Fix promotion of floats to doubles
[mesa.git] / src / mesa / program / program.c
index 4f28e2a3b547052f00c1a4972557b7d87d89e6b6..754e3d480b81fec49fa7e53a6284fe4cd4fca9e8 100644 (file)
 
 #include "main/glheader.h"
 #include "main/context.h"
+#include "main/framebuffer.h"
 #include "main/hash.h"
 #include "main/macros.h"
+#include "main/shaderobj.h"
 #include "program.h"
 #include "prog_cache.h"
 #include "prog_parameter.h"
 #include "prog_instruction.h"
+#include "util/bitscan.h"
 #include "util/ralloc.h"
+#include "util/u_atomic.h"
 
 
 /**
@@ -86,21 +90,17 @@ _mesa_init_program(struct gl_context *ctx)
    ctx->VertexProgram.PointSizeEnabled =
       (ctx->API == API_OPENGLES2) ? GL_TRUE : GL_FALSE;
    ctx->VertexProgram.TwoSideEnabled = GL_FALSE;
-   _mesa_reference_vertprog(ctx, &ctx->VertexProgram.Current,
-                            ctx->Shared->DefaultVertexProgram);
+   _mesa_reference_program(ctx, &ctx->VertexProgram.Current,
+                           ctx->Shared->DefaultVertexProgram);
    assert(ctx->VertexProgram.Current);
    ctx->VertexProgram.Cache = _mesa_new_program_cache();
 
    ctx->FragmentProgram.Enabled = GL_FALSE;
-   _mesa_reference_fragprog(ctx, &ctx->FragmentProgram.Current,
-                            ctx->Shared->DefaultFragmentProgram);
+   _mesa_reference_program(ctx, &ctx->FragmentProgram.Current,
+                           ctx->Shared->DefaultFragmentProgram);
    assert(ctx->FragmentProgram.Current);
    ctx->FragmentProgram.Cache = _mesa_new_program_cache();
-
-   ctx->GeometryProgram.Enabled = GL_FALSE;
-   /* right now by default we don't have a geometry program */
-   _mesa_reference_geomprog(ctx, &ctx->GeometryProgram.Current,
-                            NULL);
+   ctx->VertexProgram._VPMode = VP_MODE_FF;
 
    /* XXX probably move this stuff */
    ctx->ATIFragmentShader.Enabled = GL_FALSE;
@@ -116,11 +116,10 @@ _mesa_init_program(struct gl_context *ctx)
 void
 _mesa_free_program_data(struct gl_context *ctx)
 {
-   _mesa_reference_vertprog(ctx, &ctx->VertexProgram.Current, NULL);
+   _mesa_reference_program(ctx, &ctx->VertexProgram.Current, NULL);
    _mesa_delete_program_cache(ctx, ctx->VertexProgram.Cache);
-   _mesa_reference_fragprog(ctx, &ctx->FragmentProgram.Current, NULL);
+   _mesa_reference_program(ctx, &ctx->FragmentProgram.Current, NULL);
    _mesa_delete_shader_cache(ctx, ctx->FragmentProgram.Cache);
-   _mesa_reference_geomprog(ctx, &ctx->GeometryProgram.Current, NULL);
 
    /* XXX probably move this stuff */
    if (ctx->ATIFragmentShader.Current) {
@@ -142,17 +141,14 @@ _mesa_free_program_data(struct gl_context *ctx)
 void
 _mesa_update_default_objects_program(struct gl_context *ctx)
 {
-   _mesa_reference_vertprog(ctx, &ctx->VertexProgram.Current,
-                            ctx->Shared->DefaultVertexProgram);
+   _mesa_reference_program(ctx, &ctx->VertexProgram.Current,
+                           ctx->Shared->DefaultVertexProgram);
    assert(ctx->VertexProgram.Current);
 
-   _mesa_reference_fragprog(ctx, &ctx->FragmentProgram.Current,
+   _mesa_reference_program(ctx, &ctx->FragmentProgram.Current,
                             ctx->Shared->DefaultFragmentProgram);
    assert(ctx->FragmentProgram.Current);
 
-   _mesa_reference_geomprog(ctx, &ctx->GeometryProgram.Current,
-                      ctx->Shared->DefaultGeometryProgram);
-
    /* XXX probably move this stuff */
    if (ctx->ATIFragmentShader.Current) {
       ctx->ATIFragmentShader.Current->RefCount--;
@@ -181,132 +177,42 @@ _mesa_set_program_error(struct gl_context *ctx, GLint pos, const char *string)
 }
 
