prog_execute.c

Mesa is an open-source implementation of the OpenGL specification - a system for rendering interacti

/*
 * Mesa 3-D graphics library
 * Version:  7.0.3
 *
 * Copyright (C) 1999-2007  Brian Paul   All Rights Reserved.
 *
 * Permission is hereby granted, free of charge, to any person obtaining a
 * copy of this software and associated documentation files (the "Software"),
 * to deal in the Software without restriction, including without limitation
 * the rights to use, copy, modify, merge, publish, distribute, sublicense,
 * and/or sell copies of the Software, and to permit persons to whom the
 * Software is furnished to do so, subject to the following conditions:
 *
 * The above copyright notice and this permission notice shall be included
 * in all copies or substantial portions of the Software.
 *
 * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
 * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
 * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
 * BRIAN PAUL BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN
 * AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
 * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
 */

/**
 * \file prog_execute.c
 * Software interpreter for vertex/fragment programs.
 * \author Brian Paul
 */

/*
 * NOTE: we do everything in single-precision floating point; we don't
 * currently observe the single/half/fixed-precision qualifiers.
 *
 */


#include "glheader.h"
#include "colormac.h"
#include "context.h"
#include "program.h"
#include "prog_execute.h"
#include "prog_instruction.h"
#include "prog_parameter.h"
#include "prog_print.h"
#include "shader/slang/slang_library_noise.h"


/* debug predicate */
#define DEBUG_PROG 0


/**
 * Set x to positive or negative infinity.
 */
#if defined(USE_IEEE) || defined(_WIN32)
#define SET_POS_INFINITY(x)  ( *((GLuint *) (void *)&x) = 0x7F800000 )
#define SET_NEG_INFINITY(x)  ( *((GLuint *) (void *)&x) = 0xFF800000 )
#elif defined(VMS)
#define SET_POS_INFINITY(x)  x = __MAXFLOAT
#define SET_NEG_INFINITY(x)  x = -__MAXFLOAT
#else
#define SET_POS_INFINITY(x)  x = (GLfloat) HUGE_VAL
#define SET_NEG_INFINITY(x)  x = (GLfloat) -HUGE_VAL
#endif

#define SET_FLOAT_BITS(x, bits) ((fi_type *) (void *) &(x))->i = bits


static const GLfloat ZeroVec[4] = { 0.0F, 0.0F, 0.0F, 0.0F };



/**
 * Return a pointer to the 4-element float vector specified by the given
 * source register.
 */
static INLINE const GLfloat *
get_register_pointer(const struct prog_src_register *source,
                     const struct gl_program_machine *machine)
{
   if (source->RelAddr) {
      const GLint reg = source->Index + machine->AddressReg[0][0];
      if (source->File == PROGRAM_ENV_PARAM)
         if (reg < 0 || reg >= MAX_PROGRAM_ENV_PARAMS)
            return ZeroVec;
         else
            return machine->EnvParams[reg];
      else {
         const struct gl_program_parameter_list *params;
         ASSERT(source->File == PROGRAM_LOCAL_PARAM ||
                source->File == PROGRAM_CONSTANT ||
                source->File == PROGRAM_STATE_VAR ||
                source->File == PROGRAM_UNIFORM);
         params = machine->CurProgram->Parameters;
         if (reg < 0 || reg >= params->NumParameters)
            return ZeroVec;
         else
            return params->ParameterValues[reg];
      }
   }

   switch (source->File) {
   case PROGRAM_TEMPORARY:
      ASSERT(source->Index < MAX_PROGRAM_TEMPS);
      return machine->Temporaries[source->Index];

   case PROGRAM_INPUT:
      if (machine->CurProgram->Target == GL_VERTEX_PROGRAM_ARB) {
         ASSERT(source->Index < VERT_ATTRIB_MAX);
         return machine->VertAttribs[source->Index];
      }
      else {
         ASSERT(source->Index < FRAG_ATTRIB_MAX);
         return machine->Attribs[source->Index][machine->CurElement];
      }

   case PROGRAM_OUTPUT:
      ASSERT(source->Index < MAX_PROGRAM_OUTPUTS);
      return machine->Outputs[source->Index];

   case PROGRAM_LOCAL_PARAM:
      ASSERT(source->Index < MAX_PROGRAM_LOCAL_PARAMS);
      return machine->CurProgram->LocalParams[source->Index];

   case PROGRAM_ENV_PARAM:
      ASSERT(source->Index < MAX_PROGRAM_ENV_PARAMS);
      return machine->EnvParams[source->Index];

   case PROGRAM_STATE_VAR:
      /* Fallthrough */
   case PROGRAM_CONSTANT:
      /* Fallthrough */
   case PROGRAM_UNIFORM:
      /* Fallthrough */
   case PROGRAM_NAMED_PARAM:
      ASSERT(source->Index <
             (GLint) machine->CurProgram->Parameters->NumParameters);
      return machine->CurProgram->Parameters->ParameterValues[source->Index];

   default:
      _mesa_problem(NULL,
                    "Invalid input register file %d in get_register_pointer()",
                    source->File);
      return NULL;
   }
}


#if FEATURE_MESA_program_debug
static struct gl_program_machine *CurrentMachine = NULL;

