Merge branch '965-glsl'

Conflicts:

	src/mesa/drivers/dri/i965/brw_sf.h
	src/mesa/drivers/dri/i965/intel_context.c

25 files changed, 1863 insertions, 165 deletions

diff --git a/progs/glsl/CH11-toyball.frag.txt b/progs/glsl/CH11-toyball.frag.txt
index 90ec1c27fc1..f3cac62fb3e 100644
--- a/progs/glsl/CH11-toyball.frag.txt
+++ b/progs/glsl/CH11-toyball.frag.txt
@@ -49,14 +49,15 @@ void main()
     inorout += dot(distance, vec4(1.0));
 
     distance.x = dot(p, HalfSpace4);
-    distance.y = StripeWidth - abs(p.z);
+//    distance.y = StripeWidth - abs(p.z);
+    distance.y = StripeWidth - abs(p.y);
     distance = smoothstep(-FWidth, FWidth, distance);
     inorout += distance.x;
 
     inorout = clamp(inorout, 0.0, 1.0);
 
-    surfColor = mix(Yellow, Red, inorout);
-    surfColor = mix(surfColor, Blue, distance.y);
+    surfColor = mix(Yellow, Blue, distance.y);
+    surfColor = mix(surfColor, Red, inorout);
 
     // normal = point on surface for sphere at (0,0,0)
     normal = p;
diff --git a/progs/glsl/CH11-toyball.vert.txt b/progs/glsl/CH11-toyball.vert.txt
index b7da3ac839e..a3ee1b03776 100644
--- a/progs/glsl/CH11-toyball.vert.txt
+++ b/progs/glsl/CH11-toyball.vert.txt
@@ -14,10 +14,11 @@ uniform vec4 BallCenter;   // ball center in modelling coordinates
 
 void main()
 { 
-//orig:    ECposition   = gl_ModelViewMatrix * gl_Vertex;
+    ECposition   = gl_ModelViewMatrix * gl_Vertex;
 
-    ECposition = gl_TextureMatrix[0] * gl_Vertex;
-    ECposition = gl_ModelViewMatrix * ECposition;
+//  ECposition = gl_TextureMatrix[0] * gl_Vertex;
+//  ECposition = gl_MultiTexCoord0 * gl_Vertex;
+//    ECposition = gl_ModelViewMatrix * ECposition;
 
     ECballCenter = gl_ModelViewMatrix * BallCenter;
     gl_Position  = ftransform();
diff --git a/src/mesa/drivers/dri/i965/Makefile b/src/mesa/drivers/dri/i965/Makefile
index 7e07bc9c1a5..4028f1903ab 100644
--- a/src/mesa/drivers/dri/i965/Makefile
+++ b/src/mesa/drivers/dri/i965/Makefile
@@ -69,6 +69,7 @@ DRIVER_SOURCES = \
 	brw_wm_emit.c \
 	brw_wm_fp.c \
 	brw_wm_iz.c \
+	brw_wm_glsl.c \
 	brw_wm_pass0.c \
 	brw_wm_pass1.c \
 	brw_wm_pass2.c \
diff --git a/src/mesa/drivers/dri/i965/brw_clip.h b/src/mesa/drivers/dri/i965/brw_clip.h
index 49b2770a514..2a65697325b 100644
--- a/src/mesa/drivers/dri/i965/brw_clip.h
+++ b/src/mesa/drivers/dri/i965/brw_clip.h
@@ -42,7 +42,7 @@
  * up polygon offset and flatshading at this point:
  */
 struct brw_clip_prog_key {
-   GLuint attrs:16;		
+   GLuint attrs:32;		
    GLuint primitive:4;
    GLuint nr_userclip:3;
    GLuint do_flat_shading:1;
@@ -51,7 +51,7 @@ struct brw_clip_prog_key {
    GLuint fill_ccw:2;		/* includes cull information */
    GLuint offset_cw:1;
    GLuint offset_ccw:1;
-   GLuint pad0:1;
+   GLuint pad0:17;
 
    GLuint copy_bfc_cw:1;
    GLuint copy_bfc_ccw:1;
diff --git a/src/mesa/drivers/dri/i965/brw_context.c b/src/mesa/drivers/dri/i965/brw_context.c
index 0ccdd8a6610..6231cba3f93 100644
--- a/src/mesa/drivers/dri/i965/brw_context.c
+++ b/src/mesa/drivers/dri/i965/brw_context.c
@@ -43,6 +43,8 @@
 #include "api_noop.h"
 #include "vtxfmt.h"
 
+#include "shader/shader_api.h"
+
 /***************************************
  * Mesa's Driver Functions
  ***************************************/
@@ -59,12 +61,37 @@ static const struct dri_extension brw_extensions[] =
     { NULL,                                NULL }
 };
 
+static void brwUseProgram(GLcontext *ctx, GLuint program)
+{
+        struct brw_context *brw = brw_context(ctx);
+        struct gl_shader_program *sh_prog;
+        _mesa_use_program(ctx, program);
+        sh_prog = ctx->Shader.CurrentProgram;
+        if (sh_prog) {
+	    if (sh_prog->VertexProgram) {
+		brw->attribs.VertexProgram->Current = sh_prog->VertexProgram;
+		ctx->VertexProgram.Enabled = GL_TRUE;
+	    }else
+		ctx->VertexProgram.Enabled = GL_FALSE;
+		
+	    if (sh_prog->FragmentProgram) {
+		brw->attribs.FragmentProgram->Current = sh_prog->FragmentProgram;
+		ctx->FragmentProgram.Enabled = GL_TRUE;
+	    } else
+		ctx->FragmentProgram.Enabled = GL_FALSE;
+	}
+}
 
+static void brwInitProgFuncs( struct dd_function_table *functions )
+{
+   functions->UseProgram = brwUseProgram;
+}
 static void brwInitDriverFunctions( struct dd_function_table *functions )
 {
    intelInitDriverFunctions( functions );
    brwInitTextureFuncs( functions );
    brwInitFragProgFuncs( functions );
+   brwInitProgFuncs( functions );
 }
 
 
diff --git a/src/mesa/drivers/dri/i965/brw_eu.h b/src/mesa/drivers/dri/i965/brw_eu.h
index 52f89d577ca..5c98767c3e5 100644
--- a/src/mesa/drivers/dri/i965/brw_eu.h
+++ b/src/mesa/drivers/dri/i965/brw_eu.h
@@ -648,6 +648,11 @@ static __inline struct brw_reg deref_1uw(struct brw_indirect ptr, GLint offset)
    return retype(deref_1f(ptr, offset), BRW_REGISTER_TYPE_UW);
 }
 
+static __inline struct brw_reg deref_1ud(struct brw_indirect ptr, GLint offset)
+{
+   return retype(deref_1f(ptr, offset), BRW_REGISTER_TYPE_UD);
+}
+
 static __inline struct brw_reg get_addr_reg(struct brw_indirect ptr)
 {
    return brw_address_reg(ptr.addr_subnr);
@@ -668,7 +673,10 @@ static __inline struct brw_indirect brw_indirect( GLuint addr_subnr, GLint offse
    return ptr;
 }
 
-
+static __inline struct brw_instruction *current_insn( struct brw_compile *p)
+{
+	return &p->store[p->nr_insn];
+}
 
 void brw_pop_insn_state( struct brw_compile *p );
 void brw_push_insn_state( struct brw_compile *p );
@@ -808,9 +816,11 @@ void brw_ENDIF(struct brw_compile *p,
 struct brw_instruction *brw_DO(struct brw_compile *p,
 			       GLuint execute_size);
 
-void brw_WHILE(struct brw_compile *p, 
+struct brw_instruction *brw_WHILE(struct brw_compile *p, 
 	       struct brw_instruction *patch_insn);
 
+struct brw_instruction *brw_BREAK(struct brw_compile *p);
+struct brw_instruction *brw_CONT(struct brw_compile *p);
 /* Forward jumps:
  */
 void brw_land_fwd_jump(struct brw_compile *p, 
@@ -860,5 +870,6 @@ void brw_math_invert( struct brw_compile *p,
 		      struct brw_reg dst,
 		      struct brw_reg src);
 
-
+void brw_set_src1( struct brw_instruction *insn,
+                          struct brw_reg reg );
 #endif
diff --git a/src/mesa/drivers/dri/i965/brw_eu_emit.c b/src/mesa/drivers/dri/i965/brw_eu_emit.c
index 9992b47d8ae..95f9f027530 100644
--- a/src/mesa/drivers/dri/i965/brw_eu_emit.c
+++ b/src/mesa/drivers/dri/i965/brw_eu_emit.c
@@ -164,7 +164,7 @@ static void brw_set_src0( struct brw_instruction *insn,
 }
 
 
-static void brw_set_src1( struct brw_instruction *insn,
+void brw_set_src1( struct brw_instruction *insn,
 			  struct brw_reg reg )
 {
    assert(reg.file != BRW_MESSAGE_REGISTER_FILE);
@@ -186,7 +186,7 @@ static void brw_set_src1( struct brw_instruction *insn,
        * in the future:
        */
       assert (reg.address_mode == BRW_ADDRESS_DIRECT);
-      assert (reg.file == BRW_GENERAL_REGISTER_FILE);
+      //assert (reg.file == BRW_GENERAL_REGISTER_FILE);
 
       if (insn->header.access_mode == BRW_ALIGN_1) {
 	 insn->bits3.da1.src1_subreg_nr = reg.subnr;
@@ -597,6 +597,34 @@ void brw_ENDIF(struct brw_compile *p,
    }
 }
 
+struct brw_instruction *brw_BREAK(struct brw_compile *p)
+{
+   struct brw_instruction *insn;
+   insn = next_insn(p, BRW_OPCODE_BREAK);
+   brw_set_dest(insn, brw_ip_reg());
+   brw_set_src0(insn, brw_ip_reg());
+   brw_set_src1(insn, brw_imm_d(0x0));
+   insn->header.compression_control = BRW_COMPRESSION_NONE;
+   insn->header.execution_size = BRW_EXECUTE_8;
+   insn->header.mask_control = BRW_MASK_DISABLE;
+   insn->bits3.if_else.pad0 = 0;
+   return insn;
+}
+
+struct brw_instruction *brw_CONT(struct brw_compile *p)
+{
+   struct brw_instruction *insn;
+   insn = next_insn(p, BRW_OPCODE_CONTINUE);
+   brw_set_dest(insn, brw_ip_reg());
+   brw_set_src0(insn, brw_ip_reg());
+   brw_set_src1(insn, brw_imm_d(0x0));
+   insn->header.compression_control = BRW_COMPRESSION_NONE;
+   insn->header.execution_size = BRW_EXECUTE_8;
+   insn->header.mask_control = BRW_MASK_DISABLE;
+   insn->bits3.if_else.pad0 = 0;
+   return insn;
+}
+
 /* DO/WHILE loop:
  */
 struct brw_instruction *brw_DO(struct brw_compile *p, GLuint execute_size)
@@ -608,13 +636,15 @@ struct brw_instruction *brw_DO(struct brw_compile *p, GLuint execute_size)
 
       /* Override the defaults for this instruction:
        */
-      brw_set_dest(insn, retype(brw_vec1_grf(0,0), BRW_REGISTER_TYPE_UD));
-      brw_set_src0(insn, retype(brw_vec1_grf(0,0), BRW_REGISTER_TYPE_UD));
-      brw_set_src1(insn, retype(brw_vec1_grf(0,0), BRW_REGISTER_TYPE_UD));
+      brw_set_dest(insn, brw_null_reg());
+      brw_set_src0(insn, brw_null_reg());
+      brw_set_src1(insn, brw_null_reg());
 
       insn->header.compression_control = BRW_COMPRESSION_NONE;
       insn->header.execution_size = execute_size;
+      insn->header.predicate_control = BRW_PREDICATE_NONE;
       /* insn->header.mask_control = BRW_MASK_ENABLE; */
+      insn->header.mask_control = BRW_MASK_DISABLE;
 
       return insn;
    }
@@ -622,7 +652,7 @@ struct brw_instruction *brw_DO(struct brw_compile *p, GLuint execute_size)
 
 
 
-void brw_WHILE(struct brw_compile *p, 
+struct brw_instruction *brw_WHILE(struct brw_compile *p, 
 	       struct brw_instruction *do_insn)
 {
    struct brw_instruction *insn;
@@ -653,7 +683,9 @@ void brw_WHILE(struct brw_compile *p,
 