 
-/**
- * Find the line number and column for 'pos' within 'string'.
- * Return a copy of the line which contains 'pos'.  Free the line with
- * free().
- * \param string  the program string
- * \param pos     the position within the string
- * \param line    returns the line number corresponding to 'pos'.
- * \param col     returns the column number corresponding to 'pos'.
- * \return copy of the line containing 'pos'.
- */
-const GLubyte *
-_mesa_find_line_column(const GLubyte *string, const GLubyte *pos,
-                       GLint *line, GLint *col)
-{
-   const GLubyte *lineStart = string;
-   const GLubyte *p = string;
-   GLubyte *s;
-   int len;
-
-   *line = 1;
-
-   while (p != pos) {
-      if (*p == (GLubyte) '\n') {
-         (*line)++;
-         lineStart = p + 1;
-      }
-      p++;
-   }
-
-   *col = (pos - lineStart) + 1;
-
-   /* return copy of this line */
-   while (*p != 0 && *p != '\n')
-      p++;
-   len = p - lineStart;
-   s = malloc(len + 1);
-   memcpy(s, lineStart, len);
-   s[len] = 0;
-
-   return s;
-}
-
-
 /**
  * Initialize a new gl_program object.
  */
-static void
-init_program_struct(struct gl_program *prog, GLenum target, GLuint id)
+struct gl_program *
+_mesa_init_gl_program(struct gl_program *prog, GLenum target, GLuint id,
+                      bool is_arb_asm)
 {
-   GLuint i;
-
-   assert(prog);
+   if (!prog)
+      return NULL;
 
    memset(prog, 0, sizeof(*prog));
    prog->Id = id;
    prog->Target = target;
    prog->RefCount = 1;
    prog->Format = GL_PROGRAM_FORMAT_ASCII_ARB;
+   prog->info.stage = _mesa_program_enum_to_shader_stage(target);
+   prog->is_arb_asm = is_arb_asm;
 
-   /* default mapping from samplers to texture units */
-   for (i = 0; i < MAX_SAMPLERS; i++)
-      prog->SamplerUnits[i] = i;
-}
-
-
-/**
- * Initialize a new fragment program object.
- */
-struct gl_program *
-_mesa_init_fragment_program(struct gl_context *ctx,
-                            struct gl_fragment_program *prog,
-                            GLenum target, GLuint id)
-{
-   if (prog) {
-      init_program_struct(&prog->Base, target, id);
-      return &prog->Base;
-   }
-   return NULL;
-}
-
-
-/**
- * Initialize a new vertex program object.
- */
-struct gl_program *
-_mesa_init_vertex_program(struct gl_context *ctx,
-                          struct gl_vertex_program *prog,
-                          GLenum target, GLuint id)
-{
-   if (prog) {
-      init_program_struct(&prog->Base, target, id);
-      return &prog->Base;
-   }
-   return NULL;
-}
-
-
-/**
- * Initialize a new compute program object.
- */
-struct gl_program *
-_mesa_init_compute_program(struct gl_context *ctx,
-                           struct gl_compute_program *prog,
-                           GLenum target, GLuint id)
-{
-   if (prog) {
-      init_program_struct(&prog->Base, target, id);
-      return &prog->Base;
+   /* Uniforms that lack an initializer in the shader code have an initial
+    * value of zero.  This includes sampler uniforms.
+    *
+    * Page 24 (page 30 of the PDF) of the GLSL 1.20 spec says:
+    *
+    *     "The link time initial value is either the value of the variable's
+    *     initializer, if present, or 0 if no initializer is present. Sampler
+    *     types cannot have initializers."
+    *
+    * So we only initialise ARB assembly style programs.
+    */
+   if (is_arb_asm) {
+      /* default mapping from samplers to texture units */
+      for (unsigned i = 0; i < MAX_SAMPLERS; i++)
+         prog->SamplerUnits[i] = i;
    }
-   return NULL;
-}
-
 
-/**
- * Initialize a new geometry program object.
- */
-struct gl_program *
-_mesa_init_geometry_program(struct gl_context *ctx,
-                            struct gl_geometry_program *prog,
-                            GLenum target, GLuint id)
-{
-   if (prog) {
-      init_program_struct(&prog->Base, target, id);
-      return &prog->Base;
-   }
-   return NULL;
+   return prog;
 }
 
 
@@ -323,35 +229,23 @@ _mesa_init_geometry_program(struct gl_context *ctx,
  * \return  pointer to new program object
  */
 struct gl_program *
-_mesa_new_program(struct gl_context *ctx, GLenum target, GLuint id)
+_mesa_new_program(struct gl_context *ctx, GLenum target, GLuint id,
+                  bool is_arb_asm)
 {
-   struct gl_program *prog;
    switch (target) {
    case GL_VERTEX_PROGRAM_ARB: /* == GL_VERTEX_PROGRAM_NV */
-      prog = _mesa_init_vertex_program(ctx, CALLOC_STRUCT(gl_vertex_program),
-                                       target, id );
-      break;
-   case GL_FRAGMENT_PROGRAM_NV:
+   case GL_GEOMETRY_PROGRAM_NV:
+   case GL_TESS_CONTROL_PROGRAM_NV:
+   case GL_TESS_EVALUATION_PROGRAM_NV:
    case GL_FRAGMENT_PROGRAM_ARB:
-      prog =_mesa_init_fragment_program(ctx,
-                                         CALLOC_STRUCT(gl_fragment_program),
-                                         target, id );
-      break;
-   case MESA_GEOMETRY_PROGRAM:
-      prog = _mesa_init_geometry_program(ctx,
-                                         CALLOC_STRUCT(gl_geometry_program),
-                                         target, id);
-      break;
-   case GL_COMPUTE_PROGRAM_NV:
-      prog = _mesa_init_compute_program(ctx,
-                                        CALLOC_STRUCT(gl_compute_program),
-                                        target, id);
-      break;
+   case GL_COMPUTE_PROGRAM_NV: {
+      struct gl_program *prog = rzalloc(NULL, struct gl_program);
+      return _mesa_init_gl_program(prog, target, id, is_arb_asm);
+   }
    default:
       _mesa_problem(ctx, "bad target in _mesa_new_program");
-      prog = NULL;
+      return NULL;
    }
-   return prog;
 }
 
