_wm_constants,
diff --git a/src/mesa/drivers/dri/i965/intel_structs.h b/src/mesa/drivers/dri/i965/intel_structs.h
deleted file mode 100644
index 522e3bd92c..0000000000
--- a/src/mesa/drivers/dri/i965/intel_structs.h
+++ /dev/null
@@ -1,132 +0,0 @@
-#ifndef INTEL_STRUCTS_H
-#define INTEL_STRUCTS_H
-
-struct br0 {
-   GLuint length:8;
-   GLuint pad0:3;
-   GLuint dst_tiled:1;
-   GLuint pad1:8;
-   GLuint write_rgb:1;
-   GLuint write_alpha:1;
-   GLuint opcode:7;
-   GLuint client:3;
-};
-
-   
-struct br13 {
-   GLint dest_pitch:16;
-   GLuint rop:8;
-   GLuint color_depth:2;
-   GLuint pad1:3;
-   GLuint mono_source_transparency:1;
-   GLuint clipping_enable:1;
-   GLuint pad0:1;
-};
-
-
-
-/* This is an attempt to move some of the 2D interaction in this
- * driver to using structs for packets rather than a bunch of #defines
- * and dwords.
- */
-struct xy_color_blit {
-   struct br0 br0;
-   struct br13 br13;
-
-   struct {
-      GLuint dest_x1:16;
-      GLuint dest_y1:16;
-   } dw2;
-
-   struct {
-      GLuint dest_x2:16;
-      GLuint dest_y2:16;
-   } dw3;
-   
-   GLuint dest_base_addr;
-   GLuint color;
-};
-
-struct xy_src_copy_blit {
-   struct br0 br0;
-   struct br13 br13;
-
-   struct {
-      GLuint dest_x1:16;
-      GLuint dest_y1:16;
-   } dw2;
-
-   struct {
-      GLuint dest_x2:16;
-      GLuint dest_y2:16;
-   } dw3;
-   
-   GLuint dest_base_addr;
-
-   struct {
-      GLuint src_x1:16;
-      GLuint src_y1:16;
-   } dw5;
-
-   struct {
-      GLint src_pitch:16;
-      GLuint pad:16;
-   } dw6;
-   
-   GLuint src_base_addr;
-};
-
-struct xy_setup_blit {
-   struct br0 br0;
-   struct br13 br13;
-
-   struct {
-      GLuint clip_x1:16;
-      GLuint clip_y1:16;
-   } dw2;
-
-   struct {
-      GLuint clip_x2:16;
-      GLuint clip_y2:16;
-   } dw3;
-      
-   GLuint dest_base_addr;
-   GLuint background_color;
-   GLuint foreground_color;
-   GLuint pattern_base_addr;
-};
-
-
-struct xy_text_immediate_blit {
-   struct {
-      GLuint length:8;
-      GLuint pad2:3;
-      GLuint dst_tiled:1;
-      GLuint pad1:4;
-      GLuint byte_packed:1;
-      GLuint pad0:5;
-      GLuint opcode:7;
-      GLuint client:3;
-   } dw0;
-
-   struct {
-      GLuint dest_x1:16;
-      GLuint dest_y1:16;
-   } dw1;
-
-   struct {
-      GLuint dest_x2:16;
-      GLuint dest_y2:16;
-   } dw2;   
-
-   /* Src bitmap data follows as inline dwords.
-    */
-};
-
-
-#define CLIENT_2D 0x2
-#define OPCODE_XY_SETUP_BLT 0x1
-#define OPCODE_XY_COLOR_BLT 0x50
-#define OPCODE_XY_TEXT_IMMEDIATE_BLT 0x31
-
-#endif
diff --git a/src/mesa/drivers/dri/intel/intel_batchbuffer.c b/src/mesa/drivers/dri/intel/intel_batchbuffer.c
index 9b39823917..4b498f8c5b 100644
--- a/src/mesa/drivers/dri/intel/intel_batchbuffer.c
+++ b/src/mesa/drivers/dri/intel/intel_batchbuffer.c
@@ -44,7 +44,9 @@ intel_batchbuffer_reset(struct intel_batchbuffer *batch)
 
    batch->buf = drm_intel_bo_alloc(intel->bufmgr, "batchbuffer",
 				   intel->maxBatchSize, 4096);
-   batch->map = batch->buffer;
+   drm_intel_gem_bo_map_gtt(batch->buf);
+   batch->map = batch->buf->virtual;
+
    batch->size = intel->maxBatchSize;
    batch->ptr = batch->map;
    batch->reserved_space = BATCH_RESERVED;
@@ -58,7 +60,6 @@ intel_batchbuffer_alloc(struct intel_context *intel)
    struct intel_batchbuffer *batch = calloc(sizeof(*batch), 1);
 
    batch->intel = intel;
-   batch->buffer = malloc(intel->maxBatchSize);
    intel_batchbuffer_reset(batch);
 
    return batch;
@@ -67,8 +68,11 @@ intel_batchbuffer_alloc(struct intel_context *intel)
 void
 intel_batchbuffer_free(struct intel_batchbuffer *batch)
 {
-   free (batch->buffer);
-   drm_intel_bo_unreference(batch->buf);
+   if (batch->map) {
+      drm_intel_gem_bo_unmap_gtt(batch->buf);
+      batch->map = NULL;
+   }
+   dri_bo_unreference(batch->buf);
    batch->buf = NULL;
    free(batch);
 }
@@ -84,13 +88,7 @@ do_flush_locked(struct intel_batchbuffer *batch, GLuint used)
    int ret = 0;
    int x_off = 0, y_off = 0;
 
-   drm_intel_bo_subdata(batch->buf, 0, used, batch->buffer);
-   if (batch->state_batch_offset != batch->size) {
-      drm_intel_bo_subdata(batch->buf,
-			   batch->state_batch_offset,
-			   batch->size - batch->state_batch_offset,
-			   batch->buffer + batch->state_batch_offset);
-   }
+   drm_intel_gem_bo_unmap_gtt(batch->buf);
 
    batch->ptr = NULL;
 
@@ -99,7 +97,7 @@ do_flush_locked(struct intel_batchbuffer *batch, GLuint used)
 			(x_off & 0xffff) | (y_off << 16));
    }
 
-   if (INTEL_DEBUG & DEBUG_BATCH) {
+   if (unlikely(INTEL_DEBUG & DEBUG_BATCH)) {
       drm_intel_bo_map(batch->buf, GL_FALSE);
       intel_decode(batch->buf->virtual, used / 4, batch->buf->offset,
 		   intel->intelScreen->deviceID, GL_TRUE);
@@ -130,7 +128,7 @@ _intel_batchbuffer_flush(struct intel_batchbuffer *batch, const char *file,
    if (used == 0)
       return;
 
-   if (INTEL_DEBUG & DEBUG_BATCH)
+   if (unlikely(INTEL_DEBUG & DEBUG_BATCH))
       fprintf(stderr, "%s:%d: Batchbuffer flush with %db used\n", file, line,
 	      used);
 
@@ -174,7 +172,7 @@ _intel_batchbuffer_flush(struct intel_batchbuffer *batch, const char *file,
 
    do_flush_locked(batch, used);
 
-   if (INTEL_DEBUG & DEBUG_SYNC) {
+   if (unlikely(INTEL_DEBUG & DEBUG_SYNC)) {
       fprintf(stderr, "waiting for idle\n");
       drm_intel_bo_map(batch->buf, GL_TRUE);
       drm_intel_bo_unmap(batch->buf);
diff --git a/src/mesa/drivers/dri/intel/intel_batchbuffer.h b/src/mesa/drivers/dri/intel/intel_batchbuffer.h
index ae53f45511..428c027c2f 100644
--- a/src/mesa/drivers/dri/intel/intel_batchbuffer.h
+++ b/src/mesa/drivers/dri/intel/intel_batchbuffer.h
@@ -17,8 +17,6 @@ struct intel_batchbuffer
 
    drm_intel_bo *buf;
 
-   GLubyte *buffer;
-
    GLubyte *map;
    GLubyte *ptr;
 
diff --git a/src/mesa/drivers/dri/intel/intel_clear.c b/src/mesa/drivers/dri/intel/intel_clear.c
index d7814635b7..fa451f0045 100644
--- a/src/mesa/drivers/dri/intel/intel_clear.c
+++ b/src/mesa/drivers/dri/intel/intel_clear.c
@@ -58,6 +58,21 @@ static const char *buffer_names[] = {
    [BUFFER_COLOR7] = "color7",
 };
 
+static void
+debug_mask(const char *name, GLbitfield mask)
+{
+   GLuint i;
+
+   if (unlikely(INTEL_DEBUG & DEBUG_BLIT)) {
+      DBG("%s clear:", name);
+      for (i = 0; i < BUFFER_COUNT; i++) {
+	 if (mask & (1 << i))
+	    DBG(" %s", buffer_names[i]);
+      }
+      DBG("\n");
+   }
+}
+
 /**
  * Called by ctx->Driver.Clear.
  */
@@ -70,7 +85,6 @@ intelClear(struct gl_context *ctx, GLbitfield mask)
    GLbitfield blit_mask = 0;
    GLbitfield swrast_mask = 0;
    struct gl_framebuffer *fb = ctx->DrawBuffer;
-   GLuint i;
 
    if (mask & (BUFFER_BIT_FRONT_LEFT | BUFFER_BIT_FRONT_RIGHT)) {
       intel->front_buffer_dirty = GL_TRUE;
@@ -162,39 +176,17 @@ intelClear(struct gl_context *ctx, GLbitfield mask)
    }
 
    if (blit_mask) {
-      if (INTEL_DEBUG & DEBUG_BLIT) {
-	 DBG("blit clear:");
-	 for (i = 0; i < BUFFER_COUNT; i++) {
-	    if (blit_mask & (1 << i))
-	       DBG(" %s", buffer_names[i]);
-	 }
-	 DBG("\n");
-      }
+      debug_mask("blit", blit_mask);
       intelClearWithBlit(ctx, blit_mask);
    }
 
    if (tri_mask) {
-      if (INTEL_DEBUG & DEBUG_BLIT) {
-	 DBG("tri clear:");
-	 for (i = 0; i < BUFFER_COUNT; i++) {
-	    if (tri_mask & (1 << i))
-	       DBG(" %s", buffer_names[i]);
-	 }
-	 DBG("\n");
-      }
-
+      debug_mask("tri", tri_mask);
       _mesa_meta_Clear(&intel->ctx, tri_mask);
    }
 
    if (swrast_mask) {
-      if (INTEL_DEBUG & DEBUG_BLIT) {
-	 DBG("swrast clear:");
-	 for (i = 0; i < BUFFER_COUNT; i++) {
-	    if (swrast_mask & (1 << i))
-	       DBG(" %s", buffer_names[i]);
-	 }
-	 DBG("\n");
-      }
+      debug_mask("swrast", swrast_mask);
       _swrast_Clear(ctx, swrast_mask);
    }
 }
diff --git a/src/mesa/drivers/dri/intel/intel_context.c b/src/mesa/drivers/dri/intel/intel_context.c
index 7ace50bde9..de4bcde20b 100644
--- a/src/mesa/drivers/dri/intel/intel_context.c
+++ b/src/mesa/drivers/dri/intel/intel_context.c
@@ -249,7 +249,7 @@ intel_update_renderbuffers(__DRIcontext *context, __DRIdrawable *drawable)
     * thus ignore the invalidate. */
    drawable->lastStamp = drawable->dri2.stamp;
 
-   if (INTEL_DEBUG & DEBUG_DRI)
+   if (unlikely(INTEL_DEBUG & DEBUG_DRI))
       fprintf(stderr, "enter %s, drawable %p\n", __func__, drawable);
 
    screen = intel->intelScreen->driScrnPriv;
@@ -378,14 +378,14 @@ intel_update_renderbuffers(__DRIcontext *context, __DRIdrawable *drawable)
        if (rb->region && rb->region->name == buffers[i].name)
 	     continue;
 
-       if (INTEL_DEBUG & DEBUG_DRI)
+       if (unlikely(INTEL_DEBUG & DEBUG_DRI))
 	  fprintf(stderr,
 		  "attaching buffer %d, at %d, cpp %d, pitch %d\n",
 		  buffers[i].name, buffers[i].attachment,
 		  buffers[i].cpp, buffers[i].pitch);
        
        if (buffers[i].attachment == __DRI_BUFFER_STENCIL && depth_region) {
-	  if (INTEL_DEBUG & DEBUG_DRI)
+	  if (unlikely(INTEL_DEBUG & DEBUG_DRI))
 	     fprintf(stderr, "(reusing depth buffer as stencil)\n");
 	  intel_region_reference(&region, depth_region);
        }
diff --git a/src/mesa/drivers/dri/intel/intel_context.h b/src/mesa/drivers/dri/intel/intel_context.h
index 46d10d74ba..9d5139c000 100644
--- a/src/mesa/drivers/dri/intel/intel_context.h
+++ b/src/mesa/drivers/dri/intel/intel_context.h
@@ -98,6 +98,16 @@ extern void intelFallback(struct intel_context *intel, GLbitfield bit,
 
 #define INTEL_MAX_FIXUP 64
 
+#ifndef likely
+#ifdef __GNUC__
+#define likely(expr) (__builtin_expect(expr, 1))
+#define unlikely(expr) (__builtin_expect(expr, 0))
+#else
+#define likely(expr) (expr)
+#define unlikely(expr) (expr)
+#endif
+#endif
+
 struct intel_sync_object {
    struct gl_sync_object Base;
 
@@ -180,9 +190,6 @@ struct intel_context
    } prim;
 
    GLuint stats_wm;
-   GLboolean locked;
-   char *prevLockFile;
-   int prevLockLine;
 
    /* Offsets of fields within the current vertex:
     */
@@ -359,10 +366,15 @@ extern int INTEL_DEBUG;
 #define DEBUG_CLIP      0x8000000
 
 #define DBG(...) do {						\
-	if (INTEL_DEBUG & FILE_DEBUG_FLAG)			\
+	if (unlikely(INTEL_DEBUG & FILE_DEBUG_FLAG))		\
 		printf(__VA_ARGS__);			\
 } while(0)
 
+#define fallback_debug(...) do {				\
+	if (unlikely(INTEL_DEBUG & DEBUG_FALLBACKS))		\
+		printf(__VA_ARGS__);				\
+} while(0)
+
 #define PCI_CHIP_845_G			0x2562
 #define PCI_CHIP_I830_M			0x3577
 #define PCI_CHIP_I855_GM		0x3582
diff --git a/src/mesa/drivers/dri/intel/intel_pixel.c b/src/mesa/drivers/dri/intel/intel_pixel.c
index 60583ef4c0..d5c35775ce 100644
--- a/src/mesa/drivers/dri/intel/intel_pixel.c
+++ b/src/mesa/drivers/dri/intel/intel_pixel.c
@@ -147,10 +147,9 @@ intel_check_blit_format(struct intel_region * region,
       return GL_TRUE;
    }
 
-   if (INTEL_DEBUG & DEBUG_PIXEL)
-      fprintf(stderr, "%s: bad format for blit (cpp %d, type %s format %s)\n",
-              __FUNCTION__, region->cpp,
-              _mesa_lookup_enum_by_nr(type), _mesa_lookup_enum_by_nr(format));
+   DBG("%s: bad format for blit (cpp %d, type %s format %s)\n",
+       __FUNCTION__, region->cpp,
+       _mesa_lookup_enum_by_nr(type), _mesa_lookup_enum_by_nr(format));
 
    return GL_FALSE;
 }
diff --git a/src/mesa/drivers/dri/intel/intel_pixel_bitmap.c b/src/mesa/drivers/dri/intel/intel_pixel_bitmap.c
index 63fb4b37b1..e7356a6da0 100644
--- a/src/mesa/drivers/dri/intel/intel_pixel_bitmap.c
+++ b/src/mesa/drivers/dri/intel/intel_pixel_bitmap.c
@@ -113,9 +113,8 @@ static GLuint get_bitmap_rect(GLsizei width, GLsizei height,
    GLint incr;
    GLuint count = 0;
 
-   if (INTEL_DEBUG & DEBUG_PIXEL)
-      printf("%s %d,%d %dx%d bitmap %dx%d skip %d src_offset %d mask %d\n",
-		   __FUNCTION__, x,y,w,h,width,height,unpack->SkipPixels, src_offset, mask);
+   DBG("%s %d,%d %dx%d bitmap %dx%d skip %d src_offset %d mask %d\n",
+       __FUNCTION__, x,y,w,h,width,height,unpack->SkipPixels, src_offset, mask);
 
    if (invert) {
       first = h-1;
@@ -285,7 +284,7 @@ do_blit_bitmap( struct gl_context *ctx,
    }
 out:
 
-   if (INTEL_DEBUG & DEBUG_SYNC)
+   if (unlikely(INTEL_DEBUG & DEBUG_SYNC))
       intel_batchbuffer_flush(intel->batch);
 
    if (_mesa_is_bufferobj(unpack->BufferObj)) {
@@ -299,6 +298,7 @@ out:
    return GL_TRUE;
 }
 
+
 /* There are a large number of possible ways to implement bitmap on
  * this hardware, most of them have some sort of drawback.  Here are a
  * few that spring to mind:
diff --git a/src/mesa/drivers/dri/intel/intel_pixel_copy.c b/src/mesa/drivers/dri/intel/intel_pixel_copy.c
index c6b36ed429..a7ca780e94 100644
--- a/src/mesa/drivers/dri/intel/intel_pixel_copy.c
+++ b/src/mesa/drivers/dri/intel/intel_pixel_copy.c
@@ -119,8 +119,7 @@ do_blit_copypixels(struct gl_context * ctx,
    GLboolean flip = GL_FALSE;
 
    if (type == GL_DEPTH || type == GL_STENCIL) {
-      if (INTEL_DEBUG & DEBUG_FALLBACKS)
-	 fprintf(stderr, "glCopyPixels() fallback: GL_DEPTH || GL_STENCIL\n");
+      fallback_debug("glCopyPixels() fallback: GL_DEPTH || GL_STENCIL\n");
       return GL_FALSE;
    }
 
@@ -203,8 +202,7 @@ intelCopyPixels(struct gl_context * ctx,
                 GLsizei width, GLsizei height,
                 GLint destx, GLint desty, GLenum type)
 {
-   if (INTEL_DEBUG & DEBUG_PIXEL)
-      fprintf(stderr, "%s\n", __FUNCTION__);
+   DBG("%s\n", __FUNCTION__);
 
    if (do_blit_copypixels(ctx, srcx, srcy, width, height, destx, desty, type))
       return;
diff --git a/src/mesa/drivers/dri/intel/intel_pixel_read.c b/src/mesa/drivers/dri/intel/intel_pixel_read.c
index b249f9a5a0..54da29236d 100644
--- a/src/mesa/drivers/dri/intel/intel_pixel_read.c
+++ b/src/mesa/drivers/dri/intel/intel_pixel_read.c
@@ -42,6 +42,8 @@
 #include "intel_pixel.h"
 #include "intel_buffer_objects.h"
 
+#define FILE_DEBUG_FLAG DEBUG_PIXEL
+
 /* For many applications, the new ability to pull the source buffers
  * back out of the GTT and then do the packing/conversion operations
  * in software will be as much of an improvement as trying to get the
@@ -79,8 +81,7 @@ do_blit_readpixels(struct gl_context * ctx,
    GLboolean all;
    GLint dst_x, dst_y;
 
-   if (INTEL_DEBUG & DEBUG_PIXEL)
-      printf("%s\n", __FUNCTION__);
+   DBG("%s\n", __FUNCTION__);
 
    if (!src)
       return GL_FALSE;
@@ -88,22 +89,19 @@ do_blit_readpixels(struct gl_context * ctx,
    if (!_mesa_is_bufferobj(pack->BufferObj)) {
       /* PBO only for now:
        */
-      if (INTEL_DEBUG & DEBUG_PIXEL)
-         printf("%s - not PBO\n", __FUNCTION__);
+      DBG("%s - not PBO\n", __FUNCTION__);
       return GL_FALSE;
    }
 
 
    if (ctx->_ImageTransferState ||
        !intel_check_blit_format(src, format, type)) {
-      if (INTEL_DEBUG & DEBUG_PIXEL)
-         printf("%s - bad format for blit\n", __FUNCTION__);
+      DBG("%s - bad format for blit\n", __FUNCTION__);
       return GL_FALSE;
    }
 
    if (pack->Alignment != 1 || pack->SwapBytes || pack->LsbFirst) {
-      if (INTEL_DEBUG & DEBUG_PIXEL)
-         printf("%s: bad packing params\n", __FUNCTION__);
+      DBG("%s: bad packing params\n", __FUNCTION__);
       return GL_FALSE;
    }
 
@@ -113,8 +111,7 @@ do_blit_readpixels(struct gl_context * ctx,
       rowLength = width;
 
    if (pack->Invert) {
-      if (INTEL_DEBUG & DEBUG_PIXEL)
-         printf("%s: MESA_PACK_INVERT not done yet\n", __FUNCTION__);
+      DBG("%s: MESA_PACK_INVERT not done yet\n", __FUNCTION__);
       return GL_FALSE;
    }
    else {
@@ -158,8 +155,7 @@ do_blit_readpixels(struct gl_context * ctx,
       return GL_FALSE;
    }
 
-   if (INTEL_DEBUG & DEBUG_PIXEL)
-      printf("%s - DONE\n", __FUNCTION__);
+   DBG("%s - DONE\n", __FUNCTION__);
 
    return GL_TRUE;
 }
@@ -173,8 +169,7 @@ intelReadPixels(struct gl_context * ctx,
    struct intel_context *intel = intel_context(ctx);
    GLboolean dirty;
 
-   if (INTEL_DEBUG & DEBUG_PIXEL)
-      fprintf(stderr, "%s\n", __FUNCTION__);
+   DBG("%s\n", __FUNCTION__);
 
    intel_flush(ctx);
 
@@ -188,8 +183,7 @@ intelReadPixels(struct gl_context * ctx,
        (ctx, x, y, width, height, format, type, pack, pixels))
       return;
 
-   if (INTEL_DEBUG & DEBUG_PIXEL)
-      printf("%s: fallback to swrast\n", __FUNCTION__);
+   fallback_debug("%s: fallback to swrast\n", __FUNCTION__);
 
    /* Update Mesa state before calling down into _swrast_ReadPixels, as
     * the spans code requires the computed buffer states to be up to date,
diff --git a/src/mesa/drivers/dri/intel/intel_tex.c b/src/mesa/drivers/dri/intel/intel_tex.c
index 3d9a2549db..2c21ea0576 100644
--- a/src/mesa/drivers/dri/intel/intel_tex.c
+++ b/src/mesa/drivers/dri/intel/intel_tex.c
@@ -61,88 +61,6 @@ intelFreeTextureImageData(struct gl_context * ctx, struct gl_texture_image *texI
    }
 }
 
-
-/* The system memcpy (at least on ubuntu 5.10) has problems copying
- * to agp (writecombined) memory from a source which isn't 64-byte
- * aligned - there is a 4x performance falloff.
- *
- * The x86 __memcpy is immune to this but is slightly slower
- * (10%-ish) than the system memcpy.
- *
- * The sse_memcpy seems to have a slight cliff at 64/32 bytes, but
- * isn't much faster than x86_memcpy for agp copies.
- * 
- * TODO: switch dynamically.
- */
-static void *
-do_memcpy(void *dest, const void *src, size_t n)
-{
-   if ((((unsigned long) src) & 63) || (((unsigned long) dest) & 63)) {
-      return __memcpy(dest, src, n);
-   }
-   else
-      return memcpy(dest, src, n);
-}
-
-
-#if DO_DEBUG && !defined(__ia64__)
-
-#ifndef __x86_64__
-static unsigned
-fastrdtsc(void)
-{
-   unsigned eax;
-   __asm__ volatile ("\t"
-                     "pushl  %%ebx\n\t"
-                     "cpuid\n\t" ".byte 0x0f, 0x31\n\t"
-                     "popl %%ebx\n":"=a" (eax)
-                     :"0"(0)
-                     :"ecx", "edx", "cc");
-
-   return eax;
-}
-#else
-static unsigned
-fastrdtsc(void)
-{
-   unsigned eax;
-   __asm__ volatile ("\t" "cpuid\n\t" ".byte 0x0f, 0x31\n\t":"=a" (eax)
-                     :"0"(0)
-                     :"ecx", "edx", "ebx", "cc");
-
-   return eax;
-}
-#endif
-
-static unsigned
-time_diff(unsigned t, unsigned t2)
-{
-   return ((t < t2) ? t2 - t : 0xFFFFFFFFU - (t - t2 - 1));
-}
-
-
-static void *
-timed_memcpy(void *dest, const void *src, size_t n)
-{
-   void *ret;
-   unsigned t1, t2;
-   double rate;
-
-   if ((((unsigned) src) & 63) || (((unsigned) dest) & 63))
-      printf("Warning - non-aligned texture copy!\n");
-
-   t1 = fastrdtsc();
-   ret = do_memcpy(dest, src, n);
-   t2 = fastrdtsc();
-
-   rate = time_diff(t1, t2);
-   rate /= (double) n;
-   printf("timed_memcpy: %u %u --> %f clocks/byte\n", t1, t2, rate);
-   return ret;
-}
-#endif /* DO_DEBUG */
-
-
 /**
  * Called via ctx->Driver.GenerateMipmap()
  * This is basically a wrapper for _mesa_meta_GenerateMipmap() which checks
@@ -158,8 +76,7 @@ intelGenerateMipmap(struct gl_context *ctx, GLenum target,
       struct intel_context *intel = intel_context(ctx);
       struct intel_texture_object *intelObj = intel_texture_object(texObj);
 
