Merge branch 'mesa_7_0_branch' into i915tex_branch

author: Thomas Hellstrom <thomas-at-tungstengraphics-dot-com> 2008-03-30 15:33:07 +0200
committer: Thomas Hellstrom <thomas-at-tungstengraphics-dot-com> 2008-03-30 15:33:07 +0200
commit: a0e133114bbde5c1f7e6ea20613b513f7fbbbdae (patch)
tree: b0dc6c5e287c1ae0bf3bc8357ab992cbb792df11
parent: 43b05ffda491046bf0cca7f0cdd8a337eb09f1f1 (diff)
parent: 767dfa5b9c07fe6ae267e1ff0d3bbbf093a04c44 (diff)
17 files changed, 284 insertions, 140 deletions
diff --git a/docs/relnotes-7.0.3.html b/docs/relnotes-7.0.3.html
index a2a3437442..212467d6ff 100644
--- a/docs/relnotes-7.0.3.html
+++ b/docs/relnotes-7.0.3.html
@@ -8,7 +8,7 @@
 
 <body bgcolor="#eeeeee">
 
-<H1>Mesa 7.0.3 Release Notes / February TBD, 2008</H1>
+<H1>Mesa 7.0.3 Release Notes / March TBD, 2008</H1>
 
 <p>
 Mesa 7.0.3 is a stable release with bug fixes since version 7.0.2.
@@ -40,8 +40,16 @@ Mesa 7.0.3 is a stable release with bug fixes since version 7.0.2.
 <li>Bad strings given to glProgramStringARB() didn't generate GL_INVALID_OPERATION
 <li>Fixed minor point rasterization regression (bug 11016)
 <li>state.texenv.color state var didn't work in GL_ARB_fragment_program (bug 14931)
+<li>glBitmap from a PBO didn't always work
+<li>glGetTexImage into a PBO didn't always work
 </ul>
 
+<h2>Changes</h2>
+<ul>
+<li>Updated glext.h to version 40
+</ul>
+
+
 
 <h2>Driver Status</h2>
 
diff --git a/include/GL/glext.h b/include/GL/glext.h
index 2519a6cc5e..2b22714c30 100644
--- a/include/GL/glext.h
+++ b/include/GL/glext.h
@@ -46,9 +46,9 @@ extern "C" {
 /*************************************************************/
 
 /* Header file version number, required by OpenGL ABI for Linux */
-/* glext.h last updated 2007/02/12 */
+/* glext.h last updated 2008/03/24 */
 /* Current version at http://www.opengl.org/registry/ */
-#define GL_GLEXT_VERSION 39
+#define GL_GLEXT_VERSION 40
 
 #ifndef GL_VERSION_1_2
 #define GL_UNSIGNED_BYTE_3_3_2            0x8032
@@ -3091,8 +3091,8 @@ extern "C" {
 #ifndef GL_EXT_framebuffer_blit
 #define GL_READ_FRAMEBUFFER_EXT           0x8CA8
 #define GL_DRAW_FRAMEBUFFER_EXT           0x8CA9
-#define GL_READ_FRAMEBUFFER_BINDING_EXT   GL_FRAMEBUFFER_BINDING_EXT
-#define GL_DRAW_FRAMEBUFFER_BINDING_EXT   0x8CAA
+#define GL_DRAW_FRAMEBUFFER_BINDING_EXT   GL_FRAMEBUFFER_BINDING_EXT
+#define GL_READ_FRAMEBUFFER_BINDING_EXT   0x8CAA
 #endif
 
 #ifndef GL_EXT_framebuffer_multisample
@@ -3379,6 +3379,9 @@ extern "C" {
 #define GL_RGBA_INTEGER_MODE_EXT          0x8D9E
 #endif
 
+#ifndef GL_GREMEDY_frame_terminator
+#endif
+
 
 /*************************************************************/
 
@@ -7252,6 +7255,14 @@ typedef void (APIENTRYP PFNGLCLEARCOLORIIEXTPROC) (GLint red, GLint green, GLint
 typedef void (APIENTRYP PFNGLCLEARCOLORIUIEXTPROC) (GLuint red, GLuint green, GLuint blue, GLuint alpha);
 #endif
 
+#ifndef GL_GREMEDY_frame_terminator
+#define GL_GREMEDY_frame_terminator 1
+#ifdef GL_GLEXT_PROTOTYPES
+GLAPI void APIENTRY glFrameTerminatorGREMEDY (void);
+#endif /* GL_GLEXT_PROTOTYPES */
+typedef void (APIENTRYP PFNGLFRAMETERMINATORGREMEDYPROC) (void);
+#endif
+
 
 #ifdef __cplusplus
 }
diff --git a/src/mesa/main/context.c b/src/mesa/main/context.c
index 9de515d840..e79a905346 100644
--- a/src/mesa/main/context.c
+++ b/src/mesa/main/context.c
@@ -687,10 +687,10 @@ free_shared_state( GLcontext *ctx, struct gl_shared_state *ss )
    _mesa_DeleteHashTable(ss->Programs);
 #endif
 #if FEATURE_ARB_vertex_program
-   _mesa_delete_program(ctx, ss->DefaultVertexProgram);
+   ctx->Driver.DeleteProgram(ctx, ss->DefaultVertexProgram);
 #endif
 #if FEATURE_ARB_fragment_program
-   _mesa_delete_program(ctx, ss->DefaultFragmentProgram);
+   ctx->Driver.DeleteProgram(ctx, ss->DefaultFragmentProgram);
 #endif
 