 
@@ -371,12 +265,6 @@ _mesa_delete_program(struct gl_context *ctx, struct gl_program *prog)
    if (prog == &_mesa_DummyProgram)
       return;
 
-   free(prog->String);
-   free(prog->LocalParams);
-
-   if (prog->Instructions) {
-      _mesa_free_instructions(prog->Instructions, prog->NumInstructions);
-   }
    if (prog->Parameters) {
       _mesa_free_parameter_list(prog->Parameters);
    }
@@ -385,7 +273,19 @@ _mesa_delete_program(struct gl_context *ctx, struct gl_program *prog)
       ralloc_free(prog->nir);
    }
 
-   free(prog);
+   if (prog->sh.BindlessSamplers) {
+      ralloc_free(prog->sh.BindlessSamplers);
+   }
+
+   if (prog->sh.BindlessImages) {
+      ralloc_free(prog->sh.BindlessImages);
+   }
+
+   if (prog->driver_cache_blob) {
+      ralloc_free(prog->driver_cache_blob);
+   }
+
+   ralloc_free(prog);
 }
 
 
@@ -423,31 +323,20 @@ _mesa_reference_program_(struct gl_context *ctx,
       else if ((*ptr)->Target == GL_FRAGMENT_PROGRAM_ARB)
          assert(prog->Target == GL_FRAGMENT_PROGRAM_ARB ||
                 prog->Target == GL_FRAGMENT_PROGRAM_NV);
-      else if ((*ptr)->Target == MESA_GEOMETRY_PROGRAM)
-         assert(prog->Target == MESA_GEOMETRY_PROGRAM);
+      else if ((*ptr)->Target == GL_GEOMETRY_PROGRAM_NV)
+         assert(prog->Target == GL_GEOMETRY_PROGRAM_NV);
    }
 #endif
 
    if (*ptr) {
-      GLboolean deleteFlag;
-
-      /*mtx_lock(&(*ptr)->Mutex);*/
-#if 0
-      printf("Program %p ID=%u Target=%s  Refcount-- to %d\n",
-             *ptr, (*ptr)->Id,
-             ((*ptr)->Target == GL_VERTEX_PROGRAM_ARB ? "VP" :
-              ((*ptr)->Target == MESA_GEOMETRY_PROGRAM ? "GP" : "FP")),
-             (*ptr)->RefCount - 1);
-#endif
-      assert((*ptr)->RefCount > 0);
-      (*ptr)->RefCount--;
+      struct gl_program *oldProg = *ptr;
 
-      deleteFlag = ((*ptr)->RefCount == 0);
-      /*mtx_lock(&(*ptr)->Mutex);*/
+      assert(oldProg->RefCount > 0);
 
-      if (deleteFlag) {
+      if (p_atomic_dec_zero(&oldProg->RefCount)) {
          assert(ctx);
-         ctx->Driver.DeleteProgram(ctx, *ptr);
+         _mesa_reference_shader_program_data(ctx, &oldProg->sh.data, NULL);
+         ctx->Driver.DeleteProgram(ctx, oldProg);
       }
 
       *ptr = NULL;
@@ -455,122 +344,13 @@ _mesa_reference_program_(struct gl_context *ctx,
 
    assert(!*ptr);
    if (prog) {
-      /*mtx_lock(&prog->Mutex);*/
-      prog->RefCount++;
-#if 0
-      printf("Program %p ID=%u Target=%s  Refcount++ to %d\n",
-             prog, prog->Id,
-             (prog->Target == GL_VERTEX_PROGRAM_ARB ? "VP" :
-              (prog->Target == MESA_GEOMETRY_PROGRAM ? "GP" : "FP")),
-             prog->RefCount);
-#endif
-      /*mtx_unlock(&prog->Mutex);*/
+      p_atomic_inc(&prog->RefCount);
    }
 