 /*    insn->header.mask_control = BRW_MASK_ENABLE; */
 
+   insn->header.mask_control = BRW_MASK_DISABLE;
    p->current->header.predicate_control = BRW_PREDICATE_NONE;   
+   return insn;
 }
 
 
diff --git a/src/mesa/drivers/dri/i965/brw_gs.h b/src/mesa/drivers/dri/i965/brw_gs.h
index 29a4e80ce1b..18a4537c323 100644
--- a/src/mesa/drivers/dri/i965/brw_gs.h
+++ b/src/mesa/drivers/dri/i965/brw_gs.h
@@ -40,11 +40,11 @@
 #define MAX_GS_VERTS (4)	     
 
 struct brw_gs_prog_key {
+   GLuint attrs:32;
    GLuint primitive:4;
-   GLuint attrs:16;		
    GLuint hint_gs_always:1;
    GLuint need_gs_prog:1;
-   GLuint pad:10;
+   GLuint pad:26;
 };
 
 struct brw_gs_compile {
diff --git a/src/mesa/drivers/dri/i965/brw_sf.h b/src/mesa/drivers/dri/i965/brw_sf.h
index e8946511dd4..385f1eea136 100644
--- a/src/mesa/drivers/dri/i965/brw_sf.h
+++ b/src/mesa/drivers/dri/i965/brw_sf.h
@@ -45,10 +45,10 @@
 #define SF_UNFILLED_TRIS   3
 
 struct brw_sf_prog_key {
+   GLuint attrs:32;
    GLuint primitive:2;
    GLuint do_twoside_color:1;
    GLuint do_flat_shading:1;
-   GLuint attrs:16;
    GLuint frontface_ccw:1;
    GLuint do_point_sprite:1;
    GLuint pad:10;
diff --git a/src/mesa/drivers/dri/i965/brw_vs.h b/src/mesa/drivers/dri/i965/brw_vs.h
index 8843f816f9c..41a33ffe387 100644
--- a/src/mesa/drivers/dri/i965/brw_vs.h
+++ b/src/mesa/drivers/dri/i965/brw_vs.h
@@ -67,6 +67,12 @@ struct brw_vs_compile {
    struct brw_reg r1;
    struct brw_reg regs[PROGRAM_ADDRESS+1][128];
    struct brw_reg tmp;
+   struct brw_reg stack;
+
+   struct {	
+       GLboolean used_in_src;
+       struct brw_reg reg;
+   } output_regs[128];
 
    struct brw_reg userplane[6];
 
diff --git a/src/mesa/drivers/dri/i965/brw_vs_emit.c b/src/mesa/drivers/dri/i965/brw_vs_emit.c
index 8733b470c25..e62de186ae3 100644
--- a/src/mesa/drivers/dri/i965/brw_vs_emit.c
+++ b/src/mesa/drivers/dri/i965/brw_vs_emit.c
@@ -134,6 +134,16 @@ static void brw_vs_alloc_regs( struct brw_vs_compile *c )
 					     WRITEMASK_X);
       reg++;
    }
+
+   for (i = 0; i < 128; i++) {
+       if (c->output_regs[i].used_in_src) {
+            c->output_regs[i].reg = brw_vec8_grf(reg, 0);
+            reg++;
+        }
+   }
+
+   c->stack =  brw_uw16_reg(BRW_GENERAL_REGISTER_FILE, reg, 0);
+   reg += 2;
  
    
    /* Some opcodes need an internal temporary:
@@ -213,57 +223,68 @@ static void unalias2( struct brw_vs_compile *c,
    }
 }
 
+static void emit_sop( struct brw_compile *p,
+                      struct brw_reg dst,
+                      struct brw_reg arg0,
+                      struct brw_reg arg1, 
+		      GLuint cond)
+{
+   brw_push_insn_state(p);
+   brw_CMP(p, brw_null_reg(), cond, arg0, arg1);
+   brw_set_predicate_control(p, BRW_PREDICATE_NONE);
+   brw_MOV(p, dst, brw_imm_f(1.0f));
+   brw_set_predicate_control(p, BRW_PREDICATE_NORMAL);
+   brw_MOV(p, dst, brw_imm_f(0.0f));
+   brw_pop_insn_state(p);
+}
 
+static void emit_seq( struct brw_compile *p,
+                      struct brw_reg dst,
+                      struct brw_reg arg0,
+                      struct brw_reg arg1 )
+{
+   emit_sop(p, dst, arg0, arg1, BRW_CONDITIONAL_EQ);
+}
 
-
+static void emit_sne( struct brw_compile *p,
+                      struct brw_reg dst,
+                      struct brw_reg arg0,
+                      struct brw_reg arg1 )
+{
+   emit_sop(p, dst, arg0, arg1, BRW_CONDITIONAL_NEQ);
+}
 static void emit_slt( struct brw_compile *p, 
 		      struct brw_reg dst,
 		      struct brw_reg arg0,
 		      struct brw_reg arg1 )
 {
-   /* Could be done with an if/else/endif, but this method uses half
-    * the instructions.  Note that we are careful to reference the
-    * arguments before writing the dest.  That means we emit the
-    * instructions in an odd order and have to play with the flag
-    * values.
-    */
-   brw_push_insn_state(p);
-   brw_CMP(p, brw_null_reg(), BRW_CONDITIONAL_GE, arg0, arg1);
-
-   /* Write all values to 1:
-    */
-   brw_set_predicate_control(p, BRW_PREDICATE_NONE);
-   brw_MOV(p, dst, brw_imm_f(1.0));
+   emit_sop(p, dst, arg0, arg1, BRW_CONDITIONAL_L);
+}
 
-   /* Where the test succeeded, overwite with zero:
-    */
-   brw_set_predicate_control(p, BRW_PREDICATE_NORMAL);
-   brw_MOV(p, dst, brw_imm_f(0.0));
-   brw_pop_insn_state(p);
+static void emit_sle( struct brw_compile *p, 
+		      struct brw_reg dst,
+		      struct brw_reg arg0,
+		      struct brw_reg arg1 )
+{
+   emit_sop(p, dst, arg0, arg1, BRW_CONDITIONAL_LE);
 }
 
+static void emit_sgt( struct brw_compile *p, 
+		      struct brw_reg dst,
+		      struct brw_reg arg0,
+		      struct brw_reg arg1 )
+{
+   emit_sop(p, dst, arg0, arg1, BRW_CONDITIONAL_G);
+}
 
 static void emit_sge( struct brw_compile *p, 
 		      struct brw_reg dst,
 		      struct brw_reg arg0,
 		      struct brw_reg arg1 )
 {
-   brw_push_insn_state(p);
-   brw_CMP(p, brw_null_reg(), BRW_CONDITIONAL_GE, arg0, arg1);
-
-   /* Write all values to zero:
-    */
-   brw_set_predicate_control(p, BRW_PREDICATE_NONE);
-   brw_MOV(p, dst, brw_imm_f(0));
-
-   /* Where the test succeeded, overwite with 1:
-    */
-   brw_set_predicate_control(p, BRW_PREDICATE_NORMAL);
-   brw_MOV(p, dst, brw_imm_f(1.0));
-   brw_pop_insn_state(p);
+  emit_sop(p, dst, arg0, arg1, BRW_CONDITIONAL_GE);
 }
 
-
 static void emit_max( struct brw_compile *p, 
 		      struct brw_reg dst,
 		      struct brw_reg arg0,
@@ -592,9 +613,13 @@ static struct brw_reg get_reg( struct brw_vs_compile *c,
    case PROGRAM_TEMPORARY:
    case PROGRAM_INPUT:
    case PROGRAM_OUTPUT:
-   case PROGRAM_STATE_VAR:
       assert(c->regs[file][index].nr != 0);
       return c->regs[file][index];
+   case PROGRAM_STATE_VAR:
+   case PROGRAM_CONSTANT:
+   case PROGRAM_UNIFORM:
+      assert(c->regs[PROGRAM_STATE_VAR][index].nr != 0);
+      return c->regs[PROGRAM_STATE_VAR][index];
    case PROGRAM_ADDRESS:
       assert(index == 0);
       return c->regs[file][index];
@@ -668,28 +693,28 @@ static void emit_arl( struct brw_vs_compile *c,
  * account.
  */
 static struct brw_reg get_arg( struct brw_vs_compile *c,
-			       struct prog_src_register src )
+			       struct prog_src_register *src )
 {
    struct brw_reg reg;
 
-   if (src.File == PROGRAM_UNDEFINED)
+   if (src->File == PROGRAM_UNDEFINED)
       return brw_null_reg();
 
-   if (src.RelAddr) 
-      reg = deref(c, c->regs[PROGRAM_STATE_VAR][0], src.Index);
+   if (src->RelAddr) 
+      reg = deref(c, c->regs[PROGRAM_STATE_VAR][0], src->Index);
    else
-      reg = get_reg(c, src.File, src.Index);
+      reg = get_reg(c, src->File, src->Index);
 
    /* Convert 3-bit swizzle to 2-bit.  
     */
-   reg.dw1.bits.swizzle = BRW_SWIZZLE4(GET_SWZ(src.Swizzle, 0),
-				       GET_SWZ(src.Swizzle, 1),
-				       GET_SWZ(src.Swizzle, 2),
-				       GET_SWZ(src.Swizzle, 3));
+   reg.dw1.bits.swizzle = BRW_SWIZZLE4(GET_SWZ(src->Swizzle, 0),
+				       GET_SWZ(src->Swizzle, 1),
+				       GET_SWZ(src->Swizzle, 2),
+				       GET_SWZ(src->Swizzle, 3));
 
    /* Note this is ok for non-swizzle instructions: 
     */
-   reg.negate = src.NegateBase ? 1 : 0;   
+   reg.negate = src->NegateBase ? 1 : 0;   
 
    return reg;
 }
@@ -891,17 +916,50 @@ static void emit_vertex_write( struct brw_vs_compile *c)
 
 }
 
-
-
+static void 
+post_vs_emit( struct brw_vs_compile *c, struct brw_instruction *end_inst )
+{
+   GLuint nr_insns = c->vp->program.Base.NumInstructions;
+   GLuint insn, target_insn;
+   struct prog_instruction *inst1, *inst2;
+   struct brw_instruction *brw_inst1, *brw_inst2;
+   int offset;
+   for (insn = 0; insn < nr_insns; insn++) {
+       inst1 = &c->vp->program.Base.Instructions[insn];
+       brw_inst1 = inst1->Data;
+       switch (inst1->Opcode) {
+	   case OPCODE_CAL:
+	   case OPCODE_BRA:
+	       target_insn = inst1->BranchTarget;
+	       inst2 = &c->vp->program.Base.Instructions[target_insn];
+	       brw_inst2 = inst2->Data;
+	       offset = brw_inst2 - brw_inst1;
+	       brw_set_src1(brw_inst1, brw_imm_d(offset*16));
+	       break;
+	   case OPCODE_END:
+	       offset = end_inst - brw_inst1;
+	       brw_set_src1(brw_inst1, brw_imm_d(offset*16));
+	       break;
+	   default:
+	       break;
+       }
+   }
+}
 
 /* Emit the fragment program instructions here.
  */
-void brw_vs_emit( struct brw_vs_compile *c )
+void brw_vs_emit(struct brw_vs_compile *c )
 {
+#define MAX_IFSN 32
    struct brw_compile *p = &c->func;
    GLuint nr_insns = c->vp->program.Base.NumInstructions;
-   GLuint insn;
+   GLuint insn, if_insn = 0;
+   struct brw_instruction *end_inst;
+   struct brw_instruction *if_inst[MAX_IFSN];
+   struct brw_indirect stack_index = brw_indirect(0, 0);   
 
+   GLuint index;
+   GLuint file;
 
    if (INTEL_DEBUG & DEBUG_VS) {
       _mesa_printf("\n\n\nvs-emit:\n");
@@ -912,9 +970,24 @@ void brw_vs_emit( struct brw_vs_compile *c )
    brw_set_compression_control(p, BRW_COMPRESSION_NONE);
    brw_set_access_mode(p, BRW_ALIGN_16);
    
+   /* Message registers can't be read, so copy the output into GRF register
+      if they are used in source registers */
+   for (insn = 0; insn < nr_insns; insn++) {
+       GLuint i;
+       struct prog_instruction *inst = &c->vp->program.Base.Instructions[insn];
+       for (i = 0; i < 3; i++) {
+	   struct prog_src_register *src = &inst->SrcReg[i];
+	   GLuint index = src->Index;
+	   GLuint file = src->File;	
+	   if (file == PROGRAM_OUTPUT && index != VERT_RESULT_HPOS)
+	       c->output_regs[index].used_in_src = GL_TRUE;
+       }
+   }
+
    /* Static register allocation
     */
    brw_vs_alloc_regs(c);
+   brw_MOV(p, get_addr_reg(stack_index), brw_address(c->stack));
 
    for (insn = 0; insn < nr_insns; insn++) {
 
@@ -924,17 +997,29 @@ void brw_vs_emit( struct brw_vs_compile *c )
       
       /* Get argument regs.  SWZ is special and does this itself.
        */
+      inst->Data = &p->store[p->nr_insn];
       if (inst->Opcode != OPCODE_SWZ)
-	 for (i = 0; i < 3; i++) 
-	    args[i] = get_arg(c, inst->SrcReg[i]);
+	  for (i = 0; i < 3; i++) {
+	      struct prog_src_register *src = &inst->SrcReg[i];
+	      index = src->Index;
+	      file = src->File;	
+	      if (file == PROGRAM_OUTPUT&&c->output_regs[index].used_in_src)
+		  args[i] = c->output_regs[index].reg;
+	      else
+		  args[i] = get_arg(c, src);
+	  }
 
       /* Get dest regs.  Note that it is possible for a reg to be both
        * dst and arg, given the static allocation of registers.  So
        * care needs to be taken emitting multi-operation instructions.
-       */
-      dst = get_dst(c, inst->DstReg);
+       */ 
+      index = inst->DstReg.Index;
+      file = inst->DstReg.File;
+      if (file == PROGRAM_OUTPUT && c->output_regs[index].used_in_src)
+	  dst = c->output_regs[index].reg;
+      else
+	  dst = get_dst(c, inst->DstReg);
 