 #if FEATURE_ATI_fragment_shader
diff --git a/src/mesa/main/drawpix.c b/src/mesa/main/drawpix.c
index c82abccc41..3acccf0430 100644
--- a/src/mesa/main/drawpix.c
+++ b/src/mesa/main/drawpix.c
@@ -1,8 +1,8 @@
 /*
  * Mesa 3-D graphics library
- * Version:  6.5
+ * Version:  7.0.3
  *
- * Copyright (C) 1999-2006  Brian Paul   All Rights Reserved.
+ * Copyright (C) 1999-2008  Brian Paul   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"),
@@ -341,12 +341,10 @@ _mesa_Bitmap( GLsizei width, GLsizei height,
    }
 
    if (ctx->RenderMode == GL_RENDER) {
-      if (bitmap) {
-         /* Truncate, to satisfy conformance tests (matches SGI's OpenGL). */
-         GLint x = IFLOOR(ctx->Current.RasterPos[0] - xorig);
-         GLint y = IFLOOR(ctx->Current.RasterPos[1] - yorig);
-         ctx->Driver.Bitmap( ctx, x, y, width, height, &ctx->Unpack, bitmap );
-      }
+      /* Truncate, to satisfy conformance tests (matches SGI's OpenGL). */
+      GLint x = IFLOOR(ctx->Current.RasterPos[0] - xorig);
+      GLint y = IFLOOR(ctx->Current.RasterPos[1] - yorig);
+      ctx->Driver.Bitmap( ctx, x, y, width, height, &ctx->Unpack, bitmap );
    }
 #if _HAVE_FULL_GL
    else if (ctx->RenderMode == GL_FEEDBACK) {
diff --git a/src/mesa/main/fbobject.c b/src/mesa/main/fbobject.c
index 8e7d66cb9c..bbdc46f002 100644
--- a/src/mesa/main/fbobject.c
+++ b/src/mesa/main/fbobject.c
@@ -302,7 +302,7 @@ test_attachment_completeness(const GLcontext *ctx, GLenum format,
             /* OK */
          }
          else if (ctx->Extensions.EXT_packed_depth_stencil &&
-                  att->Renderbuffer->_BaseFormat == GL_DEPTH_STENCIL_EXT) {
+                  texImage->TexFormat->BaseFormat == GL_DEPTH_STENCIL_EXT) {
             /* OK */
          }
          else {
diff --git a/src/mesa/main/get.c b/src/mesa/main/get.c
index eb81ee4a52..2cbd60539e 100644
--- a/src/mesa/main/get.c
+++ b/src/mesa/main/get.c
@@ -1864,6 +1864,10 @@ _mesa_GetBooleanv( GLenum pname, GLboolean *params )
          CHECK_EXT1(EXT_framebuffer_object, "GetBooleanv");
          params[0] = INT_TO_BOOLEAN(ctx->Const.MaxRenderbufferSize);
          break;
+      case GL_READ_FRAMEBUFFER_BINDING_EXT:
+         CHECK_EXT1(EXT_framebuffer_blit, "GetBooleanv");
+         params[0] = INT_TO_BOOLEAN(ctx->ReadBuffer->Name);
+         break;
       case GL_MAX_FRAGMENT_UNIFORM_COMPONENTS_ARB:
          CHECK_EXT1(ARB_fragment_shader, "GetBooleanv");
          params[0] = INT_TO_BOOLEAN(ctx->Const.FragmentProgram.MaxUniformComponents);
@@ -3695,6 +3699,10 @@ _mesa_GetFloatv( GLenum pname, GLfloat *params )
          CHECK_EXT1(EXT_framebuffer_object, "GetFloatv");
          params[0] = (GLfloat)(ctx->Const.MaxRenderbufferSize);
          break;
+      case GL_READ_FRAMEBUFFER_BINDING_EXT:
+         CHECK_EXT1(EXT_framebuffer_blit, "GetFloatv");
+         params[0] = (GLfloat)(ctx->ReadBuffer->Name);
+         break;
       case GL_MAX_FRAGMENT_UNIFORM_COMPONENTS_ARB:
          CHECK_EXT1(ARB_fragment_shader, "GetFloatv");
          params[0] = (GLfloat)(ctx->Const.FragmentProgram.MaxUniformComponents);
@@ -5526,6 +5534,10 @@ _mesa_GetIntegerv( GLenum pname, GLint *params )
          CHECK_EXT1(EXT_framebuffer_object, "GetIntegerv");
          params[0] = ctx->Const.MaxRenderbufferSize;
          break;
+      case GL_READ_FRAMEBUFFER_BINDING_EXT:
+         CHECK_EXT1(EXT_framebuffer_blit, "GetIntegerv");
+         params[0] = ctx->ReadBuffer->Name;
+         break;
       case GL_MAX_FRAGMENT_UNIFORM_COMPONENTS_ARB:
          CHECK_EXT1(ARB_fragment_shader, "GetIntegerv");
          params[0] = ctx->Const.FragmentProgram.MaxUniformComponents;
diff --git a/src/mesa/main/get_gen.py b/src/mesa/main/get_gen.py
index 33be768999..6ff9168020 100644
--- a/src/mesa/main/get_gen.py
+++ b/src/mesa/main/get_gen.py
@@ -977,6 +977,11 @@ StateVars = [
 	  ["ctx->Const.MaxRenderbufferSize"], "",
 	  ["EXT_framebuffer_object"] ),
 
+	# GL_EXT_framebuffer_blit
+	# NOTE: GL_DRAW_FRAMEBUFFER_BINDING_EXT == GL_FRAMEBUFFER_BINDING_EXT
+	( "GL_READ_FRAMEBUFFER_BINDING_EXT", GLint, ["ctx->ReadBuffer->Name"], "",
+	  ["EXT_framebuffer_blit"] ),
+
 	# GL_ARB_fragment_shader
 	( "GL_MAX_FRAGMENT_UNIFORM_COMPONENTS_ARB", GLint,
 	  ["ctx->Const.FragmentProgram.MaxUniformComponents"], "",
diff --git a/src/mesa/main/imports.h b/src/mesa/main/imports.h
index ebdfc452a7..d6dc725b0c 100644
--- a/src/mesa/main/imports.h
+++ b/src/mesa/main/imports.h
@@ -1,8 +1,8 @@
 /*
  * Mesa 3-D graphics library
- * Version:  6.5.2
+ * Version:  7.0.3
  *
- * Copyright (C) 1999-2006  Brian Paul   All Rights Reserved.
+ * Copyright (C) 1999-2008  Brian Paul   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"),
@@ -159,7 +159,7 @@ typedef union { GLfloat f; GLint i; } fi_type;
  ***/
 #if defined(__i386__) || defined(__386__) || defined(__sparc__) || \
     defined(__s390x__) || defined(__powerpc__) || \
-    defined(__amd64__) || \
+    defined(__amd64__) || defined(__x86_64__) || \
     defined(ia64) || defined(__ia64__) || \
     defined(__hppa__) || defined(hpux) || \
     defined(__mips) || defined(_MIPS_ARCH) || \
diff --git a/src/mesa/main/teximage.c b/src/mesa/main/teximage.c
index d857a4f3a4..26d8fbc8de 100644
--- a/src/mesa/main/teximage.c
+++ b/src/mesa/main/teximage.c
@@ -2177,9 +2177,6 @@ _mesa_GetTexImage( GLenum target, GLint level, GLenum format,
       return;
    }
 