    *ptr = prog;
 }
 
 
-/**
- * Return a copy of a program.
- * XXX Problem here if the program object is actually OO-derivation
- * made by a device driver.
- */
-struct gl_program *
-_mesa_clone_program(struct gl_context *ctx, const struct gl_program *prog)
-{
-   struct gl_program *clone;
-
-   clone = ctx->Driver.NewProgram(ctx, prog->Target, prog->Id);
-   if (!clone)
-      return NULL;
-
-   assert(clone->Target == prog->Target);
-   assert(clone->RefCount == 1);
-
-   clone->String = (GLubyte *) strdup((char *) prog->String);
-   clone->Format = prog->Format;
-   clone->Instructions = _mesa_alloc_instructions(prog->NumInstructions);
-   if (!clone->Instructions) {
-      _mesa_reference_program(ctx, &clone, NULL);
-      return NULL;
-   }
-   _mesa_copy_instructions(clone->Instructions, prog->Instructions,
-                           prog->NumInstructions);
-   clone->InputsRead = prog->InputsRead;
-   clone->OutputsWritten = prog->OutputsWritten;
-   clone->SamplersUsed = prog->SamplersUsed;
-   clone->ShadowSamplers = prog->ShadowSamplers;
-   memcpy(clone->TexturesUsed, prog->TexturesUsed, sizeof(prog->TexturesUsed));
-
-   if (prog->Parameters)
-      clone->Parameters = _mesa_clone_parameter_list(prog->Parameters);
-   if (prog->LocalParams) {
-      clone->LocalParams = malloc(MAX_PROGRAM_LOCAL_PARAMS *
-                                  sizeof(float[4]));
-      if (!clone->LocalParams) {
-         _mesa_reference_program(ctx, &clone, NULL);
-         return NULL;
-      }
-      memcpy(clone->LocalParams, prog->LocalParams,
-             MAX_PROGRAM_LOCAL_PARAMS * sizeof(float[4]));
-   }
-   clone->IndirectRegisterFiles = prog->IndirectRegisterFiles;
-   clone->NumInstructions = prog->NumInstructions;
-   clone->NumTemporaries = prog->NumTemporaries;
-   clone->NumParameters = prog->NumParameters;
-   clone->NumAttributes = prog->NumAttributes;
-   clone->NumAddressRegs = prog->NumAddressRegs;
-   clone->NumNativeInstructions = prog->NumNativeInstructions;
-   clone->NumNativeTemporaries = prog->NumNativeTemporaries;
-   clone->NumNativeParameters = prog->NumNativeParameters;
-   clone->NumNativeAttributes = prog->NumNativeAttributes;
-   clone->NumNativeAddressRegs = prog->NumNativeAddressRegs;
-   clone->NumAluInstructions = prog->NumAluInstructions;
-   clone->NumTexInstructions = prog->NumTexInstructions;
-   clone->NumTexIndirections = prog->NumTexIndirections;
-   clone->NumNativeAluInstructions = prog->NumNativeAluInstructions;
-   clone->NumNativeTexInstructions = prog->NumNativeTexInstructions;
-   clone->NumNativeTexIndirections = prog->NumNativeTexIndirections;
-
-   switch (prog->Target) {
-   case GL_VERTEX_PROGRAM_ARB:
-      {
-         const struct gl_vertex_program *vp = gl_vertex_program_const(prog);
-         struct gl_vertex_program *vpc = gl_vertex_program(clone);
-         vpc->IsPositionInvariant = vp->IsPositionInvariant;
-      }
-      break;
-   case GL_FRAGMENT_PROGRAM_ARB:
-      {
-         const struct gl_fragment_program *fp = gl_fragment_program_const(prog);
-         struct gl_fragment_program *fpc = gl_fragment_program(clone);
-         fpc->UsesKill = fp->UsesKill;
-         fpc->UsesDFdy = fp->UsesDFdy;
-         fpc->OriginUpperLeft = fp->OriginUpperLeft;
-         fpc->PixelCenterInteger = fp->PixelCenterInteger;
-      }
-      break;
-   case MESA_GEOMETRY_PROGRAM:
-      {
-         const struct gl_geometry_program *gp = gl_geometry_program_const(prog);
-         struct gl_geometry_program *gpc = gl_geometry_program(clone);
-         gpc->VerticesOut = gp->VerticesOut;
-         gpc->InputType = gp->InputType;
-         gpc->Invocations = gp->Invocations;
-         gpc->OutputType = gp->OutputType;
-         gpc->UsesEndPrimitive = gp->UsesEndPrimitive;
-         gpc->UsesStreams = gp->UsesStreams;
-      }
-      break;
-   default:
-      _mesa_problem(NULL, "Unexpected target in _mesa_clone_program");
-   }
-
-   return clone;
-}
-
-
 /**
  * Insert 'count' NOP instructions at 'start' in the given program.
  * Adjust branch targets accordingly.
@@ -578,14 +358,14 @@ _mesa_clone_program(struct gl_context *ctx, const struct gl_program *prog)
 GLboolean
 _mesa_insert_instructions(struct gl_program *prog, GLuint start, GLuint count)
 {
-   const GLuint origLen = prog->NumInstructions;
+   const GLuint origLen = prog->arb.NumInstructions;
    const GLuint newLen = origLen + count;
    struct prog_instruction *newInst;
    GLuint i;
 