-      if (INTEL_DEBUG & DEBUG_FALLBACKS)
-	 fprintf(stderr, "%s - fallback to swrast\n", __FUNCTION__);
+      fallback_debug("%s - fallback to swrast\n", __FUNCTION__);
 
       intel_tex_map_level_images(intel, intelObj, texObj->BaseLevel);
       _mesa_generate_mipmap(ctx, target, texObj);
@@ -203,11 +120,4 @@ intelInitTextureFuncs(struct dd_function_table *functions)
    functions->NewTextureImage = intelNewTextureImage;
    functions->DeleteTexture = intelDeleteTextureObject;
    functions->FreeTexImageData = intelFreeTextureImageData;
-
-#if DO_DEBUG && !defined(__ia64__)
-   if (INTEL_DEBUG & DEBUG_BUFMGR)
-      functions->TextureMemCpy = timed_memcpy;
-   else
-#endif
-      functions->TextureMemCpy = do_memcpy;
 }
diff --git a/src/mesa/drivers/dri/intel/intel_tex_copy.c b/src/mesa/drivers/dri/intel/intel_tex_copy.c
index 2d046fd52d..284ba19e8a 100644
--- a/src/mesa/drivers/dri/intel/intel_tex_copy.c
+++ b/src/mesa/drivers/dri/intel/intel_tex_copy.c
@@ -105,16 +105,15 @@ do_copy_texsubimage(struct intel_context *intel,
    const struct intel_region *src = get_teximage_source(intel, internalFormat);
 
    if (!intelImage->mt || !src || !src->buffer) {
-      if (INTEL_DEBUG & DEBUG_FALLBACKS)
+      if (unlikely(INTEL_DEBUG & DEBUG_FALLBACKS))
 	 fprintf(stderr, "%s fail %p %p (0x%08x)\n",
 		 __FUNCTION__, intelImage->mt, src, internalFormat);
       return GL_FALSE;
    }
 
    if (intelImage->mt->cpp != src->cpp) {
-      if (INTEL_DEBUG & DEBUG_FALLBACKS)
-	 fprintf(stderr, "%s fail %d vs %d cpp\n",
-		 __FUNCTION__, intelImage->mt->cpp, src->cpp);
+      fallback_debug("%s fail %d vs %d cpp\n",
+		     __FUNCTION__, intelImage->mt->cpp, src->cpp);
       return GL_FALSE;
    }
 
@@ -212,8 +211,7 @@ intelCopyTexImage1D(struct gl_context * ctx, GLenum target, GLint level,
    return;
 
  fail:
-   if (INTEL_DEBUG & DEBUG_FALLBACKS)
-      fprintf(stderr, "%s - fallback to swrast\n", __FUNCTION__);
+   fallback_debug("%s - fallback to swrast\n", __FUNCTION__);
    _mesa_meta_CopyTexImage1D(ctx, target, level, internalFormat, x, y,
                              width, border);
 }
@@ -261,8 +259,7 @@ intelCopyTexImage2D(struct gl_context * ctx, GLenum target, GLint level,
    return;
 
  fail:
-   if (INTEL_DEBUG & DEBUG_FALLBACKS)
-      fprintf(stderr, "%s - fallback to swrast\n", __FUNCTION__);
+   fallback_debug("%s - fallback to swrast\n", __FUNCTION__);
    _mesa_meta_CopyTexImage2D(ctx, target, level, internalFormat, x, y,
                              width, height, border);
 }
@@ -287,8 +284,7 @@ intelCopyTexSubImage1D(struct gl_context * ctx, GLenum target, GLint level,
    if (!do_copy_texsubimage(intel_context(ctx), target,
                             intel_texture_image(texImage),
                             internalFormat, xoffset, 0, x, y, width, 1)) {
-      if (INTEL_DEBUG & DEBUG_FALLBACKS)
-         fprintf(stderr, "%s - fallback to swrast\n", __FUNCTION__);
+      fallback_debug("%s - fallback to swrast\n", __FUNCTION__);
       _mesa_meta_CopyTexSubImage1D(ctx, target, level, xoffset, x, y, width);
    }
 }
@@ -314,8 +310,7 @@ intelCopyTexSubImage2D(struct gl_context * ctx, GLenum target, GLint level,
                             internalFormat,
                             xoffset, yoffset, x, y, width, height)) {
 
-      if (INTEL_DEBUG & DEBUG_FALLBACKS)
-         fprintf(stderr, "%s - fallback to swrast\n", __FUNCTION__);
+      fallback_debug("%s - fallback to swrast\n", __FUNCTION__);
       _mesa_meta_CopyTexSubImage2D(ctx, target, level,
                                    xoffset, yoffset, x, y, width, height);
    }
diff --git a/src/mesa/drivers/dri/intel/intel_tex_image.c b/src/mesa/drivers/dri/intel/intel_tex_image.c
index 35f3d7d382..50fe9bd9f3 100644
--- a/src/mesa/drivers/dri/intel/intel_tex_image.c
+++ b/src/mesa/drivers/dri/intel/intel_tex_image.c
@@ -66,7 +66,6 @@ guess_and_alloc_mipmap_tree(struct intel_context *intel,
    GLuint width = intelImage->base.Width;
    GLuint height = intelImage->base.Height;
    GLuint depth = intelImage->base.Depth;
-   GLuint l2width, l2height, l2depth;
    GLuint i, comp_byte = 0;
    GLuint texelBytes;
 
@@ -114,10 +113,7 @@ guess_and_alloc_mipmap_tree(struct intel_context *intel,
       lastLevel = firstLevel;
    }
    else {
-      l2width = logbase2(width);
-      l2height = logbase2(height);
-      l2depth = logbase2(depth);
-      lastLevel = firstLevel + MAX2(MAX2(l2width, l2height), l2depth);
+      lastLevel = firstLevel + logbase2(MAX2(MAX2(width, height), depth));
    }
 
    assert(!intelObj->mt);
@@ -347,21 +343,6 @@ intelTexImage(struct gl_context * ctx,
       texImage->Data = NULL;
    }
 
-   /* If this is the only texture image in the tree, could call
-    * bmBufferData with NULL data to free the old block and avoid
-    * waiting on any outstanding fences.
-    */
-   if (intelObj->mt &&
-       intelObj->mt->first_level == level &&
-       intelObj->mt->last_level == level &&
-       intelObj->mt->target != GL_TEXTURE_CUBE_MAP_ARB &&
-       !intel_miptree_match_image(intelObj->mt, &intelImage->base)) {
-
-      DBG("release it\n");
-      intel_miptree_release(intel, &intelObj->mt);
-      assert(!intelObj->mt);
-   }
-
    if (!intelObj->mt) {
       guess_and_alloc_mipmap_tree(intel, intelObj, intelImage, pixels == NULL);
       if (!intelObj->mt) {
diff --git a/src/mesa/drivers/dri/nouveau/Makefile b/src/mesa/drivers/dri/nouveau/Makefile
index 7be19b26fd..3b506a91ff 100644
--- a/src/mesa/drivers/dri/nouveau/Makefile
+++ b/src/mesa/drivers/dri/nouveau/Makefile
@@ -19,6 +19,8 @@ DRIVER_SOURCES = \
 	nouveau_bo_state.c \
 	nouveau_texture.c \
 	nouveau_surface.c \
+	nouveau_scratch.c \
+	nouveau_array.c \
 	nv04_context.c \
 	nv04_render.c \
 	nv04_state_fb.c \
diff --git a/src/mesa/drivers/dri/nouveau/nouveau_array.c b/src/mesa/drivers/dri/nouveau/nouveau_array.c
new file mode 100644
index 0000000000..17e6d163a0
--- /dev/null
+++ b/src/mesa/drivers/dri/nouveau/nouveau_array.c
@@ -0,0 +1,136 @@
+/*
+ * Copyright (C) 2009-2010 Francisco Jerez.
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining
+ * a copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sublicense, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial
+ * portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+ * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.
+ * IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE
+ * LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION
+ * OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION
+ * WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ */
+
+#include "main/bufferobj.h"
+#include "nouveau_driver.h"
+#include "nouveau_array.h"
+#include "nouveau_bufferobj.h"
+
+static void
+get_array_extract(struct nouveau_array *a, extract_u_t *extract_u,
+		  extract_f_t *extract_f)
+{
+#define EXTRACT(in_t, out_t, k)						\
+	({								\
+		auto out_t f(struct nouveau_array *, int, int);		\
+		out_t f(struct nouveau_array *a, int i, int j) {	\
+			in_t x = ((in_t *)(a->buf + i * a->stride))[j];	\
+									\
+			return (out_t)x / (k);				\
+		};							\
+		f;							\
+	});
+
+	switch (a->type) {
+	case GL_BYTE:
+		*extract_u = EXTRACT(char, unsigned, 1);
+		*extract_f = EXTRACT(char, float, SCHAR_MAX);
+		break;
+	case GL_UNSIGNED_BYTE:
+		*extract_u = EXTRACT(unsigned char, unsigned, 1);
+		*extract_f = EXTRACT(unsigned char, float, UCHAR_MAX);
+		break;
+	case GL_SHORT:
+		*extract_u = EXTRACT(short, unsigned, 1);
+		*extract_f = EXTRACT(short, float, SHRT_MAX);
+		break;
+	case GL_UNSIGNED_SHORT:
+		*extract_u = EXTRACT(unsigned short, unsigned, 1);
+		*extract_f = EXTRACT(unsigned short, float, USHRT_MAX);
+		break;
+	case GL_INT:
+		*extract_u = EXTRACT(int, unsigned, 1);
+		*extract_f = EXTRACT(int, float, INT_MAX);
+		break;
+	case GL_UNSIGNED_INT:
+		*extract_u = EXTRACT(unsigned int, unsigned, 1);
+		*extract_f = EXTRACT(unsigned int, float, UINT_MAX);
+		break;
+	case GL_FLOAT:
+		*extract_u = EXTRACT(float, unsigned, 1.0 / UINT_MAX);
+		*extract_f = EXTRACT(float, float, 1);
+		break;
+	default:
+		assert(0);
+	}
+}
+
+void
+nouveau_init_array(struct nouveau_array *a, int attr, int stride,
+		   int fields, int type, struct gl_buffer_object *obj,
+		   const void *ptr, GLboolean map)
+{
+	a->attr = attr;
+	a->stride = stride;
+	a->fields = fields;
+	a->type = type;
+	a->buf = NULL;
+
+	if (obj) {
+		if (nouveau_bufferobj_hw(obj)) {
+			struct nouveau_bufferobj *nbo =
+				to_nouveau_bufferobj(obj);
+
+			nouveau_bo_ref(nbo->bo, &a->bo);
+			a->offset = (intptr_t)ptr;
+
+			if (map) {
+				nouveau_bo_map(a->bo, NOUVEAU_BO_RD);
+				a->buf = a->bo->map + a->offset;
+			}
+
+		} else {
+			nouveau_bo_ref(NULL, &a->bo);
+			a->offset = 0;
+
+			if (map)
+				a->buf = ADD_POINTERS(
+					nouveau_bufferobj_sys(obj), ptr);
+		}
+	}
+
+	if (a->buf)
+		get_array_extract(a, &a->extract_u, &a->extract_f);
+}
+
+void
+nouveau_deinit_array(struct nouveau_array *a)
+{
+	if (a->bo) {
+		if (a->bo->map)
+			nouveau_bo_unmap(a->bo);
+	}
+
+	a->buf = NULL;
+	a->fields = 0;
+}
+
+void
+nouveau_cleanup_array(struct nouveau_array *a)
+{
+	nouveau_deinit_array(a);
+	nouveau_bo_ref(NULL, &a->bo);
+}
diff --git a/src/mesa/drivers/dri/nouveau/nouveau_array.h b/src/mesa/drivers/dri/nouveau/nouveau_array.h
new file mode 100644
index 0000000000..ad3d69b33d
--- /dev/null
+++ b/src/mesa/drivers/dri/nouveau/nouveau_array.h
@@ -0,0 +1,58 @@
+/*
+ * Copyright (C) 2009-2010 Francisco Jerez.
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining
+ * a copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sublicense, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial
+ * portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+ * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.
+ * IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE
+ * LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION
+ * OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION
+ * WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ */
+
+#ifndef __NOUVEAU_ARRAY_H__
+#define __NOUVEAU_ARRAY_H__
+
+struct nouveau_array;
+
+typedef unsigned (*extract_u_t)(struct nouveau_array *, int, int);
+typedef float (*extract_f_t)(struct nouveau_array *, int, int);
+
+struct nouveau_array {
+	int attr;
+	int stride, fields, type;
+
+	struct nouveau_bo *bo;
+	unsigned offset;
+	const void *buf;
+
+	extract_u_t extract_u;
+	extract_f_t extract_f;
+};
+
+void
+nouveau_init_array(struct nouveau_array *a, int attr, int stride,
+		   int fields, int type, struct gl_buffer_object *obj,
+		   const void *ptr, GLboolean map);
+
+void
+nouveau_deinit_array(struct nouveau_array *a);
+
+void
+nouveau_cleanup_array(struct nouveau_array *a);
+
+#endif
diff --git a/src/mesa/drivers/dri/nouveau/nouveau_bo_state.c b/src/mesa/drivers/dri/nouveau/nouveau_bo_state.c
index f31772fe1d..7eef8c1ee8 100644
--- a/src/mesa/drivers/dri/nouveau/nouveau_bo_state.c
+++ b/src/mesa/drivers/dri/nouveau/nouveau_bo_state.c
@@ -126,13 +126,13 @@ void
 nouveau_bo_context_reset(struct nouveau_bo_context *bctx)
 {
 	struct nouveau_bo_state *s = &to_nouveau_context(bctx->ctx)->bo;
-	int i;
-
-	for (i = 0; i < bctx->count; i++)
-		nouveau_bo_ref(NULL, &bctx->marker[i].bo);
+	int i, n = bctx->count;
 
-	s->count -= bctx->count;
+	s->count -= n;
 	bctx->count = 0;
+
+	for (i = 0; i < n; i++)
+		nouveau_bo_ref(NULL, &bctx->marker[i].bo);
 }
 
 GLboolean
diff --git a/src/mesa/drivers/dri/nouveau/nouveau_bo_state.h b/src/mesa/drivers/dri/nouveau/nouveau_bo_state.h
index 6119a8336e..388a16a56e 100644
--- a/src/mesa/drivers/dri/nouveau/nouveau_bo_state.h
+++ b/src/mesa/drivers/dri/nouveau/nouveau_bo_state.h
@@ -29,7 +29,7 @@
 
 enum {
 	NOUVEAU_BO_CONTEXT_FRAMEBUFFER = 0,
-	NOUVEAU_BO_CONTEXT_LMA_DEPTH,
+	NOUVEAU_BO_CONTEXT_HIERZ,
 	NOUVEAU_BO_CONTEXT_SURFACE,
 	NOUVEAU_BO_CONTEXT_TEXTURE0,
 	NOUVEAU_BO_CONTEXT_TEXTURE1,
diff --git a/src/mesa/drivers/dri/nouveau/nouveau_bufferobj.c b/src/mesa/drivers/dri/nouveau/nouveau_bufferobj.c
index ad6e5bd805..e60b91f64b 100644
--- a/src/mesa/drivers/dri/nouveau/nouveau_bufferobj.c
+++ b/src/mesa/drivers/dri/nouveau/nouveau_bufferobj.c
@@ -30,6 +30,23 @@
 
 #include "main/bufferobj.h"
 
+static inline char *
+get_bufferobj_map(struct gl_buffer_object *obj, unsigned flags)
+{
+	struct nouveau_bufferobj *nbo = to_nouveau_bufferobj(obj);
+	void *map = NULL;
+
+	if (nbo->sys) {
+		map = nbo->sys;
+	} else if (nbo->bo) {
+		nouveau_bo_map(nbo->bo, flags);
+		map = nbo->bo->map;
+		nouveau_bo_unmap(nbo->bo);
+	}
+
+	return map;
+}
+
 static struct gl_buffer_object *
 nouveau_bufferobj_new(struct gl_context *ctx, GLuint buffer, GLenum target)
 {
@@ -50,6 +67,7 @@ nouveau_bufferobj_del(struct gl_context *ctx, struct gl_buffer_object *obj)
 	struct nouveau_bufferobj *nbo = to_nouveau_bufferobj(obj);
 
 	nouveau_bo_ref(NULL, &nbo->bo);
+	FREE(nbo->sys);
 	FREE(nbo);
 }
 
@@ -64,18 +82,27 @@ nouveau_bufferobj_data(struct gl_context *ctx, GLenum target, GLsizeiptrARB size
 	obj->Size = size;
 	obj->Usage = usage;
 
+	/* Free previous storage */
 	nouveau_bo_ref(NULL, &nbo->bo);
-	ret = nouveau_bo_new(context_dev(ctx),
-			     NOUVEAU_BO_GART | NOUVEAU_BO_MAP, 0,
-			     size, &nbo->bo);
-	assert(!ret);
-
-	if (data) {
-		nouveau_bo_map(nbo->bo, NOUVEAU_BO_WR);
-		memcpy(nbo->bo->map, data, size);
-		nouveau_bo_unmap(nbo->bo);
+	FREE(nbo->sys);
+
+	if (target == GL_ELEMENT_ARRAY_BUFFER_ARB ||
+	    (size < 512 && usage == GL_DYNAMIC_DRAW_ARB) ||
+	    context_chipset(ctx) < 0x10) {
+		/* Heuristic: keep it in system ram */
+		nbo->sys = MALLOC(size);
+
+	} else {
+		/* Get a hardware BO */
+		ret = nouveau_bo_new(context_dev(ctx),
+				     NOUVEAU_BO_GART | NOUVEAU_BO_MAP, 0,
+				     size, &nbo->bo);
+		assert(!ret);
 	}
 
+	if (data)
+		memcpy(get_bufferobj_map(obj, NOUVEAU_BO_WR), data, size);
+
 	return GL_TRUE;
 }
 
@@ -84,11 +111,7 @@ nouveau_bufferobj_subdata(struct gl_context *ctx, GLenum target, GLintptrARB off
 			  GLsizeiptrARB size, const GLvoid *data,
 			  struct gl_buffer_object *obj)
 {
-	struct nouveau_bufferobj *nbo = to_nouveau_bufferobj(obj);
-
-	nouveau_bo_map(nbo->bo, NOUVEAU_BO_WR);
-	memcpy(nbo->bo->map + offset, data, size);
-	nouveau_bo_unmap(nbo->bo);
+	memcpy(get_bufferobj_map(obj, NOUVEAU_BO_WR) + offset, data, size);
 }
 
 static void
@@ -96,44 +119,48 @@ nouveau_bufferobj_get_subdata(struct gl_context *ctx, GLenum target, GLintptrARB
 			   GLsizeiptrARB size, GLvoid *data,
 			   struct gl_buffer_object *obj)
 {
-	struct nouveau_bufferobj *nbo = to_nouveau_bufferobj(obj);
-
-	nouveau_bo_map(nbo->bo, NOUVEAU_BO_RD);
-	memcpy(data, nbo->bo->map + offset, size);
-	nouveau_bo_unmap(nbo->bo);
+	memcpy(data, get_bufferobj_map(obj, NOUVEAU_BO_RD) + offset, size);
 }
 
 static void *
 nouveau_bufferobj_map(struct gl_context *ctx, GLenum target, GLenum access,
 		   struct gl_buffer_object *obj)
 {
-	return ctx->Driver.MapBufferRange(ctx, target, 0, obj->Size, access,
+	unsigned flags = 0;
+
+	if (access == GL_READ_ONLY_ARB ||
+	    access == GL_READ_WRITE_ARB)
+		flags |= GL_MAP_READ_BIT;
+	if (access == GL_WRITE_ONLY_ARB ||
+	    access == GL_READ_WRITE_ARB)
+		flags |= GL_MAP_WRITE_BIT;
+
+	return ctx->Driver.MapBufferRange(ctx, target, 0, obj->Size, flags,
 					  obj);
 }
 
 static void *
 nouveau_bufferobj_map_range(struct gl_context *ctx, GLenum target, GLintptr offset,
-			    GLsizeiptr length, GLenum access,
+			    GLsizeiptr length, GLbitfield access,
 			    struct gl_buffer_object *obj)
 {
-	struct nouveau_bufferobj *nbo = to_nouveau_bufferobj(obj);
-	uint32_t flags = 0;
+	unsigned flags = 0;
+	char *map;
 
 	assert(!obj->Pointer);
 
-	if (!nbo->bo)
-		return NULL;
-
-	if (access == GL_READ_ONLY_ARB ||
-	    access == GL_READ_WRITE_ARB)
+	if (access & GL_MAP_READ_BIT)
 		flags |= NOUVEAU_BO_RD;
-	if (access == GL_WRITE_ONLY_ARB ||
-	    access == GL_READ_WRITE_ARB)
+	if (access & GL_MAP_WRITE_BIT)
 		flags |= NOUVEAU_BO_WR;
+	if (access & GL_MAP_UNSYNCHRONIZED_BIT)
+		flags |= NOUVEAU_BO_NOSYNC;
 
-	nouveau_bo_map_range(nbo->bo, offset, length, flags);
+	map = get_bufferobj_map(obj, flags);
+	if (!map)
+		return NULL;
 
-	obj->Pointer = nbo->bo->map;
+	obj->Pointer = map + offset;
 	obj->Offset = offset;
 	obj->Length = length;
 	obj->AccessFlags = access;
@@ -144,12 +171,8 @@ nouveau_bufferobj_map_range(struct gl_context *ctx, GLenum target, GLintptr offs
 static GLboolean
 nouveau_bufferobj_unmap(struct gl_context *ctx, GLenum target, struct gl_buffer_object *obj)
 {
-	struct nouveau_bufferobj *nbo = to_nouveau_bufferobj(obj);
-
 	assert(obj->Pointer);
 
-	nouveau_bo_unmap(nbo->bo);
-
 	obj->Pointer = NULL;
 	obj->Offset = 0;
 	obj->Length = 0;
diff --git a/src/mesa/drivers/dri/nouveau/nouveau_bufferobj.h b/src/mesa/drivers/dri/nouveau/nouveau_bufferobj.h
index acfc4cb9a9..01ef0bad0f 100644
--- a/src/mesa/drivers/dri/nouveau/nouveau_bufferobj.h
+++ b/src/mesa/drivers/dri/nouveau/nouveau_bufferobj.h
@@ -30,9 +30,16 @@
 struct nouveau_bufferobj {
 	struct gl_buffer_object base;
 	struct nouveau_bo *bo;
+	void *sys;
 };
 #define to_nouveau_bufferobj(x) ((struct nouveau_bufferobj *)(x))
 