-      
       switch (inst->Opcode) {
       case OPCODE_ABS:
 	 brw_MOV(p, dst, brw_abs(args[0]));
@@ -1003,12 +1088,25 @@ void brw_vs_emit( struct brw_vs_compile *c )
       case OPCODE_RSQ:
 	 emit_math1(c, BRW_MATH_FUNCTION_RSQ, dst, args[0], BRW_MATH_PRECISION_FULL);
 	 break;
+
+      case OPCODE_SEQ:
+         emit_seq(p, dst, args[0], args[1]);
+         break;
+      case OPCODE_SNE:
+         emit_sne(p, dst, args[0], args[1]);
+         break;
       case OPCODE_SGE:
 	 emit_sge(p, dst, args[0], args[1]);
 	 break;
+      case OPCODE_SGT:
+         emit_sgt(p, dst, args[0], args[1]);
+        break;
       case OPCODE_SLT:
 	 emit_slt(p, dst, args[0], args[1]);
 	 break;
+      case OPCODE_SLE:
+         emit_sle(p, dst, args[0], args[1]);
+         break;
       case OPCODE_SUB:
 	 brw_ADD(p, dst, args[0], negate(args[1]));
 	 break;
@@ -1021,21 +1119,60 @@ void brw_vs_emit( struct brw_vs_compile *c )
       case OPCODE_XPD:
 	 emit_xpd(p, dst, args[0], args[1]);
 	 break;
+      case OPCODE_IF:
+	 assert(if_insn < MAX_IFSN);
+         if_inst[if_insn++] = brw_IF(p, BRW_EXECUTE_8);
+	 break;
+      case OPCODE_ELSE:
+	 if_inst[if_insn-1] = brw_ELSE(p, if_inst[if_insn-1]);
+	 break;
+      case OPCODE_ENDIF:
+         assert(if_insn > 0);
+	 brw_ENDIF(p, if_inst[--if_insn]);
+	 break;			
+      case OPCODE_BRA:
+         brw_set_predicate_control(p, BRW_PREDICATE_NORMAL);
+         brw_ADD(p, brw_ip_reg(), brw_ip_reg(), brw_imm_d(1*16));
+         brw_set_predicate_control_flag_value(p, 0xff);
+        break;
+      case OPCODE_CAL:
+	 brw_set_access_mode(p, BRW_ALIGN_1);
+	 brw_ADD(p, deref_1uw(stack_index, 0), brw_ip_reg(), brw_imm_d(3*16));
+	 brw_set_access_mode(p, BRW_ALIGN_16);
+	 brw_ADD(p, get_addr_reg(stack_index),
+			 get_addr_reg(stack_index), brw_imm_d(4));
+	 inst->Data = &p->store[p->nr_insn];
+	 brw_ADD(p, brw_ip_reg(), brw_ip_reg(), brw_imm_d(1*16));
+        break;
+      case OPCODE_RET:
+	 brw_ADD(p, get_addr_reg(stack_index),
+			 get_addr_reg(stack_index), brw_imm_d(-4));
+	 brw_set_access_mode(p, BRW_ALIGN_1);
+         brw_MOV(p, brw_ip_reg(), deref_1uw(stack_index, 0));
+	 brw_set_access_mode(p, BRW_ALIGN_16);
       case OPCODE_END:	
+         brw_ADD(p, brw_ip_reg(), brw_ip_reg(), brw_imm_d(1*16));
+        break;
       case OPCODE_PRINT:
+      case OPCODE_BGNSUB:
+      case OPCODE_ENDSUB:
 	 break;
       default:
+	 _mesa_printf("Unsupport opcode %d in vertex shader\n", inst->Opcode);
 	 break;
       }
 
+      if (inst->DstReg.File == PROGRAM_OUTPUT
+	      &&inst->DstReg.Index != VERT_RESULT_HPOS
+	      &&c->output_regs[inst->DstReg.Index].used_in_src)
+	  brw_MOV(p, get_dst(c, inst->DstReg), dst);
+
       release_tmps(c);
    }
 
+   end_inst = &p->store[p->nr_insn];
    emit_vertex_write(c);
-
+   post_vs_emit(c, end_inst);
+   for (insn = 0; insn < nr_insns; insn++)
+       c->vp->program.Base.Instructions[insn].Data = NULL;
 }
-
-
-
-
-
diff --git a/src/mesa/drivers/dri/i965/brw_vs_tnl.c b/src/mesa/drivers/dri/i965/brw_vs_tnl.c
index 27210d1a374..aecdd4bc31a 100644
--- a/src/mesa/drivers/dri/i965/brw_vs_tnl.c
+++ b/src/mesa/drivers/dri/i965/brw_vs_tnl.c
@@ -855,7 +855,7 @@ static struct ureg calculate_light_attenuation( struct tnl_program *p,
       struct ureg slt = get_temp(p);
 
       emit_op2(p, OPCODE_DP3, spot, 0, ureg_negate(VPpli), spot_dir_norm);
-      emit_op2(p, OPCODE_SLT, slt, 0, swizzle1(spot_dir_norm,W), spot);
+      emit_op2(p, OPCODE_SLT, slt, 0, spot, swizzle1(spot_dir_norm,W));
       emit_op2(p, OPCODE_POW, spot, 0, spot, swizzle1(attenuation, W));
       emit_op2(p, OPCODE_MUL, att, 0, slt, spot);
 
diff --git a/src/mesa/drivers/dri/i965/brw_wm.c b/src/mesa/drivers/dri/i965/brw_wm.c
index 904c00bef87..2d6249e3b56 100644
--- a/src/mesa/drivers/dri/i965/brw_wm.c
+++ b/src/mesa/drivers/dri/i965/brw_wm.c
@@ -66,7 +66,11 @@ GLuint brw_wm_nr_args( GLuint opcode )
    case OPCODE_POW:
    case OPCODE_SUB:
    case OPCODE_SGE:
+   case OPCODE_SGT:
+   case OPCODE_SLE:
    case OPCODE_SLT:
+   case OPCODE_SEQ:
+   case OPCODE_SNE:
    case OPCODE_ADD:
    case OPCODE_MAX:
    case OPCODE_MIN:
@@ -150,47 +154,49 @@ static void do_wm_prog( struct brw_context *brw,
    c->fp = fp;
    c->env_param = brw->intel.ctx.FragmentProgram.Parameters;
 
-
-   /* Augment fragment program.  Add instructions for pre- and
-    * post-fragment-program tasks such as interpolation and fogging.
-    */
-   brw_wm_pass_fp(c);
-   
-   /* Translate to intermediate representation.  Build register usage
-    * chains.
-    */
-   brw_wm_pass0(c);
-
-   /* Dead code removal.
-    */
-   brw_wm_pass1(c);
-
-   /* Hal optimization
-    */
-   brw_wm_pass_hal (c);
-   
-   /* Register allocation.
-    */
-   c->grf_limit = BRW_WM_MAX_GRF/2;
-
-   /* This is where we start emitting gen4 code:
-    */
-   brw_init_compile(&c->func);    
-
-   brw_wm_pass2(c);
-
-   c->prog_data.total_grf = c->max_wm_grf;
-   if (c->last_scratch) {
-      c->prog_data.total_scratch =
-	 c->last_scratch + 0x40;
+   if (brw_wm_is_glsl(&c->fp->program)) {
+       brw_wm_glsl_emit(c);
    } else {
-      c->prog_data.total_scratch = 0;
+       /* Augment fragment program.  Add instructions for pre- and
+	* post-fragment-program tasks such as interpolation and fogging.
+	*/
+       brw_wm_pass_fp(c);
+
+       /* Translate to intermediate representation.  Build register usage
+	* chains.
+	*/
+       brw_wm_pass0(c);
+
+       /* Dead code removal.
+	*/
+       brw_wm_pass1(c);
+
+       /* Hal optimization
+	*/
+       brw_wm_pass_hal (c);
+
+       /* Register allocation.
+	*/
+       c->grf_limit = BRW_WM_MAX_GRF/2;
+
+       /* This is where we start emitting gen4 code:
+	*/
+       brw_init_compile(&c->func);    
+
+       brw_wm_pass2(c);
+
+       c->prog_data.total_grf = c->max_wm_grf;
+       if (c->last_scratch) {
+	   c->prog_data.total_scratch =
+	       c->last_scratch + 0x40;
+       } else {
+	   c->prog_data.total_scratch = 0;
+       }
+
+       /* Emit GEN4 code.
+	*/
+       brw_wm_emit(c);
    }
-
-   /* Emit GEN4 code.
-    */
-   brw_wm_emit(c);
-
    /* get the program
     */
    program = brw_get_program(&c->func, &program_size);
diff --git a/src/mesa/drivers/dri/i965/brw_wm.h b/src/mesa/drivers/dri/i965/brw_wm.h
index 6dcf4732a9f..440b5357d5a 100644
--- a/src/mesa/drivers/dri/i965/brw_wm.h
+++ b/src/mesa/drivers/dri/i965/brw_wm.h
@@ -194,6 +194,7 @@ struct brw_wm_compile {
    GLuint nr_fp_insns;
    GLuint fp_temp;
    GLuint fp_interp_emitted;
+   GLuint fp_deriv_emitted;
 
    struct prog_src_register pixel_xy;
    struct prog_src_register delta_xy;
@@ -231,6 +232,15 @@ struct brw_wm_compile {
    GLuint grf_limit;
    GLuint max_wm_grf;
    GLuint last_scratch;
+
+   struct {
+	GLboolean inited;
+	struct brw_reg reg;
+   } wm_regs[PROGRAM_PAYLOAD+1][256][4];
+   struct brw_reg stack;
+   struct brw_reg emit_mask_reg;
+   GLuint reg_index;
+   GLuint tmp_index;
 };
 
 
@@ -259,4 +269,6 @@ void brw_wm_lookup_iz( GLuint line_aa,
 		       GLuint lookup,
 		       struct brw_wm_prog_key *key );
 
+GLboolean brw_wm_is_glsl(struct gl_fragment_program *fp);
+void brw_wm_glsl_emit(struct brw_wm_compile *c);
 #endif
diff --git a/src/mesa/drivers/dri/i965/brw_wm_emit.c b/src/mesa/drivers/dri/i965/brw_wm_emit.c
index fd605159727..1f7158f7a78 100644
--- a/src/mesa/drivers/dri/i965/brw_wm_emit.c
+++ b/src/mesa/drivers/dri/i965/brw_wm_emit.c
@@ -229,20 +229,20 @@ static void emit_cinterp( struct brw_compile *p,
 			 GLuint mask,
 			 const struct brw_reg *arg0 )
 {
-   struct brw_reg interp[4];
-   GLuint nr = arg0[0].nr;
-   GLuint i;
-
-   interp[0] = brw_vec1_grf(nr, 0);
-   interp[1] = brw_vec1_grf(nr, 4);
-   interp[2] = brw_vec1_grf(nr+1, 0);
-   interp[3] = brw_vec1_grf(nr+1, 4);
-
-   for(i = 0; i < 4; i++ ) {
-      if (mask & (1<<i)) {
-	 brw_MOV(p, dst[i], suboffset(interp[i],3));	/* TODO: optimize away like other moves */
-      }
-   }
+	struct brw_reg interp[4];
+	GLuint nr = arg0[0].nr;
+	GLuint i;
+
+	interp[0] = brw_vec1_grf(nr, 0);
+	interp[1] = brw_vec1_grf(nr, 4);
+	interp[2] = brw_vec1_grf(nr+1, 0);
+	interp[3] = brw_vec1_grf(nr+1, 4);
+
+	for(i = 0; i < 4; i++ ) {
+		if (mask & (1<<i)) {
+			brw_MOV(p, dst[i], suboffset(interp[i],3));	/* TODO: optimize away like other moves */
+		}
+	}
 }
 
 
@@ -343,11 +343,10 @@ static void emit_lrp( struct brw_compile *p,
       }
    }
 }
-
-
-static void emit_slt( struct brw_compile *p, 
+static void emit_sop( struct brw_compile *p, 
 		      const struct brw_reg *dst,
 		      GLuint mask,
+		      GLuint cond,
 		      const struct brw_reg *arg0,
 		      const struct brw_reg *arg1 )
 {
@@ -356,34 +355,66 @@ static void emit_slt( struct brw_compile *p,
    for (i = 0; i < 4; i++) {
       if (mask & (1<<i)) {	
 	 brw_MOV(p, dst[i], brw_imm_f(0));
-	 brw_CMP(p, brw_null_reg(), BRW_CONDITIONAL_L, arg0[i], arg1[i]);
+	 brw_CMP(p, brw_null_reg(), cond, arg0[i], arg1[i]);
 	 brw_MOV(p, dst[i], brw_imm_f(1.0));
 	 brw_set_predicate_control_flag_value(p, 0xff);
       }
    }
 }
 
-/* Isn't this just the same as the above with the args swapped?
- */
-static void emit_sge( struct brw_compile *p, 
+static void emit_slt( struct brw_compile *p, 
 		      const struct brw_reg *dst,
 		      GLuint mask,
 		      const struct brw_reg *arg0,
 		      const struct brw_reg *arg1 )
 {
-   GLuint i;
+	 emit_sop(p, dst, mask, BRW_CONDITIONAL_L, arg0, arg1);
+}
 
-   for (i = 0; i < 4; i++) {
-      if (mask & (1<<i)) {	
-	 brw_MOV(p, dst[i], brw_imm_f(0));
-	 brw_CMP(p, brw_null_reg(), BRW_CONDITIONAL_GE, arg0[i], arg1[i]);
-	 brw_MOV(p, dst[i], brw_imm_f(1.0));
-	 brw_set_predicate_control_flag_value(p, 0xff);
-      }
-   }
+static void emit_sle( struct brw_compile *p, 
+		      const struct brw_reg *dst,
+		      GLuint mask,
+		      const struct brw_reg *arg0,
+		      const struct brw_reg *arg1 )
+{
+	 emit_sop(p, dst, mask, BRW_CONDITIONAL_LE, arg0, arg1);
+}
+
+static void emit_sgt( struct brw_compile *p, 
+		      const struct brw_reg *dst,
+		      GLuint mask,
+		      const struct brw_reg *arg0,
+		      const struct brw_reg *arg1 )
+{
+	 emit_sop(p, dst, mask, BRW_CONDITIONAL_G, arg0, arg1);
+}
+
+static void emit_sge( struct brw_compile *p, 
+		      const struct brw_reg *dst,
+		      GLuint mask,
+		      const struct brw_reg *arg0,
+		      const struct brw_reg *arg1 )
+{
+	 emit_sop(p, dst, mask, BRW_CONDITIONAL_GE, arg0, arg1);
 }
 
+static void emit_seq( struct brw_compile *p, 
+		      const struct brw_reg *dst,
+		      GLuint mask,
+		      const struct brw_reg *arg0,
+		      const struct brw_reg *arg1 )
+{
+	 emit_sop(p, dst, mask, BRW_CONDITIONAL_EQ, arg0, arg1);
+}
 