    /* adjust branches */
-   for (i = 0; i < prog->NumInstructions; i++) {
-      struct prog_instruction *inst = prog->Instructions + i;
+   for (i = 0; i < prog->arb.NumInstructions; i++) {
+      struct prog_instruction *inst = prog->arb.Instructions + i;
       if (inst->BranchTarget > 0) {
          if ((GLuint)inst->BranchTarget >= start) {
             inst->BranchTarget += count;
@@ -594,28 +374,28 @@ _mesa_insert_instructions(struct gl_program *prog, GLuint start, GLuint count)
    }
 
    /* Alloc storage for new instructions */
-   newInst = _mesa_alloc_instructions(newLen);
+   newInst = rzalloc_array(prog, struct prog_instruction, newLen);
    if (!newInst) {
       return GL_FALSE;
    }
 
    /* Copy 'start' instructions into new instruction buffer */
-   _mesa_copy_instructions(newInst, prog->Instructions, start);
+   _mesa_copy_instructions(newInst, prog->arb.Instructions, start);
 
    /* init the new instructions */
    _mesa_init_instructions(newInst + start, count);
 
    /* Copy the remaining/tail instructions to new inst buffer */
    _mesa_copy_instructions(newInst + start + count,
-                           prog->Instructions + start,
+                           prog->arb.Instructions + start,
                            origLen - start);
 
    /* free old instructions */
-   _mesa_free_instructions(prog->Instructions, origLen);
+   ralloc_free(prog->arb.Instructions);
 
    /* install new instructions */
-   prog->Instructions = newInst;
-   prog->NumInstructions = newLen;
+   prog->arb.Instructions = newInst;
+   prog->arb.NumInstructions = newLen;
 
    return GL_TRUE;
 }
@@ -625,16 +405,17 @@ _mesa_insert_instructions(struct gl_program *prog, GLuint start, GLuint count)
  * Adjust branch targets accordingly.
  */
 GLboolean
-_mesa_delete_instructions(struct gl_program *prog, GLuint start, GLuint count)
+_mesa_delete_instructions(struct gl_program *prog, GLuint start, GLuint count,
+                          void *mem_ctx)
 {
-   const GLuint origLen = prog->NumInstructions;
+   const GLuint origLen = prog->arb.NumInstructions;
    const GLuint newLen = origLen - count;
    struct prog_instruction *newInst;
    GLuint i;
 
    /* adjust branches */
-   for (i = 0; i < prog->NumInstructions; i++) {
-      struct prog_instruction *inst = prog->Instructions + i;
+   for (i = 0; i < prog->arb.NumInstructions; i++) {
+      struct prog_instruction *inst = prog->arb.Instructions + i;
       if (inst->BranchTarget > 0) {
          if (inst->BranchTarget > (GLint) start) {
             inst->BranchTarget -= count;
@@ -643,214 +424,30 @@ _mesa_delete_instructions(struct gl_program *prog, GLuint start, GLuint count)
    }
 
    /* Alloc storage for new instructions */
-   newInst = _mesa_alloc_instructions(newLen);
+   newInst = rzalloc_array(mem_ctx, struct prog_instruction, newLen);
    if (!newInst) {
       return GL_FALSE;
    }
 
    /* Copy 'start' instructions into new instruction buffer */
-   _mesa_copy_instructions(newInst, prog->Instructions, start);
+   _mesa_copy_instructions(newInst, prog->arb.Instructions, start);
 
    /* Copy the remaining/tail instructions to new inst buffer */
    _mesa_copy_instructions(newInst + start,
-                           prog->Instructions + start + count,
+                           prog->arb.Instructions + start + count,
                            newLen - start);
 
    /* free old instructions */
-   _mesa_free_instructions(prog->Instructions, origLen);
+   ralloc_free(prog->arb.Instructions);
 
    /* install new instructions */
-   prog->Instructions = newInst;
-   prog->NumInstructions = newLen;
+   prog->arb.Instructions = newInst;
+   prog->arb.NumInstructions = newLen;
 