+#define nouveau_bufferobj_hw(x) \
+	(_mesa_is_bufferobj(x) ? to_nouveau_bufferobj(x)->bo : NULL)
+
+#define nouveau_bufferobj_sys(x) \
+	(_mesa_is_bufferobj(x) ? to_nouveau_bufferobj(x)->sys : NULL)
+
 void
 nouveau_bufferobj_functions_init(struct dd_function_table *functions);
 
diff --git a/src/mesa/drivers/dri/nouveau/nouveau_class.h b/src/mesa/drivers/dri/nouveau/nouveau_class.h
index d41d431f79..687b847797 100644
--- a/src/mesa/drivers/dri/nouveau/nouveau_class.h
+++ b/src/mesa/drivers/dri/nouveau/nouveau_class.h
@@ -4954,6 +4954,8 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
 #define  NV25TCL_DMA_IN_MEMORY5								0x000001a0
 #define  NV25TCL_DMA_IN_MEMORY8								0x000001ac
 #define  NV25TCL_DMA_IN_MEMORY9								0x000001b0
+#define  NV25TCL_HIERZ_PITCH								0x0000022c
+#define  NV25TCL_HIERZ_OFFSET								0x00000230
 
 #endif /* NOUVEAU_REG_H */
 
diff --git a/src/mesa/drivers/dri/nouveau/nouveau_context.c b/src/mesa/drivers/dri/nouveau/nouveau_context.c
index 0ace139b88..f80aaedb25 100644
--- a/src/mesa/drivers/dri/nouveau/nouveau_context.c
+++ b/src/mesa/drivers/dri/nouveau/nouveau_context.c
@@ -119,6 +119,7 @@ nouveau_context_init(struct gl_context *ctx, struct nouveau_screen *screen,
 
 	nouveau_state_init(ctx);
 	nouveau_bo_state_init(ctx);
+	nouveau_scratch_init(ctx);
 	_mesa_meta_init(ctx);
 	_swrast_CreateContext(ctx);
 	_vbo_CreateContext(ctx);
@@ -163,6 +164,7 @@ nouveau_context_deinit(struct gl_context *ctx)
 	if (nctx->hw.chan)
 		nouveau_channel_free(&nctx->hw.chan);
 
+	nouveau_scratch_destroy(ctx);
 	nouveau_bo_state_destroy(ctx);
 	_mesa_free_context_data(ctx);
 }
@@ -325,10 +327,12 @@ nouveau_fallback(struct gl_context *ctx, enum nouveau_fallback mode)
 
 	nctx->fallback = MAX2(HWTNL, mode);
 
-	if (mode < SWRAST)
+	if (mode < SWRAST) {
 		nouveau_state_emit(ctx);
-	else
+		nouveau_bo_state_emit(ctx);
+	} else {
 		FIRE_RING(context_chan(ctx));
+	}
 }
 
 static void
@@ -365,5 +369,6 @@ nouveau_validate_framebuffer(struct gl_context *ctx)
 		validate_framebuffer(dri_ctx, dri_read,
 				     &dri_ctx->dri2.read_stamp);
 
-	nouveau_state_emit(ctx);
+	if (ctx->NewState & _NEW_BUFFERS)
+		_mesa_update_state(ctx);
 }
diff --git a/src/mesa/drivers/dri/nouveau/nouveau_context.h b/src/mesa/drivers/dri/nouveau/nouveau_context.h
index 23a8725672..7ebc676379 100644
--- a/src/mesa/drivers/dri/nouveau/nouveau_context.h
+++ b/src/mesa/drivers/dri/nouveau/nouveau_context.h
@@ -30,6 +30,7 @@
 #include "nouveau_screen.h"
 #include "nouveau_state.h"
 #include "nouveau_bo_state.h"
+#include "nouveau_scratch.h"
 #include "nouveau_render.h"
 
 #include "main/bitset.h"
@@ -67,6 +68,7 @@ struct nouveau_context {
 	struct nouveau_hw_state hw;
 	struct nouveau_bo_state bo;
 	struct nouveau_render_state render;
+	struct nouveau_scratch_state scratch;
 
 	struct {
 		GLboolean clear_blocked;
diff --git a/src/mesa/drivers/dri/nouveau/nouveau_render.h b/src/mesa/drivers/dri/nouveau/nouveau_render.h
index 81c6119fcc..0539c37758 100644
--- a/src/mesa/drivers/dri/nouveau/nouveau_render.h
+++ b/src/mesa/drivers/dri/nouveau/nouveau_render.h
@@ -28,46 +28,22 @@
 #define __NOUVEAU_RENDER_H__
 
 #include "vbo/vbo_context.h"
-
-struct nouveau_array_state;
+#include "nouveau_array.h"
 
 typedef void (*dispatch_t)(struct gl_context *, unsigned int, int, unsigned int);
-typedef unsigned (*extract_u_t)(struct nouveau_array_state *, int, int);
-typedef float (*extract_f_t)(struct nouveau_array_state *, int, int);
+typedef void (*emit_t)(struct gl_context *, struct nouveau_array *, const void *);
 
 struct nouveau_attr_info {
 	int vbo_index;
 	int imm_method;
 	int imm_fields;
 
-	void (*emit)(struct gl_context *, struct nouveau_array_state *, const void *);
-};
-
-struct nouveau_array_state {
-	int attr;
-	int stride, fields, type;
-
-	struct nouveau_bo *bo;
-	unsigned offset;
-	const void *buf;
-
-	extract_u_t extract_u;
-	extract_f_t extract_f;
-};
-
-#define RENDER_SCRATCH_COUNT 2
-#define RENDER_SCRATCH_SIZE 2*1024*1024
-
-struct nouveau_scratch_state {
-	struct nouveau_bo *bo[RENDER_SCRATCH_COUNT];
-
-	int index;
-	int offset;
-	void *buf;
+	emit_t emit;
 };
 
 struct nouveau_swtnl_state {
 	struct nouveau_bo *vbo;
+	unsigned offset;
 	void *buf;
 	unsigned vertex_count;
 	GLenum primitive;
@@ -79,8 +55,8 @@ struct nouveau_render_state {
 		IMM
 	} mode;
 
-	struct nouveau_array_state ib;
-	struct nouveau_array_state attrs[VERT_ATTRIB_MAX];
+	struct nouveau_array ib;
+	struct nouveau_array attrs[VERT_ATTRIB_MAX];
 
 	/* Maps a HW VBO index or IMM emission order to an index in
 	 * the attrs array above (or -1 if unused). */
@@ -89,10 +65,16 @@ struct nouveau_render_state {
 	int attr_count;
 	int vertex_size;
 
-	struct nouveau_scratch_state scratch;
 	struct nouveau_swtnl_state swtnl;
 };
 
 #define to_render_state(ctx) (&to_nouveau_context(ctx)->render)
 
+#define FOR_EACH_ATTR(render, i, attr)					\
+	for (i = 0; attr = (render)->map[i], i < NUM_VERTEX_ATTRS; i++)
+
+#define FOR_EACH_BOUND_ATTR(render, i, attr)				\
+	for (i = 0; attr = (render)->map[i], i < render->attr_count; i++) \
+		if (attr >= 0)
+
 #endif
diff --git a/src/mesa/drivers/dri/nouveau/nouveau_render_t.c b/src/mesa/drivers/dri/nouveau/nouveau_render_t.c
index dd38c14aa7..43164ec34f 100644
--- a/src/mesa/drivers/dri/nouveau/nouveau_render_t.c
+++ b/src/mesa/drivers/dri/nouveau/nouveau_render_t.c
@@ -100,8 +100,8 @@
 /*
  * Select an appropriate dispatch function for the given index buffer.
  */
-static void
-get_array_dispatch(struct nouveau_array_state *a, dispatch_t *dispatch)
+static dispatch_t
+get_array_dispatch(struct nouveau_array *a)
 {
 	if (!a->fields) {
 		auto void f(struct gl_context *, unsigned int, int, unsigned int);
@@ -114,7 +114,7 @@ get_array_dispatch(struct nouveau_array_state *a, dispatch_t *dispatch)
 			EMIT_VBO(L, ctx, start, delta, n);
 		};
 
-		*dispatch = f;
+		return f;
 
 	} else if (a->type == GL_UNSIGNED_INT) {
 		auto void f(struct gl_context *, unsigned int, int, unsigned int);
@@ -127,7 +127,7 @@ get_array_dispatch(struct nouveau_array_state *a, dispatch_t *dispatch)
 			EMIT_VBO(I32, ctx, start, delta, n);
 		};
 
-		*dispatch = f;
+		return f;
 
 	} else {
 		auto void f(struct gl_context *, unsigned int, int, unsigned int);
@@ -141,115 +141,11 @@ get_array_dispatch(struct nouveau_array_state *a, dispatch_t *dispatch)
 			EMIT_VBO(I16, ctx, start, delta, n & ~1);
 		};
 
-		*dispatch = f;
+		return f;
 	}
 }
 
 /*
- * Select appropriate element extraction functions for the given
- * array.
- */
-static void
-get_array_extract(struct nouveau_array_state *a,
-		  extract_u_t *extract_u, extract_f_t *extract_f)
-{
-#define EXTRACT(in_t, out_t, k)						\
-	({								\
-		auto out_t f(struct nouveau_array_state *, int, int);	\
-		out_t f(struct nouveau_array_state *a, int i, int j) {	\
-			in_t x = ((in_t *)(a->buf + i * a->stride))[j];	\
-									\
-			return (out_t)x / (k);				\
-		};							\
-		f;							\
-	});
-
-	switch (a->type) {
-	case GL_BYTE:
-		*extract_u = EXTRACT(char, unsigned, 1);
-		*extract_f = EXTRACT(char, float, SCHAR_MAX);
-		break;
-	case GL_UNSIGNED_BYTE:
-		*extract_u = EXTRACT(unsigned char, unsigned, 1);
-		*extract_f = EXTRACT(unsigned char, float, UCHAR_MAX);
-		break;
-	case GL_SHORT:
-		*extract_u = EXTRACT(short, unsigned, 1);
-		*extract_f = EXTRACT(short, float, SHRT_MAX);
-		break;
-	case GL_UNSIGNED_SHORT:
-		*extract_u = EXTRACT(unsigned short, unsigned, 1);
-		*extract_f = EXTRACT(unsigned short, float, USHRT_MAX);
-		break;
-	case GL_INT:
-		*extract_u = EXTRACT(int, unsigned, 1);
-		*extract_f = EXTRACT(int, float, INT_MAX);
-		break;
-	case GL_UNSIGNED_INT:
-		*extract_u = EXTRACT(unsigned int, unsigned, 1);
-		*extract_f = EXTRACT(unsigned int, float, UINT_MAX);
-		break;
-	case GL_FLOAT:
-		*extract_u = EXTRACT(float, unsigned, 1.0 / UINT_MAX);
-		*extract_f = EXTRACT(float, float, 1);
-		break;
-
-	default:
-		assert(0);
-	}
-}
-
-/*
- * Returns a pointer to a chunk of <size> bytes long GART memory. <bo>
- * will be updated with the buffer object the memory is located in.
- *
- * If <offset> is provided, it will be updated with the offset within
- * <bo> of the allocated memory. Otherwise the returned memory will
- * always be located right at the beginning of <bo>.
- */
-static inline void *
-get_scratch_vbo(struct gl_context *ctx, unsigned size, struct nouveau_bo **bo,
-		unsigned *offset)
-{
-	struct nouveau_scratch_state *scratch = &to_render_state(ctx)->scratch;
-	void *buf;
-
-	if (scratch->buf && offset &&
-	    size <= RENDER_SCRATCH_SIZE - scratch->offset) {
-		nouveau_bo_ref(scratch->bo[scratch->index], bo);
-
-		buf = scratch->buf + scratch->offset;
-		*offset = scratch->offset;
-		scratch->offset += size;
-
-	} else if (size <= RENDER_SCRATCH_SIZE) {
-		scratch->index = (scratch->index + 1) % RENDER_SCRATCH_COUNT;
-		nouveau_bo_ref(scratch->bo[scratch->index], bo);
-
-		nouveau_bo_map(*bo, NOUVEAU_BO_WR);
-		buf = scratch->buf = (*bo)->map;
-		nouveau_bo_unmap(*bo);
-
-		if (offset)
-			*offset = 0;
-		scratch->offset = size;
-
-	} else {
-		nouveau_bo_new(context_dev(ctx),
-			       NOUVEAU_BO_MAP | NOUVEAU_BO_GART, 0, size, bo);
-
-		nouveau_bo_map(*bo, NOUVEAU_BO_WR);
-		buf = (*bo)->map;
-		nouveau_bo_unmap(*bo);
-
-		if (offset)
-			*offset = 0;
-	}
-
-	return buf;
-}
-
-/*
  * Returns how many vertices you can draw using <n> pushbuf dwords.
  */
 static inline unsigned
@@ -277,6 +173,11 @@ get_max_vertices(struct gl_context *ctx, const struct _mesa_index_buffer *ib,
 			case GL_UNSIGNED_BYTE:
 				max_out = MAX_OUT_I16;
 				break;
+
+			default:
+				assert(0);
+				max_out = 0;
+				break;
 			}
 		} else {
 			max_out = MAX_OUT_L;
@@ -290,7 +191,7 @@ get_max_vertices(struct gl_context *ctx, const struct _mesa_index_buffer *ib,
 #include "nouveau_swtnl_t.c"
 
 static void
-TAG(emit_material)(struct gl_context *ctx, struct nouveau_array_state *a,
+TAG(emit_material)(struct gl_context *ctx, struct nouveau_array *a,
 		   const void *v)
 {
 	const int attr = a->attr - VERT_ATTRIB_GENERIC0;
@@ -337,21 +238,14 @@ void
 TAG(render_init)(struct gl_context *ctx)
 {
 	struct nouveau_render_state *render = to_render_state(ctx);
-	struct nouveau_scratch_state *scratch = &render->scratch;
-	int ret, i;
-
-	for (i = 0; i < RENDER_SCRATCH_COUNT; i++) {
-		ret = nouveau_bo_new(context_dev(ctx),
-				     NOUVEAU_BO_MAP | NOUVEAU_BO_GART,
-				     0, RENDER_SCRATCH_SIZE, &scratch->bo[i]);
-		assert(!ret);
-	}
+	int i;
 
 	for (i = 0; i < VERT_ATTRIB_MAX; i++)
 		render->map[i] = -1;
 
 	TAG(swtnl_init)(ctx);
 	vbo_set_draw_func(ctx, TAG(render_prims));
+	vbo_use_buffer_objects(ctx);
 }
 
 void
diff --git a/src/mesa/drivers/dri/nouveau/nouveau_scratch.c b/src/mesa/drivers/dri/nouveau/nouveau_scratch.c
new file mode 100644
index 0000000000..ddda67b2f1
--- /dev/null
+++ b/src/mesa/drivers/dri/nouveau/nouveau_scratch.c
@@ -0,0 +1,98 @@
+/*
+ * Copyright (C) 2009-2010 Francisco Jerez.
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining
+ * a copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sublicense, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial
+ * portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+ * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.
+ * IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE
+ * LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION
+ * OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION
+ * WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ */
+
+#include "nouveau_driver.h"
+#include "nouveau_context.h"
+
+/*
+ * Returns a pointer to a chunk of 'size' bytes long GART memory. 'bo'
+ * and 'offset' will point to the returned memory.
+ */
+void *
+nouveau_get_scratch(struct gl_context *ctx, unsigned size,
+		    struct nouveau_bo **bo, unsigned *offset)
+{
+	struct nouveau_scratch_state *scratch =
+		&to_nouveau_context(ctx)->scratch;
+	void *buf;
+
+	if (scratch->buf && size <= NOUVEAU_SCRATCH_SIZE - scratch->offset) {
+		nouveau_bo_ref(scratch->bo[scratch->index], bo);
+
+		buf = scratch->buf + scratch->offset;
+		*offset = scratch->offset;
+		scratch->offset += size;
+
+	} else if (size <= NOUVEAU_SCRATCH_SIZE) {
+		scratch->index = (scratch->index + 1) % NOUVEAU_SCRATCH_COUNT;
+		nouveau_bo_ref(scratch->bo[scratch->index], bo);
+
+		nouveau_bo_map(*bo, NOUVEAU_BO_WR);
+		buf = scratch->buf = (*bo)->map;
+		nouveau_bo_unmap(*bo);
+
+		*offset = 0;
+		scratch->offset = size;
+
+	} else {
+		nouveau_bo_new(context_dev(ctx),
+			       NOUVEAU_BO_MAP | NOUVEAU_BO_GART, 0, size, bo);
+
+		nouveau_bo_map(*bo, NOUVEAU_BO_WR);
+		buf = (*bo)->map;
+		nouveau_bo_unmap(*bo);
+
+		*offset = 0;
+	}
+
+	return buf;
+}
+
+void
+nouveau_scratch_init(struct gl_context *ctx)
+{
+	struct nouveau_scratch_state *scratch =
+		&to_nouveau_context(ctx)->scratch;
+	int ret, i;
+
+	for (i = 0; i < NOUVEAU_SCRATCH_COUNT; i++) {
+		ret = nouveau_bo_new(context_dev(ctx),
+				     NOUVEAU_BO_MAP | NOUVEAU_BO_GART,
+				     0, NOUVEAU_SCRATCH_SIZE, &scratch->bo[i]);
+		assert(!ret);
+	}
+}
+
+void
+nouveau_scratch_destroy(struct gl_context *ctx)
+{
+	struct nouveau_scratch_state *scratch =
+		&to_nouveau_context(ctx)->scratch;
+	int i;
+
+	for (i = 0; i < NOUVEAU_SCRATCH_COUNT; i++)
+		nouveau_bo_ref(NULL, &scratch->bo[i]);
+}
diff --git a/src/mesa/drivers/dri/nouveau/nouveau_scratch.h b/src/mesa/drivers/dri/nouveau/nouveau_scratch.h
new file mode 100644
index 0000000000..b60b33dd1a
--- /dev/null
+++ b/src/mesa/drivers/dri/nouveau/nouveau_scratch.h
@@ -0,0 +1,51 @@
+/*
+ * Copyright (C) 2009-2010 Francisco Jerez.
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining
+ * a copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sublicense, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial
+ * portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+ * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.
+ * IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE
+ * LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION
+ * OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION
+ * WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ */
+
+#ifndef __NOUVEAU_SCRATCH_H__
+#define __NOUVEAU_SCRATCH_H__
+
+#define NOUVEAU_SCRATCH_COUNT 2
+#define NOUVEAU_SCRATCH_SIZE 3*1024*1024
+
+struct nouveau_scratch_state {
+	struct nouveau_bo *bo[NOUVEAU_SCRATCH_COUNT];
+
+	int index;
+	int offset;
+	void *buf;
+};
+
+void *
+nouveau_get_scratch(struct gl_context *ctx, unsigned size,
+		    struct nouveau_bo **bo, unsigned *offset);
+
+void
+nouveau_scratch_init(struct gl_context *ctx);
+
+void
+nouveau_scratch_destroy(struct gl_context *ctx);
+
+#endif
diff --git a/src/mesa/drivers/dri/nouveau/nouveau_state.c b/src/mesa/drivers/dri/nouveau/nouveau_state.c
index 7b7ddd2f54..1579d29efc 100644
--- a/src/mesa/drivers/dri/nouveau/nouveau_state.c
+++ b/src/mesa/drivers/dri/nouveau/nouveau_state.c
@@ -113,6 +113,12 @@ nouveau_depth_range(struct gl_context *ctx, GLclampd nearval, GLclampd farval)
 }
 
 static void
+nouveau_read_buffer(struct gl_context *ctx, GLenum buffer)
+{
+	nouveau_validate_framebuffer(ctx);
+}
+
+static void
 nouveau_draw_buffers(struct gl_context *ctx, GLsizei n, const GLenum *buffers)
 {
 	nouveau_validate_framebuffer(ctx);
@@ -512,6 +518,7 @@ nouveau_state_init(struct gl_context *ctx)
 	ctx->Driver.DepthFunc = nouveau_depth_func;
 	ctx->Driver.DepthMask = nouveau_depth_mask;
 	ctx->Driver.DepthRange = nouveau_depth_range;
+	ctx->Driver.ReadBuffer = nouveau_read_buffer;
 	ctx->Driver.DrawBuffers = nouveau_draw_buffers;
 	ctx->Driver.Enable = nouveau_enable;
 	ctx->Driver.Fogfv = nouveau_fog;
diff --git a/src/mesa/drivers/dri/nouveau/nouveau_swtnl_t.c b/src/mesa/drivers/dri/nouveau/nouveau_swtnl_t.c
index b3588e8fd3..0377c7ffc9 100644
--- a/src/mesa/drivers/dri/nouveau/nouveau_swtnl_t.c
+++ b/src/mesa/drivers/dri/nouveau/nouveau_swtnl_t.c
@@ -28,6 +28,8 @@
 #include "tnl/t_pipeline.h"
 #include "tnl/t_vertex.h"
 
+#define SWTNL_VBO_SIZE 65536
+
 static enum tnl_attr_format
 swtnl_get_format(int type, int fields) {
 	switch (type) {
@@ -105,7 +107,7 @@ swtnl_choose_attrs(struct gl_context *ctx)
 	TNLcontext *tnl = TNL_CONTEXT(ctx);
 	struct tnl_clipspace *vtx = &tnl->clipspace;
 	static struct tnl_attr_map map[NUM_VERTEX_ATTRS];
-	int fields, i, n = 0;
+	int fields, attr, i, n = 0;
 
 	render->mode = VBO;
 	render->attr_count = NUM_VERTEX_ATTRS;
@@ -116,7 +118,7 @@ swtnl_choose_attrs(struct gl_context *ctx)
 	for (i = 0; i < VERT_ATTRIB_MAX; i++) {
 		struct nouveau_attr_info *ha = &TAG(vertex_attrs)[i];
 		struct swtnl_attr_info *sa = &swtnl_attrs[i];
-		struct nouveau_array_state *a = &render->attrs[i];
+		struct nouveau_array *a = &render->attrs[i];
 
 		if (!sa->fields)
 			continue; /* Unsupported attribute. */
@@ -141,13 +143,8 @@ swtnl_choose_attrs(struct gl_context *ctx)
 
 	_tnl_install_attrs(ctx, map, n, NULL, 0);
 
-	for (i = 0; i < vtx->attr_count; i++) {
-		struct tnl_clipspace_attr *ta = &vtx->attr[i];
-		struct nouveau_array_state *a = &render->attrs[ta->attrib];
-
-		a->stride = vtx->vertex_size;
-		a->offset = ta->vertoffset;
-	}
+	FOR_EACH_BOUND_ATTR(render, i, attr)
+		render->attrs[attr].stride = vtx->vertex_size;
 
 	TAG(render_set_format)(ctx);
 }
@@ -158,8 +155,8 @@ swtnl_alloc_vertices(struct gl_context *ctx)
 	struct nouveau_swtnl_state *swtnl = &to_render_state(ctx)->swtnl;
 
 	nouveau_bo_ref(NULL, &swtnl->vbo);
-	swtnl->buf = get_scratch_vbo(ctx, RENDER_SCRATCH_SIZE,
-				     &swtnl->vbo, NULL);
+	swtnl->buf = nouveau_get_scratch(ctx, SWTNL_VBO_SIZE, &swtnl->vbo,
+					 &swtnl->offset);
 	swtnl->vertex_count = 0;
 }
 
@@ -168,14 +165,15 @@ swtnl_bind_vertices(struct gl_context *ctx)
 {
 	struct nouveau_render_state *render = to_render_state(ctx);
 	struct nouveau_swtnl_state *swtnl = &render->swtnl;
+	struct tnl_clipspace *vtx = &TNL_CONTEXT(ctx)->clipspace;
 	int i;
 
-	for (i = 0; i < render->attr_count; i++) {
-		int attr = render->map[i];
+	for (i = 0; i < vtx->attr_count; i++) {
+		struct tnl_clipspace_attr *ta = &vtx->attr[i];
+		struct nouveau_array *a = &render->attrs[ta->attrib];
 
-		if (attr >= 0)
-			nouveau_bo_ref(swtnl->vbo,
-				       &render->attrs[attr].bo);
+		nouveau_bo_ref(swtnl->vbo, &a->bo);
+		a->offset = swtnl->offset + ta->vertoffset;
 	}
 
 	TAG(render_bind_vertices)(ctx);
@@ -185,15 +183,11 @@ static void
 swtnl_unbind_vertices(struct gl_context *ctx)
 {
 	struct nouveau_render_state *render = to_render_state(ctx);
-	int i;
+	int i, attr;
 