+static void emit_sne( struct brw_compile *p, 
+		      const struct brw_reg *dst,
+		      GLuint mask,
+		      const struct brw_reg *arg0,
+		      const struct brw_reg *arg1 )
+{
+	 emit_sop(p, dst, mask, BRW_CONDITIONAL_NEQ, arg0, arg1);
+}
 
 static void emit_cmp( struct brw_compile *p, 
 		      const struct brw_reg *dst,
@@ -543,8 +574,8 @@ static void emit_math1( struct brw_compile *p,
 			GLuint mask,
 			const struct brw_reg *arg0 )
 {
-   assert((mask & WRITEMASK_XYZW) == WRITEMASK_X ||
-	  function == BRW_MATH_FUNCTION_SINCOS);
+   //assert((mask & WRITEMASK_XYZW) == WRITEMASK_X ||
+   //	  function == BRW_MATH_FUNCTION_SINCOS);
    
    brw_MOV(p, brw_message_reg(2), arg0[0]);
 
@@ -671,6 +702,8 @@ static void emit_tex( struct brw_wm_compile *c,
 	      msgLength,
 	      0);	
 
+   if (shadow)
+       brw_MOV(p, dst[3], brw_imm_f(1.0));
 }
 
 
@@ -1208,9 +1241,21 @@ void brw_wm_emit( struct brw_wm_compile *c )
 	 emit_slt(p, dst, dst_flags, args[0], args[1]);
 	 break;
 
+      case OPCODE_SLE:
+	 emit_sle(p, dst, dst_flags, args[0], args[1]);
+	break;
+      case OPCODE_SGT:
+	 emit_sgt(p, dst, dst_flags, args[0], args[1]);
+	break;
       case OPCODE_SGE:
 	 emit_sge(p, dst, dst_flags, args[0], args[1]);
 	 break;
+      case OPCODE_SEQ:
+	 emit_seq(p, dst, dst_flags, args[0], args[1]);
+	break;
+      case OPCODE_SNE:
+	 emit_sne(p, dst, dst_flags, args[0], args[1]);
+	break;
 
       case OPCODE_LIT:
 	 emit_lit(p, dst, dst_flags, args[0]);
@@ -1231,7 +1276,8 @@ void brw_wm_emit( struct brw_wm_compile *c )
 	 break;
 
       default:
-	 assert(0);
+	_mesa_printf("unsupport opcode %d in fragment program\n", 
+		inst->opcode);
       }
       
       for (i = 0; i < 4; i++)
diff --git a/src/mesa/drivers/dri/i965/brw_wm_fp.c b/src/mesa/drivers/dri/i965/brw_wm_fp.c
index 67154c1b13a..1b26005169a 100644
--- a/src/mesa/drivers/dri/i965/brw_wm_fp.c
+++ b/src/mesa/drivers/dri/i965/brw_wm_fp.c
@@ -176,6 +176,7 @@ static struct prog_instruction *emit_insn(struct brw_wm_compile *c,
 {
    struct prog_instruction *inst = get_fp_inst(c);
    *inst = *inst0;
+   inst->Data = (void *)inst0;
    return inst;
 }
 
@@ -201,7 +202,6 @@ static struct prog_instruction * emit_op(struct brw_wm_compile *c,
    inst->SrcReg[0] = src0;
    inst->SrcReg[1] = src1;
    inst->SrcReg[2] = src2;
-   
    return inst;
 }
    
@@ -361,6 +361,37 @@ static void emit_interp( struct brw_wm_compile *c,
    c->fp_interp_emitted |= 1<<idx;
 }
 
+static void emit_ddx( struct brw_wm_compile *c,
+        const struct prog_instruction *inst )
+{
+    GLuint idx = inst->SrcReg[0].Index;
+    struct prog_src_register interp = src_reg(PROGRAM_PAYLOAD, idx);
+
+    c->fp_deriv_emitted |= 1<<idx;
+    emit_op(c,
+            OPCODE_DDX,
+            inst->DstReg,
+            0, 0, 0,
+            interp,
+            get_pixel_w(c),
+            src_undef());
+}
+
+static void emit_ddy( struct brw_wm_compile *c,
+        const struct prog_instruction *inst )
+{
+    GLuint idx = inst->SrcReg[0].Index;
+    struct prog_src_register interp = src_reg(PROGRAM_PAYLOAD, idx);
+
+    c->fp_deriv_emitted |= 1<<idx;
+    emit_op(c,
+            OPCODE_DDY,
+            inst->DstReg,
+            0, 0, 0,
+            interp,
+            get_pixel_w(c),
+            src_undef());
+}
 
 /***********************************************************************
  * Hacks to extend the program parameter and constant lists.
@@ -957,8 +988,16 @@ void brw_wm_pass_fp( struct brw_wm_compile *c )
 	  */
 	 out->DstReg.WriteMask = 0;
 	 break;
-
+      case OPCODE_DDX:
+	 emit_ddx(c, inst);
+	 break;
+      case OPCODE_DDY:
+         emit_ddy(c, inst);
+	break;
       case OPCODE_END:
+	 emit_fog(c);
+	 emit_fb_write(c);
+	 break;
       case OPCODE_PRINT:
 	 break;
 	 
@@ -967,15 +1006,11 @@ void brw_wm_pass_fp( struct brw_wm_compile *c )
 	 break;
       }
    }
-   
-   emit_fog(c);
-   emit_fb_write(c);
-
 
    if (INTEL_DEBUG & DEBUG_WM) {
-      _mesa_printf("\n\n\npass_fp:\n");
-      print_insns( c->prog_instructions, c->nr_fp_insns );
-      _mesa_printf("\n");
+	   _mesa_printf("\n\n\npass_fp:\n");
+	   print_insns( c->prog_instructions, c->nr_fp_insns );
+	   _mesa_printf("\n");
    }
 }
 
