X-Git-Url: https://git.libre-soc.org/?a=blobdiff_plain;f=src%2Fgallium%2Fdrivers%2Fsvga%2Fsvga_tgsi_insn.c;h=c7804b2680f4edc1022b64fa26a007c45f3c3c05;hb=81371a59093d59963a43b7f1becbed9d3c657e45;hp=2143546229370abd8829d0f0dde751ca36bf3348;hpb=384fd64ab1533df8256600733a1fc8413af3514a;p=mesa.git

diff --git a/src/gallium/drivers/svga/svga_tgsi_insn.c b/src/gallium/drivers/svga/svga_tgsi_insn.c
index 21435462293..c7804b2680f 100644
--- a/src/gallium/drivers/svga/svga_tgsi_insn.c
+++ b/src/gallium/drivers/svga/svga_tgsi_insn.c
@@ -29,6 +29,7 @@
 #include "tgsi/tgsi_parse.h"
 #include "util/u_memory.h"
 #include "util/u_math.h"
+#include "util/u_pstipple.h"
 
 #include "svga_tgsi_emit.h"
 #include "svga_context.h"
@@ -38,13 +39,11 @@ static boolean emit_vs_postamble( struct svga_shader_emitter *emit );
 static boolean emit_ps_postamble( struct svga_shader_emitter *emit );
 
 
-static unsigned
-translate_opcode(uint opcode)
+static SVGA3dShaderOpCodeType
+translate_opcode(enum tgsi_opcode opcode)
 {
    switch (opcode) {
-   case TGSI_OPCODE_ABS:        return SVGA3DOP_ABS;
    case TGSI_OPCODE_ADD:        return SVGA3DOP_ADD;
-   case TGSI_OPCODE_DP2A:       return SVGA3DOP_DP2ADD;
    case TGSI_OPCODE_DP3:        return SVGA3DOP_DP3;
    case TGSI_OPCODE_DP4:        return SVGA3DOP_DP4;
    case TGSI_OPCODE_FRC:        return SVGA3DOP_FRC;
@@ -54,17 +53,15 @@ translate_opcode(uint opcode)
    case TGSI_OPCODE_MOV:        return SVGA3DOP_MOV;
    case TGSI_OPCODE_MUL:        return SVGA3DOP_MUL;
    case TGSI_OPCODE_NOP:        return SVGA3DOP_NOP;
-   case TGSI_OPCODE_NRM4:       return SVGA3DOP_NRM;
    default:
-      debug_printf("Unkown opcode %u\n", opcode);
-      assert( 0 );
+      assert(!"svga: unexpected opcode in translate_opcode()");
       return SVGA3DOP_LAST_INST;
    }
 }
 
 
-static unsigned
-translate_file(unsigned file)
+static SVGA3dShaderRegType
+translate_file(enum tgsi_file_type file)
 {
    switch (file) {
    case TGSI_FILE_TEMPORARY: return SVGA3DREG_TEMP;
@@ -75,12 +72,17 @@ translate_file(unsigned file)
    case TGSI_FILE_SAMPLER:   return SVGA3DREG_SAMPLER;
    case TGSI_FILE_ADDRESS:   return SVGA3DREG_ADDR;
    default:
-      assert( 0 );
+      assert(!"svga: unexpected register file in translate_file()");
       return SVGA3DREG_TEMP;
    }
 }
 
 
+/**
+ * Translate a TGSI destination register to an SVGA3DShaderDestToken.
+ * \param insn  the TGSI instruction
+ * \param idx  which TGSI dest register to translate (usually (always?) zero)
+ */
 static SVGA3dShaderDestToken
 translate_dst_register( struct svga_shader_emitter *emit,
                         const struct tgsi_full_instruction *insn,
@@ -95,6 +97,7 @@ translate_dst_register( struct svga_shader_emitter *emit,
        * Need to lookup a table built at decl time:
        */
       dest = emit->output_map[reg->Register.Index];
+      emit->num_output_writes++;
       break;
 
    default:
@@ -161,7 +164,7 @@ scalar(struct src_register src, unsigned comp)
 static boolean
 svga_arl_needs_adjustment( const struct svga_shader_emitter *emit )
 {
-   int i;
+   unsigned i;
 
    for (i = 0; i < emit->num_arl_consts; ++i) {
       if (emit->arl_consts[i].arl_num == emit->current_arl)
@@ -174,7 +177,7 @@ svga_arl_needs_adjustment( const struct svga_shader_emitter *emit )
 static int
 svga_arl_adjustment( const struct svga_shader_emitter *emit )
 {
-   int i;
+   unsigned i;
 
    for (i = 0; i < emit->num_arl_consts; ++i) {
       if (emit->arl_consts[i].arl_num == emit->current_arl)
@@ -184,6 +187,9 @@ svga_arl_adjustment( const struct svga_shader_emitter *emit )
 }
 
 
+/**
+ * Translate a TGSI src register to a src_register.
+ */
 static struct src_register
 translate_src_register( const struct svga_shader_emitter *emit,
                         const struct tgsi_full_src_register *reg )
@@ -280,8 +286,10 @@ static SVGA3dShaderDestToken
 get_temp( struct svga_shader_emitter *emit )
 {
    int i = emit->nr_hw_temp + emit->internal_temp_count++;
-   assert(i < SVGA3D_TEMPREG_MAX);
-   i = MIN2(i, SVGA3D_TEMPREG_MAX - 1);
+   if (i >= SVGA3D_TEMPREG_MAX) {
+      debug_warn_once("svga: Too many temporary registers used in shader\n");
+      i = SVGA3D_TEMPREG_MAX - 1;
+   }
    return dst_register( SVGA3DREG_TEMP, i );
 }
 
@@ -300,6 +308,9 @@ release_temp( struct svga_shader_emitter *emit,
 }
 
 
+/**
+ * Release all temps.
+ */
 static void
 reset_temp_regs(struct svga_shader_emitter *emit)
 {
@@ -472,6 +483,9 @@ emit_repl(struct svga_shader_emitter *emit,
 }
 
 
+/**
+ * Submit/emit an instruction with zero operands.
+ */
 static boolean
 submit_op0(struct svga_shader_emitter *emit,
            SVGA3dShaderInstToken inst,
@@ -482,6 +496,9 @@ submit_op0(struct svga_shader_emitter *emit,
 }
 
 
+/**
+ * Submit/emit an instruction with one operand.
+ */
 static boolean
 submit_op1(struct svga_shader_emitter *emit,
            SVGA3dShaderInstToken inst,
@@ -493,6 +510,8 @@ submit_op1(struct svga_shader_emitter *emit,
 
 
 /**
+ * Submit/emit an instruction with two operands.
+ *
  * SVGA shaders may not refer to >1 constant register in a single
  * instruction.  This function checks for that usage and inserts a
  * move to temporary if detected.
@@ -543,6 +562,8 @@ submit_op2(struct svga_shader_emitter *emit,
 
 
 /**
+ * Submit/emit an instruction with three operands.
+ *
  * SVGA shaders may not refer to >1 constant register in a single
  * instruction.  This function checks for that usage and inserts a
  * move to temporary if detected.
@@ -613,6 +634,8 @@ submit_op3(struct svga_shader_emitter *emit,
 
 
 /**
+ * Submit/emit an instruction with four operands.
+ *
  * SVGA shaders may not refer to >1 constant register in a single
  * instruction.  This function checks for that usage and inserts a
  * move to temporary if detected.
@@ -644,6 +667,7 @@ submit_op4(struct svga_shader_emitter *emit,
     * in one slot at least:
     */
    assert(type1 == SVGA3DREG_SAMPLER);
+   (void) type1;
 
    if (type0 == SVGA3DREG_CONST &&
        ((type3 == SVGA3DREG_CONST && src0.base.num != src3.base.num) ||
@@ -706,42 +730,10 @@ alias_src_dst(struct src_register src,
 }
 
 
-static boolean
-submit_lrp(struct svga_shader_emitter *emit,
-           SVGA3dShaderDestToken dst,
-           struct src_register src0,
-           struct src_register src1,
-           struct src_register src2)
-{
-   SVGA3dShaderDestToken tmp;
-   boolean need_dst_tmp = FALSE;
-
-   /* The dst reg must be a temporary, and not be the same as src0 or src2 */
-   if (SVGA3dShaderGetRegType(dst.value) != SVGA3DREG_TEMP ||
-       alias_src_dst(src0, dst) ||
-       alias_src_dst(src2, dst))
-      need_dst_tmp = TRUE;
-
-   if (need_dst_tmp) {
-      tmp = get_temp( emit );
-      tmp.mask = dst.mask;
-   }
-   else {
-      tmp = dst;
-   }
-
-   if (!submit_op3(emit, inst_token( SVGA3DOP_LRP ), tmp, src0, src1, src2))
-      return FALSE;
-
-   if (need_dst_tmp) {
-      if (!submit_op1(emit, inst_token( SVGA3DOP_MOV ), dst, src( tmp )))
-         return FALSE;
-   }
-
-   return TRUE;
-}
-
-
+/**
+ * Helper for emitting SVGA immediate values using the SVGA3DOP_DEF[I]
+ * instructions.
+ */
 static boolean
 emit_def_const(struct svga_shader_emitter *emit,
                SVGA3dShaderConstType type,
@@ -774,28 +766,9 @@ emit_def_const(struct svga_shader_emitter *emit,
    }
 
    if (!emit_instruction(emit, opcode) ||
-       !svga_shader_emit_dwords( emit, def.values, Elements(def.values)))
-      return FALSE;
-
-   return TRUE;
-}
-
-
-static boolean
-create_zero_immediate( struct svga_shader_emitter *emit )
-{
-   unsigned idx = emit->nr_hw_float_const++;
-
-   /* Emit the constant (0, 0.5, -1, 1) and use swizzling to generate
-    * other useful vectors.
-    */
-   if (!emit_def_const( emit, SVGA3D_CONST_TYPE_FLOAT,
-                        idx, 0, 0.5, -1, 1 ))
+       !svga_shader_emit_dwords( emit, def.values, ARRAY_SIZE(def.values)))
       return FALSE;
 