-   if (!pixels)
-      return;
-
    _mesa_lock_texture(ctx, texObj);
    {
       texImage = _mesa_select_tex_image(ctx, texObj, target, level);
diff --git a/src/mesa/shader/arbprogparse.c b/src/mesa/shader/arbprogparse.c
index 4de8ccf087..f74f727b24 100644
--- a/src/mesa/shader/arbprogparse.c
+++ b/src/mesa/shader/arbprogparse.c
@@ -1609,8 +1609,6 @@ parse_attrib_binding(GLcontext * ctx, const GLubyte ** inst,
       program_error(ctx, Program->Position, "Bad attribute binding");
    }
 
-   Program->Base.InputsRead |= (1 << *inputReg);
-
    return err;
 }
 
@@ -2565,6 +2563,11 @@ parse_src_reg (GLcontext * ctx, const GLubyte ** inst,
          return 1;
    }
 
+   /* Add attributes to InputsRead only if they are used the program.
+    * This avoids the handling of unused ATTRIB declarations in the drivers. */
+   if (*File == PROGRAM_INPUT)
+      Program->Base.InputsRead |= (1 << *Index);
+
    return 0;
 }
 
diff --git a/src/mesa/shader/prog_statevars.c b/src/mesa/shader/prog_statevars.c
index 76295568ac..05daa8011d 100644
--- a/src/mesa/shader/prog_statevars.c
+++ b/src/mesa/shader/prog_statevars.c
@@ -181,7 +181,7 @@ _mesa_fetch_state(GLcontext *ctx, const gl_state_index state[],
                      ctx->Light.Material.Attrib[MAT_ATTRIB_FRONT_AMBIENT+face][i];
                }
                /* [3] = material alpha */
-               value[3] = ctx->Light.Material.Attrib[MAT_ATTRIB_FRONT_DIFFUSE+face][3];
+               value[3] = ctx->Light.Material.Attrib[MAT_ATTRIB_FRONT_AMBIENT+face][3];
                return;
             case STATE_DIFFUSE:
                for (i = 0; i < 3; i++) {
@@ -197,7 +197,7 @@ _mesa_fetch_state(GLcontext *ctx, const gl_state_index state[],
                      ctx->Light.Material.Attrib[MAT_ATTRIB_FRONT_SPECULAR+face][i];
                }
                /* [3] = material alpha */
-               value[3] = ctx->Light.Material.Attrib[MAT_ATTRIB_FRONT_DIFFUSE+face][3];
+               value[3] = ctx->Light.Material.Attrib[MAT_ATTRIB_FRONT_SPECULAR+face][3];
                return;
             default:
                _mesa_problem(ctx, "Invalid lightprod state in fetch_state");
diff --git a/src/mesa/shader/program.c b/src/mesa/shader/program.c
index 1f227390af..95cabe7323 100644
--- a/src/mesa/shader/program.c
+++ b/src/mesa/shader/program.c
@@ -343,7 +343,7 @@ _mesa_clone_program(GLcontext *ctx, const struct gl_program *prog)
    clone->Format = prog->Format;
    clone->Instructions = _mesa_alloc_instructions(prog->NumInstructions);
    if (!clone->Instructions) {
-      _mesa_delete_program(ctx, clone);
+      ctx->Driver.DeleteProgram(ctx, clone);
       return NULL;
    }
    _mesa_copy_instructions(clone->Instructions, prog->Instructions,
diff --git a/src/mesa/shader/shader_api.c b/src/mesa/shader/shader_api.c
index 96940be5df..5f9e2b84a4 100644
--- a/src/mesa/shader/shader_api.c
+++ b/src/mesa/shader/shader_api.c
@@ -80,7 +80,7 @@ _mesa_clear_shader_program_data(GLcontext *ctx,
          /* to prevent a double-free in the next call */
          shProg->VertexProgram->Base.Parameters = NULL;
       }
-      _mesa_delete_program(ctx, &shProg->VertexProgram->Base);
+      ctx->Driver.DeleteProgram(ctx, &shProg->VertexProgram->Base);
       shProg->VertexProgram = NULL;
    }
 
@@ -89,7 +89,7 @@ _mesa_clear_shader_program_data(GLcontext *ctx,
          /* to prevent a double-free in the next call */
          shProg->FragmentProgram->Base.Parameters = NULL;
       }
-      _mesa_delete_program(ctx, &shProg->FragmentProgram->Base);
+      ctx->Driver.DeleteProgram(ctx, &shProg->FragmentProgram->Base);
       shProg->FragmentProgram = NULL;
    }
 