diff --git a/src/mesa/drivers/dri/i965/brw_wm_glsl.c b/src/mesa/drivers/dri/i965/brw_wm_glsl.c
new file mode 100644
index 00000000000..4b273fefe92
--- /dev/null
+++ b/src/mesa/drivers/dri/i965/brw_wm_glsl.c
@@ -0,0 +1,1354 @@
+#include "macros.h"
+#include "shader/prog_parameter.h"
+#include "brw_context.h"
+#include "brw_eu.h"
+#include "brw_wm.h"
+
+/* Only guess, need a flag in gl_fragment_program later */
+GLboolean brw_wm_is_glsl(struct gl_fragment_program *fp)
+{
+    int i;
+    for (i = 0; i < fp->Base.NumInstructions; i++) {
+	struct prog_instruction *inst = &fp->Base.Instructions[i];
+	switch (inst->Opcode) {
+	    case OPCODE_IF:
+	    case OPCODE_INT:
+	    case OPCODE_ENDIF:
+	    case OPCODE_CAL:
+	    case OPCODE_BRK:
+	    case OPCODE_RET:
+	    case OPCODE_DDX:
+	    case OPCODE_DDY:
+	    case OPCODE_BGNLOOP:
+		return GL_TRUE; 
+	    default:
+		break;
+	}
+    }
+    return GL_FALSE; 
+}
+
+static void set_reg(struct brw_wm_compile *c, int file, int index, 
+	int component, struct brw_reg reg)
+{
+    c->wm_regs[file][index][component].reg = reg;
+    c->wm_regs[file][index][component].inited = GL_TRUE;
+}
+
+static int get_scalar_dst_index(struct prog_instruction *inst)
+{
+    int i;
+    for (i = 0; i < 4; i++)
+	if (inst->DstReg.WriteMask & (1<<i))
+	    break;
+    return i;
+}
+
+static struct brw_reg alloc_tmp(struct brw_wm_compile *c)
+{
+    struct brw_reg reg;
+    reg = brw_vec8_grf(c->tmp_index--, 0);
+    return reg;
+}
+
+static void release_tmps(struct brw_wm_compile *c)
+{
+    c->tmp_index = 127;
+}
+
+static struct brw_reg 
+get_reg(struct brw_wm_compile *c, int file, int index, int component, int nr, GLuint neg, GLuint abs)
+{
+    struct brw_reg reg;
+    switch (file) {
+	case PROGRAM_STATE_VAR:
+	case PROGRAM_CONSTANT:
+	case PROGRAM_UNIFORM:
+	    file = PROGRAM_STATE_VAR;
+	    break;
+	case PROGRAM_UNDEFINED:
+	    return brw_null_reg();	
+	default:
+	    break;
+    }
+
+    if(c->wm_regs[file][index][component].inited)
+	reg = c->wm_regs[file][index][component].reg;
+    else 
+	reg = brw_vec8_grf(c->reg_index, 0);
+
+    if(!c->wm_regs[file][index][component].inited) {
+	set_reg(c, file, index, component, reg);
+	c->reg_index++;
+    }
+
+    if (neg & (1<< component)) {
+	reg = negate(reg);
+    }
+    if (abs)
+	reg = brw_abs(reg);
+    return reg;
+}
+
+static void prealloc_reg(struct brw_wm_compile *c)
+{
+    int i, j;
+    struct brw_reg reg;
+    int nr_interp_regs = 0;
+    GLuint inputs = FRAG_BIT_WPOS | c->fp_interp_emitted | c->fp_deriv_emitted;
+
+    for (i = 0; i < 4; i++) {
+	reg = (i < c->key.nr_depth_regs) 
+	    ? brw_vec8_grf(i*2, 0) : brw_vec8_grf(0, 0);
+	set_reg(c, PROGRAM_PAYLOAD, PAYLOAD_DEPTH, i, reg);
+    }
+    c->reg_index += 2*c->key.nr_depth_regs;
+    {
+	int nr_params = c->fp->program.Base.Parameters->NumParameters;
+	struct gl_program_parameter_list *plist = 
+	    c->fp->program.Base.Parameters;
+	int index = 0;
+	c->prog_data.nr_params = 4*nr_params;
+	for (i = 0; i < nr_params; i++) {
+	    for (j = 0; j < 4; j++, index++) {
+		reg = brw_vec1_grf(c->reg_index + index/8, 
+			index%8);
+		c->prog_data.param[index] = 
+		    &plist->ParameterValues[i][j];
+		set_reg(c, PROGRAM_STATE_VAR, i, j, reg);
+	    }
+	}
+	c->nr_creg = 2*((4*nr_params+15)/16);
+	c->reg_index += c->nr_creg;
+    }
+    for (i = 0; i < FRAG_ATTRIB_MAX; i++) {
+	if (inputs & (1<<i)) {
+	    nr_interp_regs++;
+	    reg = brw_vec8_grf(c->reg_index, 0);
+	    for (j = 0; j < 4; j++)
+		set_reg(c, PROGRAM_PAYLOAD, i, j, reg);
+	    c->reg_index += 2;
+
+	}
+    }
+    c->prog_data.first_curbe_grf = c->key.nr_depth_regs * 2;
+    c->prog_data.urb_read_length = nr_interp_regs * 2;
+    c->prog_data.curb_read_length = c->nr_creg;
+    c->emit_mask_reg = brw_uw1_reg(BRW_GENERAL_REGISTER_FILE, c->reg_index, 0);
+    c->reg_index++;
+    c->stack =  brw_uw16_reg(BRW_GENERAL_REGISTER_FILE, c->reg_index, 0);
+    c->reg_index += 2;
+}
+
+static struct brw_reg get_dst_reg(struct brw_wm_compile *c, 
+	struct prog_instruction *inst, int component, int nr)
+{
+    return get_reg(c, inst->DstReg.File, inst->DstReg.Index, component, nr,
+	    0, 0);
+}
+
+static struct brw_reg get_src_reg(struct brw_wm_compile *c, 
+	struct prog_src_register *src, int index, int nr)
+{
+    int component = GET_SWZ(src->Swizzle, index);
+    return get_reg(c, src->File, src->Index, component, nr, 
+	    src->NegateBase, src->Abs);
+}
+
+static void emit_abs( struct brw_wm_compile *c,
+		struct prog_instruction *inst)
+{
+    int i;
+    struct brw_compile *p = &c->func;
+    brw_set_saturate(p, inst->SaturateMode != SATURATE_OFF);
+    for (i = 0; i < 4; i++) {
+	if (inst->DstReg.WriteMask & (1<<i)) {
+	    struct brw_reg src, dst;
+	    dst = get_dst_reg(c, inst, i, 1);
+	    src = get_src_reg(c, &inst->SrcReg[0], i, 1);
+	    brw_MOV(p, dst, brw_abs(src));
+	}
+    }
+    brw_set_saturate(p, 0);
+}
+
+static void emit_int( struct brw_wm_compile *c,
+		struct prog_instruction *inst)
+{
+    int i;
+    struct brw_compile *p = &c->func;
+    GLuint mask = inst->DstReg.WriteMask;
+    brw_set_saturate(p, inst->SaturateMode != SATURATE_OFF);
+    for (i = 0; i < 4; i++) {
+	if (mask & (1<<i)) {
+	    struct brw_reg src, dst;
+	    dst = get_dst_reg(c, inst, i, 1) ;
+	    src = get_src_reg(c, &inst->SrcReg[0], i, 1);
+	    brw_RNDD(p, dst, src);
+	}
+    }
+    brw_set_saturate(p, 0);
+}
+
+static void emit_mov( struct brw_wm_compile *c,
+		struct prog_instruction *inst)
+{
+    int i;
+    struct brw_compile *p = &c->func;
+    GLuint mask = inst->DstReg.WriteMask;
+    brw_set_saturate(p, inst->SaturateMode != SATURATE_OFF);
+    for (i = 0; i < 4; i++) {
+	if (mask & (1<<i)) {
+	    struct brw_reg src, dst;
+	    dst = get_dst_reg(c, inst, i, 1);
+	    src = get_src_reg(c, &inst->SrcReg[0], i, 1);
+	    brw_MOV(p, dst, src);
+	}
+    }
+    brw_set_saturate(p, 0);
+}
+
+static void emit_pixel_xy(struct brw_wm_compile *c,
+		struct prog_instruction *inst)
+{
+    struct brw_reg r1 = brw_vec1_grf(1, 0);
+    struct brw_reg r1_uw = retype(r1, BRW_REGISTER_TYPE_UW);
+
+    struct brw_reg dst0, dst1;
+    struct brw_compile *p = &c->func;
+    GLuint mask = inst->DstReg.WriteMask;
+
+    dst0 = get_dst_reg(c, inst, 0, 1);
+    dst1 = get_dst_reg(c, inst, 1, 1);
+    /* Calculate pixel centers by adding 1 or 0 to each of the
+     * micro-tile coordinates passed in r1.
+     */
+    if (mask & WRITEMASK_X) {
+	brw_ADD(p,
+		vec8(retype(dst0, BRW_REGISTER_TYPE_UW)),
+		stride(suboffset(r1_uw, 4), 2, 4, 0),
+		brw_imm_v(0x10101010));
+    }
+
+    if (mask & WRITEMASK_Y) {
+	brw_ADD(p,
+		vec8(retype(dst1, BRW_REGISTER_TYPE_UW)),
+		stride(suboffset(r1_uw, 5), 2, 4, 0),
+		brw_imm_v(0x11001100));
+    }
+
+}
+
+static void emit_delta_xy(struct brw_wm_compile *c,
+		struct prog_instruction *inst)
+{
+    struct brw_reg r1 = brw_vec1_grf(1, 0);
+    struct brw_reg dst0, dst1, src0, src1;
+    struct brw_compile *p = &c->func;
+    GLuint mask = inst->DstReg.WriteMask;
+
+    dst0 = get_dst_reg(c, inst, 0, 1);
+    dst1 = get_dst_reg(c, inst, 1, 1);
+    src0 = get_src_reg(c, &inst->SrcReg[0], 0, 1);
+    src1 = get_src_reg(c, &inst->SrcReg[0], 1, 1);
+    /* Calc delta X,Y by subtracting origin in r1 from the pixel
+     * centers.
+     */
+    if (mask & WRITEMASK_X) {
+	brw_ADD(p,
+		dst0,
+		retype(src0, BRW_REGISTER_TYPE_UW),
+		negate(r1));
+    }
+
+    if (mask & WRITEMASK_Y) {
+	brw_ADD(p,
+		dst1,
+		retype(src1, BRW_REGISTER_TYPE_UW),
+		negate(suboffset(r1,1)));
+
+    }
+
+}
+
+
+static void fire_fb_write( struct brw_wm_compile *c,
+                           GLuint base_reg,
+                           GLuint nr )
+{
+    struct brw_compile *p = &c->func;
+
+    /* Pass through control information:
+     */
+    /*  mov (8) m1.0<1>:ud   r1.0<8;8,1>:ud   { Align1 NoMask } */
+    {
+	brw_push_insn_state(p);
+	brw_set_mask_control(p, BRW_MASK_DISABLE); /* ? */
+	brw_MOV(p,
+		brw_message_reg(base_reg + 1),
+		brw_vec8_grf(1, 0));
+	brw_pop_insn_state(p);
+    }
+    /* Send framebuffer write message: */
+    brw_fb_WRITE(p,
+	    retype(vec8(brw_null_reg()), BRW_REGISTER_TYPE_UW),
+	    base_reg,
+	    retype(brw_vec8_grf(0, 0), BRW_REGISTER_TYPE_UW),
+	    0,              /* render surface always 0 */
+	    nr,
+	    0,
+	    1);
+}
+
+static void emit_fb_write(struct brw_wm_compile *c,
+		struct prog_instruction *inst)
+{
+    struct brw_compile *p = &c->func;
+    int nr = 2;
+    int channel;
+    struct brw_reg src0;//, src1, src2, dst;
+
+    /* Reserve a space for AA - may not be needed:
+     */
+    if (c->key.aa_dest_stencil_reg)
+	nr += 1;
+    {
+	brw_push_insn_state(p);
+	for (channel = 0; channel < 4; channel++) {
+	    src0 = get_src_reg(c,  &inst->SrcReg[0], channel, 1);
+	    /*  mov (8) m2.0<1>:ud   r28.0<8;8,1>:ud  { Align1 } */
+	    /*  mov (8) m6.0<1>:ud   r29.0<8;8,1>:ud  { Align1 SecHalf } */
+	    brw_MOV(p, brw_message_reg(nr + channel), src0);
+	}
+	/* skip over the regs populated above: */
+	nr += 8;
+	brw_pop_insn_state(p);
+    }
+    fire_fb_write(c, 0, nr);
+}
+
+static void emit_pixel_w( struct brw_wm_compile *c,
+		struct prog_instruction *inst)
+{
+    struct brw_compile *p = &c->func;
+    GLuint mask = inst->DstReg.WriteMask;
+    if (mask & WRITEMASK_W) {
+	struct brw_reg dst, src0, delta0, delta1;
+	struct brw_reg interp3;
+
+	dst = get_dst_reg(c, inst, 3, 1);
+	src0 = get_src_reg(c, &inst->SrcReg[0], 0, 1);
+	delta0 = get_src_reg(c, &inst->SrcReg[1], 0, 1);
+	delta1 = get_src_reg(c, &inst->SrcReg[1], 1, 1);
+
+	interp3 = brw_vec1_grf(src0.nr+1, 4);
+	/* Calc 1/w - just linterp wpos[3] optimized by putting the
+	 * result straight into a message reg.
+	 */
+	brw_LINE(p, brw_null_reg(), interp3, delta0);
+	brw_MAC(p, brw_message_reg(2), suboffset(interp3, 1), delta1);
+
+	/* Calc w */
+	brw_math_16( p, dst,
+		BRW_MATH_FUNCTION_INV,
+		BRW_MATH_SATURATE_NONE,
+		2, brw_null_reg(),
+		BRW_MATH_PRECISION_FULL);
+    }
+}
+
+static void emit_linterp(struct brw_wm_compile *c,
+		struct prog_instruction *inst)
+{
+    struct brw_compile *p = &c->func;
+    GLuint mask = inst->DstReg.WriteMask;
+    struct brw_reg interp[4];
+    struct brw_reg dst, delta0, delta1;
+    struct brw_reg src0;
+
+    src0 = get_src_reg(c, &inst->SrcReg[0], 0, 1);
+    delta0 = get_src_reg(c, &inst->SrcReg[1], 0, 1);
+    delta1 = get_src_reg(c, &inst->SrcReg[1], 1, 1);
+    GLuint nr = src0.nr;
+    int i;
+
+    interp[0] = brw_vec1_grf(nr, 0);
+    interp[1] = brw_vec1_grf(nr, 4);
+    interp[2] = brw_vec1_grf(nr+1, 0);
+    interp[3] = brw_vec1_grf(nr+1, 4);
+
+    for(i = 0; i < 4; i++ ) {
+	if (mask & (1<<i)) {
+	    dst = get_dst_reg(c, inst, i, 1);
+	    brw_LINE(p, brw_null_reg(), interp[i], delta0);
+	    brw_MAC(p, dst, suboffset(interp[i],1), delta1);
+	}
+    }
+}
+
+static void emit_cinterp(struct brw_wm_compile *c,
+		struct prog_instruction *inst)
+{
+    struct brw_compile *p = &c->func;
+    GLuint mask = inst->DstReg.WriteMask;
+
+    struct brw_reg interp[4];
+    struct brw_reg dst, src0;
+
+    src0 = get_src_reg(c, &inst->SrcReg[0], 0, 1);
+    GLuint nr = src0.nr;
+    int i;
+
+    interp[0] = brw_vec1_grf(nr, 0);
+    interp[1] = brw_vec1_grf(nr, 4);
+    interp[2] = brw_vec1_grf(nr+1, 0);
+    interp[3] = brw_vec1_grf(nr+1, 4);
+
+    for(i = 0; i < 4; i++ ) {
+	if (mask & (1<<i)) {
+	    dst = get_dst_reg(c, inst, i, 1);
+	    brw_MOV(p, dst, suboffset(interp[i],3));
+	}
+    }
+}
+
+static void emit_pinterp(struct brw_wm_compile *c,
+		struct prog_instruction *inst)
+{
+    struct brw_compile *p = &c->func;
+    GLuint mask = inst->DstReg.WriteMask;
+
+    struct brw_reg interp[4];
+    struct brw_reg dst, delta0, delta1;
+    struct brw_reg src0, w;
+
+    src0 = get_src_reg(c, &inst->SrcReg[0], 0, 1);
+    delta0 = get_src_reg(c, &inst->SrcReg[1], 0, 1);
+    delta1 = get_src_reg(c, &inst->SrcReg[1], 1, 1);
+    w = get_src_reg(c, &inst->SrcReg[2], 3, 1);
+    GLuint nr = src0.nr;
+    int i;
+
+    interp[0] = brw_vec1_grf(nr, 0);
+    interp[1] = brw_vec1_grf(nr, 4);
+    interp[2] = brw_vec1_grf(nr+1, 0);
+    interp[3] = brw_vec1_grf(nr+1, 4);
+
+    for(i = 0; i < 4; i++ ) {
+	if (mask & (1<<i)) {
+	    dst = get_dst_reg(c, inst, i, 1);
+	    brw_LINE(p, brw_null_reg(), interp[i], delta0);
+	    brw_MAC(p, dst, suboffset(interp[i],1), 
+		    delta1);
+	    brw_MUL(p, dst, dst, w);
+	}
+    }
+}
+
+static void emit_xpd(struct brw_wm_compile *c,
+		struct prog_instruction *inst)
+{
+    int i;
+    struct brw_compile *p = &c->func;
+    GLuint mask = inst->DstReg.WriteMask;
+    for (i = 0; i < 4; i++) {
+	GLuint i2 = (i+2)%3;
+	GLuint i1 = (i+1)%3;
+	if (mask & (1<<i)) {
+	    struct brw_reg src0, src1, dst;
+	    dst = get_dst_reg(c, inst, i, 1);
+	    src0 = negate(get_src_reg(c, &inst->SrcReg[0], i2, 1));
