Import i915 and i965 dri drivers from mesa 7.7.0.7.7.0

15 files changed, 355 insertions, 96 deletions

diff --git a/i915/Makefile.am b/i915/Makefile.am
index 451e9fd..c994f96 100644
--- a/i915/Makefile.am
+++ b/i915/Makefile.am
@@ -5,7 +5,7 @@ I915_CFLAGS = -I../shared -I../shared/server -DI915
 i915_dri_la_LTLIBRARIES = i915_dri.la
 i915_dri_la_CFLAGS = $(AM_CFLAGS) $(DRM_CFLAGS) $(DRI_CFLAGS) $(I915_CFLAGS)
 i915_dri_la_LDFLAGS = -module -noprefix -avoid-version -lm -ldl \
-			 $(DRM_LIBS) -ldrm_intel $(DRI_LIBS)
+			 $(DRM_LIBS) $(DRI_LIBS)
 i915_dri_ladir = @libdir@/dri
 i915_dri_la_SOURCES = \
 	i830_context.c \
@@ -20,7 +20,6 @@ i915_dri_la_SOURCES = \
 	../shared/intel_batchbuffer.c \
 	../shared/intel_clear.c \
 	../shared/intel_extensions.c \
-	../shared/intel_generatemipmap.c \
 	../shared/intel_mipmap_tree.c \
 	../shared/intel_tex_layout.c \
 	../shared/intel_tex_image.c \
diff --git a/i915/i830_texblend.c b/i915/i830_texblend.c
index 09f7f37..3f64be8 100644
--- a/i915/i830_texblend.c
+++ b/i915/i830_texblend.c
@@ -30,7 +30,6 @@
 #include "main/mtypes.h"
 #include "main/simple_list.h"
 #include "main/enums.h"
-#include "main/texformat.h"
 #include "main/texstore.h"
 #include "main/mm.h"
 
diff --git a/i915/i830_texstate.c b/i915/i830_texstate.c
index 6f998fa..ce409b3 100644
--- a/i915/i830_texstate.c
+++ b/i915/i830_texstate.c
@@ -27,7 +27,6 @@
 
 #include "main/mtypes.h"
 #include "main/enums.h"
-#include "main/texformat.h"
 
 #include "intel_mipmap_tree.h"
 #include "intel_tex.h"
@@ -56,10 +55,9 @@ translate_texture_format(GLuint mesa_format, GLuint internal_format)
    case MESA_FORMAT_ARGB4444:
       return MAPSURF_16BIT | MT_16BIT_ARGB4444;
    case MESA_FORMAT_ARGB8888:
-      if (internal_format == GL_RGB)
-	 return MAPSURF_32BIT | MT_32BIT_XRGB8888;
-      else
-	 return MAPSURF_32BIT | MT_32BIT_ARGB8888;
+      return MAPSURF_32BIT | MT_32BIT_ARGB8888;
+   case MESA_FORMAT_XRGB8888:
+      return MAPSURF_32BIT | MT_32BIT_XRGB8888;
    case MESA_FORMAT_YCBCR_REV:
       return (MAPSURF_422 | MT_422_YCRCB_NORMAL);
    case MESA_FORMAT_YCBCR:
@@ -160,13 +158,20 @@ i830_update_tex_unit(struct intel_context *intel, GLuint unit, GLuint ss3)
 
       pitch = intelObj->pitchOverride;
    } else {
+      GLuint dst_x, dst_y;
+
+      intel_miptree_get_image_offset(intelObj->mt, intelObj->firstLevel, 0, 0,
+				     &dst_x, &dst_y);
+
       dri_bo_reference(intelObj->mt->region->buffer);
       i830->state.tex_buffer[unit] = intelObj->mt->region->buffer;
-      i830->state.tex_offset[unit] = intel_miptree_image_offset(intelObj->mt,
-								0, intelObj->
-								firstLevel);
+      /* XXX: This calculation is probably broken for tiled images with
+       * a non-page-aligned offset.
+       */
+      i830->state.tex_offset[unit] = (dst_x + dst_y * intelObj->mt->pitch) *
+	 intelObj->mt->cpp;
 
-      format = translate_texture_format(firstImage->TexFormat->MesaFormat,
+      format = translate_texture_format(firstImage->TexFormat,
 					firstImage->InternalFormat);
       pitch = intelObj->mt->pitch * intelObj->mt->cpp;
    }
diff --git a/i915/i830_vtbl.c b/i915/i830_vtbl.c
index 983f672..e8c8d5a 100644
--- a/i915/i830_vtbl.c
+++ b/i915/i830_vtbl.c
@@ -26,7 +26,6 @@
  **************************************************************************/
 
 #include "glapi/glapi.h"
-#include "main/texformat.h"
 
 #include "i830_context.h"
 #include "i830_reg.h"
@@ -646,8 +645,9 @@ i830_state_draw_region(struct intel_context *intel,
             DSTORG_VERT_BIAS(0x8) | DEPTH_IS_Z);    /* .5 */
 
    if (irb != NULL) {
-      switch (irb->texformat->MesaFormat) {
+      switch (irb->Base.Format) {
       case MESA_FORMAT_ARGB8888:
+      case MESA_FORMAT_XRGB8888:
 	 value |= DV_PF_8888;
 	 break;
       case MESA_FORMAT_RGB565:
@@ -661,7 +661,7 @@ i830_state_draw_region(struct intel_context *intel,
 	 break;
       default:
 	 _mesa_problem(ctx, "Bad renderbuffer format: %d\n",
-		       irb->texformat->MesaFormat);
+		       irb->Base.Format);
       }
    }
 