@@ -247,7 +247,7 @@ _mesa_free_shader(GLcontext *ctx, struct gl_shader *sh)
       _mesa_free(sh->InfoLog);
    for (i = 0; i < sh->NumPrograms; i++) {
       assert(sh->Programs[i]);
-      _mesa_delete_program(ctx, sh->Programs[i]);
+      ctx->Driver.DeleteProgram(ctx, sh->Programs[i]);
    }
    if (sh->Programs)
       _mesa_free(sh->Programs);
diff --git a/src/mesa/shader/slang/slang_emit.c b/src/mesa/shader/slang/slang_emit.c
index 9947544a08..9c307c6275 100644
--- a/src/mesa/shader/slang/slang_emit.c
+++ b/src/mesa/shader/slang/slang_emit.c
@@ -677,6 +677,7 @@ static struct prog_instruction *
 emit_clamp(slang_emit_info *emitInfo, slang_ir_node *n)
 {
    struct prog_instruction *inst;
+   slang_ir_node tmpNode;
 
    assert(n->Opcode == IR_CLAMP);
    /* ch[0] = value
@@ -722,18 +723,27 @@ emit_clamp(slang_emit_info *emitInfo, slang_ir_node *n)
    emit(emitInfo, n->Children[1]);
    emit(emitInfo, n->Children[2]);
 
+   /* Some GPUs don't allow reading from output registers.  So if the
+    * dest for this clamp() is an output reg, we can't use that reg for
+    * the intermediate result.  Use a temp register instead.
+    */
+   _mesa_bzero(&tmpNode, sizeof(tmpNode));
+   alloc_temp_storage(emitInfo, &tmpNode, n->Store->Size);
+
    /* tmp = max(ch[0], ch[1]) */
    inst = new_instruction(emitInfo, OPCODE_MAX);
-   storage_to_dst_reg(&inst->DstReg, n->Store, n->Writemask);
+   storage_to_dst_reg(&inst->DstReg, tmpNode.Store, n->Writemask);
    storage_to_src_reg(&inst->SrcReg[0], n->Children[0]->Store);
    storage_to_src_reg(&inst->SrcReg[1], n->Children[1]->Store);
 
-   /* tmp = min(tmp, ch[2]) */
+   /* n->dest = min(tmp, ch[2]) */
    inst = new_instruction(emitInfo, OPCODE_MIN);
    storage_to_dst_reg(&inst->DstReg, n->Store, n->Writemask);
-   storage_to_src_reg(&inst->SrcReg[0], n->Store);
+   storage_to_src_reg(&inst->SrcReg[0], tmpNode.Store);
    storage_to_src_reg(&inst->SrcReg[1], n->Children[2]->Store);
 
+   free_temp_storage(emitInfo->vt, &tmpNode);
+
    return inst;
 }
 
diff --git a/src/mesa/tnl/t_vertex_sse.c b/src/mesa/tnl/t_vertex_sse.c
index 9515d9f81f..f1c98fe2d1 100644
--- a/src/mesa/tnl/t_vertex_sse.c
+++ b/src/mesa/tnl/t_vertex_sse.c
@@ -648,12 +648,12 @@ void _tnl_generate_sse_emit( GLcontext *ctx )
 
    p.ctx = ctx;
    p.inputs_safe = 0;		/* for now */
-   p.outputs_safe = 1;		/* for now */
+   p.outputs_safe = 0;		/* for now */
    p.have_sse2 = cpu_has_xmm2;
    p.identity = x86_make_reg(file_XMM, 6);
    p.chan0 = x86_make_reg(file_XMM, 7);
 
-   if (!x86_init_func(&p.func, MAX_SSE_CODE_SIZE)) {
+   if (!x86_init_func_size(&p.func, MAX_SSE_CODE_SIZE)) {
       vtx->emit = NULL;
       return;
    }
diff --git a/src/mesa/x86/rtasm/x86sse.c b/src/mesa/x86/rtasm/x86sse.c
index 612cd51a6e..772471c723 100644
--- a/src/mesa/x86/rtasm/x86sse.c
+++ b/src/mesa/x86/rtasm/x86sse.c
@@ -1,4 +1,4 @@
-#if defined(USE_X86_ASM) || defined(SLANG_X86)
+#if defined(__i386__) || defined(__386__)
 
 #include "imports.h"
 #include "x86sse.h"
@@ -6,54 +6,78 @@
 #define DISASSEM 0
 #define X86_TWOB 0x0f
 
-/* Emit bytes to the instruction stream:
- */
-static void emit_1b( struct x86_function *p, GLbyte b0 )
+static unsigned char *cptr( void (*label)() )
 {
-   *(GLbyte *)(p->csr++) = b0;
+   return (unsigned char *)(unsigned long)label;
 }
 
-static void emit_1i( struct x86_function *p, GLint i0 )
+
+static void do_realloc( struct x86_function *p )
 {
-   *(GLint *)(p->csr) = i0;
-   p->csr += 4;
+   if (p->size == 0) {
+      p->size = 1024;
+      p->store = _mesa_exec_malloc(p->size);
+      p->csr = p->store;
+   }
+   else {
+      unsigned used = p->csr - p->store;
+      unsigned char *tmp = p->store;
+      p->size *= 2;
+      p->store = _mesa_exec_malloc(p->size);
+      memcpy(p->store, tmp, used);
+      p->csr = p->store + used;
+      _mesa_exec_free(tmp);
+   }
 }
 
-static void disassem( struct x86_function *p, const char *fn )
+/* Emit bytes to the instruction stream:
+ */
+static unsigned char *reserve( struct x86_function *p, int bytes )
 {
-#if DISASSEM && 0
-   if (fn && fn != p->fn) {
-      _mesa_printf("0x%x: %s\n", p->csr, fn);
-      p->fn = fn;
+   if (p->csr + bytes - p->store > p->size)
+      do_realloc(p);
+
+   {
+      unsigned char *csr = p->csr;
+      p->csr += bytes;
+      return csr;
    }
-#endif
 }
 
-static void emit_1ub_fn( struct x86_function *p, GLubyte b0, const char *fn )
+
+
+static void emit_1b( struct x86_function *p, char b0 )
 {
-   disassem(p, fn);
-   *(p->csr++) = b0;
+   char *csr = (char *)reserve(p, 1);
+   *csr = b0;
 }
 