-	for (i = 0; i < render->attr_count; i++) {
-		int *attr = &render->map[i];
-
-		if (*attr >= 0) {
-			nouveau_bo_ref(NULL, &render->attrs[*attr].bo);
-			*attr = -1;
-		}
+	FOR_EACH_BOUND_ATTR(render, i, attr) {
+		nouveau_bo_ref(NULL, &render->attrs[attr].bo);
+		render->map[i] = -1;
 	}
 
 	render->attr_count = 0;
@@ -260,7 +254,7 @@ swtnl_reset_stipple(struct gl_context *ctx)
 	struct nouveau_swtnl_state *swtnl = &to_render_state(ctx)->swtnl; \
 	int vertex_len = TNL_CONTEXT(ctx)->clipspace.vertex_size;	\
 									\
-	if (swtnl->vertex_count + (n) > swtnl->vbo->size/vertex_len	\
+	if (swtnl->vertex_count + (n) > SWTNL_VBO_SIZE/vertex_len	\
 	    || (swtnl->vertex_count && swtnl->primitive != p))		\
 		swtnl_flush_vertices(ctx);				\
 									\
@@ -280,7 +274,7 @@ swtnl_points(struct gl_context *ctx, GLuint first, GLuint last)
 	while (first < last) {
 		BEGIN_PRIMITIVE(GL_POINTS, last - first);
 
-		count = MIN2(swtnl->vbo->size / vertex_len, last - first);
+		count = MIN2(SWTNL_VBO_SIZE / vertex_len, last - first);
 		for (i = 0; i < count; i++)
 			OUT_VERTEX(first + i);
 
diff --git a/src/mesa/drivers/dri/nouveau/nouveau_texture.c b/src/mesa/drivers/dri/nouveau/nouveau_texture.c
index cd063702af..060c2c5bcc 100644
--- a/src/mesa/drivers/dri/nouveau/nouveau_texture.c
+++ b/src/mesa/drivers/dri/nouveau/nouveau_texture.c
@@ -79,26 +79,65 @@ nouveau_teximage_free(struct gl_context *ctx, struct gl_texture_image *ti)
 }
 
 static void
-nouveau_teximage_map(struct gl_context *ctx, struct gl_texture_image *ti)
+nouveau_teximage_map(struct gl_context *ctx, struct gl_texture_image *ti,
+		     int access, int x, int y, int w, int h)
 {
-	struct nouveau_surface *s = &to_nouveau_teximage(ti)->surface;
-	int ret;
+	struct nouveau_teximage *nti = to_nouveau_teximage(ti);
+	struct nouveau_surface *s = &nti->surface;
+	struct nouveau_surface *st = &nti->transfer.surface;
 
 	if (s->bo) {
-		ret = nouveau_bo_map(s->bo, NOUVEAU_BO_RDWR);
-		assert(!ret);
-
-		ti->Data = s->bo->map;
+		if (!(access & GL_MAP_READ_BIT) &&
+		    nouveau_bo_pending(s->bo)) {
+			/*
+			 * Heuristic: use a bounce buffer to pipeline
+			 * teximage transfers.
+			 */
+			st->layout = LINEAR;
+			st->format = s->format;
+			st->cpp = s->cpp;
+			st->width = w;
+			st->height = h;
+			st->pitch = s->pitch;
+			nti->transfer.x = x;
+			nti->transfer.y = y;
+
+			ti->Data = nouveau_get_scratch(ctx, st->pitch * h,
+						       &st->bo, &st->offset);
+
+		} else {
+			int ret, flags = 0;
+
+			if (access & GL_MAP_READ_BIT)
+				flags |= NOUVEAU_BO_RD;
+			if (access & GL_MAP_WRITE_BIT)
+				flags |= NOUVEAU_BO_WR;
+
+			ret = nouveau_bo_map(s->bo, flags);
+			assert(!ret);
+
+			ti->Data = s->bo->map + y * s->pitch + x * s->cpp;
+		}
 	}
 }
 
 static void
 nouveau_teximage_unmap(struct gl_context *ctx, struct gl_texture_image *ti)
 {
-	struct nouveau_surface *s = &to_nouveau_teximage(ti)->surface;
+	struct nouveau_teximage *nti = to_nouveau_teximage(ti);
+	struct nouveau_surface *s = &nti->surface;
+	struct nouveau_surface *st = &nti->transfer.surface;
 
-	if (s->bo)
+	if (st->bo) {
+		context_drv(ctx)->surface_copy(ctx, s, st, nti->transfer.x,
+					       nti->transfer.y, 0, 0,
+					       st->width, st->height);
+		nouveau_surface_ref(NULL, st);
+
+	} else if (s->bo) {
 		nouveau_bo_unmap(s->bo);
+	}
+
 	ti->Data = NULL;
 }
 
@@ -115,6 +154,7 @@ nouveau_choose_tex_format(struct gl_context *ctx, GLint internalFormat,
 	case GL_RGBA12:
 	case GL_RGBA16:
 	case GL_RGB10_A2:
+	case GL_COMPRESSED_RGBA:
 		return MESA_FORMAT_ARGB8888;
 	case GL_RGB5_A1:
 		return MESA_FORMAT_ARGB1555;
@@ -124,6 +164,7 @@ nouveau_choose_tex_format(struct gl_context *ctx, GLint internalFormat,
 	case GL_RGB10:
 	case GL_RGB12:
 	case GL_RGB16:
+	case GL_COMPRESSED_RGB:
 		return MESA_FORMAT_XRGB8888;
 	case 3:
 	case GL_R3_G3_B2:
@@ -139,6 +180,7 @@ nouveau_choose_tex_format(struct gl_context *ctx, GLint internalFormat,
 	case GL_LUMINANCE12_ALPHA12:
 	case GL_LUMINANCE16_ALPHA16:
 	case GL_LUMINANCE8_ALPHA8:
+	case GL_COMPRESSED_LUMINANCE_ALPHA:
 		return MESA_FORMAT_ARGB8888;
 
 	case 1:
@@ -147,6 +189,7 @@ nouveau_choose_tex_format(struct gl_context *ctx, GLint internalFormat,
 	case GL_LUMINANCE12:
 	case GL_LUMINANCE16:
 	case GL_LUMINANCE8:
+	case GL_COMPRESSED_LUMINANCE:
 		return MESA_FORMAT_L8;
 
 	case GL_ALPHA:
@@ -154,6 +197,7 @@ nouveau_choose_tex_format(struct gl_context *ctx, GLint internalFormat,
 	case GL_ALPHA12:
 	case GL_ALPHA16:
 	case GL_ALPHA8:
+	case GL_COMPRESSED_ALPHA:
 		return MESA_FORMAT_A8;
 
 	case GL_INTENSITY:
@@ -356,7 +400,8 @@ nouveau_teximage(struct gl_context *ctx, GLint dims, GLenum target, GLint level,
 					     "glTexImage");
 	if (pixels) {
 		/* Store the pixel data. */
-		nouveau_teximage_map(ctx, ti);
+		nouveau_teximage_map(ctx, ti, GL_MAP_WRITE_BIT,
+				     0, 0, width, height);
 
 		ret = _mesa_texstore(ctx, dims, ti->_BaseFormat,
 				     ti->TexFormat, ti->Data,
@@ -443,13 +488,13 @@ nouveau_texsubimage(struct gl_context *ctx, GLint dims, GLenum target, GLint lev
 					     format, type, pixels, packing,
 					     "glTexSubImage");
 	if (pixels) {
-		nouveau_teximage_map(ctx, ti);
+		nouveau_teximage_map(ctx, ti, GL_MAP_WRITE_BIT,
+				     xoffset, yoffset, width, height);
 
 		ret = _mesa_texstore(ctx, 3, ti->_BaseFormat, ti->TexFormat,
-				     ti->Data, xoffset, yoffset, zoffset,
-				     s->pitch, ti->ImageOffsets,
-				     width, height, depth, format, type,
-				     pixels, packing);
+				     ti->Data, 0, 0, 0, s->pitch,
+				     ti->ImageOffsets, width, height, depth,
+				     format, type, pixels, packing);
 		assert(ret);
 
 		nouveau_teximage_unmap(ctx, ti);
@@ -508,7 +553,8 @@ nouveau_get_teximage(struct gl_context *ctx, GLenum target, GLint level,
 		     struct gl_texture_object *t,
 		     struct gl_texture_image *ti)
 {
-	nouveau_teximage_map(ctx, ti);
+	nouveau_teximage_map(ctx, ti, GL_MAP_READ_BIT,
+			     0, 0, ti->Width, ti->Height);
 	_mesa_get_teximage(ctx, target, level, format, type, pixels,
 			   t, ti);
 	nouveau_teximage_unmap(ctx, ti);
@@ -579,8 +625,11 @@ nouveau_texture_map(struct gl_context *ctx, struct gl_texture_object *t)
 	int i;
 
 	for (i = t->BaseLevel; i < t->_MaxLevel; i++) {
-		if (t->Image[0][i])
-			nouveau_teximage_map(ctx, t->Image[0][i]);
+		struct gl_texture_image *ti = t->Image[0][i];
+
+		if (ti)
+			nouveau_teximage_map(ctx, ti, GL_MAP_READ_BIT,
+					     0, 0, ti->Width, ti->Height);
 	}
 }
 
@@ -630,7 +679,8 @@ nouveau_generate_mipmap(struct gl_context *ctx, GLenum target,
 	if (_mesa_meta_check_generate_mipmap_fallback(ctx, target, t)) {
 		struct gl_texture_image *base = t->Image[0][t->BaseLevel];
 
-		nouveau_teximage_map(ctx, base);
+		nouveau_teximage_map(ctx, base, GL_MAP_READ_BIT,
+				     0, 0, base->Width, base->Height);
 		_mesa_generate_mipmap(ctx, target, t);
 		nouveau_teximage_unmap(ctx, base);
 
diff --git a/src/mesa/drivers/dri/nouveau/nouveau_texture.h b/src/mesa/drivers/dri/nouveau/nouveau_texture.h
index fc170215f3..56e61c7337 100644
--- a/src/mesa/drivers/dri/nouveau/nouveau_texture.h
+++ b/src/mesa/drivers/dri/nouveau/nouveau_texture.h
@@ -30,6 +30,10 @@
 struct nouveau_teximage {
 	struct gl_texture_image base;
 	struct nouveau_surface surface;
+	struct {
+		struct nouveau_surface surface;
+		int x, y;
+	} transfer;
 };
 #define to_nouveau_teximage(x) ((struct nouveau_teximage *)(x))
 
diff --git a/src/mesa/drivers/dri/nouveau/nouveau_vbo_t.c b/src/mesa/drivers/dri/nouveau/nouveau_vbo_t.c
index 394f3c9b50..3d8f43ffb4 100644
--- a/src/mesa/drivers/dri/nouveau/nouveau_vbo_t.c
+++ b/src/mesa/drivers/dri/nouveau/nouveau_vbo_t.c
@@ -31,59 +31,11 @@
 #include "main/image.h"
 
 /* Arbitrary pushbuf length we can assume we can get with a single
- * WAIT_RING. */
+ * call to WAIT_RING. */
 #define PUSHBUF_DWORDS 65536
 
-/* Functions to set up struct nouveau_array_state from something like
- * a GL array or index buffer. */
-
-static void
-vbo_init_array(struct nouveau_array_state *a, int attr, int stride,
-	       int fields, int type, struct gl_buffer_object *obj,
-	       const void *ptr, GLboolean map)
-{
-	a->attr = attr;
-	a->stride = stride;
-	a->fields = fields;
-	a->type = type;
-
-	if (_mesa_is_bufferobj(obj)) {
-		nouveau_bo_ref(to_nouveau_bufferobj(obj)->bo, &a->bo);
-		a->offset = (intptr_t)ptr;
-
-		if (map) {
-			nouveau_bo_map(a->bo, NOUVEAU_BO_RD);
-			a->buf = a->bo->map + a->offset;
-		} else {
-			a->buf = NULL;
-		}
-
-	} else {
-		nouveau_bo_ref(NULL, &a->bo);
-		a->offset = 0;
-
-		if (map)
-			a->buf = ptr;
-		else
-			a->buf = NULL;
-	}
-
-	if (a->buf)
-		get_array_extract(a, &a->extract_u, &a->extract_f);
-}
-
-static void
-vbo_deinit_array(struct nouveau_array_state *a)
-{
-	if (a->bo) {
-		if (a->bo->map)
-			nouveau_bo_unmap(a->bo);
-		nouveau_bo_ref(NULL, &a->bo);
-	}
-
-	a->buf = NULL;
-	a->fields = 0;
-}
+/* Functions to turn GL arrays or index buffers into nouveau_array
+ * structures. */
 
 static int
 get_array_stride(struct gl_context *ctx, const struct gl_client_array *a)
@@ -102,48 +54,45 @@ vbo_init_arrays(struct gl_context *ctx, const struct _mesa_index_buffer *ib,
 		const struct gl_client_array **arrays)
 {
 	struct nouveau_render_state *render = to_render_state(ctx);
-	int i;
+	GLboolean imm = (render->mode == IMM);
+	int i, attr;
 
 	if (ib)
-		vbo_init_array(&render->ib, 0, 0, ib->count, ib->type,
-			       ib->obj, ib->ptr, GL_TRUE);
+		nouveau_init_array(&render->ib, 0, 0, ib->count, ib->type,
+				   ib->obj, ib->ptr, GL_TRUE);
 
-	for (i = 0; i < render->attr_count; i++) {
-		int attr = render->map[i];
+	FOR_EACH_BOUND_ATTR(render, i, attr) {
+		const struct gl_client_array *array = arrays[attr];
 
-		if (attr >= 0) {
-			const struct gl_client_array *array = arrays[attr];
-
-			vbo_init_array(&render->attrs[attr], attr,
-				       get_array_stride(ctx, array),
-				       array->Size, array->Type,
-				       array->BufferObj, array->Ptr,
-				       render->mode == IMM);
-		}
+		nouveau_init_array(&render->attrs[attr], attr,
+				   get_array_stride(ctx, array),
+				   array->Size, array->Type,
+				   imm ? array->BufferObj : NULL,
+				   array->Ptr, imm);
 	}
 }
 
 static void
 vbo_deinit_arrays(struct gl_context *ctx, const struct _mesa_index_buffer *ib,
-		const struct gl_client_array **arrays)
+		  const struct gl_client_array **arrays)
 {
 	struct nouveau_render_state *render = to_render_state(ctx);
-	int i;
+	int i, attr;
 
 	if (ib)
-		vbo_deinit_array(&render->ib);
+		nouveau_cleanup_array(&render->ib);
 
-	for (i = 0; i < render->attr_count; i++) {
-		int *attr = &render->map[i];
+	FOR_EACH_BOUND_ATTR(render, i, attr) {
+		struct nouveau_array *a = &render->attrs[attr];
 
-		if (*attr >= 0) {
-			vbo_deinit_array(&render->attrs[*attr]);
-			*attr = -1;
-		}
+		if (render->mode == IMM)
+			nouveau_bo_ref(NULL, &a->bo);
+
+		nouveau_deinit_array(a);
+		render->map[i] = -1;
 	}
 
 	render->attr_count = 0;
-	context_bctx(ctx, VERTEX);
 }
 
 /* Make some rendering decisions from the GL context. */
@@ -164,20 +113,16 @@ vbo_choose_render_mode(struct gl_context *ctx, const struct gl_client_array **ar
 			}
 		}
 	}
-
-	if (render->mode == VBO)
-		render->attr_count = NUM_VERTEX_ATTRS;
-	else
-		render->attr_count = 0;
 }
 
 static void
-vbo_emit_attr(struct gl_context *ctx, const struct gl_client_array **arrays, int attr)
+vbo_emit_attr(struct gl_context *ctx, const struct gl_client_array **arrays,
+	      int attr)
 {
 	struct nouveau_channel *chan = context_chan(ctx);
 	struct nouveau_render_state *render = to_render_state(ctx);
 	const struct gl_client_array *array = arrays[attr];
-	struct nouveau_array_state *a = &render->attrs[attr];
+	struct nouveau_array *a = &render->attrs[attr];
 	RENDER_LOCALS(ctx);
 
 	if (!array->StrideB) {
@@ -186,11 +131,11 @@ vbo_emit_attr(struct gl_context *ctx, const struct gl_client_array **arrays, int
 			return;
 
 		/* Constant attribute. */
-		vbo_init_array(a, attr, array->StrideB, array->Size,
-			       array->Type, array->BufferObj, array->Ptr,
-			       GL_TRUE);
+		nouveau_init_array(a, attr, array->StrideB, array->Size,
+				   array->Type, array->BufferObj, array->Ptr,
+				   GL_TRUE);
 		EMIT_IMM(ctx, a, 0);
-		vbo_deinit_array(a);
+		nouveau_deinit_array(a);
 
 	} else {
 		/* Varying attribute. */
@@ -199,10 +144,13 @@ vbo_emit_attr(struct gl_context *ctx, const struct gl_client_array **arrays, int
 		if (render->mode == VBO) {
 			render->map[info->vbo_index] = attr;
 			render->vertex_size += array->_ElementSize;
+			render->attr_count = MAX2(render->attr_count,
+						  info->vbo_index + 1);
 		} else {
 			render->map[render->attr_count++] = attr;
 			render->vertex_size += 4 * info->imm_fields;
 		}
+
 	}
 }
 
@@ -216,6 +164,7 @@ vbo_choose_attrs(struct gl_context *ctx, const struct gl_client_array **arrays)
 
 	/* Reset the vertex size. */
 	render->vertex_size = 0;
+	render->attr_count = 0;
 
 	vbo_emit_attr(ctx, arrays, VERT_ATTRIB_COLOR0);
 	if (ctx->Fog.ColorSumEnabled && !ctx->Light.Enabled)
@@ -233,7 +182,7 @@ vbo_choose_attrs(struct gl_context *ctx, const struct gl_client_array **arrays)
 	    (ctx->Texture._GenFlags & TEXGEN_NEED_NORMALS))
 		vbo_emit_attr(ctx, arrays, VERT_ATTRIB_NORMAL);
 
-	if (ctx->Light.Enabled) {
+	if (ctx->Light.Enabled && render->mode == IMM) {
 		vbo_emit_attr(ctx, arrays, MAT(FRONT_AMBIENT));
 		vbo_emit_attr(ctx, arrays, MAT(FRONT_DIFFUSE));
 		vbo_emit_attr(ctx, arrays, MAT(FRONT_SPECULAR));
@@ -254,17 +203,13 @@ static int
 get_max_client_stride(struct gl_context *ctx, const struct gl_client_array **arrays)
 {
 	struct nouveau_render_state *render = to_render_state(ctx);
-	int i, s = 0;
-
-	for (i = 0; i < render->attr_count; i++) {
-		int attr = render->map[i];
+	int i, attr, s = 0;
 
-		if (attr >= 0) {
-			const struct gl_client_array *a = arrays[attr];
+	FOR_EACH_BOUND_ATTR(render, i, attr) {
+		const struct gl_client_array *a = arrays[attr];
 
-			if (!_mesa_is_bufferobj(a->BufferObj))
-				s = MAX2(s, get_array_stride(ctx, a));
-		}
+		if (!_mesa_is_bufferobj(a->BufferObj))
+			s = MAX2(s, get_array_stride(ctx, a));
 	}
 
 	return s;
@@ -295,7 +240,7 @@ vbo_maybe_split(struct gl_context *ctx, const struct gl_client_array **arrays,
 	if (render->mode == VBO &&
 	    (stride = get_max_client_stride(ctx, arrays)))
 		    vert_avail = MIN2(vert_avail,
-				      RENDER_SCRATCH_SIZE / stride);
+				      NOUVEAU_SCRATCH_SIZE / stride);
 
 	if (max_index - min_index > vert_avail ||
 	    (ib && ib->count > idx_avail)) {
@@ -315,42 +260,93 @@ vbo_maybe_split(struct gl_context *ctx, const struct gl_client_array **arrays,
 
 /* VBO rendering path. */
 
+static GLboolean
+check_update_array(struct nouveau_array *a, unsigned offset,
+		   struct nouveau_bo *bo, int *pdelta)
+{
+	int delta = *pdelta;
+	GLboolean dirty;
+
+	if (a->bo == bo) {
+		if (delta < 0)
+			delta = ((int)offset - (int)a->offset) / a->stride;
+
+		dirty = (delta < 0 ||
+			 offset != (a->offset + delta * a->stride));
+	} else {
+		dirty = GL_TRUE;
+	}
+
+	*pdelta = (dirty ? 0 : delta);
+	return dirty;
+}
+
 static void
 vbo_bind_vertices(struct gl_context *ctx, const struct gl_client_array **arrays,
-		  GLint basevertex, GLuint min_index, GLuint max_index)
+		  int base, unsigned min_index, unsigned max_index, int *pdelta)
 {
 	struct nouveau_render_state *render = to_render_state(ctx);
-	int i;
+	struct nouveau_channel *chan = context_chan(ctx);
+	struct nouveau_bo *bo[NUM_VERTEX_ATTRS];
+	unsigned offset[NUM_VERTEX_ATTRS];
+	GLboolean dirty = GL_FALSE;
+	int i, j, attr;
+	RENDER_LOCALS(ctx);
 
-	for (i = 0; i < NUM_VERTEX_ATTRS; i++) {
-		int attr = render->map[i];
-
-		if (attr >= 0) {
-			const struct gl_client_array *array = arrays[attr];
-			struct nouveau_array_state *a = &render->attrs[attr];
-			unsigned delta = (basevertex + min_index)
-				* array->StrideB;
-
-			if (a->bo) {
-				/* Array in a buffer obj. */
-				a->offset = (intptr_t)array->Ptr + delta;
-			} else {
-				int j, n = max_index - min_index + 1;
-				char *sp = (char *)array->Ptr + delta;
-				char *dp = get_scratch_vbo(ctx, n * a->stride,
-							   &a->bo, &a->offset);
-
-				/* Array in client memory, move it to
-				 * a scratch buffer obj. */
-				for (j = 0; j < n; j++)
-					memcpy(dp + j * a->stride,
-					       sp + j * array->StrideB,
-					       a->stride);
-			}
+	*pdelta = -1;
+
+	FOR_EACH_BOUND_ATTR(render, i, attr) {
+		const struct gl_client_array *array = arrays[attr];
+		struct gl_buffer_object *obj = array->BufferObj;
+		struct nouveau_array *a = &render->attrs[attr];
+		unsigned delta = (base + min_index) * array->StrideB;
+
+		bo[i] = NULL;
+
+		if (nouveau_bufferobj_hw(obj)) {
+			/* Array in a buffer obj. */
+			nouveau_bo_ref(to_nouveau_bufferobj(obj)->bo, &bo[i]);
+			offset[i] = delta + (intptr_t)array->Ptr;
+
+		} else {
+			int n = max_index - min_index + 1;
+			char *sp = (char *)ADD_POINTERS(
+				nouveau_bufferobj_sys(obj), array->Ptr) + delta;
+			char *dp  = nouveau_get_scratch(ctx, n * a->stride,
+							&bo[i], &offset[i]);
+
+			/* Array in client memory, move it to a
+			 * scratch buffer obj. */
+			for (j = 0; j < n; j++)
+				memcpy(dp + j * a->stride,
+				       sp + j * array->StrideB,
+				       a->stride);
 		}
+
+		dirty |= check_update_array(a, offset[i], bo[i], pdelta);
 	}
 
-	TAG(render_bind_vertices)(ctx);
+	*pdelta -= min_index;
+
+	if (dirty) {
+		/* Buffers changed, update the attribute binding. */
+		FOR_EACH_BOUND_ATTR(render, i, attr) {
+			struct nouveau_array *a = &render->attrs[attr];
+
+			nouveau_bo_ref(NULL, &a->bo);
+			a->offset = offset[i];
+			a->bo = bo[i];
+		}
+
+		TAG(render_bind_vertices)(ctx);
+
+	} else {
+		/* Just cleanup. */
+		FOR_EACH_BOUND_ATTR(render, i, attr)
+			nouveau_bo_ref(NULL, &bo[i]);
+	}
+
+	BATCH_VALIDATE();
 }
 
 static void
@@ -360,12 +356,10 @@ vbo_draw_vbo(struct gl_context *ctx, const struct gl_client_array **arrays,
 	     GLuint max_index)
 {
 	struct nouveau_channel *chan = context_chan(ctx);
-	dispatch_t dispatch;
-	int delta = -min_index, basevertex = 0, i;
+	dispatch_t dispatch = get_array_dispatch(&to_render_state(ctx)->ib);
+	int i, delta = 0, basevertex = 0;
 	RENDER_LOCALS(ctx);
 