+	    src1 = get_src_reg(c, &inst->SrcReg[1], i1, 1);
+	    brw_MUL(p, brw_null_reg(), src0, src1);
+	    src0 = get_src_reg(c, &inst->SrcReg[0], i1, 1);
+	    src1 = get_src_reg(c, &inst->SrcReg[1], i2, 1);
+	    brw_set_saturate(p, inst->SaturateMode != SATURATE_OFF);
+	    brw_MAC(p, dst, src0, src1);
+	    brw_set_saturate(p, 0);
+	}
+    }
+    brw_set_saturate(p, 0);
+}
+
+static void emit_dp3(struct brw_wm_compile *c,
+		struct prog_instruction *inst)
+{
+    struct brw_reg src0[3], src1[3], dst;
+    int i;
+    struct brw_compile *p = &c->func;
+    for (i = 0; i < 3; i++) {
+	src0[i] = get_src_reg(c, &inst->SrcReg[0], i, 1);
+	src1[i] = get_src_reg(c, &inst->SrcReg[1], i, 1);
+    }
+
+    dst = get_dst_reg(c, inst, get_scalar_dst_index(inst), 1);
+    brw_MUL(p, brw_null_reg(), src0[0], src1[0]);
+    brw_MAC(p, brw_null_reg(), src0[1], src1[1]);
+    brw_set_saturate(p, (inst->SaturateMode != SATURATE_OFF) ? 1 : 0);
+    brw_MAC(p, dst, src0[2], src1[2]);
+    brw_set_saturate(p, 0);
+}
+
+static void emit_dp4(struct brw_wm_compile *c,
+		struct prog_instruction *inst)
+{
+    struct brw_reg src0[4], src1[4], dst;
+    int i;
+    struct brw_compile *p = &c->func;
+    for (i = 0; i < 4; i++) {
+	src0[i] = get_src_reg(c, &inst->SrcReg[0], i, 1);
+	src1[i] = get_src_reg(c, &inst->SrcReg[1], i, 1);
+    }
+    dst = get_dst_reg(c, inst, get_scalar_dst_index(inst), 1);
+    brw_MUL(p, brw_null_reg(), src0[0], src1[0]);
+    brw_MAC(p, brw_null_reg(), src0[1], src1[1]);
+    brw_MAC(p, brw_null_reg(), src0[2], src1[2]);
+    brw_set_saturate(p, (inst->SaturateMode != SATURATE_OFF) ? 1 : 0);
+    brw_MAC(p, dst, src0[3], src1[3]);
+    brw_set_saturate(p, 0);
+}
+
+static void emit_dph(struct brw_wm_compile *c,
+		struct prog_instruction *inst)
+{
+    struct brw_reg src0[4], src1[4], dst;
+    int i;
+    struct brw_compile *p = &c->func;
+    for (i = 0; i < 4; i++) {
+	src0[i] = get_src_reg(c, &inst->SrcReg[0], i, 1);
+	src1[i] = get_src_reg(c, &inst->SrcReg[1], i, 1);
+    }
+    dst = get_dst_reg(c, inst, get_scalar_dst_index(inst), 1);
+    brw_MUL(p, brw_null_reg(), src0[0], src1[0]);
+    brw_MAC(p, brw_null_reg(), src0[1], src1[1]);
+    brw_MAC(p, dst, src0[2], src1[2]);
+    brw_set_saturate(p, (inst->SaturateMode != SATURATE_OFF) ? 1 : 0);
+    brw_ADD(p, dst, src0[3], src1[3]);
+    brw_set_saturate(p, 0);
+}
+
+static void emit_math1(struct brw_wm_compile *c,
+		struct prog_instruction *inst, GLuint func)
+{
+    struct brw_compile *p = &c->func;
+    struct brw_reg src0, dst;
+
+    src0 = get_src_reg(c, &inst->SrcReg[0], 0, 1);
+    dst = get_dst_reg(c, inst, get_scalar_dst_index(inst), 1);
+    brw_MOV(p, brw_message_reg(2), src0);
+    brw_math(p,
+	    dst,
+	    func,
+	    (inst->SaturateMode != SATURATE_OFF) ? BRW_MATH_SATURATE_SATURATE : BRW_MATH_SATURATE_NONE,
+	    2,
+	    brw_null_reg(),
+	    BRW_MATH_DATA_VECTOR,
+	    BRW_MATH_PRECISION_FULL);
+}
+
+static void emit_rcp(struct brw_wm_compile *c,
+		struct prog_instruction *inst)
+{
+    emit_math1(c, inst, BRW_MATH_FUNCTION_INV);
+}
+
+static void emit_rsq(struct brw_wm_compile *c,
+		struct prog_instruction *inst)
+{
+    emit_math1(c, inst, BRW_MATH_FUNCTION_RSQ);
+}
+
+static void emit_sin(struct brw_wm_compile *c,
+		struct prog_instruction *inst)
+{
+    emit_math1(c, inst, BRW_MATH_FUNCTION_SIN);
+}
+
+static void emit_cos(struct brw_wm_compile *c,
+		struct prog_instruction *inst)
+{
+    emit_math1(c, inst, BRW_MATH_FUNCTION_COS);
+}
+
+static void emit_ex2(struct brw_wm_compile *c,
+		struct prog_instruction *inst)
+{
+    emit_math1(c, inst, BRW_MATH_FUNCTION_EXP);
+}
+
+static void emit_lg2(struct brw_wm_compile *c,
+		struct prog_instruction *inst)
+{
+    emit_math1(c, inst, BRW_MATH_FUNCTION_LOG);
+}
+
+static void emit_add(struct brw_wm_compile *c,
+		struct prog_instruction *inst)
+{
+    struct brw_compile *p = &c->func;
+    struct brw_reg src0, src1, dst;
+    GLuint mask = inst->DstReg.WriteMask;
+    int i;
+    brw_set_saturate(p, (inst->SaturateMode != SATURATE_OFF) ? 1 : 0);
+    for (i = 0 ; i < 4; i++) {
+	if (mask & (1<<i)) {
+	    dst = get_dst_reg(c, inst, i, 1);
+	    src0 = get_src_reg(c, &inst->SrcReg[0], i, 1);
+	    src1 = get_src_reg(c, &inst->SrcReg[1], i, 1);
+	    brw_ADD(p, dst, src0, src1);
+	}
+    }
+    brw_set_saturate(p, 0);
+}
+
+static void emit_sub(struct brw_wm_compile *c,
+		struct prog_instruction *inst)
+{
+    struct brw_compile *p = &c->func;
+    struct brw_reg src0, src1, dst;
+    GLuint mask = inst->DstReg.WriteMask;
+    int i;
+    brw_set_saturate(p, (inst->SaturateMode != SATURATE_OFF) ? 1 : 0);
+    for (i = 0 ; i < 4; i++) {
+	if (mask & (1<<i)) {
+	    dst = get_dst_reg(c, inst, i, 1);
+	    src0 = get_src_reg(c, &inst->SrcReg[0], i, 1);
+	    src1 = get_src_reg(c, &inst->SrcReg[1], i, 1);
+	    brw_ADD(p, dst, src0, negate(src1));
+	}
+    }
+    brw_set_saturate(p, 0);
+}
+
+static void emit_mul(struct brw_wm_compile *c,
+		struct prog_instruction *inst)
+{
+    struct brw_compile *p = &c->func;
+    struct brw_reg src0, src1, dst;
+    GLuint mask = inst->DstReg.WriteMask;
+    int i;
+    brw_set_saturate(p, (inst->SaturateMode != SATURATE_OFF) ? 1 : 0);
+    for (i = 0 ; i < 4; i++) {
+	if (mask & (1<<i)) {
+	    dst = get_dst_reg(c, inst, i, 1);
+	    src0 = get_src_reg(c, &inst->SrcReg[0], i, 1);
+	    src1 = get_src_reg(c, &inst->SrcReg[1], i, 1);
+	    brw_MUL(p, dst, src0, src1);
+	}
+    }
+    brw_set_saturate(p, 0);
+}
+
+static void emit_frc(struct brw_wm_compile *c,
+		struct prog_instruction *inst)
+{
+    struct brw_compile *p = &c->func;
+    struct brw_reg src0, dst;
+    GLuint mask = inst->DstReg.WriteMask;
+    int i;
+    brw_set_saturate(p, (inst->SaturateMode != SATURATE_OFF) ? 1 : 0);
+    for (i = 0 ; i < 4; i++) {
+	if (mask & (1<<i)) {
+	    dst = get_dst_reg(c, inst, i, 1);
+	    src0 = get_src_reg(c, &inst->SrcReg[0], i, 1);
+	    brw_FRC(p, dst, src0);
+	}
+    }
+    if (inst->SaturateMode != SATURATE_OFF)
+	brw_set_saturate(p, 0);
+}
+
+static void emit_flr(struct brw_wm_compile *c,
+		struct prog_instruction *inst)
+{
+    struct brw_compile *p = &c->func;
+    struct brw_reg src0, dst;
+    GLuint mask = inst->DstReg.WriteMask;
+    int i;
+    brw_set_saturate(p, (inst->SaturateMode != SATURATE_OFF) ? 1 : 0);
+    for (i = 0 ; i < 4; i++) {
+	if (mask & (1<<i)) {
+	    dst = get_dst_reg(c, inst, i, 1);
+	    src0 = get_src_reg(c, &inst->SrcReg[0], i, 1);
+	    brw_RNDD(p, dst, src0);
+	}
+    }
+    brw_set_saturate(p, 0);
+}
+
+static void emit_max(struct brw_wm_compile *c,
+		struct prog_instruction *inst)
+{
+    struct brw_compile *p = &c->func;
+    GLuint mask = inst->DstReg.WriteMask;
+    struct brw_reg src0, src1, dst;
+    int i;
+    brw_push_insn_state(p);
+    for (i = 0; i < 4; i++) {
+	if (mask & (1<<i)) {
+	    dst = get_dst_reg(c, inst, i, 1);
+	    src0 = get_src_reg(c, &inst->SrcReg[0], i, 1);
+	    src1 = get_src_reg(c, &inst->SrcReg[1], i, 1);
+	    brw_set_saturate(p, (inst->SaturateMode != SATURATE_OFF) ? 1 : 0);
+	    brw_MOV(p, dst, src0);
+	    brw_set_saturate(p, 0);
+
+	    brw_CMP(p, brw_null_reg(), BRW_CONDITIONAL_L, src0, src1);
+	    brw_set_saturate(p, (inst->SaturateMode != SATURATE_OFF) ? 1 : 0);
+	    brw_set_predicate_control(p, BRW_PREDICATE_NORMAL);
+	    brw_MOV(p, dst, src1);
+	    brw_set_saturate(p, 0);
+	    brw_set_predicate_control_flag_value(p, 0xff);
+	}
+    }
+    brw_pop_insn_state(p);
+}
+
+static void emit_min(struct brw_wm_compile *c,
+		struct prog_instruction *inst)
+{
+    struct brw_compile *p = &c->func;
+    GLuint mask = inst->DstReg.WriteMask;
+    struct brw_reg src0, src1, dst;
+    int i;
+    brw_push_insn_state(p);
+    for (i = 0; i < 4; i++) {
+	if (mask & (1<<i)) {
+	    dst = get_dst_reg(c, inst, i, 1);
+	    src0 = get_src_reg(c, &inst->SrcReg[0], i, 1);
+	    src1 = get_src_reg(c, &inst->SrcReg[1], i, 1);
+	    brw_set_saturate(p, (inst->SaturateMode != SATURATE_OFF) ? 1 : 0);
+	    brw_MOV(p, dst, src0);
+	    brw_set_saturate(p, 0);
+
+	    brw_CMP(p, brw_null_reg(), BRW_CONDITIONAL_L, src1, src0);
+	    brw_set_saturate(p, (inst->SaturateMode != SATURATE_OFF) ? 1 : 0);
+	    brw_set_predicate_control(p, BRW_PREDICATE_NORMAL);
+	    brw_MOV(p, dst, src1);
+	    brw_set_saturate(p, 0);
+	    brw_set_predicate_control_flag_value(p, 0xff);
+	}
+    }
+    brw_pop_insn_state(p);
+}
+
+static void emit_pow(struct brw_wm_compile *c,
+		struct prog_instruction *inst)
+{
+    struct brw_compile *p = &c->func;
+    struct brw_reg dst, src0, src1;
+    dst = get_dst_reg(c, inst, get_scalar_dst_index(inst), 1);
+    src0 = get_src_reg(c, &inst->SrcReg[0], 0, 1);
+    src1 = get_src_reg(c, &inst->SrcReg[1], 0, 1);
+
+    brw_MOV(p, brw_message_reg(2), src0);
+    brw_MOV(p, brw_message_reg(3), src1);
+
+    brw_math(p,
+	    dst,
+	    BRW_MATH_FUNCTION_POW,
+	    (inst->SaturateMode != SATURATE_OFF) ? BRW_MATH_SATURATE_SATURATE : BRW_MATH_SATURATE_NONE,
+	    2,
+	    brw_null_reg(),
+	    BRW_MATH_DATA_VECTOR,
+	    BRW_MATH_PRECISION_FULL);
+}
+
+static void emit_lrp(struct brw_wm_compile *c,
+		struct prog_instruction *inst)
+{
+    struct brw_compile *p = &c->func;
+    GLuint mask = inst->DstReg.WriteMask;
+    struct brw_reg dst, tmp1, tmp2, src0, src1, src2;
+    int i;
+    for (i = 0; i < 4; i++) {
+	if (mask & (1<<i)) {
+	    dst = get_dst_reg(c, inst, i, 1);
+	    src0 = get_src_reg(c, &inst->SrcReg[0], i, 1);
+
+	    src1 = get_src_reg(c, &inst->SrcReg[1], i, 1);
+
+	    if (src1.nr == dst.nr) {
+		tmp1 = alloc_tmp(c);
+		brw_MOV(p, tmp1, src1);
+	    } else
+		tmp1 = src1;
+
+	    src2 = get_src_reg(c, &inst->SrcReg[2], i, 1);
+	    if (src2.nr == dst.nr) {
+		tmp2 = alloc_tmp(c);
+		brw_MOV(p, tmp2, src2);
+	    } else
+		tmp2 = src2;
+
+	    brw_ADD(p, dst, negate(src0), brw_imm_f(1.0));
+	    brw_MUL(p, brw_null_reg(), dst, tmp2);
+	    brw_set_saturate(p, (inst->SaturateMode != SATURATE_OFF) ? 1 : 0);
+	    brw_MAC(p, dst, src0, tmp1);
+	    brw_set_saturate(p, 0);
+	}
+	release_tmps(c);
+    }
+}
+
+static void emit_kil(struct brw_wm_compile *c)
+{
+	struct brw_compile *p = &c->func;
+	struct brw_reg depth = retype(brw_vec1_grf(0, 0), BRW_REGISTER_TYPE_UW);
+	brw_push_insn_state(p);
+	brw_set_mask_control(p, BRW_MASK_DISABLE);
+	brw_NOT(p, c->emit_mask_reg, brw_mask_reg(1)); //IMASK
+	brw_AND(p, depth, c->emit_mask_reg, depth);
+	brw_pop_insn_state(p);
+}
+
+static void emit_mad(struct brw_wm_compile *c,
+		struct prog_instruction *inst)
+{
+    struct brw_compile *p = &c->func;
+    GLuint mask = inst->DstReg.WriteMask;
+    struct brw_reg dst, src0, src1, src2;
+    int i;
+
+    for (i = 0; i < 4; i++) {
+	if (mask & (1<<i)) {
+	    dst = get_dst_reg(c, inst, i, 1);
+	    src0 = get_src_reg(c, &inst->SrcReg[0], i, 1);
+	    src1 = get_src_reg(c, &inst->SrcReg[1], i, 1);
+	    src2 = get_src_reg(c, &inst->SrcReg[2], i, 1);
+	    brw_MUL(p, dst, src0, src1);
+
+	    brw_set_saturate(p, (inst->SaturateMode != SATURATE_OFF) ? 1 : 0);
+	    brw_ADD(p, dst, dst, src2);
+	    brw_set_saturate(p, 0);
+	}
+    }
+}
+
+static void emit_sop(struct brw_wm_compile *c,
+		struct prog_instruction *inst, GLuint cond)
+{
+    struct brw_compile *p = &c->func;
+    GLuint mask = inst->DstReg.WriteMask;
+    struct brw_reg dst, src0, src1;
+    int i;
+
+    brw_push_insn_state(p);
+    for (i = 0; i < 4; i++) {
+	if (mask & (1<<i)) {
+	    dst = get_dst_reg(c, inst, i, 1);
+	    src0 = get_src_reg(c, &inst->SrcReg[0], i, 1);
+	    src1 = get_src_reg(c, &inst->SrcReg[1], i, 1);
+	    brw_CMP(p, brw_null_reg(), cond, src0, src1);
+	    brw_set_predicate_control(p, BRW_PREDICATE_NONE);
+	    brw_MOV(p, dst, brw_imm_f(0.0));
+	    brw_set_predicate_control(p, BRW_PREDICATE_NORMAL);
+	    brw_MOV(p, dst, brw_imm_f(1.0));
+	}
+    }
+    brw_pop_insn_state(p);
+}
+
+static void emit_slt(struct brw_wm_compile *c,
+		struct prog_instruction *inst)
+{
+    emit_sop(c, inst, BRW_CONDITIONAL_L);
+}
+
+static void emit_sle(struct brw_wm_compile *c,
+		struct prog_instruction *inst)
+{
+    emit_sop(c, inst, BRW_CONDITIONAL_LE);
+}
+
+static void emit_sgt(struct brw_wm_compile *c,
+		struct prog_instruction *inst)
+{
+    emit_sop(c, inst, BRW_CONDITIONAL_G);
+}
+
+static void emit_sge(struct brw_wm_compile *c,
+		struct prog_instruction *inst)
+{
+    emit_sop(c, inst, BRW_CONDITIONAL_GE);
+}
+
+static void emit_seq(struct brw_wm_compile *c,
+		struct prog_instruction *inst)
+{
+    emit_sop(c, inst, BRW_CONDITIONAL_EQ);
+}
+
+static void emit_sne(struct brw_wm_compile *c,
+		struct prog_instruction *inst)
+{
+    emit_sop(c, inst, BRW_CONDITIONAL_NEQ);
+}
+
+static void emit_ddx(struct brw_wm_compile *c,
+                struct prog_instruction *inst)
+{
+    struct brw_compile *p = &c->func;
+    GLuint mask = inst->DstReg.WriteMask;
+    struct brw_reg interp[4];