-   emit->zero_immediate_idx = idx;
-   emit->created_zero_immediate = TRUE;
-
    return TRUE;
 }
 
@@ -871,30 +844,105 @@ get_vface( struct svga_shader_emitter *emit )
 
 
 /**
- * returns {0, 0, 0, 1} immediate
+ * Create/emit a "common" constant with values {0, 0.5, -1, 1}.
+ * We can swizzle this to produce other useful constants such as
+ * {0, 0, 0, 0}, {1, 1, 1, 1}, etc.
+ */
+static boolean
+create_common_immediate( struct svga_shader_emitter *emit )
+{
+   unsigned idx = emit->nr_hw_float_const++;
+
+   /* Emit the constant (0, 0.5, -1, 1) and use swizzling to generate
+    * other useful vectors.
+    */
+   if (!emit_def_const( emit, SVGA3D_CONST_TYPE_FLOAT,
+                        idx, 0.0f, 0.5f, -1.0f, 1.0f ))
+      return FALSE;
+   emit->common_immediate_idx[0] = idx;
+   idx++;
+
+   /* Emit constant {2, 0, 0, 0} (only the 2 is used for now) */
+   if (emit->key.vs.adjust_attrib_range) {
+      if (!emit_def_const( emit, SVGA3D_CONST_TYPE_FLOAT,
+                           idx, 2.0f, 0.0f, 0.0f, 0.0f ))
+         return FALSE;
+      emit->common_immediate_idx[1] = idx;
+   }
+   else {
+      emit->common_immediate_idx[1] = -1;
+   }
+
+   emit->created_common_immediate = TRUE;
+
+   return TRUE;
+}
+
+
+/**
+ * Return swizzle/position for the given value in the "common" immediate.
+ */
+static inline unsigned
+common_immediate_swizzle(float value)
+{
+   if (value == 0.0f)
+      return TGSI_SWIZZLE_X;
+   else if (value == 0.5f)
+      return TGSI_SWIZZLE_Y;
+   else if (value == -1.0f)
+      return TGSI_SWIZZLE_Z;
+   else if (value == 1.0f)
+      return TGSI_SWIZZLE_W;
+   else {
+      assert(!"illegal value in common_immediate_swizzle");
+      return TGSI_SWIZZLE_X;
+   }
+}
+
+
+/**
+ * Returns an immediate reg where all the terms are either 0, 1, 2 or 0.5
+ */
+static struct src_register
+get_immediate(struct svga_shader_emitter *emit,
+              float x, float y, float z, float w)
+{
+   unsigned sx = common_immediate_swizzle(x);
+   unsigned sy = common_immediate_swizzle(y);
+   unsigned sz = common_immediate_swizzle(z);
+   unsigned sw = common_immediate_swizzle(w);
+   assert(emit->created_common_immediate);
+   assert(emit->common_immediate_idx[0] >= 0);
+   return swizzle(src_register(SVGA3DREG_CONST, emit->common_immediate_idx[0]),
+                  sx, sy, sz, sw);
+}
+
+
+/**
+ * returns {0, 0, 0, 0} immediate
  */
 static struct src_register
 get_zero_immediate( struct svga_shader_emitter *emit )
 {
-   assert(emit->created_zero_immediate);
-   assert(emit->zero_immediate_idx >= 0);
+   assert(emit->created_common_immediate);
+   assert(emit->common_immediate_idx[0] >= 0);
    return swizzle(src_register( SVGA3DREG_CONST,
-                                emit->zero_immediate_idx),
-                  0, 0, 0, 3);
+                                emit->common_immediate_idx[0]),
+                  0, 0, 0, 0);
 }
 
 
 /**
- * returns {1, 1, 1, -1} immediate
+ * returns {1, 1, 1, 1} immediate
  */
 static struct src_register
-get_pos_neg_one_immediate( struct svga_shader_emitter *emit )
+get_one_immediate( struct svga_shader_emitter *emit )
 {
-   assert(emit->created_zero_immediate);
-   assert(emit->zero_immediate_idx >= 0);
+   assert(emit->created_common_immediate);
+   assert(emit->common_immediate_idx[0] >= 0);
    return swizzle(src_register( SVGA3DREG_CONST,
-                                emit->zero_immediate_idx),
-                  3, 3, 3, 2);
+                                emit->common_immediate_idx[0]),
+                  3, 3, 3, 3);
 }
 
 
@@ -904,13 +952,28 @@ get_pos_neg_one_immediate( struct svga_shader_emitter *emit )
 static struct src_register
 get_half_immediate( struct svga_shader_emitter *emit )
 {
-   assert(emit->created_zero_immediate);
-   assert(emit->zero_immediate_idx >= 0);
-   return swizzle(src_register(SVGA3DREG_CONST, emit->zero_immediate_idx),
+   assert(emit->created_common_immediate);
+   assert(emit->common_immediate_idx[0] >= 0);
+   return swizzle(src_register(SVGA3DREG_CONST, emit->common_immediate_idx[0]),
                   1, 1, 1, 1);
 }
 
 
+/**
+ * returns {2, 2, 2, 2} immediate
+ */
+static struct src_register
+get_two_immediate( struct svga_shader_emitter *emit )
+{
+   /* Note we use the second common immediate here */
+   assert(emit->created_common_immediate);
+   assert(emit->common_immediate_idx[1] >= 0);
+   return swizzle(src_register( SVGA3DREG_CONST,
+                                emit->common_immediate_idx[1]),
+                  0, 0, 0, 0);
+}
+
+
 /**
  * returns the loop const
  */
@@ -943,8 +1006,8 @@ get_fake_arl_const( struct svga_shader_emitter *emit )
 
 
 /**
- * Return the register which holds the current dimenions of the
- * texture bound to the given sampler
+ * Return a register which holds the width and height of the texture
+ * currently bound to the given sampler.
  */
 static struct src_register
 get_tex_dimensions( struct svga_shader_emitter *emit, int sampler_num )
@@ -953,7 +1016,7 @@ get_tex_dimensions( struct svga_shader_emitter *emit, int sampler_num )
    struct src_register reg;
 
    /* the width/height indexes start right after constants */
-   idx = emit->key.fkey.tex[sampler_num].width_height_idx +
+   idx = emit->key.tex[sampler_num].width_height_idx +
          emit->info.file_max[TGSI_FILE_CONSTANT] + 1;
 
    reg = src_register( SVGA3DREG_CONST, idx );
@@ -993,11 +1056,10 @@ emit_if(struct svga_shader_emitter *emit,
 {
    struct src_register src0 =
       translate_src_register(emit, &insn->Src[0]);
-   struct src_register zero = get_zero_immediate( emit );
+   struct src_register zero = get_zero_immediate(emit);
    SVGA3dShaderInstToken if_token = inst_token( SVGA3DOP_IFC );
 
    if_token.control = SVGA3DOPCOMPC_NE;
-   zero = scalar(zero, TGSI_SWIZZLE_X);
 
    if (SVGA3dShaderGetRegType(src0.base.value) == SVGA3DREG_CONST) {
       /*
@@ -1020,20 +1082,20 @@ emit_if(struct svga_shader_emitter *emit,
 
 
 static boolean
-emit_endif(struct svga_shader_emitter *emit,
-           const struct tgsi_full_instruction *insn)
+emit_else(struct svga_shader_emitter *emit,
+          const struct tgsi_full_instruction *insn)
 {
-   emit->dynamic_branching_level--;
-
-   return emit_instruction(emit, inst_token(SVGA3DOP_ENDIF));
+   return emit_instruction(emit, inst_token(SVGA3DOP_ELSE));
 }
 
 
 static boolean
-emit_else(struct svga_shader_emitter *emit,
-          const struct tgsi_full_instruction *insn)
+emit_endif(struct svga_shader_emitter *emit,
+           const struct tgsi_full_instruction *insn)
 {
-   return emit_instruction(emit, inst_token(SVGA3DOP_ELSE));
+   emit->dynamic_branching_level--;
+
+   return emit_instruction(emit, inst_token(SVGA3DOP_ENDIF));
 }
 
 
@@ -1112,7 +1174,7 @@ emit_div(struct svga_shader_emitter *emit,
    const struct src_register src1 =
       translate_src_register(emit, &insn->Src[1] );
    SVGA3dShaderDestToken temp = get_temp( emit );
-   int i;
+   unsigned i;
 
    /* For each enabled element, perform a RCP instruction.  Note that
     * RCP is scalar in SVGA3D:
@@ -1175,72 +1237,8 @@ emit_dp2(struct svga_shader_emitter *emit,
 