    return GL_TRUE;
 }
 
 
-/**
- * Search instructions for registers that match (oldFile, oldIndex),
- * replacing them with (newFile, newIndex).
- */
-static void
-replace_registers(struct prog_instruction *inst, GLuint numInst,
-                  GLuint oldFile, GLuint oldIndex,
-                  GLuint newFile, GLuint newIndex)
-{
-   GLuint i, j;
-   for (i = 0; i < numInst; i++) {
-      /* src regs */
-      for (j = 0; j < _mesa_num_inst_src_regs(inst[i].Opcode); j++) {
-         if (inst[i].SrcReg[j].File == oldFile &&
-             inst[i].SrcReg[j].Index == oldIndex) {
-            inst[i].SrcReg[j].File = newFile;
-            inst[i].SrcReg[j].Index = newIndex;
-         }
-      }
-      /* dst reg */
-      if (inst[i].DstReg.File == oldFile && inst[i].DstReg.Index == oldIndex) {
-         inst[i].DstReg.File = newFile;
-         inst[i].DstReg.Index = newIndex;
-      }
-   }
-}
-
-
-/**
- * Search instructions for references to program parameters.  When found,
- * increment the parameter index by 'offset'.
- * Used when combining programs.
- */
-static void
-adjust_param_indexes(struct prog_instruction *inst, GLuint numInst,
-                     GLuint offset)
-{
-   GLuint i, j;
-   for (i = 0; i < numInst; i++) {
-      for (j = 0; j < _mesa_num_inst_src_regs(inst[i].Opcode); j++) {
-         GLuint f = inst[i].SrcReg[j].File;
-         if (f == PROGRAM_CONSTANT ||
-             f == PROGRAM_UNIFORM ||
-             f == PROGRAM_STATE_VAR) {
-            inst[i].SrcReg[j].Index += offset;
-         }
-      }
-   }
-}
-
-
-/**
- * Combine two programs into one.  Fix instructions so the outputs of
- * the first program go to the inputs of the second program.
- */
-struct gl_program *
-_mesa_combine_programs(struct gl_context *ctx,
-                       const struct gl_program *progA,
-                       const struct gl_program *progB)
-{
-   struct prog_instruction *newInst;
-   struct gl_program *newProg;
-   const GLuint lenA = progA->NumInstructions - 1; /* omit END instr */
-   const GLuint lenB = progB->NumInstructions;
-   const GLuint numParamsA = _mesa_num_parameters(progA->Parameters);
-   const GLuint newLength = lenA + lenB;
-   GLboolean usedTemps[MAX_PROGRAM_TEMPS];
-   GLuint firstTemp = 0;
-   GLbitfield64 inputsB;
-   GLuint i;
-
-   assert(progA->Target == progB->Target);
-
-   newInst = _mesa_alloc_instructions(newLength);
-   if (!newInst)
-      return GL_FALSE;
-
-   _mesa_copy_instructions(newInst, progA->Instructions, lenA);
-   _mesa_copy_instructions(newInst + lenA, progB->Instructions, lenB);
-
-   /* adjust branch / instruction addresses for B's instructions */
-   for (i = 0; i < lenB; i++) {
-      newInst[lenA + i].BranchTarget += lenA;
-   }
-
-   newProg = ctx->Driver.NewProgram(ctx, progA->Target, 0);
-   newProg->Instructions = newInst;
-   newProg->NumInstructions = newLength;
-
-   /* find used temp regs (we may need new temps below) */
-   _mesa_find_used_registers(newProg, PROGRAM_TEMPORARY,
-                             usedTemps, MAX_PROGRAM_TEMPS);
-
-   if (newProg->Target == GL_FRAGMENT_PROGRAM_ARB) {
-      const struct gl_fragment_program *fprogA, *fprogB;
-      struct gl_fragment_program *newFprog;
-      GLbitfield64 progB_inputsRead = progB->InputsRead;
-      GLint progB_colorFile, progB_colorIndex;
-
-      fprogA = gl_fragment_program_const(progA);
-      fprogB = gl_fragment_program_const(progB);
-      newFprog = gl_fragment_program(newProg);
-
-      newFprog->UsesKill = fprogA->UsesKill || fprogB->UsesKill;
-      newFprog->UsesDFdy = fprogA->UsesDFdy || fprogB->UsesDFdy;
-
-      /* We'll do a search and replace for instances
-       * of progB_colorFile/progB_colorIndex below...
-       */
-      progB_colorFile = PROGRAM_INPUT;
-      progB_colorIndex = VARYING_SLOT_COL0;
-
-      /*
-       * The fragment program may get color from a state var rather than
-       * a fragment input (vertex output) if it's constant.
-       * See the texenvprogram.c code.
-       * So, search the program's parameter list now to see if the program
-       * gets color from a state var instead of a conventional fragment
-       * input register.
-       */
-      for (i = 0; i < progB->Parameters->NumParameters; i++) {
-         struct gl_program_parameter *p = &progB->Parameters->Parameters[i];
-         if (p->Type == PROGRAM_STATE_VAR &&
-             p->StateIndexes[0] == STATE_INTERNAL &&
-             p->StateIndexes[1] == STATE_CURRENT_ATTRIB &&
-             (int) p->StateIndexes[2] == (int) VERT_ATTRIB_COLOR0) {
-            progB_inputsRead |= VARYING_BIT_COL0;
-            progB_colorFile = PROGRAM_STATE_VAR;
-            progB_colorIndex = i;
-            break;
-         }
-      }
-
-      /* Connect color outputs of fprogA to color inputs of fprogB, via a
-       * new temporary register.
-       */
-      if ((progA->OutputsWritten & BITFIELD64_BIT(FRAG_RESULT_COLOR)) &&
-          (progB_inputsRead & VARYING_BIT_COL0)) {
-         GLint tempReg = _mesa_find_free_register(usedTemps, MAX_PROGRAM_TEMPS,
-                                                  firstTemp);
-         if (tempReg < 0) {
-            _mesa_problem(ctx, "No free temp regs found in "
-                          "_mesa_combine_programs(), using 31");
-            tempReg = 31;
-         }
-         firstTemp = tempReg + 1;
-
-         /* replace writes to result.color[0] with tempReg */
-         replace_registers(newInst, lenA,
-                           PROGRAM_OUTPUT, FRAG_RESULT_COLOR,
-                           PROGRAM_TEMPORARY, tempReg);
-         /* replace reads from the input color with tempReg */
-         replace_registers(newInst + lenA, lenB,
-                           progB_colorFile, progB_colorIndex, /* search for */
-                           PROGRAM_TEMPORARY, tempReg  /* replace with */ );
-      }
-
-      /* compute combined program's InputsRead */
-      inputsB = progB_inputsRead;
-      if (progA->OutputsWritten & BITFIELD64_BIT(FRAG_RESULT_COLOR)) {
-         inputsB &= ~(1 << VARYING_SLOT_COL0);
-      }
-      newProg->InputsRead = progA->InputsRead | inputsB;
-      newProg->OutputsWritten = progB->OutputsWritten;
-      newProg->SamplersUsed = progA->SamplersUsed | progB->SamplersUsed;
-   }
-   else {
-      /* vertex program */
-      assert(0);      /* XXX todo */
-   }
-
-   /*
-    * Merge parameters (uniforms, constants, etc)
-    */
-   newProg->Parameters = _mesa_combine_parameter_lists(progA->Parameters,
-                                                       progB->Parameters);
-
-   adjust_param_indexes(newInst + lenA, lenB, numParamsA);
-
-
-   return newProg;
-}
-
-
 /**
  * Populate the 'used' array with flags indicating which registers (TEMPs,
  * INPUTs, OUTPUTs, etc, are used by the given program.
@@ -867,8 +464,8 @@ _mesa_find_used_registers(const struct gl_program *prog,
 