@@ -719,15 +719,6 @@ i830_new_batch(struct intel_context *intel)
    assert(!intel->no_batch_wrap);
 }
 
-
-
-static GLuint
-i830_flush_cmd(void)
-{
-   return MI_FLUSH | FLUSH_MAP_CACHE;
-}
-
-
 static void 
 i830_assert_not_dirty( struct intel_context *intel )
 {
@@ -753,7 +744,6 @@ i830InitVtbl(struct i830_context *i830)
    i830->intel.vtbl.reduced_primitive_state = i830_reduced_primitive_state;
    i830->intel.vtbl.set_draw_region = i830_set_draw_region;
    i830->intel.vtbl.update_texture_state = i830UpdateTextureState;
-   i830->intel.vtbl.flush_cmd = i830_flush_cmd;
    i830->intel.vtbl.render_start = i830_render_start;
    i830->intel.vtbl.render_prevalidate = i830_render_prevalidate;
    i830->intel.vtbl.assert_not_dirty = i830_assert_not_dirty;
diff --git a/i915/i915_context.c b/i915/i915_context.c
index 3ab7d68..7d4c7cf 100644
--- a/i915/i915_context.c
+++ b/i915/i915_context.c
@@ -40,6 +40,7 @@
 
 #include "utils.h"
 #include "i915_reg.h"
+#include "i915_program.h"
 
 #include "intel_regions.h"
 #include "intel_batchbuffer.h"
@@ -80,6 +81,8 @@ i915InvalidateState(GLcontext * ctx, GLuint new_state)
       i915_update_stencil(ctx);
    if (new_state & (_NEW_LIGHT))
        i915_update_provoking_vertex(ctx);
+   if (new_state & (_NEW_PROGRAM | _NEW_PROGRAM_CONSTANTS))
+       i915_update_program(ctx);
 }
 
 
@@ -139,7 +142,7 @@ i915CreateContext(const __GLcontextModes * mesaVis,
    ctx->Const.MaxTextureUnits = I915_TEX_UNITS;
    ctx->Const.MaxTextureImageUnits = I915_TEX_UNITS;
    ctx->Const.MaxTextureCoordUnits = I915_TEX_UNITS;
-
+   ctx->Const.MaxVarying = I915_TEX_UNITS;
 
    /* Advertise the full hardware capabilities.  The new memory
     * manager should cope much better with overload situations:
diff --git a/i915/i915_context.h b/i915/i915_context.h
index 8de4a9d..25418d5 100644
--- a/i915/i915_context.h
+++ b/i915/i915_context.h
@@ -39,6 +39,7 @@
 #define I915_FALLBACK_LOGICOP		 0x20000
 #define I915_FALLBACK_POLYGON_SMOOTH	 0x40000
 #define I915_FALLBACK_POINT_SMOOTH	 0x80000
+#define I915_FALLBACK_POINT_SPRITE_COORD_ORIGIN	 0x100000
 
 #define I915_UPLOAD_CTX              0x1
 #define I915_UPLOAD_BUFFERS          0x2
@@ -121,10 +122,14 @@ enum {
 #define I915_MAX_CONSTANT      32
 #define I915_CONSTANT_SIZE     (2+(4*I915_MAX_CONSTANT))
 
+#define I915_MAX_INSN          (I915_MAX_DECL_INSN + \
+				I915_MAX_TEX_INSN + \
+				I915_MAX_ALU_INSN)
 
-#define I915_PROGRAM_SIZE      192
-
-#define I915_MAX_INSN          (I915_MAX_TEX_INSN+I915_MAX_ALU_INSN)
+/* Maximum size of the program packet, which matches the limits on
+ * decl, tex, and ALU instructions.
+ */
+#define I915_PROGRAM_SIZE      (I915_MAX_INSN * 3 + 1)
 
 /* Hardware version of a parsed fragment program.  "Derived" from the
  * mesa fragment_program struct.
@@ -154,8 +159,9 @@ struct i915_fragment_program
     */
    GLcontext *ctx;
 
-   GLuint declarations[I915_PROGRAM_SIZE];
-   GLuint program[I915_PROGRAM_SIZE];
+   /* declarations contains the packet header. */
+   GLuint declarations[I915_MAX_DECL_INSN * 3 + 1];
+   GLuint program[(I915_MAX_TEX_INSN + I915_MAX_ALU_INSN) * 3];
 