 
 /**
- * Translate the following TGSI DPH instruction.
- *    DPH  DST, SRC1, SRC2
- * To the following SVGA3D instruction sequence.
- *    DP3  TMP, SRC1, SRC2
- *    ADD  DST, TMP, SRC2.wwww
+ * Sine / Cosine helper function.
  */
-static boolean
-emit_dph(struct svga_shader_emitter *emit,
-         const struct tgsi_full_instruction *insn )
-{
-   SVGA3dShaderDestToken dst = translate_dst_register( emit, insn, 0 );
-   const struct src_register src0 = translate_src_register(
-      emit, &insn->Src[0] );
-   struct src_register src1 =
-      translate_src_register(emit, &insn->Src[1]);
-   SVGA3dShaderDestToken temp = get_temp( emit );
-
-   /* DP3  TMP, SRC1, SRC2 */
-   if (!submit_op2( emit, inst_token( SVGA3DOP_DP3 ), temp, src0, src1 ))
-      return FALSE;
-
-   src1 = scalar(src1, TGSI_SWIZZLE_W);
-
-   /* ADD  DST, TMP, SRC2.wwww */
-   if (!submit_op2( emit, inst_token( SVGA3DOP_ADD ), dst,
-                    src( temp ), src1 ))
-      return FALSE;
-
-   return TRUE;
-}
-
-
-/**
- * Translate the following TGSI DST instruction.
- *    NRM  DST, SRC
- * To the following SVGA3D instruction sequence.
- *    DP3  TMP, SRC, SRC
- *    RSQ  TMP, TMP
- *    MUL  DST, SRC, TMP
- */
-static boolean
-emit_nrm(struct svga_shader_emitter *emit,
-         const struct tgsi_full_instruction *insn)
-{
-   SVGA3dShaderDestToken dst = translate_dst_register( emit, insn, 0 );
-   const struct src_register src0 =
-      translate_src_register(emit, &insn->Src[0]);
-   SVGA3dShaderDestToken temp = get_temp( emit );
-
-   /* DP3  TMP, SRC, SRC */
-   if (!submit_op2( emit, inst_token( SVGA3DOP_DP3 ), temp, src0, src0 ))
-      return FALSE;
-
-   /* RSQ  TMP, TMP */
-   if (!submit_op1( emit, inst_token( SVGA3DOP_RSQ ), temp, src( temp )))
-      return FALSE;
-
-   /* MUL  DST, SRC, TMP */
-   if (!submit_op2( emit, inst_token( SVGA3DOP_MUL ), dst,
-                    src0, src( temp )))
-      return FALSE;
-
-   return TRUE;
-}
-
-
 static boolean
 do_emit_sincos(struct svga_shader_emitter *emit,
                SVGA3dShaderDestToken dst,
@@ -1251,27 +1249,8 @@ do_emit_sincos(struct svga_shader_emitter *emit,
 }
 
 
-static boolean
-emit_sincos(struct svga_shader_emitter *emit,
-            const struct tgsi_full_instruction *insn)
-{
-   SVGA3dShaderDestToken dst = translate_dst_register( emit, insn, 0 );
-   struct src_register src0 = translate_src_register(emit, &insn->Src[0]);
-   SVGA3dShaderDestToken temp = get_temp( emit );
-
-   /* SCS TMP SRC */
-   if (!do_emit_sincos(emit, writemask(temp, TGSI_WRITEMASK_XY), src0 ))
-      return FALSE;
-
-   /* MOV DST TMP */
-   if (!submit_op1( emit, inst_token( SVGA3DOP_MOV ), dst, src( temp ) ))
-      return FALSE;
-
-   return TRUE;
-}
-
-
 /**
+ * Translate TGSI SIN instruction into:
  * SCS TMP SRC
  * MOV DST TMP.yyyy
  */
@@ -1297,7 +1276,9 @@ emit_sin(struct svga_shader_emitter *emit,
    return TRUE;
 }
 
+
 /*
+ * Translate TGSI COS instruction into:
  * SCS TMP SRC
  * MOV DST TMP.xxxx
  */
@@ -1324,6 +1305,9 @@ emit_cos(struct svga_shader_emitter *emit,
 }
 
 
+/**
+ * Translate/emit TGSI SSG (Set Sign: -1, 0, +1) instruction.
+ */
 static boolean
 emit_ssg(struct svga_shader_emitter *emit,
          const struct tgsi_full_instruction *insn)
@@ -1341,9 +1325,8 @@ emit_ssg(struct svga_shader_emitter *emit,
                          src( temp0 ), src( temp1 ) );
    }
 
-   zero = get_zero_immediate( emit );
-   one = scalar( zero, TGSI_SWIZZLE_W );
-   zero = scalar( zero, TGSI_SWIZZLE_X );
+   one = get_one_immediate(emit);
+   zero = get_zero_immediate(emit);
 
    /* CMP  TMP0, SRC0, one, zero */
    if (!submit_op3( emit, inst_token( SVGA3DOP_CMP ),
@@ -1363,28 +1346,8 @@ emit_ssg(struct svga_shader_emitter *emit,
 
 
 /**
- * ADD DST SRC0, negate(SRC0)
+ * Translate/emit KILL_IF instruction (kill if any of X,Y,Z,W are negative).
  */
-static boolean
-emit_sub(struct svga_shader_emitter *emit,
-         const struct tgsi_full_instruction *insn)
-{
-   SVGA3dShaderDestToken dst = translate_dst_register( emit, insn, 0 );
-   struct src_register src0 = translate_src_register(
-      emit, &insn->Src[0] );
-   struct src_register src1 = translate_src_register(
-      emit, &insn->Src[1] );
-
-   src1 = negate(src1);
-
-   if (!submit_op2( emit, inst_token( SVGA3DOP_ADD ), dst,
-                    src0, src1 ))
-      return FALSE;
-
-   return TRUE;
-}
-
-
 static boolean
 emit_kill_if(struct svga_shader_emitter *emit,
              const struct tgsi_full_instruction *insn)
@@ -1426,15 +1389,15 @@ emit_kill_if(struct svga_shader_emitter *emit,
 
 
 /**
- * unconditional kill
+ * Translate/emit unconditional kill instruction (usually found inside
+ * an IF/ELSE/ENDIF block).
  */
 static boolean
 emit_kill(struct svga_shader_emitter *emit,
           const struct tgsi_full_instruction *insn)
 {
    SVGA3dShaderDestToken temp;
-   struct src_register one = scalar( get_zero_immediate( emit ),
-                                     TGSI_SWIZZLE_W );
+   struct src_register one = get_one_immediate(emit);
    SVGA3dShaderInstToken inst = inst_token( SVGA3DOP_TEXKILL );
 
    /* texkill doesn't allow negation on the operand so lets move
@@ -1461,7 +1424,8 @@ same_register(struct src_register r1, struct src_register r2)
 
 
 
-/* Implement conditionals by initializing destination reg to 'fail',
+/**
+ * Implement conditionals by initializing destination reg to 'fail',
  * then set predicate reg with UFOP_SETP, then move 'pass' to dest
  * based on predicate reg.
  *
@@ -1471,7 +1435,7 @@ same_register(struct src_register r1, struct src_register r2)
  */
 static boolean
 emit_conditional(struct svga_shader_emitter *emit,
-                 unsigned compare_func,
+                 enum pipe_compare_func compare_func,
                  SVGA3dShaderDestToken dst,
                  struct src_register src0,
                  struct src_register src1,
@@ -1479,8 +1443,7 @@ emit_conditional(struct svga_shader_emitter *emit,
                  struct src_register fail)
 {
    SVGA3dShaderDestToken pred_reg = dst_register( SVGA3DREG_PREDICATE, 0 );
-   SVGA3dShaderInstToken setp_token, mov_token;
-   setp_token = inst_token( SVGA3DOP_SETP );
+   SVGA3dShaderInstToken setp_token;
 
    switch (compare_func) {
    case PIPE_FUNC_NEVER:
@@ -1488,22 +1451,22 @@ emit_conditional(struct svga_shader_emitter *emit,
                          dst, fail );
       break;
    case PIPE_FUNC_LESS:
-      setp_token.control = SVGA3DOPCOMP_LT;
+      setp_token = inst_token_setp(SVGA3DOPCOMP_LT);
       break;
    case PIPE_FUNC_EQUAL:
-      setp_token.control = SVGA3DOPCOMP_EQ;
+      setp_token = inst_token_setp(SVGA3DOPCOMP_EQ);
       break;
    case PIPE_FUNC_LEQUAL:
-      setp_token.control = SVGA3DOPCOMP_LE;
+      setp_token = inst_token_setp(SVGA3DOPCOMP_LE);
       break;
    case PIPE_FUNC_GREATER:
-      setp_token.control = SVGA3DOPCOMP_GT;
+      setp_token = inst_token_setp(SVGA3DOPCOMP_GT);
       break;
    case PIPE_FUNC_NOTEQUAL:
-      setp_token.control = SVGA3DOPCOMPC_NE;
+      setp_token = inst_token_setp(SVGA3DOPCOMPC_NE);
       break;
    case PIPE_FUNC_GEQUAL:
-      setp_token.control = SVGA3DOPCOMP_GE;
+      setp_token = inst_token_setp(SVGA3DOPCOMP_GE);
       break;
    case PIPE_FUNC_ALWAYS:
       return submit_op1( emit, inst_token( SVGA3DOP_MOV ),
@@ -1526,11 +1489,8 @@ emit_conditional(struct svga_shader_emitter *emit,
                     src0, src1 ))
       return FALSE;
 
-   mov_token = inst_token( SVGA3DOP_MOV );
-
    /* MOV dst, fail */
-   if (!submit_op1( emit, mov_token, dst,
-                    fail ))
+   if (!submit_op1(emit, inst_token(SVGA3DOP_MOV), dst, fail))
       return FALSE;
 
    /* MOV dst, pass (predicated)
@@ -1538,18 +1498,25 @@ emit_conditional(struct svga_shader_emitter *emit,
     * Note that the predicate reg (and possible modifiers) is passed
     * as the first source argument.
     */
-   mov_token.predicated = 1;
-   if (!submit_op2( emit, mov_token, dst,
-                    src( pred_reg ), pass ))
+   if (!submit_op2(emit,
+                   inst_token_predicated(SVGA3DOP_MOV), dst,
+                   src(pred_reg), pass))
       return FALSE;
 
    return TRUE;
 }
 
 
+/**
+ * Helper for emiting 'selection' commands.  Basically:
+ * if (src0 OP src1)
+ *    dst = 1.0;
+ * else
+ *    dst = 0.0;
+ */
 static boolean
 emit_select(struct svga_shader_emitter *emit,
-            unsigned compare_func,
+            enum pipe_compare_func compare_func,
             SVGA3dShaderDestToken dst,
             struct src_register src0,
             struct src_register src1 )
@@ -1577,21 +1544,17 @@ emit_select(struct svga_shader_emitter *emit,
    {
       struct src_register one, zero;
       /* zero immediate is 0,0,0,1 */
-      zero = get_zero_immediate( emit );
-      one  = scalar( zero, TGSI_SWIZZLE_W );
-      zero = scalar( zero, TGSI_SWIZZLE_X );
+      zero = get_zero_immediate(emit);
+      one = get_one_immediate(emit);
 