-static void emit_2ub_fn( struct x86_function *p, GLubyte b0, GLubyte b1, const char *fn )
+static void emit_1i( struct x86_function *p, int i0 )
 {
-   disassem(p, fn);
-   *(p->csr++) = b0;
-   *(p->csr++) = b1;
+   int *icsr = (int *)reserve(p, sizeof(i0));
+   *icsr = i0;
 }
 
-static void emit_3ub_fn( struct x86_function *p, GLubyte b0, GLubyte b1, GLubyte b2, const char *fn )
+static void emit_1ub( struct x86_function *p, unsigned char b0 )
 {
-   disassem(p, fn);
-   *(p->csr++) = b0;
-   *(p->csr++) = b1;
-   *(p->csr++) = b2;
+   unsigned char *csr = reserve(p, 1);
+   *csr++ = b0;
 }
 
-#define emit_1ub(p, b0)         emit_1ub_fn(p, b0, __FUNCTION__)
-#define emit_2ub(p, b0, b1)     emit_2ub_fn(p, b0, b1, __FUNCTION__)
-#define emit_3ub(p, b0, b1, b2) emit_3ub_fn(p, b0, b1, b2, __FUNCTION__)
+static void emit_2ub( struct x86_function *p, unsigned char b0, unsigned char b1 )
+{
+   unsigned char *csr = reserve(p, 2);
+   *csr++ = b0;
+   *csr++ = b1;
+}
 
+static void emit_3ub( struct x86_function *p, unsigned char b0, unsigned char b1, unsigned char b2 )
+{
+   unsigned char *csr = reserve(p, 3);
+   *csr++ = b0;
+   *csr++ = b1;
+   *csr++ = b2;
+}
 
 
 /* Build a modRM byte + possible displacement.  No treatment of SIB
@@ -63,7 +87,7 @@ static void emit_modrm( struct x86_function *p,
 			struct x86_reg reg, 
 			struct x86_reg regmem )
 {
-   GLubyte val = 0;
+   unsigned char val = 0;
    
    assert(reg.mod == mod_REG);
    
@@ -71,13 +95,13 @@ static void emit_modrm( struct x86_function *p,
    val |= reg.idx << 3;		/* reg field */
    val |= regmem.idx;		/* r/m field */
    
-   emit_1ub_fn(p, val, 0);
+   emit_1ub(p, val);
 
    /* Oh-oh we've stumbled into the SIB thing.
     */
    if (regmem.file == file_REG32 &&
        regmem.idx == reg_SP) {
-      emit_1ub_fn(p, 0x24, 0);		/* simplistic! */
+      emit_1ub(p, 0x24);		/* simplistic! */
    }
 
    switch (regmem.mod) {
@@ -98,7 +122,7 @@ static void emit_modrm( struct x86_function *p,
 
 
 static void emit_modrm_noreg( struct x86_function *p,
-			      GLuint op,
+			      unsigned op,
 			      struct x86_reg regmem )
 {
    struct x86_reg dummy = x86_make_reg(file_REG32, op);
@@ -111,21 +135,21 @@ static void emit_modrm_noreg( struct x86_function *p,
  * the arguments presented.
  */
 static void emit_op_modrm( struct x86_function *p,
-			   GLubyte op_dst_is_reg, 
-			   GLubyte op_dst_is_mem,
+			   unsigned char op_dst_is_reg, 
+			   unsigned char op_dst_is_mem,
 			   struct x86_reg dst,
 			   struct x86_reg src )
 {  
    switch (dst.mod) {
    case mod_REG:
-      emit_1ub_fn(p, op_dst_is_reg, 0);
+      emit_1ub(p, op_dst_is_reg);
       emit_modrm(p, dst, src);
       break;
    case mod_INDIRECT:
    case mod_DISP32:
    case mod_DISP8:
       assert(src.mod == mod_REG);
-      emit_1ub_fn(p, op_dst_is_mem, 0);
+      emit_1ub(p, op_dst_is_mem);
       emit_modrm(p, src, dst);
       break;
    default:
@@ -156,7 +180,7 @@ struct x86_reg x86_make_reg( enum x86_reg_file file,
 }
 
 struct x86_reg x86_make_disp( struct x86_reg reg,
-			      GLint disp )
+			      int disp )
 {
    assert(reg.file == file_REG32);
 
@@ -185,7 +209,7 @@ struct x86_reg x86_get_base_reg( struct x86_reg reg )
    return x86_make_reg( reg.file, reg.idx );
 }
 
-GLubyte *x86_get_label( struct x86_function *p )
+unsigned char *x86_get_label( struct x86_function *p )
 {
    return p->csr;
 }
@@ -199,13 +223,13 @@ GLubyte *x86_get_label( struct x86_function *p )
 
 void x86_jcc( struct x86_function *p,
 	      enum x86_cc cc,
-	      GLubyte *label )
+	      unsigned char *label )
 {
-   GLint offset = label - (x86_get_label(p) + 2);
+   int offset = label - (x86_get_label(p) + 2);
    
    if (offset <= 127 && offset >= -128) {
       emit_1ub(p, 0x70 + cc);
-      emit_1b(p, (GLbyte) offset);
+      emit_1b(p, (char) offset);
    }
    else {
       offset = label - (x86_get_label(p) + 6);
@@ -216,7 +240,7 @@ void x86_jcc( struct x86_function *p,
 
 /* Always use a 32bit offset for forward jumps:
  */
-GLubyte *x86_jcc_forward( struct x86_function *p,
+unsigned char *x86_jcc_forward( struct x86_function *p,
 			  enum x86_cc cc )
 {
    emit_2ub(p, 0x0f, 0x80 + cc);
@@ -224,14 +248,14 @@ GLubyte *x86_jcc_forward( struct x86_function *p,
    return x86_get_label(p);
 }
 
-GLubyte *x86_jmp_forward( struct x86_function *p)
+unsigned char *x86_jmp_forward( struct x86_function *p)
 {
    emit_1ub(p, 0xe9);
    emit_1i(p, 0);
    return x86_get_label(p);
 }
 