    GLfloat constant[I915_MAX_CONSTANT][4];
    GLuint constant_flags[I915_MAX_CONSTANT];
diff --git a/i915/i915_debug.c b/i915/i915_debug.c
index f7bb7ea..fecfac3 100644
--- a/i915/i915_debug.c
+++ b/i915/i915_debug.c
@@ -806,6 +806,7 @@ static GLboolean i915_debug_packet( struct debug_stream *stream )
       default:
 	 return debug(stream, "", 0);
       }
+      break;
    default:
       assert(0);
       return 0;
diff --git a/i915/i915_fragprog.c b/i915/i915_fragprog.c
index 2db10c6..d9c6144 100644
--- a/i915/i915_fragprog.c
+++ b/i915/i915_fragprog.c
@@ -89,7 +89,8 @@ src_vector(struct i915_fragment_program *p,
        */
    case PROGRAM_TEMPORARY:
       if (source->Index >= I915_MAX_TEMPORARY) {
-         i915_program_error(p, "Exceeded max temporary reg");
+         i915_program_error(p, "Exceeded max temporary reg: %d/%d",
+			    source->Index, I915_MAX_TEMPORARY);
          return 0;
       }
       src = UREG(REG_TYPE_R, source->Index);
@@ -121,10 +122,23 @@ src_vector(struct i915_fragment_program *p,
          src = i915_emit_decl(p, REG_TYPE_T,
                               T_TEX0 + (source->Index - FRAG_ATTRIB_TEX0),
                               D0_CHANNEL_ALL);
+	 break;
+
+      case FRAG_ATTRIB_VAR0:
+      case FRAG_ATTRIB_VAR0 + 1:
+      case FRAG_ATTRIB_VAR0 + 2:
+      case FRAG_ATTRIB_VAR0 + 3:
+      case FRAG_ATTRIB_VAR0 + 4:
+      case FRAG_ATTRIB_VAR0 + 5:
+      case FRAG_ATTRIB_VAR0 + 6:
+      case FRAG_ATTRIB_VAR0 + 7:
+         src = i915_emit_decl(p, REG_TYPE_T,
+                              T_TEX0 + (source->Index - FRAG_ATTRIB_VAR0),
+                              D0_CHANNEL_ALL);
          break;
 
       default:
-         i915_program_error(p, "Bad source->Index");
+         i915_program_error(p, "Bad source->Index: %d", source->Index);
          return 0;
       }
       break;
@@ -146,6 +160,7 @@ src_vector(struct i915_fragment_program *p,
    case PROGRAM_CONSTANT:
    case PROGRAM_STATE_VAR:
    case PROGRAM_NAMED_PARAM:
+   case PROGRAM_UNIFORM:
       src =
          i915_emit_param4fv(p,
                             program->Base.Parameters->ParameterValues[source->
@@ -153,7 +168,7 @@ src_vector(struct i915_fragment_program *p,
       break;
 
    default:
-      i915_program_error(p, "Bad source->File");
+      i915_program_error(p, "Bad source->File: %d", source->File);
       return 0;
    }
 
@@ -186,13 +201,14 @@ get_result_vector(struct i915_fragment_program *p,
          p->depth_written = 1;
          return UREG(REG_TYPE_OD, 0);
       default:
-         i915_program_error(p, "Bad inst->DstReg.Index");
+         i915_program_error(p, "Bad inst->DstReg.Index: %d",
+			    inst->DstReg.Index);
          return 0;
       }
    case PROGRAM_TEMPORARY:
       return UREG(REG_TYPE_R, inst->DstReg.Index);
    default:
-      i915_program_error(p, "Bad inst->DstReg.File");
+      i915_program_error(p, "Bad inst->DstReg.File: %d", inst->DstReg.File);
       return 0;
    }
 }
@@ -231,7 +247,7 @@ translate_tex_src_target(struct i915_fragment_program *p, GLubyte bit)
    case TEXTURE_CUBE_INDEX:
       return D0_SAMPLE_TYPE_CUBE;
    default:
-      i915_program_error(p, "TexSrcBit");
+      i915_program_error(p, "TexSrcBit: %d", bit);
       return 0;
    }
 }
@@ -351,7 +367,7 @@ upload_program(struct i915_fragment_program *p)
 
    while (1) {
       GLuint src0, src1, src2, flags;
-      GLuint tmp = 0, consts0 = 0, consts1 = 0;
+      GLuint tmp = 0, dst, consts0 = 0, consts1 = 0;
 
       switch (inst->Opcode) {
       case OPCODE_ABS:
@@ -503,6 +519,10 @@ upload_program(struct i915_fragment_program *p)
          EMIT_1ARG_ARITH(A0_FLR);
          break;
 
+      case OPCODE_TRUNC:
+	 EMIT_1ARG_ARITH(A0_TRC);
+	 break;
+
       case OPCODE_FRC:
          EMIT_1ARG_ARITH(A0_FRC);
          break;
@@ -516,6 +536,22 @@ upload_program(struct i915_fragment_program *p)
                          0, src0, T0_TEXKILL);
          break;
 
+      case OPCODE_KIL_NV:
+	 if (inst->DstReg.CondMask == COND_TR) {
+	    tmp = i915_get_utemp(p);
+
+	    i915_emit_texld(p, get_live_regs(p, inst),
+			    tmp, A0_DEST_CHANNEL_ALL,
+			    0, /* use a dummy dest reg */
+			    swizzle(tmp, ONE, ONE, ONE, ONE), /* always */
+			    T0_TEXKILL);
+	 } else {
+	    p->error = 1;
+	    i915_program_error(p, "Unsupported KIL_NV condition code: %d",
+			       inst->DstReg.CondMask);
+	 }
+	 break;
+
       case OPCODE_LG2:
          src0 = src_vector(p, &inst->SrcReg[0], program);
 
@@ -615,6 +651,20 @@ upload_program(struct i915_fragment_program *p)
          EMIT_2ARG_ARITH(A0_MUL);
          break;
 
+      case OPCODE_NOISE1:
+      case OPCODE_NOISE2:
+      case OPCODE_NOISE3:
+      case OPCODE_NOISE4:
+	 /* Don't implement noise because we just don't have the instructions
+	  * to spare.  We aren't the first vendor to do so.
+	  */
+	 i915_program_error(p, "Stubbed-out noise functions");
+	 i915_emit_arith(p,
+			 A0_MOV,
+			 get_result_vector(p, inst),
+			 get_result_flags(inst), 0,
+			 swizzle(src0, ZERO, ZERO, ZERO, ZERO), 0, 0);
+
       case OPCODE_POW:
          src0 = src_vector(p, &inst->SrcReg[0], program);
          src1 = src_vector(p, &inst->SrcReg[1], program);
@@ -721,9 +771,38 @@ upload_program(struct i915_fragment_program *p)
          }
          break;
 