-      return emit_conditional(
-         emit,
-         compare_func,
-         dst,
-         src0,
-         src1,
-         one, zero);
+      return emit_conditional(emit, compare_func, dst, src0, src1, one, zero);
    }
 }
 
 
+/**
+ * Translate/emit a TGSI SEQ, SNE, SLT, SGE, etc. instruction.
+ */
 static boolean
 emit_select_op(struct svga_shader_emitter *emit,
                unsigned compare,
@@ -1608,7 +1571,8 @@ emit_select_op(struct svga_shader_emitter *emit,
 
 
 /**
- * Translate TGSI CMP instruction.
+ * Translate TGSI CMP instruction.  Component-wise:
+ * dst = (src0 < 0.0) ? src1 : src2
  */
 static boolean
 emit_cmp(struct svga_shader_emitter *emit,
@@ -1623,8 +1587,7 @@ emit_cmp(struct svga_shader_emitter *emit,
       translate_src_register(emit, &insn->Src[2] );
 
    if (emit->unit == PIPE_SHADER_VERTEX) {
-      struct src_register zero =
-         scalar(get_zero_immediate(emit), TGSI_SWIZZLE_X);
+      struct src_register zero = get_zero_immediate(emit);
       /* We used to simulate CMP with SLT+LRP.  But that didn't work when
        * src1 or src2 was Inf/NaN.  In particular, GLSL sqrt(0) failed
        * because it involves a CMP to handle the 0 case.
@@ -1644,7 +1607,7 @@ emit_cmp(struct svga_shader_emitter *emit,
 
 
 /**
- * Translate texture instructions to SVGA3D representation.
+ * Translate/emit 2-operand (coord, sampler) texture instructions.
  */
 static boolean
 emit_tex2(struct svga_shader_emitter *emit,
@@ -1681,7 +1644,7 @@ emit_tex2(struct svga_shader_emitter *emit,
    texcoord = translate_src_register( emit, &insn->Src[0] );
    sampler = translate_src_register( emit, &insn->Src[1] );
 
-   if (emit->key.fkey.tex[sampler.base.num].unnormalized ||
+   if (emit->key.tex[sampler.base.num].unnormalized ||
        emit->dynamic_branching_level > 0)
       tmp = get_temp( emit );
 
@@ -1691,7 +1654,7 @@ emit_tex2(struct svga_shader_emitter *emit,
    if (emit->dynamic_branching_level > 0 &&
        inst.op == SVGA3DOP_TEX &&
        SVGA3dShaderGetRegType(texcoord.base.value) == SVGA3DREG_TEMP) {
-      struct src_register zero = get_zero_immediate( emit );
+      struct src_register zero = get_zero_immediate(emit);
 
       /* MOV  tmp, texcoord */
       if (!submit_op1( emit,
@@ -1704,7 +1667,7 @@ emit_tex2(struct svga_shader_emitter *emit,
       if (!submit_op1( emit,
                        inst_token( SVGA3DOP_MOV ),
                        writemask( tmp, TGSI_WRITEMASK_W ),
-                       scalar( zero, TGSI_SWIZZLE_X )))
+                       zero ))
          return FALSE;
 
       texcoord = src( tmp );
@@ -1713,7 +1676,7 @@ emit_tex2(struct svga_shader_emitter *emit,
 
    /* Explicit normalization of texcoords:
     */
-   if (emit->key.fkey.tex[sampler.base.num].unnormalized) {
+   if (emit->key.tex[sampler.base.num].unnormalized) {
       struct src_register wh = get_tex_dimensions( emit, sampler.base.num );
 
       /* MUL  tmp, SRC0, WH */
@@ -1729,7 +1692,7 @@ emit_tex2(struct svga_shader_emitter *emit,
 
 
 /**
- * Translate texture instructions to SVGA3D representation.
+ * Translate/emit 4-operand (coord, ddx, ddy, sampler) texture instructions.
  */
 static boolean
 emit_tex4(struct svga_shader_emitter *emit,
@@ -1763,7 +1726,8 @@ emit_tex4(struct svga_shader_emitter *emit,
 
 
 /**
- * Emit texture swizzle code.
+ * Emit texture swizzle code.  We do this here since SVGA samplers don't
+ * directly support swizzles.
  */
 static boolean
 emit_tex_swizzle(struct svga_shader_emitter *emit,
@@ -1777,15 +1741,15 @@ emit_tex_swizzle(struct svga_shader_emitter *emit,
    const unsigned swizzleIn[4] = {swizzle_x, swizzle_y, swizzle_z, swizzle_w};
    unsigned srcSwizzle[4];
    unsigned srcWritemask = 0x0, zeroWritemask = 0x0, oneWritemask = 0x0;
-   int i;
+   unsigned i;
 
    /* build writemasks and srcSwizzle terms */
    for (i = 0; i < 4; i++) {
-      if (swizzleIn[i] == PIPE_SWIZZLE_ZERO) {
+      if (swizzleIn[i] == PIPE_SWIZZLE_0) {
          srcSwizzle[i] = TGSI_SWIZZLE_X + i;
          zeroWritemask |= (1 << i);
       }
-      else if (swizzleIn[i] == PIPE_SWIZZLE_ONE) {
+      else if (swizzleIn[i] == PIPE_SWIZZLE_1) {
          srcSwizzle[i] = TGSI_SWIZZLE_X + i;
          oneWritemask |= (1 << i);
       }
@@ -1813,7 +1777,7 @@ emit_tex_swizzle(struct svga_shader_emitter *emit,
       if (!submit_op1(emit,
                       inst_token(SVGA3DOP_MOV),
                       writemask(dst, zeroWritemask),
-                      scalar(get_zero_immediate(emit), TGSI_SWIZZLE_X)))
+                      get_zero_immediate(emit)))
          return FALSE;
    }
 
@@ -1822,7 +1786,7 @@ emit_tex_swizzle(struct svga_shader_emitter *emit,
       if (!submit_op1(emit,
                       inst_token(SVGA3DOP_MOV),
                       writemask(dst, oneWritemask),
-                      scalar(get_zero_immediate(emit), TGSI_SWIZZLE_W)))
+                      get_one_immediate(emit)))
          return FALSE;
    }
 
@@ -1830,6 +1794,9 @@ emit_tex_swizzle(struct svga_shader_emitter *emit,
 }
 
 
+/**
+ * Translate/emit a TGSI texture sample instruction.
+ */
 static boolean
 emit_tex(struct svga_shader_emitter *emit,
          const struct tgsi_full_instruction *insn)
@@ -1845,16 +1812,16 @@ emit_tex(struct svga_shader_emitter *emit,
    const unsigned unit = src1.base.num;
 
    /* check for shadow samplers */
-   boolean compare = (emit->key.fkey.tex[unit].compare_mode ==
+   boolean compare = (emit->key.tex[unit].compare_mode ==
                       PIPE_TEX_COMPARE_R_TO_TEXTURE);
 
    /* texture swizzle */
-   boolean swizzle = (emit->key.fkey.tex[unit].swizzle_r != PIPE_SWIZZLE_RED ||
-                      emit->key.fkey.tex[unit].swizzle_g != PIPE_SWIZZLE_GREEN ||
-                      emit->key.fkey.tex[unit].swizzle_b != PIPE_SWIZZLE_BLUE ||
-                      emit->key.fkey.tex[unit].swizzle_a != PIPE_SWIZZLE_ALPHA);
+   boolean swizzle = (emit->key.tex[unit].swizzle_r != PIPE_SWIZZLE_X ||
+                      emit->key.tex[unit].swizzle_g != PIPE_SWIZZLE_Y ||
+                      emit->key.tex[unit].swizzle_b != PIPE_SWIZZLE_Z ||
+                      emit->key.tex[unit].swizzle_a != PIPE_SWIZZLE_W);
 
-   boolean saturate = insn->Instruction.Saturate != TGSI_SAT_NONE;
+   boolean saturate = insn->Instruction.Saturate;
 
    /* If doing compare processing or tex swizzle or saturation, we need to put
     * the fetched color into a temporary so it can be used as a source later on.
@@ -1919,7 +1886,7 @@ emit_tex(struct svga_shader_emitter *emit,
 