-GLubyte *x86_call_forward( struct x86_function *p)
+unsigned char *x86_call_forward( struct x86_function *p)
 {
    emit_1ub(p, 0xe8);
    emit_1i(p, 0);
@@ -241,28 +265,41 @@ GLubyte *x86_call_forward( struct x86_function *p)
 /* Fixup offset from forward jump:
  */
 void x86_fixup_fwd_jump( struct x86_function *p,
-			 GLubyte *fixup )
+			 unsigned char *fixup )
 {
    *(int *)(fixup - 4) = x86_get_label(p) - fixup;
 }
 
-void x86_jmp( struct x86_function *p, GLubyte *label)
+void x86_jmp( struct x86_function *p, unsigned char *label)
 {
    emit_1ub(p, 0xe9);
    emit_1i(p, label - x86_get_label(p) - 4);
 }
 
-void x86_call( struct x86_function *p, GLubyte *label)
+#if 0
+/* This doesn't work once we start reallocating & copying the
+ * generated code on buffer fills, because the call is relative to the
+ * current pc.
+ */
+void x86_call( struct x86_function *p, void (*label)())
 {
    emit_1ub(p, 0xe8);
-   emit_1i(p, label - x86_get_label(p) - 4);
+   emit_1i(p, cptr(label) - x86_get_label(p) - 4);
+}
+#else
+void x86_call( struct x86_function *p, struct x86_reg reg)
+{
+   emit_1ub(p, 0xff);
+   emit_modrm(p, reg, reg);
 }
+#endif
+
 
 /* michal:
  * Temporary. As I need immediate operands, and dont want to mess with the codegen,
  * I load the immediate into general purpose register and use it.
  */
-void x86_mov_reg_imm( struct x86_function *p, struct x86_reg dst, GLint imm )
+void x86_mov_reg_imm( struct x86_function *p, struct x86_reg dst, int imm )
 {
    assert(dst.mod == mod_REG);
    emit_1ub(p, 0xb8 + dst.idx);
@@ -502,6 +539,14 @@ void sse_addss( struct x86_function *p,
    emit_modrm( p, dst, src );
 }
 
+void sse_andnps( struct x86_function *p,
+                 struct x86_reg dst,
+                 struct x86_reg src )
+{
+   emit_2ub(p, X86_TWOB, 0x55);
+   emit_modrm( p, dst, src );
+}
+
 void sse_andps( struct x86_function *p,
 		struct x86_reg dst,
 		struct x86_reg src )
@@ -510,6 +555,13 @@ void sse_andps( struct x86_function *p,
    emit_modrm( p, dst, src );
 }
 
+void sse_rsqrtps( struct x86_function *p,
+                  struct x86_reg dst,
+                  struct x86_reg src )
+{
+   emit_2ub(p, X86_TWOB, 0x52);
+   emit_modrm( p, dst, src );
+}
 
 void sse_rsqrtss( struct x86_function *p,
 		  struct x86_reg dst,
@@ -538,6 +590,21 @@ void sse_movlhps( struct x86_function *p,
    emit_modrm( p, dst, src );
 }
 
+void sse_orps( struct x86_function *p,
+               struct x86_reg dst,
+               struct x86_reg src )
+{
+   emit_2ub(p, X86_TWOB, 0x56);
+   emit_modrm( p, dst, src );
+}
+
+void sse_xorps( struct x86_function *p,
+                struct x86_reg dst,
+                struct x86_reg src )
+{
+   emit_2ub(p, X86_TWOB, 0x57);
+   emit_modrm( p, dst, src );
+}
 
 void sse_cvtps2pi( struct x86_function *p,
 		   struct x86_reg dst,
@@ -559,7 +626,7 @@ void sse_cvtps2pi( struct x86_function *p,
 void sse_shufps( struct x86_function *p,
 		 struct x86_reg dest,
 		 struct x86_reg arg0,
-		 GLubyte shuf) 
+		 unsigned char shuf) 
 {
    emit_2ub(p, X86_TWOB, 0xC6);
    emit_modrm(p, dest, arg0);
@@ -569,13 +636,21 @@ void sse_shufps( struct x86_function *p,
 void sse_cmpps( struct x86_function *p,
 		struct x86_reg dest,
 		struct x86_reg arg0,
-		GLubyte cc) 
+		unsigned char cc) 
 {
    emit_2ub(p, X86_TWOB, 0xC2);
    emit_modrm(p, dest, arg0);
    emit_1ub(p, cc); 
 }
 
+void sse_pmovmskb( struct x86_function *p,
+                   struct x86_reg dest,
+                   struct x86_reg src)
+{
+    emit_3ub(p, 0x66, X86_TWOB, 0xD7);
+    emit_modrm(p, dest, src);
+}
+
 /***********************************************************************
  * SSE2 instructions
  */
@@ -586,13 +661,21 @@ void sse_cmpps( struct x86_function *p,
 void sse2_pshufd( struct x86_function *p,
 		  struct x86_reg dest,
 		  struct x86_reg arg0,
-		  GLubyte shuf) 
+		  unsigned char shuf) 
 {
    emit_3ub(p, 0x66, X86_TWOB, 0x70);
    emit_modrm(p, dest, arg0);
    emit_1ub(p, shuf); 
 }
 
+void sse2_cvttps2dq( struct x86_function *p,
+                     struct x86_reg dst,
+                     struct x86_reg src )
+{
+   emit_3ub( p, 0xF3, X86_TWOB, 0x5B );
+   emit_modrm( p, dst, src );
+}
+
 void sse2_cvtps2dq( struct x86_function *p,
 		    struct x86_reg dst,
 		    struct x86_reg src )
@@ -625,6 +708,14 @@ void sse2_packuswb( struct x86_function *p,
    emit_modrm( p, dst, src );
 }
 