-      case OPCODE_SGE:
-         EMIT_2ARG_ARITH(A0_SGE);
-         break;
+      case OPCODE_SEQ:
+	 tmp = i915_get_utemp(p);
+	 flags = get_result_flags(inst);
+	 dst = get_result_vector(p, inst);
+
+	 /* dst = src1 >= src2 */
+	 i915_emit_arith(p,
+			 A0_SGE,
+			 dst,
+			 flags, 0,
+			 src_vector(p, &inst->SrcReg[0], program),
+			 src_vector(p, &inst->SrcReg[1], program),
+			 0);
+	 /* tmp = src1 <= src2 */
+	 i915_emit_arith(p,
+			 A0_SGE,
+			 tmp,
+			 flags, 0,
+			 negate(src_vector(p, &inst->SrcReg[0], program),
+				1, 1, 1, 1),
+			 negate(src_vector(p, &inst->SrcReg[1], program),
+				1, 1, 1, 1),
+			 0);
+	 /* dst = tmp && dst */
+	 i915_emit_arith(p,
+			 A0_MUL,
+			 dst,
+			 flags, 0,
+			 dst,
+			 tmp,
+			 0);
+	 break;
 
       case OPCODE_SIN:
          src0 = src_vector(p, &inst->SrcReg[0], program);
@@ -809,10 +888,71 @@ upload_program(struct i915_fragment_program *p)
 
          break;
 
+      case OPCODE_SGE:
+	 EMIT_2ARG_ARITH(A0_SGE);
+	 break;
+
+      case OPCODE_SGT:
+	 i915_emit_arith(p,
+			 A0_SLT,
+			 get_result_vector( p, inst ),
+			 get_result_flags( inst ), 0,
+			 negate(src_vector( p, &inst->SrcReg[0], program),
+				1, 1, 1, 1),
+			 negate(src_vector( p, &inst->SrcReg[1], program),
+				1, 1, 1, 1),
+			 0);
+         break;
+
+      case OPCODE_SLE:
+	 i915_emit_arith(p,
+			 A0_SGE,
+			 get_result_vector( p, inst ),
+			 get_result_flags( inst ), 0,
+			 negate(src_vector( p, &inst->SrcReg[0], program),
+				1, 1, 1, 1),
+			 negate(src_vector( p, &inst->SrcReg[1], program),
+				1, 1, 1, 1),
+			 0);
+         break;
+
       case OPCODE_SLT:
          EMIT_2ARG_ARITH(A0_SLT);
          break;
 
+      case OPCODE_SNE:
+	 tmp = i915_get_utemp(p);
+	 flags = get_result_flags(inst);
+	 dst = get_result_vector(p, inst);
+
+	 /* dst = src1 < src2 */
+	 i915_emit_arith(p,
+			 A0_SLT,
+			 dst,
+			 flags, 0,
+			 src_vector(p, &inst->SrcReg[0], program),
+			 src_vector(p, &inst->SrcReg[1], program),
+			 0);
+	 /* tmp = src1 > src2 */
+	 i915_emit_arith(p,
+			 A0_SLT,
+			 tmp,
+			 flags, 0,
+			 negate(src_vector(p, &inst->SrcReg[0], program),
+				1, 1, 1, 1),
+			 negate(src_vector(p, &inst->SrcReg[1], program),
+				1, 1, 1, 1),
+			 0);
+	 /* dst = tmp || dst */
+	 i915_emit_arith(p,
+			 A0_ADD,
+			 dst,
+			 flags | A0_DEST_SATURATE, 0,
+			 dst,
+			 tmp,
+			 0);
+         break;
+
       case OPCODE_SUB:
          src0 = src_vector(p, &inst->SrcReg[0], program);
          src1 = src_vector(p, &inst->SrcReg[1], program);
@@ -869,8 +1009,39 @@ upload_program(struct i915_fragment_program *p)
       case OPCODE_END:
          return;
 
+      case OPCODE_BGNLOOP:
+      case OPCODE_BGNSUB:
+      case OPCODE_BRA:
+      case OPCODE_BRK:
+      case OPCODE_CAL:
+      case OPCODE_CONT:
+      case OPCODE_DDX:
+      case OPCODE_DDY:
+      case OPCODE_ELSE:
+      case OPCODE_ENDIF:
+      case OPCODE_ENDLOOP:
+      case OPCODE_ENDSUB:
+      case OPCODE_IF:
+      case OPCODE_RET:
+	 p->error = 1;
+	 i915_program_error(p, "Unsupported opcode: %s",
+			    _mesa_opcode_string(inst->Opcode));
+	 return;
+
+      case OPCODE_EXP:
+      case OPCODE_LOG:
+	 /* These opcodes are claimed as GLSL, NV_vp, and ARB_vp in
+	  * prog_instruction.h, but apparently GLSL doesn't ever emit them.
+	  * Instead, it translates to EX2 or LG2.
+	  */
+      case OPCODE_TXD:
+      case OPCODE_TXL:
+	 /* These opcodes are claimed by GLSL in prog_instruction.h, but
+	  * only NV_vp/fp appears to emit them.
+	  */
       default:
-         i915_program_error(p, "bad opcode");
+         i915_program_error(p, "bad opcode: %s",
+			    _mesa_opcode_string(inst->Opcode));
          return;
       }
 