          /* Compare texture sample value against R component of texcoord */
          if (!emit_select(emit,
-                          emit->key.fkey.tex[unit].compare_func,
+                          emit->key.tex[unit].compare_func,
                           writemask( dst2, TGSI_WRITEMASK_XYZ ),
                           r_coord,
                           tex_src_x))
@@ -1927,8 +1894,7 @@ emit_tex(struct svga_shader_emitter *emit,
       }
 
       if (dst.mask & TGSI_WRITEMASK_W) {
-         struct src_register one =
-            scalar( get_zero_immediate( emit ), TGSI_SWIZZLE_W );
+         struct src_register one = get_one_immediate(emit);
 
         if (!submit_op1( emit, inst_token( SVGA3DOP_MOV ),
                          writemask( dst2, TGSI_WRITEMASK_W ),
@@ -1946,10 +1912,10 @@ emit_tex(struct svga_shader_emitter *emit,
       /* swizzle from tex_result to dst (handles saturation too, if any) */
       emit_tex_swizzle(emit,
                        dst, src(tex_result),
-                       emit->key.fkey.tex[unit].swizzle_r,
-                       emit->key.fkey.tex[unit].swizzle_g,
-                       emit->key.fkey.tex[unit].swizzle_b,
-                       emit->key.fkey.tex[unit].swizzle_a);
+                       emit->key.tex[unit].swizzle_r,
+                       emit->key.tex[unit].swizzle_g,
+                       emit->key.tex[unit].swizzle_b,
+                       emit->key.tex[unit].swizzle_a);
    }
 
    return TRUE;
@@ -1957,8 +1923,8 @@ emit_tex(struct svga_shader_emitter *emit,
 
 
 static boolean
-emit_bgnloop2(struct svga_shader_emitter *emit,
-              const struct tgsi_full_instruction *insn)
+emit_bgnloop(struct svga_shader_emitter *emit,
+             const struct tgsi_full_instruction *insn)
 {
    SVGA3dShaderInstToken inst = inst_token( SVGA3DOP_LOOP );
    struct src_register loop_reg = src_register( SVGA3DREG_LOOP, 0 );
@@ -1973,8 +1939,8 @@ emit_bgnloop2(struct svga_shader_emitter *emit,
 
 
 static boolean
-emit_endloop2(struct svga_shader_emitter *emit,
-              const struct tgsi_full_instruction *insn)
+emit_endloop(struct svga_shader_emitter *emit,
+             const struct tgsi_full_instruction *insn)
 {
    SVGA3dShaderInstToken inst = inst_token( SVGA3DOP_ENDLOOP );
 
@@ -1984,6 +1950,9 @@ emit_endloop2(struct svga_shader_emitter *emit,
 }
 
 
+/**
+ * Translate/emit TGSI BREAK (out of loop) instruction.
+ */
 static boolean
 emit_brk(struct svga_shader_emitter *emit,
          const struct tgsi_full_instruction *insn)
@@ -1993,9 +1962,13 @@ emit_brk(struct svga_shader_emitter *emit,
 }
 
 
+/**
+ * Emit simple instruction which operates on one scalar value (not
+ * a vector).  Ex: LG2, RCP, RSQ.
+ */
 static boolean
 emit_scalar_op1(struct svga_shader_emitter *emit,
-                unsigned opcode,
+                SVGA3dShaderOpCodeType opcode,
                 const struct tgsi_full_instruction *insn)
 {
    SVGA3dShaderInstToken inst;
@@ -2011,9 +1984,13 @@ emit_scalar_op1(struct svga_shader_emitter *emit,
 }
 
 
+/**
+ * Translate/emit a simple instruction (one which has no special-case
+ * code) such as ADD, MUL, MIN, MAX.
+ */
 static boolean
 emit_simple_instruction(struct svga_shader_emitter *emit,
-                        unsigned opcode,
+                        SVGA3dShaderOpCodeType opcode,
                         const struct tgsi_full_instruction *insn)
 {
    const struct tgsi_full_src_register *src = insn->Src;
@@ -2045,6 +2022,32 @@ emit_simple_instruction(struct svga_shader_emitter *emit,
 }
 
 
+/**
+ * TGSI_OPCODE_MOVE is only special-cased here to detect the
+ * svga_fragment_shader::constant_color_output case.
+ */
+static boolean
+emit_mov(struct svga_shader_emitter *emit,
+         const struct tgsi_full_instruction *insn)
+{
+   const struct tgsi_full_src_register *src = &insn->Src[0];
+   const struct tgsi_full_dst_register *dst = &insn->Dst[0];
+
+   if (emit->unit == PIPE_SHADER_FRAGMENT &&
+       dst->Register.File == TGSI_FILE_OUTPUT &&
+       dst->Register.Index == 0 &&
+       src->Register.File == TGSI_FILE_CONSTANT &&
+       !src->Register.Indirect) {
+      emit->constant_color_output = TRUE;
+   }
+
+   return emit_simple_instruction(emit, SVGA3DOP_MOV, insn);
+}
+
+
+/**
+ * Translate/emit TGSI DDX, DDY instructions.
+ */
 static boolean
 emit_deriv(struct svga_shader_emitter *emit,
            const struct tgsi_full_instruction *insn )
@@ -2052,7 +2055,6 @@ emit_deriv(struct svga_shader_emitter *emit,
    if (emit->dynamic_branching_level > 0 &&
        insn->Src[0].Register.File == TGSI_FILE_TEMPORARY)
    {
-      struct src_register zero = get_zero_immediate( emit );
       SVGA3dShaderDestToken dst =
          translate_dst_register( emit, insn, 0 );
 
@@ -2062,13 +2064,13 @@ emit_deriv(struct svga_shader_emitter *emit,
       if (!submit_op1(emit,
                       inst_token( SVGA3DOP_MOV ),
                       dst,
-                      scalar(zero, TGSI_SWIZZLE_X)))
+                      get_zero_immediate(emit)))
          return FALSE;
 
       return TRUE;
    }
    else {
-      unsigned opcode;
+      SVGA3dShaderOpCodeType opcode;
       const struct tgsi_full_src_register *reg = &insn->Src[0];
       SVGA3dShaderInstToken inst;
       SVGA3dShaderDestToken dst;
@@ -2104,6 +2106,11 @@ emit_deriv(struct svga_shader_emitter *emit,
 }
 
 
+/**
+ * Translate/emit ARL (Address Register Load) instruction.  Used to
+ * move a value into the special 'address' register.  Used to implement
+ * indirect/variable indexing into arrays.
+ */
 static boolean
 emit_arl(struct svga_shader_emitter *emit,
          const struct tgsi_full_instruction *insn)
@@ -2165,53 +2172,38 @@ emit_pow(struct svga_shader_emitter *emit,
 }
 
 
+/**
+ * Emit a LRP (linear interpolation) instruction.
+ */
 static boolean
-emit_xpd(struct svga_shader_emitter *emit,
-         const struct tgsi_full_instruction *insn)
+submit_lrp(struct svga_shader_emitter *emit,
+           SVGA3dShaderDestToken dst,
+           struct src_register src0,
+           struct src_register src1,
+           struct src_register src2)
 {
-   SVGA3dShaderDestToken dst = translate_dst_register( emit, insn, 0 );
-   const struct src_register src0 = translate_src_register(
-      emit, &insn->Src[0] );
-   const struct src_register src1 = translate_src_register(
-      emit, &insn->Src[1] );
+   SVGA3dShaderDestToken tmp;
    boolean need_dst_tmp = FALSE;
 
-   /* XPD can only output to a temporary */
-   if (SVGA3dShaderGetRegType(dst.value) != SVGA3DREG_TEMP)
-      need_dst_tmp = TRUE;
-
-   /* The dst reg must not be the same as src0 or src1*/
-   if (alias_src_dst(src0, dst) ||
-       alias_src_dst(src1, dst))
+   /* The dst reg must be a temporary, and not be the same as src0 or src2 */
+   if (SVGA3dShaderGetRegType(dst.value) != SVGA3DREG_TEMP ||
+       alias_src_dst(src0, dst) ||
+       alias_src_dst(src2, dst))
       need_dst_tmp = TRUE;
 
    if (need_dst_tmp) {
-      SVGA3dShaderDestToken tmp = get_temp( emit );
-
-      /* Obey DX9 restrictions on mask:
-       */
-      tmp.mask = dst.mask & TGSI_WRITEMASK_XYZ;
-
-      if (!submit_op2(emit, inst_token( SVGA3DOP_CRS ), tmp, src0, src1))
-         return FALSE;
-
-      if (!submit_op1(emit, inst_token( SVGA3DOP_MOV ), dst, src( tmp )))
-         return FALSE;
+      tmp = get_temp( emit );
+      tmp.mask = dst.mask;
    }
    else {
-      if (!submit_op2(emit, inst_token( SVGA3DOP_CRS ), dst, src0, src1))
-         return FALSE;
+      tmp = dst;
    }
 
-   /* Need to emit 1.0 to dst.w?
-    */
-   if (dst.mask & TGSI_WRITEMASK_W) {
-      struct src_register zero = get_zero_immediate( emit );
+   if (!submit_op3(emit, inst_token( SVGA3DOP_LRP ), tmp, src0, src1, src2))
+      return FALSE;
 
-      if (!submit_op1(emit,
-                      inst_token( SVGA3DOP_MOV ),
-                      writemask(dst, TGSI_WRITEMASK_W),
-                      zero))
+   if (need_dst_tmp) {
+      if (!submit_op1(emit, inst_token( SVGA3DOP_MOV ), dst, src( tmp )))
          return FALSE;
    }
 
@@ -2219,6 +2211,9 @@ emit_xpd(struct svga_shader_emitter *emit,
 }
 
 
+/**
+ * Translate/emit LRP (Linear Interpolation) instruction.
+ */
 static boolean
 emit_lrp(struct svga_shader_emitter *emit,
          const struct tgsi_full_instruction *insn)
@@ -2234,7 +2229,9 @@ emit_lrp(struct svga_shader_emitter *emit,
    return submit_lrp(emit, dst, src0, src1, src2);
 }
 
-
+/**
+ * Translate/emit DST (Distance function) instruction.
+ */
 static boolean
 emit_dst_insn(struct svga_shader_emitter *emit,
               const struct tgsi_full_instruction *insn)
@@ -2256,7 +2253,6 @@ emit_dst_insn(struct svga_shader_emitter *emit,
          emit, &insn->Src[0] );
       const struct src_register src1 = translate_src_register(
          emit, &insn->Src[1] );
-      struct src_register zero = get_zero_immediate( emit );
       boolean need_tmp = FALSE;
 
       if (SVGA3dShaderGetRegType(dst.value) != SVGA3DREG_TEMP ||
@@ -2276,7 +2272,7 @@ emit_dst_insn(struct svga_shader_emitter *emit,
       if (tmp.mask & TGSI_WRITEMASK_XW) {
          if (!submit_op1( emit, inst_token( SVGA3DOP_MOV ),
                           writemask(tmp, TGSI_WRITEMASK_XW ),
-                          scalar( zero, 3 )))
+                          get_one_immediate(emit)))
             return FALSE;
       }
 