    memset(used, 0, usedSize);
 
-   for (i = 0; i < prog->NumInstructions; i++) {
-      const struct prog_instruction *inst = prog->Instructions + i;
+   for (i = 0; i < prog->arb.NumInstructions; i++) {
+      const struct prog_instruction *inst = prog->arb.Instructions + i;
       const GLuint n = _mesa_num_inst_src_regs(inst->Opcode);
 
       if (inst->DstReg.File == file) {
@@ -913,148 +510,13 @@ _mesa_find_free_register(const GLboolean used[],
 }
 
 
-
-/**
- * Check if the given register index is valid (doesn't exceed implementation-
- * dependent limits).
- * \return GL_TRUE if OK, GL_FALSE if bad index
- */
-GLboolean
-_mesa_valid_register_index(const struct gl_context *ctx,
-                           gl_shader_stage shaderType,
-                           gl_register_file file, GLint index)
-{
-   const struct gl_program_constants *c;
-
-   assert(0 <= shaderType && shaderType < MESA_SHADER_STAGES);
-   c = &ctx->Const.Program[shaderType];
-
-   switch (file) {
-   case PROGRAM_UNDEFINED:
-      return GL_TRUE;  /* XXX or maybe false? */
-
-   case PROGRAM_TEMPORARY:
-      return index >= 0 && index < (GLint) c->MaxTemps;
-
-   case PROGRAM_UNIFORM:
-   case PROGRAM_STATE_VAR:
-      /* aka constant buffer */
-      return index >= 0 && index < (GLint) c->MaxUniformComponents / 4;
-
-   case PROGRAM_CONSTANT:
-      /* constant buffer w/ possible relative negative addressing */
-      return (index > (int) c->MaxUniformComponents / -4 &&
-              index < (int) c->MaxUniformComponents / 4);
-
-   case PROGRAM_INPUT:
-      if (index < 0)
-         return GL_FALSE;
-
-      switch (shaderType) {
-      case MESA_SHADER_VERTEX:
-         return index < VERT_ATTRIB_GENERIC0 + (GLint) c->MaxAttribs;
-      case MESA_SHADER_FRAGMENT:
-         return index < VARYING_SLOT_VAR0 + (GLint) ctx->Const.MaxVarying;
-      case MESA_SHADER_GEOMETRY:
-         return index < VARYING_SLOT_VAR0 + (GLint) ctx->Const.MaxVarying;
-      default:
-         return GL_FALSE;
-      }
-
-   case PROGRAM_OUTPUT:
-      if (index < 0)
-         return GL_FALSE;
-
-      switch (shaderType) {
-      case MESA_SHADER_VERTEX:
-         return index < VARYING_SLOT_VAR0 + (GLint) ctx->Const.MaxVarying;
-      case MESA_SHADER_FRAGMENT:
-         return index < FRAG_RESULT_DATA0 + (GLint) ctx->Const.MaxDrawBuffers;
-      case MESA_SHADER_GEOMETRY:
-         return index < VARYING_SLOT_VAR0 + (GLint) ctx->Const.MaxVarying;
-      default:
-         return GL_FALSE;
-      }
-
-   case PROGRAM_ADDRESS:
-      return index >= 0 && index < (GLint) c->MaxAddressRegs;
-
-   default:
-      _mesa_problem(ctx,
-                    "unexpected register file in _mesa_valid_register_index()");
-      return GL_FALSE;
-   }
-}
-
-
-
-/**
- * "Post-process" a GPU program.  This is intended to be used for debugging.
- * Example actions include no-op'ing instructions or changing instruction
- * behaviour.
- */
-void
-_mesa_postprocess_program(struct gl_context *ctx, struct gl_program *prog)
-{
-   static const GLfloat white[4] = { 0.5, 0.5, 0.5, 0.5 };
-   GLuint i;
-   GLuint whiteSwizzle;
-   GLint whiteIndex = _mesa_add_unnamed_constant(prog->Parameters,
-                                                 (gl_constant_value *) white,