@@ -906,7 +1077,7 @@ check_wpos(struct i915_fragment_program *p)
    p->wpos_tex = -1;
 
    for (i = 0; i < p->ctx->Const.MaxTextureCoordUnits; i++) {
-      if (inputs & FRAG_BIT_TEX(i))
+      if (inputs & (FRAG_BIT_TEX(i) | FRAG_BIT_VAR(i)))
          continue;
       else if (inputs & FRAG_BIT_WPOS) {
          p->wpos_tex = i;
@@ -1055,6 +1226,28 @@ i915ProgramStringNotify(GLcontext * ctx,
    _tnl_program_string(ctx, target, prog);
 }
 
+void
+i915_update_program(GLcontext *ctx)
+{
+   struct intel_context *intel = intel_context(ctx);
+   struct i915_context *i915 = i915_context(&intel->ctx);
+   struct i915_fragment_program *fp =
+      (struct i915_fragment_program *) ctx->FragmentProgram._Current;
+
+   if (i915->current_program != fp) {
+      if (i915->current_program) {
+         i915->current_program->on_hardware = 0;
+         i915->current_program->params_uptodate = 0;
+      }
+
+      i915->current_program = fp;
+   }
+
+   if (!fp->translated)
+      translate_program(fp);
+
+   FALLBACK(&i915->intel, I915_FALLBACK_PROGRAM, fp->error);
+}
 
 void
 i915ValidateFragmentProgram(struct i915_context *i915)
@@ -1072,16 +1265,6 @@ i915ValidateFragmentProgram(struct i915_context *i915)
    GLuint s2 = S2_TEXCOORD_NONE;
    int i, offset = 0;
 
-   if (i915->current_program != p) {
-      if (i915->current_program) {
-         i915->current_program->on_hardware = 0;
-         i915->current_program->params_uptodate = 0;
-      }
-
-      i915->current_program = p;
-   }
-
-
    /* Important:
     */
    VB->AttribPtr[VERT_ATTRIB_POS] = VB->NdcPtr;
@@ -1125,6 +1308,14 @@ i915ValidateFragmentProgram(struct i915_context *i915)
 
          EMIT_ATTR(_TNL_ATTRIB_TEX0 + i, EMIT_SZ(sz), 0, sz * 4);
       }
+      else if (inputsRead & FRAG_BIT_VAR(i)) {
+         int sz = VB->AttribPtr[_TNL_ATTRIB_GENERIC0 + i]->size;
+
+         s2 &= ~S2_TEXCOORD_FMT(i, S2_TEXCOORD_FMT0_MASK);
+         s2 |= S2_TEXCOORD_FMT(i, SZ_TO_HW(sz));
+
+         EMIT_ATTR(_TNL_ATTRIB_GENERIC0 + i, EMIT_SZ(sz), 0, sz * 4);
+      }
       else if (i == p->wpos_tex) {
 
          /* If WPOS is required, duplicate the XYZ position data in an
diff --git a/i915/i915_program.c b/i915/i915_program.c
index e87700f..e7908bd 100644
--- a/i915/i915_program.c
+++ b/i915/i915_program.c
@@ -130,6 +130,7 @@ i915_emit_decl(struct i915_fragment_program *p,
    *(p->decl++) = (D0_DCL | D0_DEST(reg) | d0_flags);
    *(p->decl++) = D1_MBZ;
    *(p->decl++) = D2_MBZ;
+   assert(p->decl <= p->declarations + ARRAY_SIZE(p->declarations));
 
    p->nr_decl_insn++;
    return reg;
@@ -186,6 +187,11 @@ i915_emit_arith(struct i915_fragment_program * p,
       p->utemp_flag = old_utemp_flag;   /* restore */
    }
 
+   if (p->csr >= p->program + ARRAY_SIZE(p->program)) {
+      i915_program_error(p, "Program contains too many instructions");
+      return UREG_BAD;
+   }
+
    *(p->csr++) = (op | A0_DEST(dest) | mask | saturate | A0_SRC0(src0));
    *(p->csr++) = (A1_SRC0(src0) | A1_SRC1(src1));
    *(p->csr++) = (A2_SRC1(src1) | A2_SRC2(src2));
@@ -270,6 +276,11 @@ GLuint i915_emit_texld( struct i915_fragment_program *p,
 	  p->register_phases[GET_UREG_NR(coord)] == p->nr_tex_indirect)
 	 p->nr_tex_indirect++;
 
+      if (p->csr >= p->program + ARRAY_SIZE(p->program)) {
+	 i915_program_error(p, "Program contains too many instructions");
+	 return UREG_BAD;
+      }
+
       *(p->csr++) = (op | 
 		     T0_DEST( dest ) |
 		     T0_SAMPLER( sampler ));
@@ -424,12 +435,21 @@ i915_emit_param4fv(struct i915_fragment_program * p, const GLfloat * values)
    return 0;
 }
 
-
-
+/* Warning the user about program errors seems to be quite valuable, from
+ * our bug reports.  It unfortunately means piglit reporting errors
+ * when we fall back to software due to an unsupportable program, though.
+ */
 void
-i915_program_error(struct i915_fragment_program *p, const char *msg)
+i915_program_error(struct i915_fragment_program *p, const char *fmt, ...)
 {
-   _mesa_problem(NULL, "i915_program_error: %s", msg);
+   va_list args;
+
+   fprintf(stderr, "i915_program_error: ");
+   va_start(args, fmt);
+   vfprintf(stderr, fmt, args);
+   va_end(args);
+
+   fprintf(stderr, "\n");
    p->error = 1;
 }
 