+void sse2_rcpps( struct x86_function *p,
+                 struct x86_reg dst,
+                 struct x86_reg src )
+{
+   emit_2ub(p, X86_TWOB, 0x53);
+   emit_modrm( p, dst, src );
+}
+
 void sse2_rcpss( struct x86_function *p,
 		struct x86_reg dst,
 		struct x86_reg src )
@@ -712,11 +803,11 @@ void x87_fclex( struct x86_function *p )
 
 
 static void x87_arith_op( struct x86_function *p, struct x86_reg dst, struct x86_reg arg,
-			  GLubyte dst0ub0,
-			  GLubyte dst0ub1,
-			  GLubyte arg0ub0,
-			  GLubyte arg0ub1,
-			  GLubyte argmem_noreg)
+			  unsigned char dst0ub0,
+			  unsigned char dst0ub1,
+			  unsigned char arg0ub0,
+			  unsigned char arg0ub1,
+			  unsigned char argmem_noreg)
 {
    assert(dst.file == file_x87);
 
@@ -729,7 +820,7 @@ static void x87_arith_op( struct x86_function *p, struct x86_reg dst, struct x86
 	 assert(0);
    }
    else if (dst.idx == 0) {
-      assert(arg.file = file_REG32);
+      assert(arg.file == file_REG32);
       emit_1ub(p, 0xd8);
       emit_modrm_noreg(p, argmem_noreg, arg);
    }
@@ -1056,44 +1147,42 @@ void mmx_movq( struct x86_function *p,
  * account any push/pop activity:
  */
 struct x86_reg x86_fn_arg( struct x86_function *p,
-			   GLuint arg )
+			   unsigned arg )
 {
    return x86_make_disp(x86_make_reg(file_REG32, reg_SP), 
 			p->stack_offset + arg * 4);	/* ??? */
 }
 
 
-/**
- * Initialize an x86_function object, allocating space for up to
- * 'code_size' bytes of code.
- */
-GLboolean x86_init_func( struct x86_function *p, GLuint code_size )
+void x86_init_func( struct x86_function *p )
 {
-   assert(!p->store);
+   p->size = 0;
+   p->store = NULL;
+   p->csr = p->store;
+}
+
+int x86_init_func_size( struct x86_function *p, unsigned code_size )
+{
+   p->size = code_size;
    p->store = _mesa_exec_malloc(code_size);
-   if (p->store) {
-      p->csr = p->store;
-      return GL_TRUE;
-   }
-   else {
-      p->csr = NULL;
-      return GL_FALSE;
-   }
+   p->csr = p->store;
+   return p->store != NULL;
 }
 
 void x86_release_func( struct x86_function *p )
 {
-   if (p->store)
-      _mesa_exec_free(p->store);
-   p->store = p->csr = NULL;
+   _mesa_exec_free(p->store);
+   p->store = NULL;
+   p->csr = NULL;
+   p->size = 0;
 }
 
 
 void (*x86_get_func( struct x86_function *p ))(void)
 {
-   if (DISASSEM)
+   if (DISASSEM && p->store)
       _mesa_printf("disassemble %p %p\n", p->store, p->csr);
-   return (void (*)(void))p->store;
+   return (void (*)(void)) (unsigned long) p->store;
 }
 
 #else
diff --git a/src/mesa/x86/rtasm/x86sse.h b/src/mesa/x86/rtasm/x86sse.h
index 42b09937bc..f6282f5bd4 100644
--- a/src/mesa/x86/rtasm/x86sse.h
+++ b/src/mesa/x86/rtasm/x86sse.h
@@ -2,26 +2,25 @@
 #ifndef _X86SSE_H_
 #define _X86SSE_H_
 
-#if defined(USE_X86_ASM) || defined(SLANG_X86)
-
-#include "glheader.h"
+#if defined(__i386__) || defined(__386__)
 
 /* It is up to the caller to ensure that instructions issued are
  * suitable for the host cpu.  There are no checks made in this module
  * for mmx/sse/sse2 support on the cpu.
  */
 struct x86_reg {
-   GLuint file:3;
-   GLuint idx:3;
-   GLuint mod:2;		/* mod_REG if this is just a register */
-   GLint  disp:24;		/* only +/- 23bits of offset - should be enough... */
+   unsigned file:3;
+   unsigned idx:3;
+   unsigned mod:2;		/* mod_REG if this is just a register */
+   int      disp:24;		/* only +/- 23bits of offset - should be enough... */
 };
 
 struct x86_function {
-   GLubyte *store;
-   GLubyte *csr;
-   GLuint stack_offset;
-   GLint need_emms;
+   unsigned size;
+   unsigned char *store;
+   unsigned char *csr;
+   unsigned stack_offset;
+   int need_emms;
    const char *fn;
 };
 
@@ -80,7 +79,8 @@ enum sse_cc {
  */
 
 
-GLboolean x86_init_func( struct x86_function *p, GLuint code_size );
+void x86_init_func( struct x86_function *p );
+int x86_init_func_size( struct x86_function *p, unsigned code_size );
 void x86_release_func( struct x86_function *p );
 void (*x86_get_func( struct x86_function *p ))( void );
 
@@ -92,7 +92,7 @@ struct x86_reg x86_make_reg( enum x86_reg_file file,
 			     enum x86_reg_name idx );
 
 struct x86_reg x86_make_disp( struct x86_reg reg,
-			      GLint disp );
+			      int disp );
 
 struct x86_reg x86_deref( struct x86_reg reg );
 
@@ -101,31 +101,32 @@ struct x86_reg x86_get_base_reg( struct x86_reg reg );
 
 /* Labels, jumps and fixup:
  */
-GLubyte *x86_get_label( struct x86_function *p );
+unsigned char *x86_get_label( struct x86_function *p );
 
 void x86_jcc( struct x86_function *p,
 	      enum x86_cc cc,
-	      GLubyte *label );
+	      unsigned char *label );
 
-GLubyte *x86_jcc_forward( struct x86_function *p,
+unsigned char *x86_jcc_forward( struct x86_function *p,
 			  enum x86_cc cc );
 
-GLubyte *x86_jmp_forward( struct x86_function *p);
+unsigned char *x86_jmp_forward( struct x86_function *p);
 