-                                                 4, &whiteSwizzle);
-
-   (void) whiteIndex;
-
-   for (i = 0; i < prog->NumInstructions; i++) {
-      struct prog_instruction *inst = prog->Instructions + i;
-      const GLuint n = _mesa_num_inst_src_regs(inst->Opcode);
-
-      (void) n;
-
-      if (_mesa_is_tex_instruction(inst->Opcode)) {
-#if 0
-         /* replace TEX/TXP/TXB with MOV */
-         inst->Opcode = OPCODE_MOV;
-         inst->DstReg.WriteMask = WRITEMASK_XYZW;
-         inst->SrcReg[0].Swizzle = SWIZZLE_XYZW;
-         inst->SrcReg[0].Negate = NEGATE_NONE;
-#endif
-
-#if 0
-         /* disable shadow texture mode */
-         inst->TexShadow = 0;
-#endif
-      }
-
-      if (inst->Opcode == OPCODE_TXP) {
-#if 0
-         inst->Opcode = OPCODE_MOV;
-         inst->DstReg.WriteMask = WRITEMASK_XYZW;
-         inst->SrcReg[0].File = PROGRAM_CONSTANT;
-         inst->SrcReg[0].Index = whiteIndex;
-         inst->SrcReg[0].Swizzle = SWIZZLE_XYZW;
-         inst->SrcReg[0].Negate = NEGATE_NONE;
-#endif
-#if 0
-         inst->TexShadow = 0;
-#endif
-#if 0
-         inst->Opcode = OPCODE_TEX;
-         inst->TexShadow = 0;
-#endif
-      }
-
-   }
-}
-
 /* Gets the minimum number of shader invocations per fragment.
  * This function is useful to determine if we need to do per
  * sample shading or per fragment shading.
  */
 GLint
 _mesa_get_min_invocations_per_fragment(struct gl_context *ctx,
-                                       const struct gl_fragment_program *prog,
-                                       bool ignore_sample_qualifier)
+                                       const struct gl_program *prog)
 {
    /* From ARB_sample_shading specification:
     * "Using gl_SampleID in a fragment shader causes the entire shader
@@ -1072,17 +534,31 @@ _mesa_get_min_invocations_per_fragment(struct gl_context *ctx,
        * "Use of the "sample" qualifier on a fragment shader input
        *  forces per-sample shading"
        */
-      if (prog->IsSample && !ignore_sample_qualifier)
-         return MAX2(ctx->DrawBuffer->Visual.samples, 1);
-
-      if (prog->Base.SystemValuesRead & (SYSTEM_BIT_SAMPLE_ID |
-                                         SYSTEM_BIT_SAMPLE_POS))
-         return MAX2(ctx->DrawBuffer->Visual.samples, 1);
+      if (prog->info.fs.uses_sample_qualifier ||
+          (prog->info.system_values_read & (SYSTEM_BIT_SAMPLE_ID |
+                                            SYSTEM_BIT_SAMPLE_POS)))
+         return MAX2(_mesa_geometric_samples(ctx->DrawBuffer), 1);
       else if (ctx->Multisample.SampleShading)
-         return MAX2(ceil(ctx->Multisample.MinSampleShadingValue *
-                          ctx->DrawBuffer->Visual.samples), 1);
+         return MAX2(ceilf(ctx->Multisample.MinSampleShadingValue *
+                          _mesa_geometric_samples(ctx->DrawBuffer)), 1);
       else
          return 1;
    }
    return 1;
 }
+
+
+GLbitfield
+gl_external_samplers(const struct gl_program *prog)
+{
+   GLbitfield external_samplers = 0;
+   GLbitfield mask = prog->SamplersUsed;
+
+   while (mask) {
+      int idx = u_bit_scan(&mask);
+      if (prog->sh.SamplerTargets[idx] == TEXTURE_EXTERNAL_INDEX)
+         external_samplers |= (1 << idx);
+   }
+
+   return external_samplers;
+}