@@ -2320,7 +2316,6 @@ emit_exp(struct svga_shader_emitter *emit,
    SVGA3dShaderDestToken dst = translate_dst_register( emit, insn, 0 );
    struct src_register src0 =
       translate_src_register( emit, &insn->Src[0] );
-   struct src_register zero = get_zero_immediate( emit );
    SVGA3dShaderDestToken fraction;
 
    if (dst.mask & TGSI_WRITEMASK_Y)
@@ -2371,7 +2366,7 @@ emit_exp(struct svga_shader_emitter *emit,
    if (dst.mask & TGSI_WRITEMASK_W) {
       if (!submit_op1( emit, inst_token( SVGA3DOP_MOV ),
                        writemask(dst, TGSI_WRITEMASK_W),
-                       scalar( zero, TGSI_SWIZZLE_W ) ))
+                       get_one_immediate(emit)))
          return FALSE;
    }
 
@@ -2379,6 +2374,9 @@ emit_exp(struct svga_shader_emitter *emit,
 }
 
 
+/**
+ * Translate/emit LIT (Lighting helper) instruction.
+ */
 static boolean
 emit_lit(struct svga_shader_emitter *emit,
          const struct tgsi_full_instruction *insn)
@@ -2412,7 +2410,6 @@ emit_lit(struct svga_shader_emitter *emit,
       SVGA3dShaderDestToken tmp = get_temp( emit );
       const struct src_register src0 = translate_src_register(
          emit, &insn->Src[0] );
-      struct src_register zero = get_zero_immediate( emit );
 
       /* tmp = pow(src.y, src.w)
        */
@@ -2438,14 +2435,8 @@ emit_lit(struct svga_shader_emitter *emit,
        */
       {
          SVGA3dShaderDestToken pred_reg = dst_register( SVGA3DREG_PREDICATE, 0 );
-         SVGA3dShaderInstToken setp_token, mov_token;
          struct src_register predsrc;
 
-         setp_token = inst_token( SVGA3DOP_SETP );
-         mov_token = inst_token( SVGA3DOP_MOV );
-
-         setp_token.control = SVGA3DOPCOMP_GT;
-
          /* D3D vs GL semantics:
           */
          if (0)
@@ -2454,14 +2445,16 @@ emit_lit(struct svga_shader_emitter *emit,
             predsrc = swizzle(src0, 0, 0, 0, 0); /* GL */
 
          /* SETP src0.xxyy, GT, {0}.x */
-         if (!submit_op2( emit, setp_token, pred_reg,
+         if (!submit_op2( emit,
+                          inst_token_setp(SVGA3DOPCOMP_GT),
+                          pred_reg,
                           predsrc,
-                          swizzle(zero, 0, 0, 0, 0) ))
+                          get_zero_immediate(emit)))
             return FALSE;
 
          /* MOV dst, fail */
          if (!submit_op1( emit, inst_token( SVGA3DOP_MOV ), dst,
-                          swizzle(zero, 3, 0, 0, 3 )))
+                          get_immediate(emit, 1.0f, 0.0f, 0.0f, 1.0f)))
              return FALSE;
 
          /* MOV dst.yz, tmp (predicated)
@@ -2470,8 +2463,8 @@ emit_lit(struct svga_shader_emitter *emit,
           * as the first source argument.
           */
          if (dst.mask & TGSI_WRITEMASK_YZ) {
-            mov_token.predicated = 1;
-            if (!submit_op2( emit, mov_token,
+            if (!submit_op2( emit,
+                             inst_token_predicated(SVGA3DOP_MOV),
                              writemask(dst, TGSI_WRITEMASK_YZ),
                              src( pred_reg ), src( tmp ) ))
                return FALSE;
@@ -2518,7 +2511,6 @@ emit_log(struct svga_shader_emitter *emit,
    SVGA3dShaderDestToken dst = translate_dst_register( emit, insn, 0 );
    struct src_register src0 =
       translate_src_register( emit, &insn->Src[0] );
-   struct src_register zero = get_zero_immediate( emit );
    SVGA3dShaderDestToken abs_tmp;
    struct src_register abs_src0;
    SVGA3dShaderDestToken log2_abs;
@@ -2610,7 +2602,7 @@ emit_log(struct svga_shader_emitter *emit,
    if (dst.mask & TGSI_WRITEMASK_W) {
       if (!submit_op1( emit, inst_token( SVGA3DOP_MOV ),
                        writemask(dst, TGSI_WRITEMASK_W),
-                       scalar( zero, TGSI_SWIZZLE_W ) ))
+                       get_one_immediate(emit)))
          return FALSE;
    }
 
@@ -2692,6 +2684,9 @@ emit_trunc_round(struct svga_shader_emitter *emit,
 }
 
 
+/**
+ * Translate/emit "begin subroutine" instruction/marker/label.
+ */
 static boolean
 emit_bgnsub(struct svga_shader_emitter *emit,
             unsigned position,
@@ -2718,6 +2713,9 @@ emit_bgnsub(struct svga_shader_emitter *emit,
 }
 
 
+/**
+ * Translate/emit subroutine call instruction.
+ */
 static boolean
 emit_call(struct svga_shader_emitter *emit,
           const struct tgsi_full_instruction *insn)
@@ -2730,7 +2728,7 @@ emit_call(struct svga_shader_emitter *emit,
          break;
    }
 
-   if (emit->nr_labels == Elements(emit->label))
+   if (emit->nr_labels == ARRAY_SIZE(emit->label))
       return FALSE;
 
    if (i == emit->nr_labels) {
@@ -2759,7 +2757,9 @@ emit_end(struct svga_shader_emitter *emit)
 }
 
 
-
+/**
+ * Translate any TGSI instruction to SVGA.
+ */
 static boolean
 svga_emit_instruction(struct svga_shader_emitter *emit,
                       unsigned position,
@@ -2811,21 +2811,12 @@ svga_emit_instruction(struct svga_shader_emitter *emit,
    case TGSI_OPCODE_DP2:
       return emit_dp2( emit, insn );
 
-   case TGSI_OPCODE_DPH:
-      return emit_dph( emit, insn );
-
-   case TGSI_OPCODE_NRM:
-      return emit_nrm( emit, insn );
-
    case TGSI_OPCODE_COS:
       return emit_cos( emit, insn );
 
    case TGSI_OPCODE_SIN:
       return emit_sin( emit, insn );
 
-   case TGSI_OPCODE_SCS:
-      return emit_sincos( emit, insn );
-
    case TGSI_OPCODE_END:
       /* TGSI always finishes the main func with an END */
       return emit_end( emit );
@@ -2854,9 +2845,6 @@ svga_emit_instruction(struct svga_shader_emitter *emit,
    case TGSI_OPCODE_SLE:
       return emit_select_op( emit, PIPE_FUNC_LEQUAL, insn );
 
-   case TGSI_OPCODE_SUB:
-      return emit_sub( emit, insn );
-
    case TGSI_OPCODE_POW:
       return emit_pow( emit, insn );
 
@@ -2892,7 +2880,6 @@ svga_emit_instruction(struct svga_shader_emitter *emit,
       /* These aren't actually used by any of the frontends we care
        * about:
        */
-   case TGSI_OPCODE_CLAMP:
    case TGSI_OPCODE_AND:
    case TGSI_OPCODE_OR:
    case TGSI_OPCODE_I2F:
@@ -2910,15 +2897,12 @@ svga_emit_instruction(struct svga_shader_emitter *emit,
       return emit_endif( emit, insn );
 
    case TGSI_OPCODE_BGNLOOP:
-      return emit_bgnloop2( emit, insn );
+      return emit_bgnloop( emit, insn );
    case TGSI_OPCODE_ENDLOOP:
-      return emit_endloop2( emit, insn );
+      return emit_endloop( emit, insn );
    case TGSI_OPCODE_BRK:
       return emit_brk( emit, insn );
 
-   case TGSI_OPCODE_XPD:
-      return emit_xpd( emit, insn );
-
    case TGSI_OPCODE_KILL:
       return emit_kill( emit, insn );
 
@@ -2934,9 +2918,13 @@ svga_emit_instruction(struct svga_shader_emitter *emit,
    case TGSI_OPCODE_SSG:
       return emit_ssg( emit, insn );
 
+   case TGSI_OPCODE_MOV:
+      return emit_mov( emit, insn );
+
    default:
       {
-         unsigned opcode = translate_opcode(insn->Instruction.Opcode);
+         SVGA3dShaderOpCodeType opcode =
+            translate_opcode(insn->Instruction.Opcode);
 
          if (opcode == SVGA3DOP_LAST_INST)
             return FALSE;
@@ -2950,9 +2938,13 @@ svga_emit_instruction(struct svga_shader_emitter *emit,
 }
 
 
+/**
+ * Translate/emit a TGSI IMMEDIATE declaration.
+ * An immediate vector is a constant that's hard-coded into the shader.
+ */
 static boolean
 svga_emit_immediate(struct svga_shader_emitter *emit,
-                    struct tgsi_full_immediate *imm)
+                    const struct tgsi_full_immediate *imm)
 {
    static const float id[4] = {0,0,0,1};
    float value[4];
@@ -2964,6 +2956,9 @@ svga_emit_immediate(struct svga_shader_emitter *emit,
       value[i] = util_is_inf_or_nan(f) ? 0.0f : f;
    }
 
+   /* If the immediate has less than four values, fill in the remaining
+    * positions from id={0,0,0,1}.
+    */
    for ( ; i < 4; i++ )
       value[i] = id[i];
 