@@ -511,7 +531,8 @@ i915_upload_program(struct i915_context *i915,
    GLuint program_size = p->csr - p->program;
    GLuint decl_size = p->decl - p->declarations;
 
-   FALLBACK(&i915->intel, I915_FALLBACK_PROGRAM, p->error);
+   if (p->error)
+      return;
 
    /* Could just go straight to the batchbuffer from here:
     */
diff --git a/i915/i915_program.h b/i915/i915_program.h
index 14a3f08..0d17d04 100644
--- a/i915/i915_program.h
+++ b/i915/i915_program.h
@@ -145,7 +145,7 @@ extern GLuint i915_emit_param4fv(struct i915_fragment_program *p,
                                  const GLfloat * values);
 
 extern void i915_program_error(struct i915_fragment_program *p,
-                               const char *msg);
+                               const char *fmt, ...);
 
 extern void i915_init_program(struct i915_context *i915,
                               struct i915_fragment_program *p);
@@ -155,7 +155,6 @@ extern void i915_upload_program(struct i915_context *i915,
 
 extern void i915_fini_program(struct i915_fragment_program *p);
 
-
-
+extern void i915_update_program(GLcontext *ctx);
 
 #endif
diff --git a/i915/i915_reg.h b/i915/i915_reg.h
index b5fa7fd..7f31ff6 100644
--- a/i915/i915_reg.h
+++ b/i915/i915_reg.h
@@ -626,9 +626,9 @@
 #define    MT_32BIT_AWVU2101010	           (0xA<<3)
 #define    MT_32BIT_GR1616	           (0xB<<3)
 #define    MT_32BIT_VU1616	           (0xC<<3)
-#define    MT_32BIT_xI824	           (0xD<<3)
-#define    MT_32BIT_xA824	           (0xE<<3)
-#define    MT_32BIT_xL824	           (0xF<<3)
+#define    MT_32BIT_x8I24	           (0xD<<3)
+#define    MT_32BIT_x8L24	           (0xE<<3)
+#define    MT_32BIT_x8A24	           (0xF<<3)
 #define    MT_422_YCRCB_SWAPY	           (0<<3)       /* SURFACE_422 */
 #define    MT_422_YCRCB_NORMAL	           (1<<3)
 #define    MT_422_YCRCB_SWAPUV	           (2<<3)
diff --git a/i915/i915_state.c b/i915/i915_state.c
index b60efea..cc98d12 100644
--- a/i915/i915_state.c
+++ b/i915/i915_state.c
@@ -585,7 +585,7 @@ i915PointSize(GLcontext * ctx, GLfloat size)
 {
    struct i915_context *i915 = I915_CONTEXT(ctx);
    int lis4 = i915->state.Ctx[I915_CTXREG_LIS4] & ~S4_POINT_WIDTH_MASK;
-   GLint point_size = (int) size;
+   GLint point_size = (int) round(size);
 
    DBG("%s\n", __FUNCTION__);
    
@@ -599,6 +599,24 @@ i915PointSize(GLcontext * ctx, GLfloat size)
 }
 
 
+static void
+i915PointParameterfv(GLcontext * ctx, GLenum pname, const GLfloat *params)
+{
+   struct i915_context *i915 = I915_CONTEXT(ctx);
+
+   switch (pname) {
+   case GL_POINT_SPRITE_COORD_ORIGIN:
+      /* This could be supported, but it would require modifying the fragment
+       * program to invert the y component of the texture coordinate by
+       * inserting a 'SUB tc.y, {1.0}.xxxx, tc' instruction.
+       */
+      FALLBACK(&i915->intel, I915_FALLBACK_POINT_SPRITE_COORD_ORIGIN,
+	       (params[0] != GL_UPPER_LEFT));
+      break;
+   }
+}
+
+
 /* =============================================================
  * Color masks
  */
@@ -939,6 +957,17 @@ i915Enable(GLcontext * ctx, GLenum cap, GLboolean state)
    case GL_POLYGON_SMOOTH:
       break;
 
+   case GL_POINT_SPRITE:
+      /* This state change is handled in i915_reduced_primitive_state because
+       * the hardware bit should only be set when rendering points.
+       */
+      I915_STATECHANGE(i915, I915_UPLOAD_CTX);
+      if (state)
+	 i915->state.Ctx[I915_CTXREG_LIS4] |= S4_SPRITE_POINT_ENABLE;
+      else
+	 i915->state.Ctx[I915_CTXREG_LIS4] &= ~S4_SPRITE_POINT_ENABLE;
+      break;
+
    case GL_POINT_SMOOTH:
       break;
 
@@ -1108,6 +1137,7 @@ i915InitStateFunctions(struct dd_function_table *functions)
    functions->LineWidth = i915LineWidth;
    functions->LogicOpcode = i915LogicOp;
    functions->PointSize = i915PointSize;
+   functions->PointParameterfv = i915PointParameterfv;
    functions->PolygonStipple = i915PolygonStipple;
    functions->Scissor = i915Scissor;
    functions->ShadeModel = i915ShadeModel;
diff --git a/i915/i915_texstate.c b/i915/i915_texstate.c
index 32d4b30..de25848 100644
--- a/i915/i915_texstate.c
+++ b/i915/i915_texstate.c
@@ -27,7 +27,7 @@
 