-	get_array_dispatch(&to_render_state(ctx)->ib, &dispatch);
-
 	TAG(render_set_format)(ctx);
 
 	for (i = 0; i < nr_prims; i++) {
@@ -374,8 +368,8 @@ vbo_draw_vbo(struct gl_context *ctx, const struct gl_client_array **arrays,
 
 		if (i == 0 || basevertex != prims[i].basevertex) {
 			basevertex = prims[i].basevertex;
-			vbo_bind_vertices(ctx, arrays, basevertex,
-					  min_index, max_index);
+			vbo_bind_vertices(ctx, arrays, basevertex, min_index,
+					  max_index, &delta);
 		}
 
 		if (count > get_max_vertices(ctx, ib, AVAIL_RING(chan)))
@@ -390,7 +384,7 @@ vbo_draw_vbo(struct gl_context *ctx, const struct gl_client_array **arrays,
 /* Immediate rendering path. */
 
 static unsigned
-extract_id(struct nouveau_array_state *a, int i, int j)
+extract_id(struct nouveau_array *a, int i, int j)
 {
 	return j;
 }
@@ -404,7 +398,7 @@ vbo_draw_imm(struct gl_context *ctx, const struct gl_client_array **arrays,
 	struct nouveau_render_state *render = to_render_state(ctx);
 	struct nouveau_channel *chan = context_chan(ctx);
 	extract_u_t extract = ib ? render->ib.extract_u : extract_id;
-	int i, j, k;
+	int i, j, k, attr;
 	RENDER_LOCALS(ctx);
 
 	for (i = 0; i < nr_prims; i++) {
@@ -421,9 +415,8 @@ vbo_draw_imm(struct gl_context *ctx, const struct gl_client_array **arrays,
 			j = prims[i].basevertex +
 				extract(&render->ib, 0, start);
 
-			for (k = 0; k < render->attr_count; k++)
-				EMIT_IMM(ctx, &render->attrs[render->map[k]],
-					 j);
+			FOR_EACH_BOUND_ATTR(render, k, attr)
+				EMIT_IMM(ctx, &render->attrs[attr], j);
 		}
 
 		BATCH_END();
@@ -433,7 +426,8 @@ vbo_draw_imm(struct gl_context *ctx, const struct gl_client_array **arrays,
 /* draw_prims entry point when we're doing hw-tnl. */
 
 static void
-TAG(vbo_render_prims)(struct gl_context *ctx, const struct gl_client_array **arrays,
+TAG(vbo_render_prims)(struct gl_context *ctx,
+		      const struct gl_client_array **arrays,
 		      const struct _mesa_prim *prims, GLuint nr_prims,
 		      const struct _mesa_index_buffer *ib,
 		      GLboolean index_bounds_valid,
diff --git a/src/mesa/drivers/dri/nouveau/nv10_context.c b/src/mesa/drivers/dri/nouveau/nv10_context.c
index fdcb43b771..78be401ba2 100644
--- a/src/mesa/drivers/dri/nouveau/nv10_context.c
+++ b/src/mesa/drivers/dri/nouveau/nv10_context.c
@@ -24,6 +24,7 @@
  *
  */
 
+#include "main/state.h"
 #include "nouveau_driver.h"
 #include "nouveau_context.h"
 #include "nouveau_fbo.h"
@@ -184,6 +185,9 @@ nv10_clear(struct gl_context *ctx, GLbitfield buffers)
 			nv17_zclear(ctx, &buffers);
 		else
 			nv10_zclear(ctx, &buffers);
+
+		/* Emit the zclear state if it's dirty */
+		_mesa_update_state(ctx);
 	}
 
 	nouveau_clear(ctx, buffers);
diff --git a/src/mesa/drivers/dri/nouveau/nv10_render.c b/src/mesa/drivers/dri/nouveau/nv10_render.c
index a03ace3536..94ff3ed01e 100644
--- a/src/mesa/drivers/dri/nouveau/nv10_render.c
+++ b/src/mesa/drivers/dri/nouveau/nv10_render.c
@@ -32,7 +32,7 @@
 #define NUM_VERTEX_ATTRS 8
 
 static void
-nv10_emit_material(struct gl_context *ctx, struct nouveau_array_state *a,
+nv10_emit_material(struct gl_context *ctx, struct nouveau_array *a,
 		   const void *v);
 
 /* Vertex attribute format. */
@@ -111,13 +111,11 @@ nv10_render_set_format(struct gl_context *ctx)
 	struct nouveau_render_state *render = to_render_state(ctx);
 	struct nouveau_channel *chan = context_chan(ctx);
 	struct nouveau_grobj *celsius = context_eng3d(ctx);
-	int i, hw_format;
-
-	for (i = 0; i < NUM_VERTEX_ATTRS; i++) {
-		int attr = render->map[i];
+	int i, attr, hw_format;
 
+	FOR_EACH_ATTR(render, i, attr) {
 		if (attr >= 0) {
-			struct nouveau_array_state *a = &render->attrs[attr];
+			struct nouveau_array *a = &render->attrs[attr];
 
 			hw_format = a->stride << 8 |
 				a->fields << 4 |
@@ -140,31 +138,27 @@ nv10_render_bind_vertices(struct gl_context *ctx)
 {
 	struct nouveau_render_state *render = to_render_state(ctx);
 	struct nouveau_bo_context *bctx = context_bctx(ctx, VERTEX);
-	struct nouveau_channel *chan = context_chan(ctx);
 	struct nouveau_grobj *celsius = context_eng3d(ctx);
-	int i;
+	int i, attr;
 
-	for (i = 0; i < NUM_VERTEX_ATTRS; i++) {
-		int attr = render->map[i];
-
-		if (attr >= 0) {
-			struct nouveau_array_state *a = &render->attrs[attr];
+	FOR_EACH_BOUND_ATTR(render, i, attr) {
+		struct nouveau_array *a = &render->attrs[attr];
 
-			nouveau_bo_markl(bctx, celsius,
-					 NV10TCL_VTXBUF_ADDRESS(i),
-					 a->bo, a->offset,
-					 NOUVEAU_BO_GART | NOUVEAU_BO_RD);
-		}
+		nouveau_bo_markl(bctx, celsius,
+				 NV10TCL_VTXBUF_ADDRESS(i),
+				 a->bo, a->offset,
+				 NOUVEAU_BO_GART | NOUVEAU_BO_RD);
 	}
-
-	BEGIN_RING(chan, celsius, NV10TCL_VERTEX_ARRAY_VALIDATE, 1);
-	OUT_RING(chan, 0);
 }
 
 /* Vertex array rendering defs. */
 #define RENDER_LOCALS(ctx)					\
 	struct nouveau_grobj *celsius = context_eng3d(ctx)
 
+#define BATCH_VALIDATE()						\
+	BEGIN_RING(chan, celsius, NV10TCL_VERTEX_ARRAY_VALIDATE, 1);	\
+	OUT_RING(chan, 0)
+
 #define BATCH_BEGIN(prim)						\
 	BEGIN_RING(chan, celsius, NV10TCL_VERTEX_BUFFER_BEGIN_END, 1);	\
 	OUT_RING(chan, prim)
diff --git a/src/mesa/drivers/dri/nouveau/nv10_state_fb.c b/src/mesa/drivers/dri/nouveau/nv10_state_fb.c
index d87fe96b1c..0fda9faf49 100644
--- a/src/mesa/drivers/dri/nouveau/nv10_state_fb.c
+++ b/src/mesa/drivers/dri/nouveau/nv10_state_fb.c
@@ -51,11 +51,11 @@ get_rt_format(gl_format format)
 }
 
 static void
-setup_lma_buffer(struct gl_context *ctx)
+setup_hierz_buffer(struct gl_context *ctx)
 {
 	struct nouveau_channel *chan = context_chan(ctx);
 	struct nouveau_grobj *celsius = context_eng3d(ctx);
-	struct nouveau_bo_context *bctx = context_bctx(ctx, LMA_DEPTH);
+	struct nouveau_bo_context *bctx = context_bctx(ctx, HIERZ);
 	struct gl_framebuffer *fb = ctx->DrawBuffer;
 	struct nouveau_framebuffer *nfb = to_nouveau_framebuffer(fb);
 	unsigned pitch = align(fb->Width, 128),
@@ -135,7 +135,7 @@ nv10_emit_framebuffer(struct gl_context *ctx, int emit)
 				 s->bo, 0, bo_flags);
 
 		if (context_chipset(ctx) >= 0x17) {
-			setup_lma_buffer(ctx);
+			setup_hierz_buffer(ctx);
 			context_dirty(ctx, ZCLEAR);
 		}
 	}
diff --git a/src/mesa/drivers/dri/nouveau/nv20_context.c b/src/mesa/drivers/dri/nouveau/nv20_context.c
index c6111a2a9a..c31fd39fac 100644
--- a/src/mesa/drivers/dri/nouveau/nv20_context.c
+++ b/src/mesa/drivers/dri/nouveau/nv20_context.c
@@ -26,6 +26,8 @@
 
 #include "nouveau_driver.h"
 #include "nouveau_context.h"
+#include "nouveau_fbo.h"
+#include "nouveau_util.h"
 #include "nouveau_class.h"
 #include "nv04_driver.h"
 #include "nv10_driver.h"
@@ -40,6 +42,57 @@ static const struct dri_extension nv20_extensions[] = {
 };
 
 static void
+nv20_clear(struct gl_context *ctx, GLbitfield buffers)
+{
+	struct nouveau_channel *chan = context_chan(ctx);
+	struct nouveau_grobj *kelvin = context_eng3d(ctx);
+	struct gl_framebuffer *fb = ctx->DrawBuffer;
+	uint32_t clear = 0;
+
+	nouveau_validate_framebuffer(ctx);
+
+	if (buffers & BUFFER_BITS_COLOR) {
+		struct nouveau_surface *s = &to_nouveau_renderbuffer(
+			fb->_ColorDrawBuffers[0])->surface;
+
+		if (ctx->Color.ColorMask[0][RCOMP])
+			clear |= NV20TCL_CLEAR_BUFFERS_COLOR_R;
+		if (ctx->Color.ColorMask[0][GCOMP])
+			clear |= NV20TCL_CLEAR_BUFFERS_COLOR_G;
+		if (ctx->Color.ColorMask[0][BCOMP])
+			clear |= NV20TCL_CLEAR_BUFFERS_COLOR_B;
+		if (ctx->Color.ColorMask[0][ACOMP])
+			clear |= NV20TCL_CLEAR_BUFFERS_COLOR_A;
+
+		BEGIN_RING(chan, kelvin, NV20TCL_CLEAR_VALUE, 1);
+		OUT_RING(chan, pack_rgba_f(s->format, ctx->Color.ClearColor));
+
+		buffers &= ~BUFFER_BITS_COLOR;
+	}
+
+	if (buffers & (BUFFER_BIT_DEPTH | BUFFER_BIT_STENCIL)) {
+		struct nouveau_surface *s = &to_nouveau_renderbuffer(
+			fb->_DepthBuffer->Wrapped)->surface;
+
+		if (buffers & BUFFER_BIT_DEPTH && ctx->Depth.Mask)
+			clear |= NV20TCL_CLEAR_BUFFERS_DEPTH;
+		if (buffers & BUFFER_BIT_STENCIL && ctx->Stencil.WriteMask[0])
+			clear |= NV20TCL_CLEAR_BUFFERS_STENCIL;
+
+		BEGIN_RING(chan, kelvin, NV20TCL_CLEAR_DEPTH_VALUE, 1);
+		OUT_RING(chan, pack_zs_f(s->format, ctx->Depth.Clear,
+					 ctx->Stencil.Clear));
+
+		buffers &= ~(BUFFER_BIT_DEPTH | BUFFER_BIT_STENCIL);
+	}
+
+	BEGIN_RING(chan, kelvin, NV20TCL_CLEAR_BUFFERS, 1);
+	OUT_RING(chan, clear);
+
+	nouveau_clear(ctx, buffers);
+}
+
+static void
 nv20_hwctx_init(struct gl_context *ctx)
 {
 	struct nouveau_channel *chan = context_chan(ctx);
@@ -134,10 +187,6 @@ nv20_hwctx_init(struct gl_context *ctx)
 	OUT_RING  (chan, 2);
 
 	if (context_chipset(ctx) >= 0x25) {
-		BEGIN_RING(chan, kelvin, 0x022c, 2);
-		OUT_RING  (chan, 0x280);
-		OUT_RING  (chan, 0x07d28000);
-
 		BEGIN_RING(chan, kelvin, 0x1da4, 1);
 		OUT_RING  (chan, 0);
 	}
@@ -410,6 +459,7 @@ nv20_context_create(struct nouveau_screen *screen, const struct gl_config *visua
 	ctx->Const.MaxTextureUnits = NV20_TEXTURE_UNITS;
 	ctx->Const.MaxTextureMaxAnisotropy = 8;
 	ctx->Const.MaxTextureLodBias = 15;
+	ctx->Driver.Clear = nv20_clear;
 
 	/* 2D engine. */
 	ret = nv04_surface_init(ctx);
diff --git a/src/mesa/drivers/dri/nouveau/nv20_render.c b/src/mesa/drivers/dri/nouveau/nv20_render.c
index 6b66854462..19fc1e81fa 100644
--- a/src/mesa/drivers/dri/nouveau/nv20_render.c
+++ b/src/mesa/drivers/dri/nouveau/nv20_render.c
@@ -32,7 +32,7 @@
 #define NUM_VERTEX_ATTRS 16
 
 static void
-nv20_emit_material(struct gl_context *ctx, struct nouveau_array_state *a,
+nv20_emit_material(struct gl_context *ctx, struct nouveau_array *a,
 		   const void *v);
 
 /* Vertex attribute format. */
@@ -135,13 +135,11 @@ nv20_render_set_format(struct gl_context *ctx)
 	struct nouveau_render_state *render = to_render_state(ctx);
 	struct nouveau_channel *chan = context_chan(ctx);
 	struct nouveau_grobj *kelvin = context_eng3d(ctx);
-	int i, hw_format;
-
-	for (i = 0; i < NUM_VERTEX_ATTRS; i++) {
-		int attr = render->map[i];
+	int i, attr, hw_format;
 
+	FOR_EACH_ATTR(render, i, attr) {
 		if (attr >= 0) {
-			struct nouveau_array_state *a = &render->attrs[attr];
+			struct nouveau_array *a = &render->attrs[attr];
 
 			hw_format = a->stride << 8 |
 				a->fields << 4 |
@@ -162,33 +160,29 @@ nv20_render_bind_vertices(struct gl_context *ctx)
 {
 	struct nouveau_render_state *render = to_render_state(ctx);
 	struct nouveau_bo_context *bctx = context_bctx(ctx, VERTEX);
-	struct nouveau_channel *chan = context_chan(ctx);
 	struct nouveau_grobj *kelvin = context_eng3d(ctx);
-	int i;
+	int i, attr;
 
-	for (i = 0; i < NUM_VERTEX_ATTRS; i++) {
-		int attr = render->map[i];
+	FOR_EACH_BOUND_ATTR(render, i, attr) {
+		struct nouveau_array *a = &render->attrs[attr];
 
-		if (attr >= 0) {
-			struct nouveau_array_state *a = &render->attrs[attr];
-
-			nouveau_bo_mark(bctx, kelvin,
-					NV20TCL_VTXBUF_ADDRESS(i),
-					a->bo, a->offset, 0,
-					0, NV20TCL_VTXBUF_ADDRESS_DMA1,
-					NOUVEAU_BO_LOW | NOUVEAU_BO_OR |
-					NOUVEAU_BO_GART | NOUVEAU_BO_RD);
-		}
+		nouveau_bo_mark(bctx, kelvin,
+				NV20TCL_VTXBUF_ADDRESS(i),
+				a->bo, a->offset, 0,
+				0, NV20TCL_VTXBUF_ADDRESS_DMA1,
+				NOUVEAU_BO_LOW | NOUVEAU_BO_OR |
+				NOUVEAU_BO_GART | NOUVEAU_BO_RD);
 	}
-
-	BEGIN_RING(chan, kelvin, NV20TCL_VTX_CACHE_INVALIDATE, 1);
-	OUT_RING(chan, 0);
 }
 
 /* Vertex array rendering defs. */
 #define RENDER_LOCALS(ctx)					\
 	struct nouveau_grobj *kelvin = context_eng3d(ctx)
 
+#define BATCH_VALIDATE()						\
+	BEGIN_RING(chan, kelvin, NV20TCL_VTX_CACHE_INVALIDATE, 1);	\
+	OUT_RING(chan, 0)
+
 #define BATCH_BEGIN(prim)					\
 	BEGIN_RING(chan, kelvin, NV20TCL_VERTEX_BEGIN_END, 1);	\
 	OUT_RING(chan, prim)
diff --git a/src/mesa/drivers/dri/nouveau/nv20_state_fb.c b/src/mesa/drivers/dri/nouveau/nv20_state_fb.c
index 7822ca2a09..854392f9ff 100644
--- a/src/mesa/drivers/dri/nouveau/nv20_state_fb.c
+++ b/src/mesa/drivers/dri/nouveau/nv20_state_fb.c
@@ -51,6 +51,31 @@ get_rt_format(gl_format format)
 	}
 }
 
+static void
+setup_hierz_buffer(struct gl_context *ctx)
+{
+	struct nouveau_channel *chan = context_chan(ctx);
+	struct nouveau_grobj *kelvin = context_eng3d(ctx);
+	struct nouveau_bo_context *bctx = context_bctx(ctx, HIERZ);
+	struct gl_framebuffer *fb = ctx->DrawBuffer;
+	struct nouveau_framebuffer *nfb = to_nouveau_framebuffer(fb);
+	unsigned pitch = align(fb->Width, 128),
+		height = align(fb->Height, 2),
+		size = pitch * height;
+
+	if (!nfb->hierz.bo || nfb->hierz.bo->size != size) {
+		nouveau_bo_ref(NULL, &nfb->hierz.bo);
+		nouveau_bo_new(context_dev(ctx), NOUVEAU_BO_VRAM, 0, size,
+			       &nfb->hierz.bo);
+	}
+
+	BEGIN_RING(chan, kelvin, NV25TCL_HIERZ_PITCH, 1);
+	OUT_RING(chan, pitch);
+
+	nouveau_bo_markl(bctx, kelvin, NV25TCL_HIERZ_OFFSET, nfb->hierz.bo,
+			 0, NOUVEAU_BO_VRAM | NOUVEAU_BO_RDWR);
+}
+
 void
 nv20_emit_framebuffer(struct gl_context *ctx, int emit)
 {
@@ -88,6 +113,9 @@ nv20_emit_framebuffer(struct gl_context *ctx, int emit)
 
 		nouveau_bo_markl(bctx, kelvin, NV20TCL_ZETA_OFFSET,
 				 s->bo, 0, bo_flags);
+
+		if (context_chipset(ctx) >= 0x25)
+			setup_hierz_buffer(ctx);
 	} else {
 		rt_format |= get_rt_format(MESA_FORMAT_Z24_S8);
 		zeta_pitch = rt_pitch;
diff --git a/src/mesa/drivers/dri/r300/compiler/radeon_dataflow.c b/src/mesa/drivers/dri/r300/compiler/radeon_dataflow.c
index 5927498818..fd94194dc3 100644
--- a/src/mesa/drivers/dri/r300/compiler/radeon_dataflow.c
+++ b/src/mesa/drivers/dri/r300/compiler/radeon_dataflow.c
@@ -465,15 +465,16 @@ static void get_readers_normal_read_callback(
 {
 	struct get_readers_callback_data * d = userdata;
 	unsigned int read_mask;
+	unsigned int shared_mask;
 
 	if (src->RelAddr)
 		d->ReaderData->Abort = 1;
 
-	unsigned int shared_mask = rc_src_reads_dst_mask(src->File, src->Index,
-				src->Swizzle,
-				d->ReaderData->Writer->U.I.DstReg.File,
-				d->ReaderData->Writer->U.I.DstReg.Index,
-				d->AliveWriteMask);
+	shared_mask = rc_src_reads_dst_mask(src->File, src->Index,
+		src->Swizzle,
+		d->ReaderData->Writer->U.I.DstReg.File,
+		d->ReaderData->Writer->U.I.DstReg.Index,
+		d->AliveWriteMask);
 
 	if (shared_mask == RC_MASK_NONE)
 		return;
@@ -624,6 +625,9 @@ void  rc_get_readers_normal(
 			data->Abort = 1;
 			return;
 		case RC_OPCODE_IF:
+			/* XXX We can do better here, but this will have to
+			 * do until this dataflow analysis is more mature. */
+			data->Abort = 1;
 			branch_depth++;
 			break;
 		case RC_OPCODE_ELSE:
diff --git a/src/mesa/drivers/dri/r300/compiler/radeon_optimize.c b/src/mesa/drivers/dri/r300/compiler/radeon_optimize.c
index 5556927357..15b9c5e7dc 100644
--- a/src/mesa/drivers/dri/r300/compiler/radeon_optimize.c
+++ b/src/mesa/drivers/dri/r300/compiler/radeon_optimize.c
@@ -32,9 +32,11 @@
 #include "radeon_compiler_util.h"
 #include "radeon_swizzle.h"
 
-struct src_clobbered_data {
-	unsigned int NumSrcRegs;
-	unsigned int SrcMasks[3];
+struct src_clobbered_reads_cb_data {
+	rc_register_file File;
+	unsigned int Index;
+	unsigned int Mask;
+	struct rc_reader_data * ReaderData;
 };
 
 typedef void (*rc_presub_replace_fn)(struct rc_instruction *,
@@ -99,6 +101,25 @@ static void copy_propagate_scan_read(void * data, struct rc_instruction * inst,
 	}
 }
 
+static void src_clobbered_reads_cb(
+	void * data,
+	struct rc_instruction * inst,
+	struct rc_src_register * src)
+{
+	struct src_clobbered_reads_cb_data * sc_data = data;
+
+	if (src->File == sc_data->File
+	    && src->Index == sc_data->Index
+	    && (rc_swizzle_to_writemask(src->Swizzle) & sc_data->Mask)) {
+
+		sc_data->ReaderData->AbortOnRead = 1;
+	}
+
+	if (src->RelAddr && sc_data->File == RC_FILE_ADDRESS) {
+		sc_data->ReaderData->AbortOnRead = 1;
+	}
+}
+
 static void is_src_clobbered_scan_write(
 	void * data,
 	struct rc_instruction * inst,
@@ -106,29 +127,19 @@ static void is_src_clobbered_scan_write(
 	unsigned int index,
 	unsigned int mask)
 {
-	unsigned int i;
+	struct src_clobbered_reads_cb_data sc_data;
 	struct rc_reader_data * reader_data = data;
-	struct src_clobbered_data * d = reader_data->CbData;
-	for (i = 0; i < d->NumSrcRegs; i++) {
-		if (file == reader_data->Writer->U.I.SrcReg[i].File
-			&& index == reader_data->Writer->U.I.SrcReg[i].Index
-			&& (mask & d->SrcMasks[i])){
-
-			reader_data->AbortOnRead = 1;
-			return;
-		}
-		if (reader_data->Writer->U.I.SrcReg[i].RelAddr &&
-						file == RC_FILE_ADDRESS) {
-			reader_data->AbortOnRead = 1;
-			return;
-		}
-	}
+	sc_data.File = file;
+	sc_data.Index = index;
+	sc_data.Mask = mask;
+	sc_data.ReaderData = reader_data;
+	rc_for_all_reads_src(reader_data->Writer,
+					src_clobbered_reads_cb, &sc_data);
 }
 
 static void copy_propagate(struct radeon_compiler * c, struct rc_instruction * inst_mov)
 {
 	struct rc_reader_data reader_data;
-	struct src_clobbered_data sc_data;
 	unsigned int i;
 
 	if (inst_mov->U.I.DstReg.File != RC_FILE_TEMPORARY ||
@@ -137,12 +148,6 @@ static void copy_propagate(struct radeon_compiler * c, struct rc_instruction * i
 	    inst_mov->U.I.SaturateMode)
 		return;
 