+    struct brw_reg dst;
+    struct brw_reg src0, w;
+    GLuint nr, i;
+    src0 = get_src_reg(c, &inst->SrcReg[0], 0, 1);
+    w = get_src_reg(c, &inst->SrcReg[1], 3, 1);
+    nr = src0.nr;
+    interp[0] = brw_vec1_grf(nr, 0);
+    interp[1] = brw_vec1_grf(nr, 4);
+    interp[2] = brw_vec1_grf(nr+1, 0);
+    interp[3] = brw_vec1_grf(nr+1, 4);
+    brw_set_saturate(p, inst->SaturateMode != SATURATE_OFF);
+    for(i = 0; i < 4; i++ ) {
+        if (mask & (1<<i)) {
+            dst = get_dst_reg(c, inst, i, 1);
+            brw_MOV(p, dst, interp[i]);
+            brw_MUL(p, dst, dst, w);
+        }
+    }
+    brw_set_saturate(p, 0);
+}
+
+static void emit_ddy(struct brw_wm_compile *c,
+                struct prog_instruction *inst)
+{
+    struct brw_compile *p = &c->func;
+    GLuint mask = inst->DstReg.WriteMask;
+    struct brw_reg interp[4];
+    struct brw_reg dst;
+    struct brw_reg src0, w;
+    GLuint nr, i;
+
+    src0 = get_src_reg(c, &inst->SrcReg[0], 0, 1);
+    nr = src0.nr;
+    w = get_src_reg(c, &inst->SrcReg[1], 3, 1);
+    interp[0] = brw_vec1_grf(nr, 0);
+    interp[1] = brw_vec1_grf(nr, 4);
+    interp[2] = brw_vec1_grf(nr+1, 0);
+    interp[3] = brw_vec1_grf(nr+1, 4);
+    brw_set_saturate(p, inst->SaturateMode != SATURATE_OFF);
+    for(i = 0; i < 4; i++ ) {
+        if (mask & (1<<i)) {
+            dst = get_dst_reg(c, inst, i, 1);
+            brw_MOV(p, dst, suboffset(interp[i], 1));
+            brw_MUL(p, dst, dst, w);
+        }
+    }
+    brw_set_saturate(p, 0);
+}
+
+static void emit_wpos_xy(struct brw_wm_compile *c,
+                struct prog_instruction *inst)
+{
+    struct brw_compile *p = &c->func;
+    GLuint mask = inst->DstReg.WriteMask;
+    struct brw_reg src0[2], dst[2];
+
+    dst[0] = get_dst_reg(c, inst, 0, 1);
+    dst[1] = get_dst_reg(c, inst, 1, 1);
+
+    src0[0] = get_src_reg(c, &inst->SrcReg[0], 0, 1);
+    src0[1] = get_src_reg(c, &inst->SrcReg[0], 1, 1);
+
+    /* Calc delta X,Y by subtracting origin in r1 from the pixel
+     * centers.
+     */
+    if (mask & WRITEMASK_X) {
+	brw_MOV(p,
+		dst[0],
+		retype(src0[0], BRW_REGISTER_TYPE_UW));
+    }
+
+    if (mask & WRITEMASK_Y) {
+	/* TODO -- window_height - Y */
+	brw_MOV(p,
+		dst[1],
+		retype(src0[1], BRW_REGISTER_TYPE_UW));
+
+    }
+}
+
+/* TODO
+   BIAS on SIMD8 not workind yet...
+ */	
+static void emit_txb(struct brw_wm_compile *c,
+		struct prog_instruction *inst)
+{
+    struct brw_compile *p = &c->func;
+    struct brw_reg dst[4], src[4], payload_reg;
+    GLuint i;
+    payload_reg = get_reg(c, PROGRAM_PAYLOAD, PAYLOAD_DEPTH, 0, 1, 0, 0);
+    for (i = 0; i < 4; i++) 
+	dst[i] = get_dst_reg(c, inst, i, 1);
+    for (i = 0; i < 4; i++)
+	src[i] = get_src_reg(c, &inst->SrcReg[0], i, 1);
+
+    switch (inst->TexSrcTarget) {
+	case TEXTURE_1D_INDEX:
+	    brw_MOV(p, brw_message_reg(2), src[0]);
+	    brw_MOV(p, brw_message_reg(3), brw_imm_f(0));
+	    brw_MOV(p, brw_message_reg(4), brw_imm_f(0));
+	    break;
+	case TEXTURE_2D_INDEX:
+	case TEXTURE_RECT_INDEX:
+	    brw_MOV(p, brw_message_reg(2), src[0]);
+	    brw_MOV(p, brw_message_reg(3), src[1]);
+	    brw_MOV(p, brw_message_reg(4), brw_imm_f(0));
+	    break;
+	default:
+	    brw_MOV(p, brw_message_reg(2), src[0]);
+	    brw_MOV(p, brw_message_reg(3), src[1]);
+	    brw_MOV(p, brw_message_reg(4), src[2]);
+	    break;
+    }
+    brw_MOV(p, brw_message_reg(5), src[3]);
+    brw_MOV(p, brw_message_reg(6), brw_imm_f(0));
+    brw_SAMPLE(p,
+	    retype(vec8(dst[0]), BRW_REGISTER_TYPE_UW),
+	    1,
+	    retype(payload_reg, BRW_REGISTER_TYPE_UW),
+	    inst->TexSrcUnit + 1, /* surface */
+	    inst->TexSrcUnit,     /* sampler */
+	    inst->DstReg.WriteMask,
+	    BRW_SAMPLER_MESSAGE_SIMD16_SAMPLE_BIAS,
+	    4,
+	    4,
+	    0);
+}
+
+static void emit_tex(struct brw_wm_compile *c,
+		struct prog_instruction *inst)
+{
+    struct brw_compile *p = &c->func;
+    struct brw_reg dst[4], src[4], payload_reg;
+    GLuint msg_len;
+    GLuint i, nr;
+    GLuint emit;
+    GLboolean shadow = (c->key.shadowtex_mask & (1<<inst->TexSrcUnit)) ? 1 : 0;
+
+    payload_reg = get_reg(c, PROGRAM_PAYLOAD, PAYLOAD_DEPTH, 0, 1, 0, 0);
+
+    for (i = 0; i < 4; i++) 
+	dst[i] = get_dst_reg(c, inst, i, 1);
+    for (i = 0; i < 4; i++)
+	src[i] = get_src_reg(c, &inst->SrcReg[0], i, 1);
+
+
+    switch (inst->TexSrcTarget) {
+	case TEXTURE_1D_INDEX:
+	    emit = WRITEMASK_X;
+	    nr = 1;
+	    break;
+	case TEXTURE_2D_INDEX:
+	case TEXTURE_RECT_INDEX:
+	    emit = WRITEMASK_XY;
+	    nr = 2;
+	    break;
+	default:
+	    emit = WRITEMASK_XYZ;
+	    nr = 3;
+	    break;
+    }
+    msg_len = 1;
+
+    for (i = 0; i < nr; i++) {
+	static const GLuint swz[4] = {0,1,2,2};
+	if (emit & (1<<i))
+	    brw_MOV(p, brw_message_reg(msg_len+1), src[swz[i]]);
+	else
+	    brw_MOV(p, brw_message_reg(msg_len+1), brw_imm_f(0));
+	msg_len += 1;
+    }
+
+    if (shadow) {
+	brw_MOV(p, brw_message_reg(5), brw_imm_f(0));
+	brw_MOV(p, brw_message_reg(6), src[2]);
+    }
+
+    brw_SAMPLE(p,
+	    retype(vec8(dst[0]), BRW_REGISTER_TYPE_UW),
+	    1,
+	    retype(payload_reg, BRW_REGISTER_TYPE_UW),
+	    inst->TexSrcUnit + 1, /* surface */
+	    inst->TexSrcUnit,     /* sampler */
+	    inst->DstReg.WriteMask,
+	    BRW_SAMPLER_MESSAGE_SIMD8_SAMPLE,
+	    4,
+	    shadow ? 6 : 4,
+	    0);
+
+    if (shadow)
+	brw_MOV(p, dst[3], brw_imm_f(1.0));
+}
+
+static void post_wm_emit( struct brw_wm_compile *c )
+{
+    GLuint nr_insns = c->fp->program.Base.NumInstructions;
+    GLuint insn, target_insn;
+    struct prog_instruction *inst1, *inst2;
+    struct brw_instruction *brw_inst1, *brw_inst2;
+    int offset;
+    for (insn = 0; insn < nr_insns; insn++) {
+	inst1 = &c->fp->program.Base.Instructions[insn];
+	brw_inst1 = inst1->Data;
+	switch (inst1->Opcode) {
+	    case OPCODE_CAL:
+		target_insn = inst1->BranchTarget;
+		inst2 = &c->fp->program.Base.Instructions[target_insn];
+		brw_inst2 = inst2->Data;
+		offset = brw_inst2 - brw_inst1;
+		brw_set_src1(brw_inst1, brw_imm_d(offset*16));
+		break;
+	    default:
+		break;
+	}
+    }
+}
+
+static void brw_wm_emit_glsl(struct brw_wm_compile *c)
+
+{
+#define MAX_IFSN 32
+#define MAX_LOOP_DEPTH 32
+    struct brw_instruction *if_inst[MAX_IFSN], *loop_inst[MAX_LOOP_DEPTH];
+    struct brw_instruction *inst0, *inst1;
+    int i, if_insn = 0, loop_insn = 0;
+    struct brw_compile *p = &c->func;
+    struct brw_indirect stack_index = brw_indirect(0, 0);
+
+    brw_init_compile(&c->func);
+    c->reg_index = 0;
+    prealloc_reg(c);
+    brw_set_compression_control(p, BRW_COMPRESSION_NONE);
+    brw_MOV(p, get_addr_reg(stack_index), brw_address(c->stack));
+
+    for (i = 0; i < c->nr_fp_insns; i++) {
+	struct prog_instruction *inst = &c->prog_instructions[i];
+	struct prog_instruction *orig_inst;
+
+	if ((orig_inst = inst->Data) != 0)
+	    orig_inst->Data = current_insn(p);
+
+	if (inst->CondUpdate)
+	    brw_set_conditionalmod(p, BRW_CONDITIONAL_NZ);
+	else
+	    brw_set_conditionalmod(p, BRW_CONDITIONAL_NONE);
+
+	switch (inst->Opcode) {
+	    case WM_PIXELXY:
+		emit_pixel_xy(c, inst);
+		break;
+	    case WM_DELTAXY: 
+		emit_delta_xy(c, inst);
+		break;
+	    case WM_PIXELW:
+		emit_pixel_w(c, inst);
+		break;	
+	    case WM_LINTERP:
+		emit_linterp(c, inst);
+		break;
+	    case WM_PINTERP:
+		emit_pinterp(c, inst);
+		break;
+	    case WM_CINTERP:
+		emit_cinterp(c, inst);
+		break;
+	    case WM_WPOSXY:
+		emit_wpos_xy(c, inst);
+		break;
+	    case WM_FB_WRITE:
+		emit_fb_write(c, inst);
+		break;
+	    case OPCODE_ABS:
+		emit_abs(c, inst);
+		break;
+	    case OPCODE_ADD:
+		emit_add(c, inst);
+		break;
+	    case OPCODE_SUB:
+		emit_sub(c, inst);
+		break;
+	    case OPCODE_FRC:
+		emit_frc(c, inst);
+		break;
+	    case OPCODE_FLR:
+		emit_flr(c, inst);
+		break;
+	    case OPCODE_LRP:
+		emit_lrp(c, inst);
+		break;
+	    case OPCODE_INT:
+		emit_int(c, inst);
+		break;
+	    case OPCODE_MOV:
+		emit_mov(c, inst);
+		break;
+	    case OPCODE_DP3:
+		emit_dp3(c, inst);
+		break;
+	    case OPCODE_DP4:
+		emit_dp4(c, inst);
+		break;
+	    case OPCODE_XPD:
+		emit_xpd(c, inst);
+		break;
+	    case OPCODE_DPH:
+		emit_dph(c, inst);
+		break;
+	    case OPCODE_RCP:
+		emit_rcp(c, inst);
+		break;
+	    case OPCODE_RSQ:
+		emit_rsq(c, inst);
+		break;
+	    case OPCODE_SIN:
+		emit_sin(c, inst);
+		break;
+	    case OPCODE_COS:
+		emit_cos(c, inst);
+		break;
+	    case OPCODE_EX2:
+		emit_ex2(c, inst);
+		break;
+	    case OPCODE_LG2:
+		emit_lg2(c, inst);
+		break;
+	    case OPCODE_MAX:	
+		emit_max(c, inst);
+		break;
+	    case OPCODE_MIN:	
+		emit_min(c, inst);
+		break;
+	    case OPCODE_DDX:
+		emit_ddx(c, inst);
+		break;
+	    case OPCODE_DDY:
+                emit_ddy(c, inst);
+                break;
+	    case OPCODE_SLT:
+		emit_slt(c, inst);
+		break;
+	    case OPCODE_SLE:
+		emit_sle(c, inst);
+		break;
+	    case OPCODE_SGT:
+		emit_sgt(c, inst);
+		break;
+	    case OPCODE_SGE:
+		emit_sge(c, inst);
+		break;
+	    case OPCODE_SEQ:
+		emit_seq(c, inst);
+		break;
+	    case OPCODE_SNE:
+		emit_sne(c, inst);
+		break;
+	    case OPCODE_MUL:
+		emit_mul(c, inst);
+		break;
+	    case OPCODE_POW:
+		emit_pow(c, inst);
+		break;
+	    case OPCODE_MAD:
+		emit_mad(c, inst);
+		break;
+	    case OPCODE_TEX:
+		emit_tex(c, inst);
+		break;
+	    case OPCODE_TXB:
+		emit_txb(c, inst);
+		break;
+	    case OPCODE_KIL_NV:
+		emit_kil(c);
+		break;
+	    case OPCODE_IF:
+		assert(if_insn < MAX_IFSN);
+		if_inst[if_insn++] = brw_IF(p, BRW_EXECUTE_8);
+		break;
+	    case OPCODE_ELSE:
+		if_inst[if_insn-1]  = brw_ELSE(p, if_inst[if_insn-1]);
+		break;
+	    case OPCODE_ENDIF:
+		assert(if_insn > 0);
+		brw_ENDIF(p, if_inst[--if_insn]);
+		break;
+	    case OPCODE_BGNSUB:
+	    case OPCODE_ENDSUB:
+		break;
+	    case OPCODE_CAL: 
+		brw_push_insn_state(p);
+		brw_set_mask_control(p, BRW_MASK_DISABLE);
+                brw_set_access_mode(p, BRW_ALIGN_1);
+                brw_ADD(p, deref_1ud(stack_index, 0), brw_ip_reg(), brw_imm_d(3*16));
+                brw_set_access_mode(p, BRW_ALIGN_16);
+                brw_ADD(p, get_addr_reg(stack_index),
+                         get_addr_reg(stack_index), brw_imm_d(4));
+                orig_inst = inst->Data;
+                orig_inst->Data = &p->store[p->nr_insn];
+                brw_ADD(p, brw_ip_reg(), brw_ip_reg(), brw_imm_d(1*16));
+                brw_pop_insn_state(p);
+		break;
+
+	    case OPCODE_RET:
+		brw_push_insn_state(p);
+		brw_set_mask_control(p, BRW_MASK_DISABLE);
+                brw_ADD(p, get_addr_reg(stack_index),
+                        get_addr_reg(stack_index), brw_imm_d(-4));
+                brw_set_access_mode(p, BRW_ALIGN_1);
+                brw_MOV(p, brw_ip_reg(), deref_1ud(stack_index, 0));
+                brw_set_access_mode(p, BRW_ALIGN_16);
+		brw_pop_insn_state(p);
+
+		break;
+	    case OPCODE_BGNLOOP:
+		loop_inst[loop_insn++] = brw_DO(p, BRW_EXECUTE_8);
+		break;
+	    case OPCODE_BRK:
+		brw_BREAK(p);
+		brw_set_predicate_control(p, BRW_PREDICATE_NONE);
+		break;
+	    case OPCODE_CONT:
+		brw_CONT(p);
+		brw_set_predicate_control(p, BRW_PREDICATE_NONE);
+		break;
+	    case OPCODE_ENDLOOP: 
+		loop_insn--;
+		inst0 = inst1 = brw_WHILE(p, loop_inst[loop_insn]);
+		/* patch all the BREAK instructions from
+		   last BEGINLOOP */
+		while (inst0 > loop_inst[loop_insn]) {
+		    inst0--;
+		    if (inst0->header.opcode == BRW_OPCODE_BREAK) {
+			inst0->bits3.if_else.jump_count = inst1 - inst0 + 1;
+			inst0->bits3.if_else.pop_count = 0;
+		    } else if (inst0->header.opcode == BRW_OPCODE_CONTINUE) {
+                        inst0->bits3.if_else.jump_count = inst1 - inst0;
+                        inst0->bits3.if_else.pop_count = 0;
+                    }
+		}
+		break;
+	    default:
+		_mesa_printf("unsupported IR in fragment shader %d\n",
+			inst->Opcode);
+	}
+	if (inst->CondUpdate)
+	    brw_set_predicate_control(p, BRW_PREDICATE_NORMAL);
+	else
+	    brw_set_predicate_control(p, BRW_PREDICATE_NONE);
+    }
+    post_wm_emit(c);
+    for (i = 0; i < c->fp->program.Base.NumInstructions; i++)
+	c->fp->program.Base.Instructions[i].Data = NULL;
+}
+
+void brw_wm_glsl_emit(struct brw_wm_compile *c)
+{
+    brw_wm_pass_fp(c);
+    c->tmp_index = 127;
+    brw_wm_emit_glsl(c);
+    c->prog_data.total_grf = c->reg_index;
+    c->prog_data.total_scratch = 0;
+}
diff --git a/src/mesa/drivers/dri/i965/brw_wm_pass0.c b/src/mesa/drivers/dri/i965/brw_wm_pass0.c
index 00f6f6b9a4f..1bfae5a069b 100644
--- a/src/mesa/drivers/dri/i965/brw_wm_pass0.c
+++ b/src/mesa/drivers/dri/i965/brw_wm_pass0.c
@@ -168,6 +168,7 @@ static const struct brw_wm_ref *pass0_get_reg( struct brw_wm_compile *c,
       case PROGRAM_PAYLOAD:
       case PROGRAM_TEMPORARY:
       case PROGRAM_OUTPUT:
+      case PROGRAM_VARYING:
 	 break;
 