@@ -2990,10 +2985,13 @@ make_immediate(struct svga_shader_emitter *emit,
 }
 
 
+/**
+ * Emit special VS instructions at top of shader.
+ */
 static boolean
 emit_vs_preamble(struct svga_shader_emitter *emit)
 {
-   if (!emit->key.vkey.need_prescale) {
+   if (!emit->key.vs.need_prescale) {
       if (!make_immediate( emit, 0, 0, .5, .5,
                            &emit->imm_0055))
          return FALSE;
@@ -3003,6 +3001,9 @@ emit_vs_preamble(struct svga_shader_emitter *emit)
 }
 
 
+/**
+ * Emit special PS instructions at top of shader.
+ */
 static boolean
 emit_ps_preamble(struct svga_shader_emitter *emit)
 {
@@ -3040,6 +3041,9 @@ emit_ps_preamble(struct svga_shader_emitter *emit)
 }
 
 
+/**
+ * Emit special PS instructions at end of shader.
+ */
 static boolean
 emit_ps_postamble(struct svga_shader_emitter *emit)
 {
@@ -3064,9 +3068,8 @@ emit_ps_postamble(struct svga_shader_emitter *emit)
           * logicop workaround.
           */
          if (emit->unit == PIPE_SHADER_FRAGMENT &&
-             emit->key.fkey.white_fragments) {
-            struct src_register one = scalar( get_zero_immediate( emit ),
-                                              TGSI_SWIZZLE_W );
+             emit->key.fs.white_fragments) {
+            struct src_register one = get_one_immediate(emit);
 
             if (!submit_op1( emit,
                              inst_token(SVGA3DOP_MOV),
@@ -3074,6 +3077,15 @@ emit_ps_postamble(struct svga_shader_emitter *emit)
                              one ))
                return FALSE;
          }
+         else if (emit->unit == PIPE_SHADER_FRAGMENT &&
+                  i < emit->key.fs.write_color0_to_n_cbufs) {
+            /* Write temp color output [0] to true output [i] */
+            if (!submit_op1(emit, inst_token(SVGA3DOP_MOV),
+                            emit->true_color_output[i],
+                            src(emit->temp_color_output[0]))) {
+               return FALSE;
+            }
+         }
          else {
             if (!submit_op1( emit,
                              inst_token(SVGA3DOP_MOV),
@@ -3088,6 +3100,9 @@ emit_ps_postamble(struct svga_shader_emitter *emit)
 }
 
 
+/**
+ * Emit special VS instructions at end of shader.
+ */
 static boolean
 emit_vs_postamble(struct svga_shader_emitter *emit)
 {
@@ -3107,7 +3122,7 @@ emit_vs_postamble(struct svga_shader_emitter *emit)
    /* Need to perform various manipulations on vertex position to cope
     * with the different GL and D3D clip spaces.
     */
-   if (emit->key.vkey.need_prescale) {
+   if (emit->key.vs.need_prescale) {
       SVGA3dShaderDestToken temp_pos = emit->temp_pos;
       SVGA3dShaderDestToken depth = emit->depth_pos;
       SVGA3dShaderDestToken pos = emit->true_pos;
@@ -3190,6 +3205,7 @@ emit_vs_postamble(struct svga_shader_emitter *emit)
 /**
  * For the pixel shader: emit the code which chooses the front
  * or back face color depending on triangle orientation.
+ * This happens at the top of the fragment shader.
  *
  *  0: IF VFACE :4
  *  1:   COLOR = FrontColor;
@@ -3205,14 +3221,14 @@ emit_light_twoside(struct svga_shader_emitter *emit)
    struct src_register back[2];
    SVGA3dShaderDestToken color[2];
    int count = emit->internal_color_count;
-   int i;
+   unsigned i;
    SVGA3dShaderInstToken if_token;
 
    if (count == 0)
       return TRUE;
 
    vface = get_vface( emit );
-   zero = get_zero_immediate( emit );
+   zero = get_zero_immediate(emit);
 
    /* Can't use get_temp() to allocate the color reg as such
     * temporaries will be reclaimed after each instruction by the call
@@ -3234,13 +3250,11 @@ emit_light_twoside(struct svga_shader_emitter *emit)
 
    if_token = inst_token( SVGA3DOP_IFC );
 
-   if (emit->key.fkey.front_ccw)
+   if (emit->key.fs.front_ccw)
       if_token.control = SVGA3DOPCOMP_LT;
    else
       if_token.control = SVGA3DOPCOMP_GT;
 
-   zero = scalar(zero, TGSI_SWIZZLE_X);
-
    if (!(emit_instruction( emit, if_token ) &&
          emit_src( emit, vface ) &&
          emit_src( emit, zero ) ))
@@ -3267,18 +3281,18 @@ emit_light_twoside(struct svga_shader_emitter *emit)
 
 
 /**
+ * Emit special setup code for the front/back face register in the FS.
  *  0: SETP_GT TEMP, VFACE, 0
  *  where TEMP is a fake frontface register
  */
 static boolean
 emit_frontface(struct svga_shader_emitter *emit)
 {
-   struct src_register vface, zero;
+   struct src_register vface;
    SVGA3dShaderDestToken temp;
    struct src_register pass, fail;
 
    vface = get_vface( emit );
-   zero = get_zero_immediate( emit );
 
    /* Can't use get_temp() to allocate the fake frontface reg as such
     * temporaries will be reclaimed after each instruction by the call
@@ -3287,16 +3301,16 @@ emit_frontface(struct svga_shader_emitter *emit)
    temp = dst_register( SVGA3DREG_TEMP,
                         emit->nr_hw_temp++ );
 
-   if (emit->key.fkey.front_ccw) {
-      pass = scalar( zero, TGSI_SWIZZLE_X );
-      fail = scalar( zero, TGSI_SWIZZLE_W );
+   if (emit->key.fs.front_ccw) {
+      pass = get_zero_immediate(emit);
+      fail = get_one_immediate(emit);
    } else {
-      pass = scalar( zero, TGSI_SWIZZLE_W );
-      fail = scalar( zero, TGSI_SWIZZLE_X );
+      pass = get_one_immediate(emit);
+      fail = get_zero_immediate(emit);
    }
 
    if (!emit_conditional(emit, PIPE_FUNC_GREATER,
-                         temp, vface, scalar( zero, TGSI_SWIZZLE_X ),
+                         temp, vface, get_zero_immediate(emit),
                          pass, fail))
       return FALSE;
 
@@ -3316,8 +3330,6 @@ emit_frontface(struct svga_shader_emitter *emit)
 static boolean
 emit_inverted_texcoords(struct svga_shader_emitter *emit)
 {
-   struct src_register zero = get_zero_immediate(emit);
-   struct src_register pos_neg_one = get_pos_neg_one_immediate( emit );
    unsigned inverted_texcoords = emit->inverted_texcoords;
 
    while (inverted_texcoords) {
@@ -3325,20 +3337,20 @@ emit_inverted_texcoords(struct svga_shader_emitter *emit)
 
       assert(emit->inverted_texcoords & (1 << unit));
 
-      assert(unit < Elements(emit->ps_true_texcoord));
+      assert(unit < ARRAY_SIZE(emit->ps_true_texcoord));
 
-      assert(unit < Elements(emit->ps_inverted_texcoord_input));
+      assert(unit < ARRAY_SIZE(emit->ps_inverted_texcoord_input));
 
       assert(emit->ps_inverted_texcoord_input[unit]
-             < Elements(emit->input_map));
+             < ARRAY_SIZE(emit->input_map));
 
       /* inverted = coord * (1, -1, 1, 1) + (0, 1, 0, 0) */
       if (!submit_op3(emit,
                       inst_token(SVGA3DOP_MAD),
                       dst(emit->ps_inverted_texcoord[unit]),
                       emit->ps_true_texcoord[unit],
-                      swizzle(pos_neg_one, 0, 3, 0, 0),  /* (1, -1, 1, 1) */
-                      swizzle(zero, 0, 3, 0, 0)))  /* (0, 1, 0, 0) */
+                      get_immediate(emit, 1.0f, -1.0f, 1.0f, 1.0f),
+                      get_immediate(emit, 0.0f, 1.0f, 0.0f, 0.0f)))
          return FALSE;
 
       /* Reassign the input_map entry to the new texcoord register */
@@ -3352,16 +3364,112 @@ emit_inverted_texcoords(struct svga_shader_emitter *emit)
 }
 