-	sc_data.NumSrcRegs = 1;
-	sc_data.SrcMasks[0] = rc_swizzle_to_writemask(
-					inst_mov->U.I.SrcReg[0].Swizzle);
-
-	reader_data.CbData = &sc_data;
-
 	/* Get a list of all the readers of this MOV instruction. */
 	rc_get_readers_normal(c, inst_mov, &reader_data,
 			copy_propagate_scan_read, is_src_clobbered_scan_write);
@@ -203,8 +208,8 @@ static int is_src_uniform_constant(struct rc_src_register src,
 
 static void constant_folding_mad(struct rc_instruction * inst)
 {
-	rc_swizzle swz;
-	unsigned int negate;
+	rc_swizzle swz = 0;
+	unsigned int negate= 0;
 
 	if (is_src_uniform_constant(inst->U.I.SrcReg[2], &swz, &negate)) {
 		if (swz == RC_SWIZZLE_ZERO) {
@@ -244,8 +249,8 @@ static void constant_folding_mad(struct rc_instruction * inst)
 
 static void constant_folding_mul(struct rc_instruction * inst)
 {
-	rc_swizzle swz;
-	unsigned int negate;
+	rc_swizzle swz = 0;
+	unsigned int negate = 0;
 
 	if (is_src_uniform_constant(inst->U.I.SrcReg[0], &swz, &negate)) {
 		if (swz == RC_SWIZZLE_ONE) {
@@ -277,8 +282,8 @@ static void constant_folding_mul(struct rc_instruction * inst)
 
 static void constant_folding_add(struct rc_instruction * inst)
 {
-	rc_swizzle swz;
-	unsigned int negate;
+	rc_swizzle swz = 0;
+	unsigned int negate = 0;
 
 	if (is_src_uniform_constant(inst->U.I.SrcReg[0], &swz, &negate)) {
 		if (swz == RC_SWIZZLE_ZERO) {
@@ -448,15 +453,8 @@ static int presub_helper(
 	rc_presub_replace_fn presub_replace)
 {
 	struct rc_reader_data reader_data;
-	struct src_clobbered_data sc_data;
 	unsigned int i;
 
-	sc_data.NumSrcRegs = 2;
-	sc_data.SrcMasks[0] = rc_swizzle_to_writemask(
-					inst_add->U.I.SrcReg[0].Swizzle);
-	sc_data.SrcMasks[1] = rc_swizzle_to_writemask(
-					inst_add->U.I.SrcReg[1].Swizzle);
-	reader_data.CbData = &sc_data;
 	rc_get_readers_normal(c, inst_add, &reader_data, presub_scan_read,
 						is_src_clobbered_scan_write);
 
diff --git a/src/mesa/drivers/dri/r300/compiler/radeon_pair_schedule.c b/src/mesa/drivers/dri/r300/compiler/radeon_pair_schedule.c
index d4a38607d9..553e9dcf7c 100644
--- a/src/mesa/drivers/dri/r300/compiler/radeon_pair_schedule.c
+++ b/src/mesa/drivers/dri/r300/compiler/radeon_pair_schedule.c
@@ -290,6 +290,7 @@ static int merge_presub_sources(
 {
 	unsigned int srcp_src, srcp_regs, is_rgb, is_alpha;
 	struct rc_pair_sub_instruction * dst_sub;
+	const struct rc_opcode_info * info;
 
 	assert(dst_full->Alpha.Opcode == RC_OPCODE_NOP);
 
@@ -309,8 +310,8 @@ static int merge_presub_sources(
 		return 0;
 	}
 
-	const struct rc_opcode_info * info =
-					rc_get_opcode_info(dst_full->RGB.Opcode);
+	info = rc_get_opcode_info(dst_full->RGB.Opcode);
+
 	if (dst_sub->Src[RC_PAIR_PRESUB_SRC].Used)
 		return 0;
 
diff --git a/src/mesa/drivers/dri/savage/savage_xmesa.c b/src/mesa/drivers/dri/savage/savage_xmesa.c
index b3aaa0e504..92fb4f4488 100644
--- a/src/mesa/drivers/dri/savage/savage_xmesa.c
+++ b/src/mesa/drivers/dri/savage/savage_xmesa.c
@@ -50,7 +50,6 @@
 #include "savagespan.h"
 #include "savagetris.h"
 #include "savageioctl.h"
-#include "savage_bci.h"
 
 #include "savage_dri.h"
 
diff --git a/src/mesa/drivers/osmesa/osmesa.c b/src/mesa/drivers/osmesa/osmesa.c
index 37dc35cbed..10ba6b7931 100644
--- a/src/mesa/drivers/osmesa/osmesa.c
+++ b/src/mesa/drivers/osmesa/osmesa.c
@@ -1327,10 +1327,12 @@ OSMesaMakeCurrent( OSMesaContext osmesa, void *buffer, GLenum type,
     * that converts rendering from CHAN_BITS to the user-requested channel
     * size.
     */
-   osmesa->rb = new_osmesa_renderbuffer(&osmesa->mesa, osmesa->format, type);
-   _mesa_remove_renderbuffer(osmesa->gl_buffer, BUFFER_FRONT_LEFT);
-   _mesa_add_renderbuffer(osmesa->gl_buffer, BUFFER_FRONT_LEFT, osmesa->rb);
-   assert(osmesa->rb->RefCount == 2);
+   if (!osmesa->rb) {
+      osmesa->rb = new_osmesa_renderbuffer(&osmesa->mesa, osmesa->format, type);
+      _mesa_remove_renderbuffer(osmesa->gl_buffer, BUFFER_FRONT_LEFT);
+      _mesa_add_renderbuffer(osmesa->gl_buffer, BUFFER_FRONT_LEFT, osmesa->rb);
+      assert(osmesa->rb->RefCount == 2);
+   }
 
    /* Set renderbuffer fields.  Set width/height = 0 to force 
     * osmesa_renderbuffer_storage() being called by _mesa_resize_framebuffer()
diff --git a/src/mesa/main/api_exec.c b/src/mesa/main/api_exec.c
index 5f8ce3927c..25ece5c95e 100644
--- a/src/mesa/main/api_exec.c
+++ b/src/mesa/main/api_exec.c
@@ -105,46 +105,6 @@
 #if FEATURE_GL
 
 
-#ifdef _GLAPI_USE_REMAP_TABLE
-
-#define need_MESA_remap_table
-#include "main/remap.h"
-#include "main/remap_helper.h"
-
-/* This is shared across all APIs but We define this here since
- * desktop GL has the biggest remap table. */
-int driDispatchRemapTable[driDispatchRemapTable_size];
-
-/**
- * Map the functions which are already static.
- *
- * When a extension function are incorporated into the ABI, the
- * extension suffix is usually stripped.  Mapping such functions
- * makes sure the alternative names are available.
- *
- * Note that functions mapped by _mesa_init_remap_table() are
- * excluded.
- */
-void
-_mesa_map_static_functions(void)
-{
-   /* Remap static functions which have alternative names and are in the ABI.
-    * This is to be on the safe side.  glapi should have defined those names.
-    */
-   _mesa_map_function_array(MESA_alt_functions);
-}
-
-void
-_mesa_init_remap_table(void)
-{
-   _mesa_do_init_remap_table(_mesa_function_pool,
-			     driDispatchRemapTable_size,
-			     MESA_remap_table_functions);
-}
-
-#endif /* _GLAPI_USE_REMAP_TABLE */
-
-
 /**
  * Initialize a dispatch table with pointers to Mesa's immediate-mode
  * commands.
diff --git a/src/mesa/main/context.c b/src/mesa/main/context.c
index 81e862275d..b60875b712 100644
--- a/src/mesa/main/context.c
+++ b/src/mesa/main/context.c
@@ -131,6 +131,7 @@
 #if _HAVE_FULL_GL
 #include "math/m_matrix.h"
 #endif
+#include "main/dispatch.h" /* for _gloffset_COUNT */
 
 #ifdef USE_SPARC_ASM
 #include "sparc/sparc.h"
@@ -379,10 +380,12 @@ _glthread_DECLARE_STATIC_MUTEX(OneTimeLock);
 static void
 one_time_init( struct gl_context *ctx )
 {
-   static GLboolean alreadyCalled = GL_FALSE;
-   (void) ctx;
+   static GLbitfield api_init_mask = 0x0;
+
    _glthread_LOCK_MUTEX(OneTimeLock);
-   if (!alreadyCalled) {
+
+   /* truly one-time init */
+   if (!api_init_mask) {
       GLuint i;
 
       /* do some implementation tests */
@@ -395,27 +398,9 @@ one_time_init( struct gl_context *ctx )
 
       _mesa_get_cpu_features();
 
-      switch (ctx->API) {
-#if FEATURE_GL
-      case API_OPENGL:
-	 _mesa_init_remap_table();
-	 break;
-#endif
-#if FEATURE_ES1
-      case API_OPENGLES:
-	 _mesa_init_remap_table_es1();
-	 break;
-#endif
-#if FEATURE_ES2
-      case API_OPENGLES2:
-	 _mesa_init_remap_table_es2();
-	 break;
-#endif
-      default:
-	 break;
-      }
-
       _mesa_init_sqrt_table();
+
+      /* context dependence is never a one-time thing... */
       _mesa_init_get_hash(ctx);
 
       for (i = 0; i < 256; i++) {
@@ -426,9 +411,22 @@ one_time_init( struct gl_context *ctx )
       _mesa_debug(ctx, "Mesa %s DEBUG build %s %s\n",
                   MESA_VERSION_STRING, __DATE__, __TIME__);
 #endif
+   }
 
-      alreadyCalled = GL_TRUE;
+   /* per-API one-time init */
+   if (!(api_init_mask & (1 << ctx->API))) {
+      /*
+       * This is fine as ES does not use the remap table, but it may not be
+       * future-proof.  We cannot always initialize the remap table because
+       * when an app is linked to libGLES*, there are not enough dynamic
+       * entries.
+       */
+      if (ctx->API == API_OPENGL)
+         _mesa_init_remap_table();
    }
+
+   api_init_mask |= 1 << ctx->API;
+
    _glthread_UNLOCK_MUTEX(OneTimeLock);
 
    /* Hopefully atexit() is widely available.  If not, we may need some
@@ -811,9 +809,13 @@ _mesa_alloc_dispatch_table(int size)
     * Mesa we do this to accomodate different versions of libGL and various
     * DRI drivers.
     */
-   GLint numEntries = MAX2(_glapi_get_dispatch_table_size(), size);
-   struct _glapi_table *table =
-      (struct _glapi_table *) malloc(numEntries * sizeof(_glapi_proc));
+   GLint numEntries = MAX2(_glapi_get_dispatch_table_size(), _gloffset_COUNT);
+   struct _glapi_table *table;
+
+   /* should never happen, but just in case */
+   numEntries = MAX2(numEntries, size);
+
+   table = (struct _glapi_table *) malloc(numEntries * sizeof(_glapi_proc));
    if (table) {
       _glapi_proc *entry = (_glapi_proc *) table;
       GLint i;
@@ -1290,9 +1292,6 @@ check_compatible(const struct gl_context *ctx, const struct gl_framebuffer *buff
    const struct gl_config *ctxvis = &ctx->Visual;
    const struct gl_config *bufvis = &buffer->Visual;
 
-   if (ctxvis == bufvis)
-      return GL_TRUE;
-
    if (buffer == _mesa_get_incomplete_framebuffer())
       return GL_TRUE;
 
diff --git a/src/mesa/main/dispatch.h b/src/mesa/main/dispatch.h
index 552384c500..a597959cf8 100644
--- a/src/mesa/main/dispatch.h
+++ b/src/mesa/main/dispatch.h
@@ -26,7 +26,9 @@
 #ifndef _DISPATCH_H
 #define _DISPATCH_H
 
-#ifdef IN_DRI_DRIVER
+#include "main/mfeatures.h"
+
+#if FEATURE_remap_table
 #define _GLAPI_USE_REMAP_TABLE
 #endif
 
diff --git a/src/mesa/main/es_generator.py b/src/mesa/main/es_generator.py
index 8f28da16df..aa8dab7a74 100644
--- a/src/mesa/main/es_generator.py
+++ b/src/mesa/main/es_generator.py
@@ -191,6 +191,8 @@ print """
 #include "%s"
 #include "%s"
 #include "main/mfeatures.h"
+#include "main/compiler.h"
+#include "main/api_exec.h"
 
 #if FEATURE_%s
 """ % (versionHeader, versionExtHeader, shortname.upper())
@@ -206,48 +208,7 @@ typedef double GLclampd;
 /* Mesa error handling requires these */
 extern void *_mesa_get_current_context(void);
 extern void _mesa_error(void *ctx, GLenum error, const char *fmtString, ... );
-
-#include "main/compiler.h"
-#include "main/api_exec.h"
-#include "main/remap.h"
-
-/* cannot include main/dispatch.h here */
-#ifdef IN_DRI_DRIVER
-#define _GLAPI_USE_REMAP_TABLE
-#endif
-/* glapi uses GLAPIENTRY while GLES headers define GL_APIENTRY */
-#ifndef GLAPIENTRY
-#define GLAPIENTRY GL_APIENTRY
-#endif
-#include "%sapi/main/glapidispatch.h"
-
-#if FEATURE_remap_table
-
-#if !FEATURE_GL
-int driDispatchRemapTable[driDispatchRemapTable_size];
-#endif
-
-#define need_MESA_remap_table
-
-#include "%sapi/main/remap_helper.h"
-
-void
-_mesa_init_remap_table_%s(void)
-{
-   _mesa_do_init_remap_table(_mesa_function_pool,
-                             driDispatchRemapTable_size,
-                             MESA_remap_table_functions);
-}
-
-void
-_mesa_map_static_functions_%s(void)
-{
-}
-
-#endif
-
-typedef void (*_glapi_proc)(void); /* generic function pointer */
-""" % (shortname, shortname, shortname, shortname);
+"""
 
 # Finally we get to the all-important functions
 print """/*************************************************************
@@ -711,15 +672,73 @@ for funcName in keys:
 # end for each function
 
 print """
+#include "glapi/glapi.h"
+
+#if FEATURE_remap_table
+
+/* cannot include main/dispatch.h here */
+#define _GLAPI_USE_REMAP_TABLE
+#include "%sapi/main/glapidispatch.h"
+
+#define need_MESA_remap_table
+#include "%sapi/main/remap_helper.h"
+
+/* force SET_* macros to use the local remap table */
+#define driDispatchRemapTable remap_table
+static int remap_table[driDispatchRemapTable_size];
+
+static void
+init_remap_table(void)
+{
+   _glthread_DECLARE_STATIC_MUTEX(mutex);
+   static GLboolean initialized = GL_FALSE;
+   const struct gl_function_pool_remap *remap = MESA_remap_table_functions;
+   int i;
+
+   _glthread_LOCK_MUTEX(mutex);
+   if (initialized) {
+      _glthread_UNLOCK_MUTEX(mutex);
+      return;
+   }
+
+   for (i = 0; i < driDispatchRemapTable_size; i++) {
+      GLint offset;
+      const char *spec;
+
+      /* sanity check */
+      ASSERT(i == remap[i].remap_index);
+      spec = _mesa_function_pool + remap[i].pool_index;
+
+      offset = _mesa_map_function_spec(spec);
+      remap_table[i] = offset;
+   }
+   initialized = GL_TRUE;
+   _glthread_UNLOCK_MUTEX(mutex);
+}
+
+#else /* FEATURE_remap_table */
+
+/* cannot include main/dispatch.h here */
+#include "%sapi/main/glapidispatch.h"
+
+static INLINE void
+init_remap_table(void)
+{
+}
+
+#endif /* FEATURE_remap_table */
+
 struct _glapi_table *
 _mesa_create_exec_table_%s(void)
 {
    struct _glapi_table *exec;
+
    exec = _mesa_alloc_dispatch_table(_gloffset_COUNT);
    if (exec == NULL)
       return NULL;
 
-""" % shortname
+   init_remap_table();
+""" % (shortname, shortname, shortname, shortname)
 
 for func in keys:
     prefix = "_es_" if func not in allSpecials else "_check_"
diff --git a/src/mesa/main/mfeatures.h b/src/mesa/main/mfeatures.h
index 5afd65d976..911b30dda2 100644
--- a/src/mesa/main/mfeatures.h
+++ b/src/mesa/main/mfeatures.h
@@ -78,7 +78,7 @@
 #define FEATURE_GL !FEATURE_ES
 #endif
 
-#ifdef IN_DRI_DRIVER
+#if defined(IN_DRI_DRIVER) || (FEATURE_GL + FEATURE_ES1 + FEATURE_ES2 > 1)
 #define FEATURE_remap_table               1
 #else
 #define FEATURE_remap_table               0
diff --git a/src/mesa/main/mtypes.h b/src/mesa/main/mtypes.h
index f8060f32de..51378f7ac8 100644
--- a/src/mesa/main/mtypes.h
+++ b/src/mesa/main/mtypes.h
@@ -454,6 +454,68 @@ typedef enum
 
 
 /**
+ * Framebuffer configuration (aka visual / pixelformat)
+ * Note: some of these fields should be boolean, but it appears that
+ * code in drivers/dri/common/util.c requires int-sized fields.
+ */
+struct gl_config
+{
+   GLboolean rgbMode;
+   GLboolean floatMode;
+   GLboolean colorIndexMode;  /* XXX is this used anywhere? */
+   GLuint doubleBufferMode;
+   GLuint stereoMode;
+
+   GLboolean haveAccumBuffer;
+   GLboolean haveDepthBuffer;
+   GLboolean haveStencilBuffer;
+
+   GLint redBits, greenBits, blueBits, alphaBits;	/* bits per comp */
+   GLuint redMask, greenMask, blueMask, alphaMask;
+   GLint rgbBits;		/* total bits for rgb */
+   GLint indexBits;		/* total bits for colorindex */
+
+   GLint accumRedBits, accumGreenBits, accumBlueBits, accumAlphaBits;
+   GLint depthBits;
+   GLint stencilBits;
+
+   GLint numAuxBuffers;
+
+   GLint level;
+
+   /* EXT_visual_rating / GLX 1.2 */
+   GLint visualRating;
+
+   /* EXT_visual_info / GLX 1.2 */
+   GLint transparentPixel;
+   /*    colors are floats scaled to ints */
+   GLint transparentRed, transparentGreen, transparentBlue, transparentAlpha;
+   GLint transparentIndex;
+
+   /* ARB_multisample / SGIS_multisample */
+   GLint sampleBuffers;
+   GLint samples;
+
+   /* SGIX_pbuffer / GLX 1.3 */
+   GLint maxPbufferWidth;
+   GLint maxPbufferHeight;
+   GLint maxPbufferPixels;
+   GLint optimalPbufferWidth;   /* Only for SGIX_pbuffer. */
+   GLint optimalPbufferHeight;  /* Only for SGIX_pbuffer. */
+
+   /* OML_swap_method */
+   GLint swapMethod;
+
+   /* EXT_texture_from_pixmap */
+   GLint bindToTextureRgb;
+   GLint bindToTextureRgba;
+   GLint bindToMipmapTexture;
+   GLint bindToTextureTargets;
+   GLint yInverted;
+};
+
+
+/**
  * Data structure for color tables
  */
 struct gl_color_table
@@ -547,60 +609,6 @@ struct gl_shine_tab
    GLuint refcount;
 };
 
-struct gl_config {
-   GLboolean rgbMode;
-   GLboolean floatMode;
-   GLboolean colorIndexMode;
-   GLuint doubleBufferMode;
-   GLuint stereoMode;
-
-   GLboolean haveAccumBuffer;
-   GLboolean haveDepthBuffer;
-   GLboolean haveStencilBuffer;
-
-   GLint redBits, greenBits, blueBits, alphaBits;	/* bits per comp */
-   GLuint redMask, greenMask, blueMask, alphaMask;
-   GLint rgbBits;		/* total bits for rgb */
-   GLint indexBits;		/* total bits for colorindex */
-
-   GLint accumRedBits, accumGreenBits, accumBlueBits, accumAlphaBits;
-   GLint depthBits;
-   GLint stencilBits;
-
-   GLint numAuxBuffers;
-
-   GLint level;
-
-   /* EXT_visual_rating / GLX 1.2 */
-   GLint visualRating;
-
-   /* EXT_visual_info / GLX 1.2 */
-   GLint transparentPixel;
-   /*    colors are floats scaled to ints */
-   GLint transparentRed, transparentGreen, transparentBlue, transparentAlpha;
-   GLint transparentIndex;
-
-   /* ARB_multisample / SGIS_multisample */
-   GLint sampleBuffers;
-   GLint samples;
-
-   /* SGIX_pbuffer / GLX 1.3 */
-   GLint maxPbufferWidth;
-   GLint maxPbufferHeight;
-   GLint maxPbufferPixels;
-   GLint optimalPbufferWidth;   /* Only for SGIX_pbuffer. */
-   GLint optimalPbufferHeight;  /* Only for SGIX_pbuffer. */
-
-   /* OML_swap_method */
-   GLint swapMethod;
-
-   /* EXT_texture_from_pixmap */
-   GLint bindToTextureRgb;
-   GLint bindToTextureRgba;
-   GLint bindToMipmapTexture;
-   GLint bindToTextureTargets;
-   GLint yInverted;
-};
 