 #include "main/mtypes.h"
 #include "main/enums.h"
-#include "main/texformat.h"
+#include "main/macros.h"
 
 #include "intel_mipmap_tree.h"
 #include "intel_tex.h"
@@ -37,7 +37,7 @@
 
 
 static GLuint
-translate_texture_format(GLuint mesa_format, GLuint internal_format,
+translate_texture_format(gl_format mesa_format, GLuint internal_format,
 			 GLenum DepthMode)
 {
    switch (mesa_format) {
@@ -56,10 +56,9 @@ translate_texture_format(GLuint mesa_format, GLuint internal_format,
    case MESA_FORMAT_ARGB4444:
       return MAPSURF_16BIT | MT_16BIT_ARGB4444;
    case MESA_FORMAT_ARGB8888:
-      if (internal_format == GL_RGB)
-	 return MAPSURF_32BIT | MT_32BIT_XRGB8888;
-      else
-	 return MAPSURF_32BIT | MT_32BIT_ARGB8888;
+      return MAPSURF_32BIT | MT_32BIT_ARGB8888;
+   case MESA_FORMAT_XRGB8888:
+      return MAPSURF_32BIT | MT_32BIT_XRGB8888;
    case MESA_FORMAT_YCBCR_REV:
       return (MAPSURF_422 | MT_422_YCRCB_NORMAL);
    case MESA_FORMAT_YCBCR:
@@ -82,7 +81,12 @@ translate_texture_format(GLuint mesa_format, GLuint internal_format,
    case MESA_FORMAT_RGBA_DXT5:
       return (MAPSURF_COMPRESSED | MT_COMPRESS_DXT4_5);
    case MESA_FORMAT_S8_Z24:
-      return (MAPSURF_32BIT | MT_32BIT_xI824);
+      if (DepthMode == GL_ALPHA)
+	 return (MAPSURF_32BIT | MT_32BIT_x8A24);
+      else if (DepthMode == GL_INTENSITY)
+	 return (MAPSURF_32BIT | MT_32BIT_x8I24);
+      else
+	 return (MAPSURF_32BIT | MT_32BIT_x8L24);
    default:
       fprintf(stderr, "%s: bad image format %x\n", __FUNCTION__, mesa_format);
       abort();
@@ -134,6 +138,7 @@ i915_update_tex_unit(struct intel_context *intel, GLuint unit, GLuint ss3)
    GLuint *state = i915->state.Tex[unit], format, pitch;
    GLint lodbias, aniso = 0;
    GLubyte border[4];
+   GLfloat maxlod;
 
    memset(state, 0, sizeof(state));
 
@@ -173,11 +178,9 @@ i915_update_tex_unit(struct intel_context *intel, GLuint unit, GLuint ss3)
    } else {
       dri_bo_reference(intelObj->mt->region->buffer);
       i915->state.tex_buffer[unit] = intelObj->mt->region->buffer;
-      i915->state.tex_offset[unit] =  intel_miptree_image_offset(intelObj->mt,
-								 0, intelObj->
-								 firstLevel);
+      i915->state.tex_offset[unit] = 0; /* Always the origin of the miptree */
 
-      format = translate_texture_format(firstImage->TexFormat->MesaFormat, 
+      format = translate_texture_format(firstImage->TexFormat,
 					firstImage->InternalFormat,
 					tObj->DepthMode);
       pitch = intelObj->mt->pitch * intelObj->mt->cpp;
@@ -193,11 +196,15 @@ i915_update_tex_unit(struct intel_context *intel, GLuint unit, GLuint ss3)
 	 state[I915_TEXREG_MS3] |= MS3_TILE_WALK;
    }
 
+   /* We get one field with fraction bits to cover the maximum addressable (smallest
+    * resolution) LOD.  Use it to cover both MAX_LEVEL and MAX_LOD.
+    */
+   maxlod = MIN2(tObj->MaxLod, tObj->MaxLevel - tObj->BaseLevel);
    state[I915_TEXREG_MS4] =
-     ((((pitch / 4) - 1) << MS4_PITCH_SHIFT) | MS4_CUBE_FACE_ENA_MASK |
-       ((((intelObj->lastLevel - intelObj->firstLevel) * 4)) <<
-	MS4_MAX_LOD_SHIFT) | ((firstImage->Depth - 1) <<
-			      MS4_VOLUME_DEPTH_SHIFT));
+      ((((pitch / 4) - 1) << MS4_PITCH_SHIFT) |
+       MS4_CUBE_FACE_ENA_MASK |
+       (U_FIXED(CLAMP(maxlod, 0.0, 11.0), 2) << MS4_MAX_LOD_SHIFT) |
+       ((firstImage->Depth - 1) << MS4_VOLUME_DEPTH_SHIFT));
 
 
    {
@@ -263,8 +270,8 @@ i915_update_tex_unit(struct intel_context *intel, GLuint unit, GLuint ss3)
 
       /* YUV conversion:
        */
-      if (firstImage->TexFormat->MesaFormat == MESA_FORMAT_YCBCR ||
-          firstImage->TexFormat->MesaFormat == MESA_FORMAT_YCBCR_REV)
+      if (firstImage->TexFormat == MESA_FORMAT_YCBCR ||
+          firstImage->TexFormat == MESA_FORMAT_YCBCR_REV)
          state[I915_TEXREG_SS2] |= SS2_COLORSPACE_CONVERSION;
 