 
+/**
+ * Emit code to adjust vertex shader inputs/attributes:
+ * - Change range from [0,1] to [-1,1] (for normalized byte/short attribs).
+ * - Set attrib W component = 1.
+ */
 static boolean
-needs_to_create_zero( struct svga_shader_emitter *emit )
+emit_adjusted_vertex_attribs(struct svga_shader_emitter *emit)
+{
+   unsigned adjust_mask = (emit->key.vs.adjust_attrib_range |
+                           emit->key.vs.adjust_attrib_w_1);
+ 
+   while (adjust_mask) {
+      /* Adjust vertex attrib range and/or set W component = 1 */
+      const unsigned index = u_bit_scan(&adjust_mask);
+      struct src_register tmp;
+
+      /* allocate a temp reg */
+      tmp = src_register(SVGA3DREG_TEMP, emit->nr_hw_temp);
+      emit->nr_hw_temp++;
+
+      if (emit->key.vs.adjust_attrib_range & (1 << index)) {
+         /* The vertex input/attribute is supposed to be a signed value in
+          * the range [-1,1] but we actually fetched/converted it to the
+          * range [0,1].  This most likely happens when the app specifies a
+          * signed byte attribute but we interpreted it as unsigned bytes.
+          * See also svga_translate_vertex_format().
+          *
+          * Here, we emit some extra instructions to adjust
+          * the attribute values from [0,1] to [-1,1].
+          *
+          * The adjustment we implement is:
+          *   new_attrib = attrib * 2.0;
+          *   if (attrib >= 0.5)
+          *      new_attrib = new_attrib - 2.0;
+          * This isn't exactly right (it's off by a bit or so) but close enough.
+          */
+         SVGA3dShaderDestToken pred_reg = dst_register(SVGA3DREG_PREDICATE, 0);
+
+         /* tmp = attrib * 2.0 */
+         if (!submit_op2(emit,
+                         inst_token(SVGA3DOP_MUL),
+                         dst(tmp),
+                         emit->input_map[index],
+                         get_two_immediate(emit)))
+            return FALSE;
+
+         /* pred = (attrib >= 0.5) */
+         if (!submit_op2(emit,
+                         inst_token_setp(SVGA3DOPCOMP_GE),
+                         pred_reg,
+                         emit->input_map[index],  /* vert attrib */
+                         get_half_immediate(emit)))  /* 0.5 */
+            return FALSE;
+
+         /* sub(pred) tmp, tmp, 2.0 */
+         if (!submit_op3(emit,
+                         inst_token_predicated(SVGA3DOP_SUB),
+                         dst(tmp),
+                         src(pred_reg),
+                         tmp,
+                         get_two_immediate(emit)))
+            return FALSE;
+      }
+      else {
+         /* just copy the vertex input attrib to the temp register */
+         if (!submit_op1(emit,
+                         inst_token(SVGA3DOP_MOV),
+                         dst(tmp),
+                         emit->input_map[index]))
+            return FALSE;
+      }
+
+      if (emit->key.vs.adjust_attrib_w_1 & (1 << index)) {
+         /* move 1 into W position of tmp */
+         if (!submit_op1(emit,
+                         inst_token(SVGA3DOP_MOV),
+                         writemask(dst(tmp), TGSI_WRITEMASK_W),
+                         get_one_immediate(emit)))
+            return FALSE;
+      }
+
+      /* Reassign the input_map entry to the new tmp register */
+      emit->input_map[index] = tmp;
+   }
+
+   return TRUE;
+}
+
+
+/**
+ * Determine if we need to create the "common" immediate value which is
+ * used for generating useful vector constants such as {0,0,0,0} and
+ * {1,1,1,1}.
+ * We could just do this all the time except that we want to conserve
+ * registers whenever possible.
+ */
+static boolean
+needs_to_create_common_immediate(const struct svga_shader_emitter *emit)
 {
    unsigned i;
 
    if (emit->unit == PIPE_SHADER_FRAGMENT) {
-      if (emit->key.fkey.light_twoside)
+      if (emit->key.fs.light_twoside)
          return TRUE;
 
-      if (emit->key.fkey.white_fragments)
+      if (emit->key.fs.white_fragments)
          return TRUE;
 
       if (emit->emit_frontface)
@@ -3375,25 +3483,27 @@ needs_to_create_zero( struct svga_shader_emitter *emit )
       if (emit->inverted_texcoords)
          return TRUE;
 
-      /* look for any PIPE_SWIZZLE_ZERO/ONE terms */
-      for (i = 0; i < emit->key.fkey.num_textures; i++) {
-         if (emit->key.fkey.tex[i].swizzle_r > PIPE_SWIZZLE_ALPHA ||
-             emit->key.fkey.tex[i].swizzle_g > PIPE_SWIZZLE_ALPHA ||
-             emit->key.fkey.tex[i].swizzle_b > PIPE_SWIZZLE_ALPHA ||
-             emit->key.fkey.tex[i].swizzle_a > PIPE_SWIZZLE_ALPHA)
+      /* look for any PIPE_SWIZZLE_0/ONE terms */
+      for (i = 0; i < emit->key.num_textures; i++) {
+         if (emit->key.tex[i].swizzle_r > PIPE_SWIZZLE_W ||
+             emit->key.tex[i].swizzle_g > PIPE_SWIZZLE_W ||
+             emit->key.tex[i].swizzle_b > PIPE_SWIZZLE_W ||
+             emit->key.tex[i].swizzle_a > PIPE_SWIZZLE_W)
             return TRUE;
       }
 
-      for (i = 0; i < emit->key.fkey.num_textures; i++) {
-         if (emit->key.fkey.tex[i].compare_mode
+      for (i = 0; i < emit->key.num_textures; i++) {
+         if (emit->key.tex[i].compare_mode
              == PIPE_TEX_COMPARE_R_TO_TEXTURE)
             return TRUE;
       }
    }
-
-   if (emit->unit == PIPE_SHADER_VERTEX) {
+   else if (emit->unit == PIPE_SHADER_VERTEX) {
       if (emit->info.opcode_count[TGSI_OPCODE_CMP] >= 1)
          return TRUE;
+      if (emit->key.vs.adjust_attrib_range ||
+          emit->key.vs.adjust_attrib_w_1)
+         return TRUE;
    }
 
    if (emit->info.opcode_count[TGSI_OPCODE_IF] >= 1 ||
@@ -3409,7 +3519,6 @@ needs_to_create_zero( struct svga_shader_emitter *emit )
        emit->info.opcode_count[TGSI_OPCODE_SEQ] >= 1 ||
        emit->info.opcode_count[TGSI_OPCODE_EXP] >= 1 ||
        emit->info.opcode_count[TGSI_OPCODE_LOG] >= 1 ||
-       emit->info.opcode_count[TGSI_OPCODE_XPD] >= 1 ||
        emit->info.opcode_count[TGSI_OPCODE_KILL] >= 1)
       return TRUE;
 
@@ -3417,15 +3526,18 @@ needs_to_create_zero( struct svga_shader_emitter *emit )
 }
 
 
+/**
+ * Do we need to create a looping constant?
+ */
 static boolean
-needs_to_create_loop_const( struct svga_shader_emitter *emit )
+needs_to_create_loop_const(const struct svga_shader_emitter *emit)
 {
    return (emit->info.opcode_count[TGSI_OPCODE_BGNLOOP] >= 1);
 }
 
 
 static boolean
-needs_to_create_arl_consts( struct svga_shader_emitter *emit )
+needs_to_create_arl_consts(const struct svga_shader_emitter *emit)
 {
    return (emit->num_arl_consts > 0);
 }
@@ -3435,7 +3547,7 @@ static boolean
 pre_parse_add_indirect( struct svga_shader_emitter *emit,
                         int num, int current_arl)
 {
-   int i;
+   unsigned i;
    assert(num < 0);
 
    for (i = 0; i < emit->num_arl_consts; ++i) {
@@ -3523,8 +3635,8 @@ pre_parse_tokens( struct svga_shader_emitter *emit,
 static boolean
 svga_shader_emit_helpers(struct svga_shader_emitter *emit)
 {
-   if (needs_to_create_zero( emit )) {
-      create_zero_immediate( emit );
+   if (needs_to_create_common_immediate( emit )) {
+      create_common_immediate( emit );
    }
    if (needs_to_create_loop_const( emit )) {
       create_loop_const( emit );
@@ -3534,10 +3646,13 @@ svga_shader_emit_helpers(struct svga_shader_emitter *emit)
    }
 
    if (emit->unit == PIPE_SHADER_FRAGMENT) {
+      if (!svga_shader_emit_samplers_decl( emit ))
+         return FALSE;
+
       if (!emit_ps_preamble( emit ))
          return FALSE;
 
-      if (emit->key.fkey.light_twoside) {
+      if (emit->key.fs.light_twoside) {
          if (!emit_light_twoside( emit ))
             return FALSE;
       }
@@ -3550,20 +3665,54 @@ svga_shader_emit_helpers(struct svga_shader_emitter *emit)
             return FALSE;
       }
    }
+   else {
+      assert(emit->unit == PIPE_SHADER_VERTEX);
+      if (emit->key.vs.adjust_attrib_range) {
+         if (!emit_adjusted_vertex_attribs(emit) ||
+             emit->key.vs.adjust_attrib_w_1) {
+            return FALSE;
+         }
+      }
+   }
 
    return TRUE;
 }
 
 
+/**
+ * This is the main entrypoint into the TGSI instruction translater.
+ * Translate TGSI shader tokens into an SVGA shader.
+ */
 boolean
 svga_shader_emit_instructions(struct svga_shader_emitter *emit,
                               const struct tgsi_token *tokens)
 {
    struct tgsi_parse_context parse;
+   const struct tgsi_token *new_tokens = NULL;
    boolean ret = TRUE;
    boolean helpers_emitted = FALSE;
    unsigned line_nr = 0;
 
+   if (emit->unit == PIPE_SHADER_FRAGMENT && emit->key.fs.pstipple) {
+      unsigned unit;
+
+      new_tokens = util_pstipple_create_fragment_shader(tokens, &unit, 0,
+                                                        TGSI_FILE_INPUT);
+
+      if (new_tokens) {
+         /* Setup texture state for stipple */
+         emit->sampler_target[unit] = TGSI_TEXTURE_2D;
+         emit->key.tex[unit].swizzle_r = TGSI_SWIZZLE_X;
+         emit->key.tex[unit].swizzle_g = TGSI_SWIZZLE_Y;
+         emit->key.tex[unit].swizzle_b = TGSI_SWIZZLE_Z;
+         emit->key.tex[unit].swizzle_a = TGSI_SWIZZLE_W;
+
+         emit->pstipple_sampler_unit = unit;
+
+         tokens = new_tokens;
+      }
+   }
+
    tgsi_parse_init( &parse, tokens );
    emit->internal_imm_count = 0;
 
@@ -3630,5 +3779,9 @@ svga_shader_emit_instructions(struct svga_shader_emitter *emit,
 
 done:
    tgsi_parse_free( &parse );
+   if (new_tokens) {
+      tgsi_free_tokens(new_tokens);
+   }
+
    return ret;
 }