 /**
  * Light source state.
diff --git a/src/mesa/main/pack.c b/src/mesa/main/pack.c
index 877d27ce1e..542fcc4f53 100644
--- a/src/mesa/main/pack.c
+++ b/src/mesa/main/pack.c
@@ -33,6 +33,7 @@
 #include "colormac.h"
 #include "enums.h"
 #include "image.h"
+#include "imports.h"
 #include "pack.h"
 #include "pixeltransfer.h"
 #include "imports.h"
@@ -2693,9 +2694,589 @@ extract_float_rgba(GLuint n, GLfloat rgba[][4],
          _mesa_problem(NULL, "bad srcType in extract float data");
          break;
    }
+#undef PROCESS
+}
+
+
+static INLINE GLuint
+clamp_byte_to_uint(GLbyte b)
+{
+   return b < 0 ? 0 : b;
+}
+
+
+static INLINE GLuint
+clamp_short_to_uint(GLshort s)
+{
+   return s < 0 ? 0 : s;
+}
+
+
+static INLINE GLuint
+clamp_int_to_uint(GLint i)
+{
+   return i < 0 ? 0 : i;
+}
+
+
+static INLINE GLuint
+clamp_float_to_uint(GLfloat f)
+{
+   return f < 0.0F ? 0 : IROUND(f);
+}
+
+
+static INLINE GLuint
+clamp_half_to_uint(GLhalfARB h)
+{
+   GLfloat f = _mesa_half_to_float(h);
+   return f < 0.0F ? 0 : IROUND(f);
+}
+
+
+/**
+ * \sa extract_float_rgba()
+ */
+static void
+extract_uint_rgba(GLuint n, GLuint rgba[][4],
+                  GLenum srcFormat, GLenum srcType, const GLvoid *src,
+                  GLboolean swapBytes)
+{
+   GLint redIndex, greenIndex, blueIndex, alphaIndex;
+   GLint stride;
+   GLint rComp, bComp, gComp, aComp;
+   GLboolean intFormat;
+
+   ASSERT(srcFormat == GL_RED ||
+          srcFormat == GL_GREEN ||
+          srcFormat == GL_BLUE ||
+          srcFormat == GL_ALPHA ||
+          srcFormat == GL_LUMINANCE ||
+          srcFormat == GL_LUMINANCE_ALPHA ||
+          srcFormat == GL_INTENSITY ||
+          srcFormat == GL_RG ||
+          srcFormat == GL_RGB ||
+          srcFormat == GL_BGR ||
+          srcFormat == GL_RGBA ||
+          srcFormat == GL_BGRA ||
+          srcFormat == GL_ABGR_EXT ||
+          srcFormat == GL_DU8DV8_ATI ||
+          srcFormat == GL_DUDV_ATI ||
+          srcFormat == GL_RED_INTEGER_EXT ||
+          srcFormat == GL_GREEN_INTEGER_EXT ||
+          srcFormat == GL_BLUE_INTEGER_EXT ||
+          srcFormat == GL_ALPHA_INTEGER_EXT ||
+          srcFormat == GL_RGB_INTEGER_EXT ||
+          srcFormat == GL_RGBA_INTEGER_EXT ||
+          srcFormat == GL_BGR_INTEGER_EXT ||
+          srcFormat == GL_BGRA_INTEGER_EXT ||
+          srcFormat == GL_LUMINANCE_INTEGER_EXT ||
+          srcFormat == GL_LUMINANCE_ALPHA_INTEGER_EXT);
+
+   ASSERT(srcType == GL_UNSIGNED_BYTE ||
+          srcType == GL_BYTE ||
+          srcType == GL_UNSIGNED_SHORT ||
+          srcType == GL_SHORT ||
+          srcType == GL_UNSIGNED_INT ||
+          srcType == GL_INT ||
+          srcType == GL_HALF_FLOAT_ARB ||
+          srcType == GL_FLOAT ||
+          srcType == GL_UNSIGNED_BYTE_3_3_2 ||
+          srcType == GL_UNSIGNED_BYTE_2_3_3_REV ||
+          srcType == GL_UNSIGNED_SHORT_5_6_5 ||
+          srcType == GL_UNSIGNED_SHORT_5_6_5_REV ||
+          srcType == GL_UNSIGNED_SHORT_4_4_4_4 ||
+          srcType == GL_UNSIGNED_SHORT_4_4_4_4_REV ||
+          srcType == GL_UNSIGNED_SHORT_5_5_5_1 ||
+          srcType == GL_UNSIGNED_SHORT_1_5_5_5_REV ||
+          srcType == GL_UNSIGNED_INT_8_8_8_8 ||
+          srcType == GL_UNSIGNED_INT_8_8_8_8_REV ||
+          srcType == GL_UNSIGNED_INT_10_10_10_2 ||
+          srcType == GL_UNSIGNED_INT_2_10_10_10_REV);
+
+   rComp = gComp = bComp = aComp = -1;
+
+   switch (srcFormat) {
+      case GL_RED:
+      case GL_RED_INTEGER_EXT:
+         redIndex = 0;
+         greenIndex = blueIndex = alphaIndex = -1;
+         stride = 1;
+         break;
+      case GL_GREEN:
+      case GL_GREEN_INTEGER_EXT:
+         greenIndex = 0;
+         redIndex = blueIndex = alphaIndex = -1;
+         stride = 1;
+         break;
+      case GL_BLUE:
+      case GL_BLUE_INTEGER_EXT:
+         blueIndex = 0;
+         redIndex = greenIndex = alphaIndex = -1;
+         stride = 1;
+         break;
+      case GL_ALPHA:
+      case GL_ALPHA_INTEGER_EXT:
+         redIndex = greenIndex = blueIndex = -1;
+         alphaIndex = 0;
+         stride = 1;
+         break;
+      case GL_LUMINANCE:
+      case GL_LUMINANCE_INTEGER_EXT:
+         redIndex = greenIndex = blueIndex = 0;
+         alphaIndex = -1;
+         stride = 1;
+         break;
+      case GL_LUMINANCE_ALPHA:
+      case GL_LUMINANCE_ALPHA_INTEGER_EXT:
+         redIndex = greenIndex = blueIndex = 0;
+         alphaIndex = 1;
+         stride = 2;
+         break;
+      case GL_INTENSITY:
+         redIndex = greenIndex = blueIndex = alphaIndex = 0;
+         stride = 1;
+         break;
+      case GL_RG:
+      case GL_RG_INTEGER:
+         redIndex = 0;
+         greenIndex = 1;
+         blueIndex = -1;
+         alphaIndex = -1;
+         rComp = 0;
+         gComp = 1;
+         bComp = 2;
+         aComp = 3;
+         stride = 2;
+         break;
+      case GL_RGB:
+      case GL_RGB_INTEGER:
+         redIndex = 0;
+         greenIndex = 1;
+         blueIndex = 2;
+         alphaIndex = -1;
+         rComp = 0;
+         gComp = 1;
+         bComp = 2;
+         aComp = 3;
+         stride = 3;
+         break;
+      case GL_BGR:
+         redIndex = 2;
+         greenIndex = 1;
+         blueIndex = 0;
+         alphaIndex = -1;
+         rComp = 2;
+         gComp = 1;
+         bComp = 0;
+         aComp = 3;
+         stride = 3;
+         break;
+      case GL_RGBA:
+      case GL_RGBA_INTEGER:
+         redIndex = 0;
+         greenIndex = 1;
+         blueIndex = 2;
+         alphaIndex = 3;
+         rComp = 0;
+         gComp = 1;
+         bComp = 2;
+         aComp = 3;
+         stride = 4;
+         break;
+      case GL_BGRA:
+         redIndex = 2;
+         greenIndex = 1;
+         blueIndex = 0;
+         alphaIndex = 3;
+         rComp = 2;
+         gComp = 1;
+         bComp = 0;
+         aComp = 3;
+         stride = 4;
+         break;
+      case GL_ABGR_EXT:
+         redIndex = 3;
+         greenIndex = 2;
+         blueIndex = 1;
+         alphaIndex = 0;
+         rComp = 3;
+         gComp = 2;
+         bComp = 1;
+         aComp = 0;
+         stride = 4;
+         break;
+      default:
+         _mesa_problem(NULL, "bad srcFormat %s in extract float data",
+                       _mesa_lookup_enum_by_nr(srcFormat));
+         return;
+   }
+
+   intFormat = _mesa_is_integer_format(srcFormat);
+
+#define PROCESS(INDEX, CHANNEL, DEFAULT, TYPE, CONVERSION)		\
+   if ((INDEX) < 0) {							\
+      GLuint i;								\
+      for (i = 0; i < n; i++) {						\
+         rgba[i][CHANNEL] = DEFAULT;					\
+      }									\
+   }									\
+   else if (swapBytes) {						\
+      const TYPE *s = (const TYPE *) src;				\
+      GLuint i;								\
+      for (i = 0; i < n; i++) {						\
+         TYPE value = s[INDEX];						\
+         if (sizeof(TYPE) == 2) {					\
+            SWAP2BYTE(value);						\
+         }								\
+         else if (sizeof(TYPE) == 4) {					\
+            SWAP4BYTE(value);						\
+         }								\
+         rgba[i][CHANNEL] = CONVERSION(value);                          \
+         s += stride;							\
+      }									\
+   }									\
+   else {								\
+      const TYPE *s = (const TYPE *) src;				\
+      GLuint i;								\
+      for (i = 0; i < n; i++) {						\
+         rgba[i][CHANNEL] = CONVERSION(s[INDEX]);			\
+         s += stride;							\
+      }									\
+   }
+
+   switch (srcType) {
+      case GL_UNSIGNED_BYTE:
+         PROCESS(redIndex,   RCOMP, 0, GLubyte, (GLuint));
+         PROCESS(greenIndex, GCOMP, 0, GLubyte, (GLuint));
+         PROCESS(blueIndex,  BCOMP, 0, GLubyte, (GLuint));
+         PROCESS(alphaIndex, ACOMP, 1, GLubyte, (GLuint));
+         break;
+      case GL_BYTE:
+         PROCESS(redIndex,   RCOMP, 0, GLbyte, clamp_byte_to_uint);
+         PROCESS(greenIndex, GCOMP, 0, GLbyte, clamp_byte_to_uint);
+         PROCESS(blueIndex,  BCOMP, 0, GLbyte, clamp_byte_to_uint);
+         PROCESS(alphaIndex, ACOMP, 1, GLbyte, clamp_byte_to_uint);
+         break;
+      case GL_UNSIGNED_SHORT:
+         PROCESS(redIndex,   RCOMP, 0, GLushort, (GLuint));
+         PROCESS(greenIndex, GCOMP, 0, GLushort, (GLuint));
+         PROCESS(blueIndex,  BCOMP, 0, GLushort, (GLuint));
+         PROCESS(alphaIndex, ACOMP, 1, GLushort, (GLuint));
+         break;
+      case GL_SHORT:
+         PROCESS(redIndex,   RCOMP, 0, GLshort, clamp_short_to_uint);
+         PROCESS(greenIndex, GCOMP, 0, GLshort, clamp_short_to_uint);
+         PROCESS(blueIndex,  BCOMP, 0, GLshort, clamp_short_to_uint);
+         PROCESS(alphaIndex, ACOMP, 1, GLshort, clamp_short_to_uint);
+         break;
+      case GL_UNSIGNED_INT:
+         PROCESS(redIndex,   RCOMP, 0, GLuint, (GLuint));
+         PROCESS(greenIndex, GCOMP, 0, GLuint, (GLuint));
+         PROCESS(blueIndex,  BCOMP, 0, GLuint, (GLuint));
+         PROCESS(alphaIndex, ACOMP, 1, GLuint, (GLuint));
+         break;
+      case GL_INT:
+         PROCESS(redIndex,   RCOMP, 0, GLint, clamp_int_to_uint);
+         PROCESS(greenIndex, GCOMP, 0, GLint, clamp_int_to_uint);
+         PROCESS(blueIndex,  BCOMP, 0, GLint, clamp_int_to_uint);
+         PROCESS(alphaIndex, ACOMP, 1, GLint, clamp_int_to_uint);
+         break;
+      case GL_FLOAT:
+         PROCESS(redIndex,   RCOMP, 0, GLfloat, clamp_float_to_uint);
+         PROCESS(greenIndex, GCOMP, 0, GLfloat, clamp_float_to_uint);
+         PROCESS(blueIndex,  BCOMP, 0, GLfloat, clamp_float_to_uint);
+         PROCESS(alphaIndex, ACOMP, 1, GLfloat, clamp_float_to_uint);
+         break;
+      case GL_HALF_FLOAT_ARB:
+         PROCESS(redIndex,   RCOMP, 0, GLhalfARB, clamp_half_to_uint);
+         PROCESS(greenIndex, GCOMP, 0, GLhalfARB, clamp_half_to_uint);
+         PROCESS(blueIndex,  BCOMP, 0, GLhalfARB, clamp_half_to_uint);
+         PROCESS(alphaIndex, ACOMP, 1, GLhalfARB, clamp_half_to_uint);
+         break;
+      case GL_UNSIGNED_BYTE_3_3_2:
+         {
+            const GLubyte *ubsrc = (const GLubyte *) src;
+            GLuint i;
+            for (i = 0; i < n; i ++) {
+               GLubyte p = ubsrc[i];
+               rgba[i][rComp] = ((p >> 5)      );
+               rgba[i][gComp] = ((p >> 2) & 0x7);
+               rgba[i][bComp] = ((p     ) & 0x3);
+               rgba[i][aComp] = 1;
+            }
+         }
+         break;
+      case GL_UNSIGNED_BYTE_2_3_3_REV:
+         {
+            const GLubyte *ubsrc = (const GLubyte *) src;
+            GLuint i;
+            for (i = 0; i < n; i ++) {
+               GLubyte p = ubsrc[i];
+               rgba[i][rComp] = ((p     ) & 0x7);
+               rgba[i][gComp] = ((p >> 3) & 0x7);
+               rgba[i][bComp] = ((p >> 6)      );
+               rgba[i][aComp] = 1;
+            }
+         }
+         break;
+      case GL_UNSIGNED_SHORT_5_6_5:
+         if (swapBytes) {
+            const GLushort *ussrc = (const GLushort *) src;
+            GLuint i;
+            for (i = 0; i < n; i ++) {
+               GLushort p = ussrc[i];
+               SWAP2BYTE(p);
+               rgba[i][rComp] = ((p >> 11)       );
+               rgba[i][gComp] = ((p >>  5) & 0x3f);
+               rgba[i][bComp] = ((p      ) & 0x1f);
+               rgba[i][aComp] = 1;
+            }
+         }
+         else {
+            const GLushort *ussrc = (const GLushort *) src;
+            GLuint i;
+            for (i = 0; i < n; i ++) {
+               GLushort p = ussrc[i];
+               rgba[i][rComp] = ((p >> 11)       );
+               rgba[i][gComp] = ((p >>  5) & 0x3f);
+               rgba[i][bComp] = ((p      ) & 0x1f);
+               rgba[i][aComp] = 1;
+            }
+         }
+         break;
+      case GL_UNSIGNED_SHORT_5_6_5_REV:
+         if (swapBytes) {
+            const GLushort *ussrc = (const GLushort *) src;
+            GLuint i;
+            for (i = 0; i < n; i ++) {
+               GLushort p = ussrc[i];
+               SWAP2BYTE(p);
+               rgba[i][rComp] = ((p      ) & 0x1f);
+               rgba[i][gComp] = ((p >>  5) & 0x3f);
+               rgba[i][bComp] = ((p >> 11)       );
+               rgba[i][aComp] = 1;
+            }
+         }
+         else {
+            const GLushort *ussrc = (const GLushort *) src;
+            GLuint i;
+            for (i = 0; i < n; i ++) {
+               GLushort p = ussrc[i];
+               rgba[i][rComp] = ((p      ) & 0x1f);
+               rgba[i][gComp] = ((p >>  5) & 0x3f);
+               rgba[i][bComp] = ((p >> 11)       );
+               rgba[i][aComp] = 1;
+            }
+         }
+         break;
+      case GL_UNSIGNED_SHORT_4_4_4_4:
+         if (swapBytes) {
+            const GLushort *ussrc = (const GLushort *) src;
+            GLuint i;
+            for (i = 0; i < n; i ++) {
+               GLushort p = ussrc[i];
+               SWAP2BYTE(p);
+               rgba[i][rComp] = ((p >> 12)      );
+               rgba[i][gComp] = ((p >>  8) & 0xf);
+               rgba[i][bComp] = ((p >>  4) & 0xf);
+               rgba[i][aComp] = ((p      ) & 0xf);
+            }
+         }
+         else {
+            const GLushort *ussrc = (const GLushort *) src;
+            GLuint i;
+            for (i = 0; i < n; i ++) {
+               GLushort p = ussrc[i];
+               rgba[i][rComp] = ((p >> 12)      );
+               rgba[i][gComp] = ((p >>  8) & 0xf);
+               rgba[i][bComp] = ((p >>  4) & 0xf);
+               rgba[i][aComp] = ((p      ) & 0xf);
+            }
+         }
+         break;
+      case GL_UNSIGNED_SHORT_4_4_4_4_REV:
+         if (swapBytes) {
+            const GLushort *ussrc = (const GLushort *) src;
+            GLuint i;
+            for (i = 0; i < n; i ++) {
+               GLushort p = ussrc[i];
+               SWAP2BYTE(p);
+               rgba[i][rComp] = ((p      ) & 0xf);
+               rgba[i][gComp] = ((p >>  4) & 0xf);
+               rgba[i][bComp] = ((p >>  8) & 0xf);
+               rgba[i][aComp] = ((p >> 12)      );
+            }
+         }
+         else {
+            const GLushort *ussrc = (const GLushort *) src;
+            GLuint i;
+            for (i = 0; i < n; i ++) {
+               GLushort p = ussrc[i];
+               rgba[i][rComp] = ((p      ) & 0xf);
+               rgba[i][gComp] = ((p >>  4) & 0xf);
+               rgba[i][bComp] = ((p >>  8) & 0xf);
+               rgba[i][aComp] = ((p >> 12)      );
+            }
+         }
+         break;
+      case GL_UNSIGNED_SHORT_5_5_5_1:
+         if (swapBytes) {
+            const GLushort *ussrc = (const GLushort *) src;
+            GLuint i;
+            for (i = 0; i < n; i ++) {
+               GLushort p = ussrc[i];
+               SWAP2BYTE(p);
+               rgba[i][rComp] = ((p >> 11)       );
+               rgba[i][gComp] = ((p >>  6) & 0x1f);
+               rgba[i][bComp] = ((p >>  1) & 0x1f);
+               rgba[i][aComp] = ((p      ) & 0x1 );
+            }
+         }
+         else {
+            const GLushort *ussrc = (const GLushort *) src;
+            GLuint i;
+            for (i = 0; i < n; i ++) {
+               GLushort p = ussrc[i];
+               rgba[i][rComp] = ((p >> 11)       );
+               rgba[i][gComp] = ((p >>  6) & 0x1f);
+               rgba[i][bComp] = ((p >>  1) & 0x1f);
+               rgba[i][aComp] = ((p      ) & 0x1 );
+            }
+         }
+         break;
+      case GL_UNSIGNED_SHORT_1_5_5_5_REV:
+         if (swapBytes) {
+            const GLushort *ussrc = (const GLushort *) src;
+            GLuint i;
+            for (i = 0; i < n; i ++) {
+               GLushort p = ussrc[i];
+               SWAP2BYTE(p);
+               rgba[i][rComp] = ((p      ) & 0x1f);
+               rgba[i][gComp] = ((p >>  5) & 0x1f);
+               rgba[i][bComp] = ((p >> 10) & 0x1f);
+               rgba[i][aComp] = ((p >> 15)       );
+            }
+         }
+         else {
+            const GLushort *ussrc = (const GLushort *) src;
+            GLuint i;
+            for (i = 0; i < n; i ++) {
+               GLushort p = ussrc[i];
+               rgba[i][rComp] = ((p      ) & 0x1f);
+               rgba[i][gComp] = ((p >>  5) & 0x1f);
+               rgba[i][bComp] = ((p >> 10) & 0x1f);
+               rgba[i][aComp] = ((p >> 15)       );
+            }
+         }
+         break;
+      case GL_UNSIGNED_INT_8_8_8_8:
+         if (swapBytes) {
+            const GLuint *uisrc = (const GLuint *) src;
+            GLuint i;
+            for (i = 0; i < n; i ++) {
+               GLuint p = uisrc[i];
+               rgba[i][rComp] = ((p      ) & 0xff);
+               rgba[i][gComp] = ((p >>  8) & 0xff);
+               rgba[i][bComp] = ((p >> 16) & 0xff);
+               rgba[i][aComp] = ((p >> 24)       );
+            }
+         }
+         else {
+            const GLuint *uisrc = (const GLuint *) src;
+            GLuint i;
+            for (i = 0; i < n; i ++) {
+               GLuint p = uisrc[i];
+               rgba[i][rComp] = ((p >> 24)       );
+               rgba[i][gComp] = ((p >> 16) & 0xff);
+               rgba[i][bComp] = ((p >>  8) & 0xff);
+               rgba[i][aComp] = ((p      ) & 0xff);
+            }
+         }
+         break;
+      case GL_UNSIGNED_INT_8_8_8_8_REV:
+         if (swapBytes) {
+            const GLuint *uisrc = (const GLuint *) src;
+            GLuint i;
+            for (i = 0; i < n; i ++) {
+               GLuint p = uisrc[i];
+               rgba[i][rComp] = ((p >> 24)       );
+               rgba[i][gComp] = ((p >> 16) & 0xff);
+               rgba[i][bComp] = ((p >>  8) & 0xff);
+               rgba[i][aComp] = ((p      ) & 0xff);
+            }
+         }
+         else {
+            const GLuint *uisrc = (const GLuint *) src;
+            GLuint i;
+            for (i = 0; i < n; i ++) {
+               GLuint p = uisrc[i];
+               rgba[i][rComp] = ((p      ) & 0xff);
+               rgba[i][gComp] = ((p >>  8) & 0xff);
+               rgba[i][bComp] = ((p >> 16) & 0xff);
+               rgba[i][aComp] = ((p >> 24)       );
+            }
+         }
+         break;
+      case GL_UNSIGNED_INT_10_10_10_2:
+         if (swapBytes) {
+            const GLuint *uisrc = (const GLuint *) src;
+            GLuint i;
+            for (i = 0; i < n; i ++) {
+               GLuint p = uisrc[i];
+               SWAP4BYTE(p);
+               rgba[i][rComp] = ((p >> 22)        );
+               rgba[i][gComp] = ((p >> 12) & 0x3ff);
+               rgba[i][bComp] = ((p >>  2) & 0x3ff);
+               rgba[i][aComp] = ((p      ) & 0x3  );
+            }
+         }
+         else {
+            const GLuint *uisrc = (const GLuint *) src;
+            GLuint i;
+            for (i = 0; i < n; i ++) {
+               GLuint p = uisrc[i];
+               rgba[i][rComp] = ((p >> 22)        );
+               rgba[i][gComp] = ((p >> 12) & 0x3ff);
+               rgba[i][bComp] = ((p >>  2) & 0x3ff);
+               rgba[i][aComp] = ((p      ) & 0x3  );
+            }
+         }
+         break;
+      case GL_UNSIGNED_INT_2_10_10_10_REV:
+         if (swapBytes) {
+            const GLuint *uisrc = (const GLuint *) src;
+            GLuint i;
+            for (i = 0; i < n; i ++) {
+               GLuint p = uisrc[i];
+               SWAP4BYTE(p);
+               rgba[i][rComp] = ((p      ) & 0x3ff);
+               rgba[i][gComp] = ((p >> 10) & 0x3ff);
+               rgba[i][bComp] = ((p >> 20) & 0x3ff);
+               rgba[i][aComp] = ((p >> 30)        );
+            }
+         }
+         else {
+            const GLuint *uisrc = (const GLuint *) src;
+            GLuint i;
+            for (i = 0; i < n; i ++) {
+               GLuint p = uisrc[i];
+               rgba[i][rComp] = ((p      ) & 0x3ff);
+               rgba[i][gComp] = ((p >> 10) & 0x3ff);
+               rgba[i][bComp] = ((p >> 20) & 0x3ff);
+               rgba[i][aComp] = ((p >> 30)        );
+            }
+         }
+         break;
+      default:
+         _mesa_problem(NULL, "bad srcType in extract uint data");
+         break;
+   }
+#undef PROCESS
 }
 
 
+
 /*
  * Unpack a row of color image data from a client buffer according to
  * the pixel unpacking parameters.
@@ -3295,6 +3876,205 @@ _mesa_unpack_color_span_float( struct gl_context *ctx,
    }
 }
 