       case PROGRAM_LOCAL_PARAM:
@@ -179,6 +180,8 @@ static const struct brw_wm_ref *pass0_get_reg( struct brw_wm_compile *c,
 	 break;
 
       case PROGRAM_STATE_VAR:
+      case PROGRAM_UNIFORM:
+      case PROGRAM_CONSTANT:
       case PROGRAM_NAMED_PARAM: {
 	 struct gl_program_parameter_list *plist = c->fp->program.Base.Parameters;
 	 
@@ -197,6 +200,7 @@ static const struct brw_wm_ref *pass0_get_reg( struct brw_wm_compile *c,
 	    break;
 	    
 	 case PROGRAM_STATE_VAR:
+	 case PROGRAM_UNIFORM:
 	    /* These may change from run to run:
 	     */
 	    ref = get_param_ref(c, &plist->ParameterValues[idx][component] );
diff --git a/src/mesa/drivers/dri/i965/brw_wm_pass1.c b/src/mesa/drivers/dri/i965/brw_wm_pass1.c
index d668def7007..26c044d4002 100644
--- a/src/mesa/drivers/dri/i965/brw_wm_pass1.c
+++ b/src/mesa/drivers/dri/i965/brw_wm_pass1.c
@@ -155,7 +155,11 @@ void brw_wm_pass1( struct brw_wm_compile *c )
 
       case OPCODE_SUB:
       case OPCODE_SLT:
+      case OPCODE_SLE:
       case OPCODE_SGE:
+      case OPCODE_SGT:
+      case OPCODE_SEQ:
+      case OPCODE_SNE:
       case OPCODE_ADD:
       case OPCODE_MAX:
       case OPCODE_MIN:
@@ -257,7 +261,6 @@ void brw_wm_pass1( struct brw_wm_compile *c )
       case OPCODE_DST:
       case OPCODE_TXP:
       default:
-	 assert(0);
 	 break;
       }
 
diff --git a/src/mesa/drivers/dri/i965/brw_wm_pass2.c b/src/mesa/drivers/dri/i965/brw_wm_pass2.c
index a1edbd6168d..c1ce6a9b6b5 100644
--- a/src/mesa/drivers/dri/i965/brw_wm_pass2.c
+++ b/src/mesa/drivers/dri/i965/brw_wm_pass2.c
@@ -328,7 +328,7 @@ void brw_wm_pass2( struct brw_wm_compile *c )
    c->state = PASS2_DONE;
 
    if (INTEL_DEBUG & DEBUG_WM) {
-      brw_wm_print_program(c, "pass2/done");
+       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 351de6d90ef..7856da0896f 100644
--- a/src/mesa/drivers/dri/i965/brw_wm_state.c
+++ b/src/mesa/drivers/dri/i965/brw_wm_state.c
@@ -34,6 +34,7 @@
 #include "brw_context.h"
 #include "brw_state.h"
 #include "brw_defines.h"
+#include "brw_wm.h"
 #include "bufmgr.h"
 
 /***********************************************************************
@@ -134,9 +135,13 @@ static void upload_wm_unit(struct brw_context *brw )
       if (fp->UsesKill || 
 	  brw->attribs.Color->AlphaEnabled) 
 	 wm.wm5.program_uses_killpixel = 1; 
+      
+      if (brw_wm_is_glsl(fp))
+	  wm.wm5.enable_8_pix = 1;
+      else
+	  wm.wm5.enable_16_pix = 1;
    }
 
-   wm.wm5.enable_16_pix = 1;
    wm.wm5.thread_dispatch_enable = 1;	/* AKA: color_write */
    wm.wm5.legacy_line_rast = 0;
    wm.wm5.legacy_global_depth_bias = 0;
diff --git a/src/mesa/drivers/dri/i965/intel_context.c b/src/mesa/drivers/dri/i965/intel_context.c
index 7ec316aa8ae..9850997aad1 100644
--- a/src/mesa/drivers/dri/i965/intel_context.c
+++ b/src/mesa/drivers/dri/i965/intel_context.c
@@ -84,6 +84,11 @@ int INTEL_DEBUG = (0);
 #define need_GL_EXT_multi_draw_arrays
 #define need_GL_EXT_secondary_color
 #define need_GL_EXT_point_parameters
+#define need_GL_VERSION_2_0
+#define need_GL_VERSION_2_1
+#define need_GL_ARB_shader_objects
+#define need_GL_ARB_vertex_shader
+
 #include "extension_helper.h"
 
 #ifndef VERBOSE
@@ -190,6 +195,13 @@ const struct dri_extension card_extensions[] =
     { "GL_MESA_ycbcr_texture",             NULL },
     { "GL_NV_blend_square",                NULL },
     { "GL_SGIS_generate_mipmap",           NULL },
+    { "GL_ARB_shading_language_100",       GL_VERSION_2_0_functions},
+    { "GL_ARB_shading_language_120",       GL_VERSION_2_1_functions},
+    { "GL_ARB_shader_objects",             GL_ARB_shader_objects_functions},
+    { "GL_ARB_vertex_shader",              GL_ARB_vertex_shader_functions},
+    { "GL_ARB_fragment_shader",            NULL },
+    /* XXX not implement yet, to compile builtin glsl lib */
+    { "GL_ARB_draw_buffers",               NULL },
     { NULL,                                NULL }
 };
 
@@ -559,6 +571,8 @@ void intelDestroyContext(__DRIcontextPrivate *driContextPriv)
 #endif
 
       /* free the Mesa context */
+      intel->ctx.VertexProgram.Current = NULL;
+      intel->ctx.FragmentProgram.Current = NULL;
       _mesa_destroy_context(&intel->ctx);
    }
 
diff --git a/src/mesa/main/texenvprogram.c b/src/mesa/main/texenvprogram.c
index 38be0390841..a2b9657bbb7 100644
--- a/src/mesa/main/texenvprogram.c
+++ b/src/mesa/main/texenvprogram.c
@@ -1242,7 +1242,7 @@ _mesa_UpdateTexEnvProgram( GLcontext *ctx )
 
    /* If a conventional fragment program/shader isn't in effect... */
    if (!ctx->FragmentProgram._Enabled &&
-       !ctx->Shader.CurrentProgram) {
+       (!ctx->Shader.CurrentProgram || !ctx->Shader.CurrentProgram->FragmentProgram)) {
       make_state_key(ctx, &key);
       hash = hash_key(&key);
       
diff --git a/src/mesa/shader/program.c b/src/mesa/shader/program.c
index 2097c395918..11330e24322 100644
--- a/src/mesa/shader/program.c
+++ b/src/mesa/shader/program.c
@@ -285,6 +285,8 @@ _mesa_init_vertex_program( GLcontext *ctx, struct gl_vertex_program *prog,
 struct gl_program *
 _mesa_new_program(GLcontext *ctx, GLenum target, GLuint id)
 {
+   if (ctx->Driver.NewProgram)
+        return ctx->Driver.NewProgram(ctx, target, id);
    switch (target) {
    case GL_VERTEX_PROGRAM_ARB: /* == GL_VERTEX_PROGRAM_NV */
       return _mesa_init_vertex_program(ctx, CALLOC_STRUCT(gl_vertex_program),
diff --git a/src/mesa/shader/slang/slang_codegen.c b/src/mesa/shader/slang/slang_codegen.c
index a44e42b6d6d..5191e0c222f 100644
--- a/src/mesa/shader/slang/slang_codegen.c
+++ b/src/mesa/shader/slang/slang_codegen.c
@@ -1847,6 +1847,7 @@ _slang_gen_var_decl(slang_assemble_ctx *A, slang_variable *var)
 
       n->Store->File = PROGRAM_TEMPORARY;
       n->Store->Size = _slang_sizeof_type_specifier(&n->Var->type.specifier);
+      A->program->NumTemporaries++;
       assert(n->Store->Size > 0);
    }
    return n;