-GLubyte *x86_call_forward( struct x86_function *p);
+unsigned char *x86_call_forward( struct x86_function *p);
 
 void x86_fixup_fwd_jump( struct x86_function *p,
-			 GLubyte *fixup );
+			 unsigned char *fixup );
 
-void x86_jmp( struct x86_function *p, GLubyte *label );
+void x86_jmp( struct x86_function *p, unsigned char *label );
 
-void x86_call( struct x86_function *p, GLubyte *label );
+/* void x86_call( struct x86_function *p, void (*label)() ); */
+void x86_call( struct x86_function *p, struct x86_reg reg);
 
 /* michal:
  * Temporary. As I need immediate operands, and dont want to mess with the codegen,
  * I load the immediate into general purpose register and use it.
  */
-void x86_mov_reg_imm( struct x86_function *p, struct x86_reg dst, GLint imm );
+void x86_mov_reg_imm( struct x86_function *p, struct x86_reg dst, int imm );
 
 
 /* Macro for sse_shufps() and sse2_pshufd():
@@ -141,19 +142,24 @@ void mmx_packssdw( struct x86_function *p, struct x86_reg dst, struct x86_reg sr
 void mmx_packuswb( struct x86_function *p, struct x86_reg dst, struct x86_reg src );
 
 void sse2_cvtps2dq( struct x86_function *p, struct x86_reg dst, struct x86_reg src );
+void sse2_cvttps2dq( struct x86_function *p, struct x86_reg dst, struct x86_reg src );
 void sse2_movd( struct x86_function *p, struct x86_reg dst, struct x86_reg src );
 void sse2_packssdw( struct x86_function *p, struct x86_reg dst, struct x86_reg src );
 void sse2_packsswb( struct x86_function *p, struct x86_reg dst, struct x86_reg src );
 void sse2_packuswb( struct x86_function *p, struct x86_reg dst, struct x86_reg src );
-void sse2_pshufd( struct x86_function *p, struct x86_reg dest, struct x86_reg arg0, GLubyte shuf );
+void sse2_pshufd( struct x86_function *p, struct x86_reg dest, struct x86_reg arg0,
+                  unsigned char shuf );
+void sse2_rcpps( struct x86_function *p, struct x86_reg dst, struct x86_reg src );
 void sse2_rcpss( struct x86_function *p, struct x86_reg dst, struct x86_reg src );
 
 void sse_addps( struct x86_function *p, struct x86_reg dst, struct x86_reg src );
 void sse_addss( struct x86_function *p, struct x86_reg dst, struct x86_reg src );
 void sse_cvtps2pi( struct x86_function *p, struct x86_reg dst, struct x86_reg src );
 void sse_divss( struct x86_function *p, struct x86_reg dst, struct x86_reg src );
+void sse_andnps( struct x86_function *p, struct x86_reg dst, struct x86_reg src );
 void sse_andps( struct x86_function *p, struct x86_reg dst, struct x86_reg src );
-void sse_cmpps( struct x86_function *p, struct x86_reg dst, struct x86_reg src, GLubyte cc );
+void sse_cmpps( struct x86_function *p, struct x86_reg dst, struct x86_reg src,
+                unsigned char cc );
 void sse_maxps( struct x86_function *p, struct x86_reg dst, struct x86_reg src );
 void sse_maxss( struct x86_function *p, struct x86_reg dst, struct x86_reg src );
 void sse_minps( struct x86_function *p, struct x86_reg dst, struct x86_reg src );
@@ -166,9 +172,14 @@ void sse_movss( struct x86_function *p, struct x86_reg dst, struct x86_reg src )
 void sse_movups( struct x86_function *p, struct x86_reg dst, struct x86_reg src );
 void sse_mulps( struct x86_function *p, struct x86_reg dst, struct x86_reg src );
 void sse_mulss( struct x86_function *p, struct x86_reg dst, struct x86_reg src );
+void sse_orps( struct x86_function *p, struct x86_reg dst, struct x86_reg src );
+void sse_xorps( struct x86_function *p, struct x86_reg dst, struct x86_reg src );
 void sse_subps( struct x86_function *p, struct x86_reg dst, struct x86_reg src );
+void sse_rsqrtps( struct x86_function *p, struct x86_reg dst, struct x86_reg src );
 void sse_rsqrtss( struct x86_function *p, struct x86_reg dst, struct x86_reg src );
-void sse_shufps( struct x86_function *p, struct x86_reg dest, struct x86_reg arg0, GLubyte shuf );
+void sse_shufps( struct x86_function *p, struct x86_reg dest, struct x86_reg arg0,
+                 unsigned char shuf );
+void sse_pmovmskb( struct x86_function *p, struct x86_reg dest, struct x86_reg src );
 
 void x86_add( struct x86_function *p, struct x86_reg dst, struct x86_reg src );
 void x86_and( struct x86_function *p, struct x86_reg dst, struct x86_reg src );
@@ -239,7 +250,7 @@ void x87_fucom( struct x86_function *p, struct x86_reg arg );
  * account any push/pop activity.  Note - doesn't track explict
  * manipulation of ESP by other instructions.
  */
-struct x86_reg x86_fn_arg( struct x86_function *p, GLuint arg );
+struct x86_reg x86_fn_arg( struct x86_function *p, unsigned arg );
 
 #endif
 #endif
author	Thomas Hellstrom <thomas-at-tungstengraphics-dot-com>	2008-03-30 15:33:07 +0200
committer	Thomas Hellstrom <thomas-at-tungstengraphics-dot-com>	2008-03-30 15:33:07 +0200
commit	a0e133114bbde5c1f7e6ea20613b513f7fbbbdae (patch)
tree	b0dc6c5e287c1ae0bf3bc8357ab992cbb792df11
parent	43b05ffda491046bf0cca7f0cdd8a337eb09f1f1 (diff)
parent	767dfa5b9c07fe6ae267e1ff0d3bbbf093a04c44 (diff)