+
+
+
+
+/**
+ * Same as _mesa_unpack_color_span_chan(), but return GLuint data
+ * instead of GLchan.
+ * No pixel transfer ops are applied.
+ */
+void
+_mesa_unpack_color_span_uint(struct gl_context *ctx,
+                             GLuint n, GLenum dstFormat, GLuint *dest,
+                             GLenum srcFormat, GLenum srcType,
+                             const GLvoid *source,
+                             const struct gl_pixelstore_attrib *srcPacking)
+{
+   GLuint rgba[MAX_WIDTH][4];
+
+   ASSERT(n <= MAX_WIDTH);
+
+   ASSERT(dstFormat == GL_ALPHA ||
+          dstFormat == GL_LUMINANCE ||
+          dstFormat == GL_LUMINANCE_ALPHA ||
+          dstFormat == GL_INTENSITY ||
+          dstFormat == GL_RED ||
+          dstFormat == GL_RG ||
+          dstFormat == GL_RGB ||
+          dstFormat == GL_RGBA);
+
+   ASSERT(srcFormat == GL_RED ||
+          srcFormat == GL_GREEN ||
+          srcFormat == GL_BLUE ||
+          srcFormat == GL_ALPHA ||
+          srcFormat == GL_LUMINANCE ||
+          srcFormat == GL_LUMINANCE_ALPHA ||
+          srcFormat == GL_INTENSITY ||
+          srcFormat == GL_RG ||
+          srcFormat == GL_RGB ||
+          srcFormat == GL_BGR ||
+          srcFormat == GL_RGBA ||
+          srcFormat == GL_BGRA ||
+          srcFormat == GL_ABGR_EXT ||
+          srcFormat == GL_RED_INTEGER_EXT ||
+          srcFormat == GL_GREEN_INTEGER_EXT ||
+          srcFormat == GL_BLUE_INTEGER_EXT ||
+          srcFormat == GL_ALPHA_INTEGER_EXT ||
+          srcFormat == GL_RGB_INTEGER_EXT ||
+          srcFormat == GL_RGBA_INTEGER_EXT ||
+          srcFormat == GL_BGR_INTEGER_EXT ||
+          srcFormat == GL_BGRA_INTEGER_EXT ||
+          srcFormat == GL_LUMINANCE_INTEGER_EXT ||
+          srcFormat == GL_LUMINANCE_ALPHA_INTEGER_EXT);
+
+   ASSERT(srcType == GL_UNSIGNED_BYTE ||
+          srcType == GL_BYTE ||
+          srcType == GL_UNSIGNED_SHORT ||
+          srcType == GL_SHORT ||
+          srcType == GL_UNSIGNED_INT ||
+          srcType == GL_INT ||
+          srcType == GL_HALF_FLOAT_ARB ||
+          srcType == GL_FLOAT ||
+          srcType == GL_UNSIGNED_BYTE_3_3_2 ||
+          srcType == GL_UNSIGNED_BYTE_2_3_3_REV ||
+          srcType == GL_UNSIGNED_SHORT_5_6_5 ||
+          srcType == GL_UNSIGNED_SHORT_5_6_5_REV ||
+          srcType == GL_UNSIGNED_SHORT_4_4_4_4 ||
+          srcType == GL_UNSIGNED_SHORT_4_4_4_4_REV ||
+          srcType == GL_UNSIGNED_SHORT_5_5_5_1 ||
+          srcType == GL_UNSIGNED_SHORT_1_5_5_5_REV ||
+          srcType == GL_UNSIGNED_INT_8_8_8_8 ||
+          srcType == GL_UNSIGNED_INT_8_8_8_8_REV ||
+          srcType == GL_UNSIGNED_INT_10_10_10_2 ||
+          srcType == GL_UNSIGNED_INT_2_10_10_10_REV);
+
+
+   /* Extract image data as uint[4] pixels */
+   extract_uint_rgba(n, rgba, srcFormat, srcType, source,
+                     srcPacking->SwapBytes);
+
+   if (dstFormat == GL_RGBA) {
+      /* simple case */
+      memcpy(dest, rgba, 4 * sizeof(GLuint) * n);
+   }
+   else {
+      /* general case */
+      GLint dstRedIndex = -1;
+      GLint dstGreenIndex = -1;
+      GLint dstBlueIndex = -1;
+      GLint dstAlphaIndex = -1;
+      GLint dstLuminanceIndex = -1;
+      GLint dstIntensityIndex = -1;
+      GLint dstComponents = _mesa_components_in_format( dstFormat );
+
+      assert(dstComponents > 0);
+
+      /* Now determine which color channels we need to produce.
+       * And determine the dest index (offset) within each color tuple.
+       */
+      switch (dstFormat) {
+      case GL_ALPHA:
+         dstAlphaIndex = 0;
+         break;
+      case GL_LUMINANCE:
+         dstLuminanceIndex = 0;
+         break;
+      case GL_LUMINANCE_ALPHA:
+         dstLuminanceIndex = 0;
+         dstAlphaIndex = 1;
+         break;
+      case GL_INTENSITY:
+         dstIntensityIndex = 0;
+         break;
+      case GL_RED:
+         dstRedIndex = 0;
+         break;
+      case GL_RG:
+         dstRedIndex = 0;
+         dstGreenIndex = 1;
+         break;
+      case GL_RGB:
+         dstRedIndex = 0;
+         dstGreenIndex = 1;
+         dstBlueIndex = 2;
+         break;
+      case GL_RGBA:
+         dstRedIndex = 0;
+         dstGreenIndex = 1;
+         dstBlueIndex = 2;
+         dstAlphaIndex = 3;
+         break;
+      default:
+         _mesa_problem(ctx,
+                       "bad dstFormat in _mesa_unpack_color_span_uint()");
+         return;
+      }
+
+      /* Now pack values in the requested dest format */
+      if (dstRedIndex >= 0) {
+         GLuint *dst = dest;
+         GLuint i;
+         for (i = 0; i < n; i++) {
+            dst[dstRedIndex] = rgba[i][RCOMP];
+            dst += dstComponents;
+         }
+      }
+
+      if (dstGreenIndex >= 0) {
+         GLuint *dst = dest;
+         GLuint i;
+         for (i = 0; i < n; i++) {
+            dst[dstGreenIndex] = rgba[i][GCOMP];
+            dst += dstComponents;
+         }
+      }
+
+      if (dstBlueIndex >= 0) {
+         GLuint *dst = dest;
+         GLuint i;
+         for (i = 0; i < n; i++) {
+            dst[dstBlueIndex] = rgba[i][BCOMP];
+            dst += dstComponents;
+         }
+      }
+
+      if (dstAlphaIndex >= 0) {
+         GLuint *dst = dest;
+         GLuint i;
+         for (i = 0; i < n; i++) {
+            dst[dstAlphaIndex] = rgba[i][ACOMP];
+            dst += dstComponents;
+         }
+      }
+
+      if (dstIntensityIndex >= 0) {
+         GLuint *dst = dest;
+         GLuint i;
+         assert(dstIntensityIndex == 0);
+         assert(dstComponents == 1);
+         for (i = 0; i < n; i++) {
+            /* Intensity comes from red channel */
+            dst[i] = rgba[i][RCOMP];
+         }
+      }
+
+      if (dstLuminanceIndex >= 0) {
+         GLuint *dst = dest;
+         GLuint i;
+         assert(dstLuminanceIndex == 0);
+         for (i = 0; i < n; i++) {
+            /* Luminance comes from red channel */
+            dst[0] = rgba[i][RCOMP];
+            dst += dstComponents;
+         }
+      }
+   }
+}
+
+
+
 /**
  * Similar to _mesa_unpack_color_span_float(), but for dudv data instead of rgba,
  * directly return GLbyte data, no transfer ops apply.
diff --git a/src/mesa/main/pack.h b/src/mesa/main/pack.h
index 3b24159fc2..65d3f7a0fb 100644
--- a/src/mesa/main/pack.h
+++ b/src/mesa/main/pack.h
@@ -75,6 +75,13 @@ _mesa_unpack_color_span_float(struct gl_context *ctx,
                               GLbitfield transferOps);
 
 extern void
+_mesa_unpack_color_span_uint(struct gl_context *ctx,
+                             GLuint n, GLenum dstFormat, GLuint *dest,
+                             GLenum srcFormat, GLenum srcType,
+                             const GLvoid *source,
+                             const struct gl_pixelstore_attrib *srcPacking);
+
+extern void
 _mesa_unpack_dudv_span_byte(struct gl_context *ctx,
                             GLuint n, GLenum dstFormat, GLbyte dest[],
                             GLenum srcFormat, GLenum srcType,
diff --git a/src/mesa/main/remap.c b/src/mesa/main/remap.c
index 2341f8488d..c89fba4530 100644
--- a/src/mesa/main/remap.c
+++ b/src/mesa/main/remap.c
@@ -44,10 +44,15 @@
 #include "imports.h"
 #include "glapi/glapi.h"
 
-#define ARRAY_SIZE(a) (sizeof (a) / sizeof ((a)[0]))
 #define MAX_ENTRY_POINTS 16
 
-static const char *_mesa_function_pool;
+#define need_MESA_remap_table
+#include "main/remap_helper.h"
+
+
+/* this is global for quick access */
+int driDispatchRemapTable[driDispatchRemapTable_size];
+
 
 /**
  * Return the spec string associated with the given function index.
@@ -60,7 +65,10 @@ static const char *_mesa_function_pool;
 const char *
 _mesa_get_function_spec(GLint func_index)
 {
-   return _mesa_function_pool + func_index;
+   if (func_index < Elements(_mesa_function_pool))
+      return _mesa_function_pool + func_index;
+   else
+      return NULL;
 }
 
 
@@ -152,11 +160,31 @@ _mesa_map_function_array(const struct gl_function_remap *func_array)
 
 
 /**
+ * Map the functions which are already static.
+ *
+ * When a extension function are incorporated into the ABI, the
+ * extension suffix is usually stripped.  Mapping such functions
+ * makes sure the alternative names are available.
+ *
+ * Note that functions mapped by _mesa_init_remap_table() are
+ * excluded.
+ */
+void
+_mesa_map_static_functions(void)
+{
+   /* Remap static functions which have alternative names and are in the ABI.
+    * This is to be on the safe side.  glapi should have defined those names.
+    */
+   _mesa_map_function_array(MESA_alt_functions);
+}
+
+
+/**
  * Initialize the remap table.  This is called in one_time_init().
  * The remap table needs to be initialized before calling the
  * CALL/GET/SET macros defined in main/dispatch.h.
  */
-void
+static void
 _mesa_do_init_remap_table(const char *pool,
 			  int size,
 			  const struct gl_function_pool_remap *remap)
@@ -167,7 +195,6 @@ _mesa_do_init_remap_table(const char *pool,
    if (initialized)
       return;
    initialized = GL_TRUE;
-   _mesa_function_pool = pool;
 
    /* initialize the remap table */
    for (i = 0; i < size; i++) {
@@ -187,4 +214,13 @@ _mesa_do_init_remap_table(const char *pool,
 }
 
 
+void
+_mesa_init_remap_table(void)
+{
+   _mesa_do_init_remap_table(_mesa_function_pool,
+			     driDispatchRemapTable_size,
+			     MESA_remap_table_functions);
+}
+
+
 #endif /* FEATURE_remap_table */
diff --git a/src/mesa/main/remap.h b/src/mesa/main/remap.h
index a2a55f615d..5fee300529 100644
--- a/src/mesa/main/remap.h
+++ b/src/mesa/main/remap.h
@@ -60,25 +60,8 @@ extern void
 _mesa_map_static_functions(void);
 
 extern void
-_mesa_map_static_functions_es1(void);
-
-extern void
-_mesa_map_static_functions_es2(void);
-
-extern void
-_mesa_do_init_remap_table(const char *pool,
-			  int size,
-			  const struct gl_function_pool_remap *remap);
-
-extern void
 _mesa_init_remap_table(void);
 
-extern void
-_mesa_init_remap_table_es1(void);
-
-extern void
-_mesa_init_remap_table_es2(void);
-
 #else /* FEATURE_remap_table */
 
 static INLINE const char *
@@ -105,37 +88,10 @@ _mesa_map_static_functions(void)
 
 
 static INLINE void
-_mesa_map_static_functions_es1(void)
-{
-}
-
-static INLINE void
-_mesa_map_static_functions_es2(void)
-{
-}
-
-static INLINE void
-_mesa_do_init_remap_table(const char *pool,
-			  int size,
-			  const struct gl_function_pool_remap *remap)
-{
-}
-
-static INLINE void
 _mesa_init_remap_table(void)
 {
 }
 
-static INLINE void
-_mesa_init_remap_table_es1(void)
-{
-}
-
-static INLINE void
-_mesa_init_remap_table_es2(void)
-{
-}
-
 #endif /* FEATURE_remap_table */
 
 
diff --git a/src/mesa/main/renderbuffer.c b/src/mesa/main/renderbuffer.c
index dc8bc74787..3d1a8f8592 100644
--- a/src/mesa/main/renderbuffer.c
+++ b/src/mesa/main/renderbuffer.c
@@ -1766,7 +1766,7 @@ _mesa_add_accum_renderbuffer(struct gl_context *ctx, struct gl_framebuffer *fb,
 
 
 /**
- * Add a software-based accumulation renderbuffer to the given framebuffer.
+ * Add a software-based aux renderbuffer to the given framebuffer.
  * This is a helper routine for device drivers when creating a
  * window system framebuffer (not a user-created render/framebuffer).
  * Once this function is called, you can basically forget about this
@@ -1795,7 +1795,7 @@ _mesa_add_aux_renderbuffers(struct gl_context *ctx, struct gl_framebuffer *fb,
       assert(fb->Attachment[BUFFER_AUX0 + i].Renderbuffer == NULL);
 
       if (!rb) {
-         _mesa_error(ctx, GL_OUT_OF_MEMORY, "Allocating accum buffer");
+         _mesa_error(ctx, GL_OUT_OF_MEMORY, "Allocating aux buffer");
          return GL_FALSE;
       }
 
diff --git a/src/mesa/main/shaderapi.c b/src/mesa/main/shaderapi.c
index 622285debb..9717e00559 100644
--- a/src/mesa/main/shaderapi.c
+++ b/src/mesa/main/shaderapi.c
@@ -936,9 +936,9 @@ print_shader_info(const struct gl_shader_program *shProg)
 /**
  * Use the named shader program for subsequent glUniform calls
  */
-static void
-active_program(struct gl_context *ctx, struct gl_shader_program *shProg,
-	       const char *caller)
+void
+_mesa_active_program(struct gl_context *ctx, struct gl_shader_program *shProg,
+		     const char *caller)
 {
    if ((shProg != NULL) && !shProg->LinkStatus) {
       _mesa_error(ctx, GL_INVALID_OPERATION,
@@ -1003,45 +1003,16 @@ use_shader_program(struct gl_context *ctx, GLenum type,
  * Use the named shader program for subsequent rendering.
  */
 void
-_mesa_use_program(struct gl_context *ctx, GLuint program)
+_mesa_use_program(struct gl_context *ctx, struct gl_shader_program *shProg)
 {
-   struct gl_shader_program *shProg;
-   struct gl_transform_feedback_object *obj =
-      ctx->TransformFeedback.CurrentObject;
    bool changed = false;
 
-   if (obj->Active) {
-      _mesa_error(ctx, GL_INVALID_OPERATION,
-                  "glUseProgram(transform feedback active)");
-      return;
-   }
-
-   if (program) {
-      shProg = _mesa_lookup_shader_program_err(ctx, program, "glUseProgram");
-      if (!shProg) {
-         return;
-      }
-      if (!shProg->LinkStatus) {
-         _mesa_error(ctx, GL_INVALID_OPERATION,
-                     "glUseProgram(program %u not linked)", program);
-         return;
-      }
-
-      /* debug code */
-      if (ctx->Shader.Flags & GLSL_USE_PROG) {
-         print_shader_info(shProg);
-      }
-   }
-   else {
-      shProg = NULL;
-   }
-
    changed = use_shader_program(ctx, GL_VERTEX_SHADER, shProg);
    changed = use_shader_program(ctx, GL_GEOMETRY_SHADER_ARB, shProg)
       || changed;
    changed = use_shader_program(ctx, GL_FRAGMENT_SHADER, shProg)
       || changed;
-   active_program(ctx, shProg, "glUseProgram");
+   _mesa_active_program(ctx, shProg, "glUseProgram");
 
    if (changed) {
       FLUSH_VERTICES(ctx, _NEW_PROGRAM | _NEW_PROGRAM_CONSTANTS);
@@ -1614,8 +1585,39 @@ void GLAPIENTRY
 _mesa_UseProgramObjectARB(GLhandleARB program)
 {
    GET_CURRENT_CONTEXT(ctx);
+   struct gl_shader_program *shProg;
+   struct gl_transform_feedback_object *obj =
+      ctx->TransformFeedback.CurrentObject;
+
    FLUSH_VERTICES(ctx, _NEW_PROGRAM);
-   _mesa_use_program(ctx, program);
+
+   if (obj->Active) {
+      _mesa_error(ctx, GL_INVALID_OPERATION,
+                  "glUseProgram(transform feedback active)");
+      return;
+   }
+
+   if (program) {
+      shProg = _mesa_lookup_shader_program_err(ctx, program, "glUseProgram");
+      if (!shProg) {
+         return;
+      }
+      if (!shProg->LinkStatus) {
+         _mesa_error(ctx, GL_INVALID_OPERATION,
+                     "glUseProgram(program %u not linked)", program);
+         return;
+      }
+
+      /* debug code */
+      if (ctx->Shader.Flags & GLSL_USE_PROG) {
+         print_shader_info(shProg);
+      }
+   }
+   else {
+      shProg = NULL;
+   }
+
+   _mesa_use_program(ctx, shProg);
 }
 
 
@@ -1731,12 +1733,25 @@ _mesa_ProgramParameteriARB(GLuint program, GLenum pname,
 
 #endif
 
+void
+_mesa_use_shader_program(struct gl_context *ctx, GLenum type,
+			 struct gl_shader_program *shProg)
+{
+   bool changed;
+
+   changed = use_shader_program(ctx, type, shProg);
+   if (changed)
+      FLUSH_VERTICES(ctx, _NEW_PROGRAM | _NEW_PROGRAM_CONSTANTS);
+
+   if (ctx->Driver.UseProgram)
+      ctx->Driver.UseProgram(ctx, shProg);
+}
+
 void GLAPIENTRY
 _mesa_UseShaderProgramEXT(GLenum type, GLuint program)
 {
    GET_CURRENT_CONTEXT(ctx);
    struct gl_shader_program *shProg = NULL;
-   bool changed = false;
 
    if (!validate_shader_target(ctx, type)) {
       _mesa_error(ctx, GL_INVALID_ENUM, "glUseShaderProgramEXT(type)");
@@ -1762,13 +1777,7 @@ _mesa_UseShaderProgramEXT(GLenum type, GLuint program)
       }
    }
 
-   changed = use_shader_program(ctx, type, shProg);
-   if (changed)
-      FLUSH_VERTICES(ctx, _NEW_PROGRAM | _NEW_PROGRAM_CONSTANTS);
-
-   if (ctx->Driver.UseProgram)
-      ctx->Driver.UseProgram(ctx, shProg);
-   return;
+   _mesa_use_shader_program(ctx, type, shProg);
 }
 
 void GLAPIENTRY
@@ -1779,7 +1788,7 @@ _mesa_ActiveProgramEXT(GLuint program)
       ? _mesa_lookup_shader_program_err(ctx, program, "glActiveProgramEXT")
       : NULL;
 
-   active_program(ctx, shProg, "glActiveProgramEXT");
+   _mesa_active_program(ctx, shProg, "glActiveProgramEXT");
    return;
 }
 
diff --git a/src/mesa/main/shaderapi.h b/src/mesa/main/shaderapi.h
index db316ab6a7..df78351411 100644
--- a/src/mesa/main/shaderapi.h
+++ b/src/mesa/main/shaderapi.h
@@ -39,8 +39,11 @@ _mesa_copy_string(GLchar *dst, GLsizei maxLength,
                   GLsizei *length, const GLchar *src);
 
 extern void
-_mesa_use_program(struct gl_context *ctx, GLuint program);
+_mesa_use_program(struct gl_context *ctx, struct gl_shader_program *shProg);
 
+extern void
+_mesa_active_program(struct gl_context *ctx, struct gl_shader_program *shProg,
+		     const char *caller);
 
 extern void
 _mesa_init_shader_dispatch(struct _glapi_table *exec);
@@ -172,6 +175,9 @@ _mesa_ShaderBinary(GLint n, const GLuint *shaders, GLenum binaryformat,
 extern void GLAPIENTRY
 _mesa_ProgramParameteriARB(GLuint program, GLenum pname,
                            GLint value);
+void
+_mesa_use_shader_program(struct gl_context *ctx, GLenum type,
+			 struct gl_shader_program *shProg);
 
 extern void GLAPIENTRY
 _mesa_UseShaderProgramEXT(GLenum type, GLuint program);
diff --git a/src/mesa/main/texobj.c b/src/mesa/main/texobj.c
index e08df0f7fe..1326760961 100644
--- a/src/mesa/main/texobj.c
+++ b/src/mesa/main/texobj.c
@@ -380,15 +380,20 @@ _mesa_reference_texobj(struct gl_texture_object **ptr,
  *
  * \note For debug purposes only.
  */
-#if 0
 static void
-incomplete(const struct gl_texture_object *t, const char *why)
+incomplete(const struct gl_texture_object *t, const char *fmt, ...)
 {
-   printf("Texture Obj %d incomplete because: %s\n", t->Name, why);
-}
-#else
-#define incomplete(t, why)
+#if 0
+   va_list args;
+   char s[100];
+
+   va_start(args, fmt);
+   vsnprintf(s, sizeof(s), fmt, args);
+   va_end(args);
+
+   printf("Texture Obj %d incomplete because: %s\n", t->Name, s);
 #endif
+}
 
 
 /**
@@ -416,18 +421,14 @@ _mesa_test_texobj_completeness( const struct gl_context *ctx,
     * value.
     */
    if ((baseLevel < 0) || (baseLevel >= MAX_TEXTURE_LEVELS)) {
-      char s[100];
-      _mesa_snprintf(s, sizeof(s), "base level = %d is invalid", baseLevel);
-      incomplete(t, s);
+      incomplete(t, "base level = %d is invalid", baseLevel);
       t->_Complete = GL_FALSE;
       return;
    }
 
    /* Always need the base level image */
    if (!t->Image[0][baseLevel]) {
-      char s[100];
-      _mesa_snprintf(s, sizeof(s), "Image[baseLevel=%d] == NULL", baseLevel);
-      incomplete(t, s);
+      incomplete(t, "Image[baseLevel=%d] == NULL", baseLevel);
       t->_Complete = GL_FALSE;
       return;
    }
diff --git a/src/mesa/main/varray.c b/src/mesa/main/varray.c
index 73172c5779..6cb9638c9f 100644
--- a/src/mesa/main/varray.c
+++ b/src/mesa/main/varray.c
@@ -321,7 +321,7 @@ _mesa_TexCoordPointer(GLint size, GLenum type, GLsizei stride,
 void GLAPIENTRY
 _mesa_EdgeFlagPointer(GLsizei stride, const GLvoid *ptr)
 {
-   const GLbitfield legalTypes = BOOL_BIT;
+   const GLbitfield legalTypes = UNSIGNED_BYTE_BIT;
    /* see table 2.4 edits in GL_EXT_gpu_shader4 spec: */
    const GLboolean integer = GL_TRUE;
    GET_CURRENT_CONTEXT(ctx);
@@ -330,7 +330,7 @@ _mesa_EdgeFlagPointer(GLsizei stride, const GLvoid *ptr)
    update_array(ctx, "glEdgeFlagPointer",
                 &ctx->Array.ArrayObj->EdgeFlag, _NEW_ARRAY_EDGEFLAG,
                 legalTypes, 1, 1,
-                1, GL_BOOL, stride, GL_FALSE, integer, ptr);
+                1, GL_UNSIGNED_BYTE, stride, GL_FALSE, integer, ptr);
 }
 
 
diff --git a/src/mesa/program/sampler.cpp b/src/mesa/program/sampler.cpp
index 0e58aef9c9..9a813c8795 100644
--- a/src/mesa/program/sampler.cpp
+++ b/src/mesa/program/sampler.cpp
@@ -23,15 +23,15 @@
  * DEALINGS IN THE SOFTWARE.
  */
 
-#include <stdio.h>
+#include <cstdio>
+#include "ir.h"
+#include "glsl_types.h"
+#include "ir_visitor.h"
 
 extern "C" {
 #include "main/compiler.h"
 #include "main/mtypes.h"
 #include "program/prog_parameter.h"
-#include "ir.h"
-#include "ir_visitor.h"
-#include "glsl_types.h"
 }
 
 static void fail_link(struct gl_shader_program *prog, const char *fmt, ...) PRINTFLIKE(2, 3);
diff --git a/src/mesa/state_tracker/st_atom_constbuf.c b/src/mesa/state_tracker/st_atom_constbuf.c
index 6f9d71e845..8d1dc792bc 100644
--- a/src/mesa/state_tracker/st_atom_constbuf.c
+++ b/src/mesa/state_tracker/st_atom_constbuf.c
@@ -90,8 +90,11 @@ void st_upload_constants( struct st_context *st,
 
       st->pipe->set_constant_buffer(st->pipe, shader_type, 0, *cbuf);
    }
-   else {
+   else if (*cbuf) {
       st->constants.tracked_state[shader_type].dirty.mesa = 0x0;
+
+      pipe_resource_reference(cbuf, NULL);
+      st->pipe->set_constant_buffer(st->pipe, shader_type, 0, NULL);
    }
 }
 
diff --git a/src/mesa/state_tracker/st_cb_eglimage.c b/src/mesa/state_tracker/st_cb_eglimage.c
index 298f8a5b12..b39a624a61 100644
--- a/src/mesa/state_tracker/st_cb_eglimage.c
+++ b/src/mesa/state_tracker/st_cb_eglimage.c
@@ -129,6 +129,7 @@ st_bind_surface(struct gl_context *ctx, GLenum target,
 
    /* FIXME create a non-default sampler view from the pipe_surface? */
    pipe_resource_reference(&stObj->pt, ps->texture);
+   pipe_sampler_view_reference(&stObj->sampler_view, NULL);
    pipe_resource_reference(&stImage->pt, stObj->pt);
 
    stObj->width0 = ps->width;
diff --git a/src/mesa/state_tracker/st_context.c b/src/mesa/state_tracker/st_context.c
index 75fd69540f..d0dcdd4e29 100644
--- a/src/mesa/state_tracker/st_context.c
+++ b/src/mesa/state_tracker/st_context.c
@@ -236,6 +236,13 @@ void st_destroy_context( struct st_context *st )
    }
    pipe_surface_reference(&st->state.framebuffer.zsbuf, NULL);
 
+   pipe->set_index_buffer(pipe, NULL);
+
+   for (i = 0; i < PIPE_SHADER_TYPES; i++) {
+      pipe->set_constant_buffer(pipe, PIPE_SHADER_VERTEX, 0, NULL);
+      pipe_resource_reference(&st->state.constants[PIPE_SHADER_VERTEX], NULL);
+   }
+
    _mesa_delete_program_cache(st->ctx, st->pixel_xfer.cache);
 
    _vbo_DestroyContext(st->ctx);