       /* Shadow:
@@ -293,6 +300,12 @@ i915_update_tex_unit(struct intel_context *intel, GLuint unit, GLuint ss3)
       GLenum wt = tObj->WrapT;
       GLenum wr = tObj->WrapR;
 
+      /* We program 1D textures as 2D textures, so the 2D texcoord could
+       * result in sampling border values if we don't set the T wrap to
+       * repeat.
+       */
+      if (tObj->Target == GL_TEXTURE_1D)
+	 wt = GL_REPEAT;
 
       /* 3D textures don't seem to respect the border color.
        * Fallback if there's ever a danger that they might refer to
@@ -327,6 +340,9 @@ i915_update_tex_unit(struct intel_context *intel, GLuint unit, GLuint ss3)
           (translate_wrap_mode(wr) << SS3_TCZ_ADDR_MODE_SHIFT));
 
       state[I915_TEXREG_SS3] |= (unit << SS3_TEXTUREMAP_INDEX_SHIFT);
+      state[I915_TEXREG_SS3] |= (U_FIXED(CLAMP(tObj->MinLod, 0.0, 11.0), 4) <<
+				 SS3_MIN_LOD_SHIFT);
+
    }
 
    /* convert border color from float to ubyte */
diff --git a/i915/i915_vtbl.c b/i915/i915_vtbl.c
index 9a723d3..ff97e5a 100644
--- a/i915/i915_vtbl.c
+++ b/i915/i915_vtbl.c
@@ -32,7 +32,6 @@
 #include "main/imports.h"
 #include "main/macros.h"
 #include "main/colormac.h"
-#include "main/texformat.h"
 
 #include "tnl/t_context.h"
 #include "tnl/t_vertex.h"
@@ -54,8 +53,7 @@ i915_render_prevalidate(struct intel_context *intel)
 {
    struct i915_context *i915 = i915_context(&intel->ctx);
 
-   if (!intel->Fallback)
-       i915ValidateFragmentProgram(i915);
+   i915ValidateFragmentProgram(i915);
 }
 
 static void
@@ -589,8 +587,9 @@ i915_state_draw_region(struct intel_context *intel,
             DSTORG_VERT_BIAS(0x8) |     /* .5 */
             LOD_PRECLAMP_OGL | TEX_DEFAULT_COLOR_OGL);
    if (irb != NULL) {
-      switch (irb->texformat->MesaFormat) {
+      switch (irb->Base.Format) {
       case MESA_FORMAT_ARGB8888:
+      case MESA_FORMAT_XRGB8888:
 	 value |= DV_PF_8888;
 	 break;
       case MESA_FORMAT_RGB565:
@@ -604,7 +603,7 @@ i915_state_draw_region(struct intel_context *intel,
 	 break;
       default:
 	 _mesa_problem(ctx, "Bad renderbuffer format: %d\n",
-		       irb->texformat->MesaFormat);
+		       irb->Base.Format);
       }
    }
 
@@ -673,12 +672,6 @@ i915_new_batch(struct intel_context *intel)
    assert(!intel->no_batch_wrap);
 }
 
-static GLuint
-i915_flush_cmd(void)
-{
-   return MI_FLUSH | FLUSH_MAP_CACHE;
-}
-
 static void 
 i915_assert_not_dirty( struct intel_context *intel )
 {
@@ -700,7 +693,6 @@ i915InitVtbl(struct i915_context *i915)
    i915->intel.vtbl.render_prevalidate = i915_render_prevalidate;
    i915->intel.vtbl.set_draw_region = i915_set_draw_region;
    i915->intel.vtbl.update_texture_state = i915UpdateTextureState;
-   i915->intel.vtbl.flush_cmd = i915_flush_cmd;
    i915->intel.vtbl.assert_not_dirty = i915_assert_not_dirty;
    i915->intel.vtbl.finish_batch = intel_finish_vb;
 }
diff --git a/i915/intel_tris.c b/i915/intel_tris.c
index a905455..bc527aa 100644
--- a/i915/intel_tris.c
+++ b/i915/intel_tris.c
@@ -1076,7 +1076,9 @@ intelRunPipeline(GLcontext * ctx)
       intel->NewGLState = 0;
    }
 
+   intel_map_vertex_shader_textures(ctx);
    _tnl_run_pipeline(ctx);
+   intel_unmap_vertex_shader_textures(ctx);
 
    _mesa_unlock_context_textures(ctx);
 }
@@ -1086,6 +1088,7 @@ intelRenderStart(GLcontext * ctx)
 {
    struct intel_context *intel = intel_context(ctx);
 
+   intel_check_front_buffer_rendering(intel);
    intel->vtbl.render_start(intel_context(ctx));
    intel->vtbl.emit_state(intel);
 }
@@ -1194,12 +1197,16 @@ getFallbackString(GLuint bit)
 
 
 
+/**
+ * Enable/disable a fallback flag.
+ * \param bit  one of INTEL_FALLBACK_x flags.
+ */
 void
-intelFallback(struct intel_context *intel, GLuint bit, GLboolean mode)
+intelFallback(struct intel_context *intel, GLbitfield bit, GLboolean mode)
 {
    GLcontext *ctx = &intel->ctx;
    TNLcontext *tnl = TNL_CONTEXT(ctx);
-   GLuint oldfallback = intel->Fallback;
+   const GLbitfield oldfallback = intel->Fallback;
 
    if (mode) {
       intel->Fallback |= bit;