/**
 * For GL_MESA_program_debug.
 * Return current value (4*GLfloat) of a program register.
 * Called via ctx->Driver.GetProgramRegister().
 */
void
_mesa_get_program_register(GLcontext *ctx, enum register_file file,
                           GLuint index, GLfloat val[4])
{
   if (CurrentMachine) {
      struct prog_src_register src;
      const GLfloat *reg;
      src.File = file;
      src.Index = index;
      reg = get_register_pointer(&src, CurrentMachine);
      COPY_4V(val, reg);
   }
}
#endif /* FEATURE_MESA_program_debug */


/**
 * Fetch a 4-element float vector from the given source register.
 * Apply swizzling and negating as needed.
 */
static void
fetch_vector4(const struct prog_src_register *source,
              const struct gl_program_machine *machine, GLfloat result[4])
{
   const GLfloat *src = get_register_pointer(source, machine);
   ASSERT(src);

   if (source->Swizzle == SWIZZLE_NOOP) {
      /* no swizzling */
      COPY_4V(result, src);
   }
   else {
      ASSERT(GET_SWZ(source->Swizzle, 0) <= 3);
      ASSERT(GET_SWZ(source->Swizzle, 1) <= 3);
      ASSERT(GET_SWZ(source->Swizzle, 2) <= 3);
      ASSERT(GET_SWZ(source->Swizzle, 3) <= 3);
      result[0] = src[GET_SWZ(source->Swizzle, 0)];
      result[1] = src[GET_SWZ(source->Swizzle, 1)];
      result[2] = src[GET_SWZ(source->Swizzle, 2)];
      result[3] = src[GET_SWZ(source->Swizzle, 3)];
   }

   if (source->NegateBase) {
      result[0] = -result[0];
      result[1] = -result[1];
      result[2] = -result[2];
      result[3] = -result[3];
   }
   if (source->Abs) {
      result[0] = FABSF(result[0]);
      result[1] = FABSF(result[1]);
      result[2] = FABSF(result[2]);
      result[3] = FABSF(result[3]);
   }
   if (source->NegateAbs) {
      result[0] = -result[0];
      result[1] = -result[1];
      result[2] = -result[2];
      result[3] = -result[3];
   }
}


/**
 * Fetch the derivative with respect to X or Y for the given register.
 * XXX this currently only works for fragment program input attribs.
 */
static void
fetch_vector4_deriv(GLcontext * ctx,
                    const struct prog_src_register *source,
                    const struct gl_program_machine *machine,
                    char xOrY, GLfloat result[4])
{
   if (source->File == PROGRAM_INPUT && source->Index < machine->NumDeriv) {
      const GLint col = machine->CurElement;
      const GLfloat w = machine->Attribs[FRAG_ATTRIB_WPOS][col][3];
      const GLfloat invQ = 1.0f / w;
      GLfloat deriv[4];

      if (xOrY == 'X') {
         deriv[0] = machine->DerivX[source->Index][0] * invQ;
         deriv[1] = machine->DerivX[source->Index][1] * invQ;
         deriv[2] = machine->DerivX[source->Index][2] * invQ;
         deriv[3] = machine->DerivX[source->Index][3] * invQ;
      }
      else {
         deriv[0] = machine->DerivY[source->Index][0] * invQ;
         deriv[1] = machine->DerivY[source->Index][1] * invQ;
         deriv[2] = machine->DerivY[source->Index][2] * invQ;
         deriv[3] = machine->DerivY[source->Index][3] * invQ;
      }

      result[0] = deriv[GET_SWZ(source->Swizzle, 0)];
      result[1] = deriv[GET_SWZ(source->Swizzle, 1)];
      result[2] = deriv[GET_SWZ(source->Swizzle, 2)];
      result[3] = deriv[GET_SWZ(source->Swizzle, 3)];
      
      if (source->NegateBase) {
         result[0] = -result[0];
         result[1] = -result[1];
         result[2] = -result[2];
         result[3] = -result[3];
      }
      if (source->Abs) {
         result[0] = FABSF(result[0]);
         result[1] = FABSF(result[1]);
         result[2] = FABSF(result[2]);
         result[3] = FABSF(result[3]);
      }
      if (source->NegateAbs) {
         result[0] = -result[0];
         result[1] = -result[1];
         result[2] = -result[2];
         result[3] = -result[3];
      }
   }
   else {
      ASSIGN_4V(result, 0.0, 0.0, 0.0, 0.0);
   }
}


/**
 * As above, but only return result[0] element.
 */
static void
fetch_vector1(const struct prog_src_register *source,
              const struct gl_program_machine *machine, GLfloat result[4])
{
   const GLfloat *src = get_register_pointer(source, machine);
   ASSERT(src);

   result[0] = src[GET_SWZ(source->Swizzle, 0)];

   if (source->NegateBase) {
      result[0] = -result[0];
   }
   if (source->Abs) {
      result[0] = FABSF(result[0]);
   }
   if (source->NegateAbs) {
      result[0] = -result[0];
   }
}


/**
 * Fetch texel from texture.  Use partial derivatives when possible.
 */
static INLINE void
fetch_texel(GLcontext *ctx,
            const struct gl_program_machine *machine,
            const struct prog_instruction *inst,
            const GLfloat texcoord[4], GLfloat lodBias,
            GLfloat color[4])
{
   const GLuint unit = machine->Samplers[inst->TexSrcUnit];

   /* Note: we only have the right derivatives for fragment input attribs.
    */
   if (machine->NumDeriv > 0 &&
       inst->SrcReg[0].File == PROGRAM_INPUT &&
       inst->SrcReg[0].Index == FRAG_ATTRIB_TEX0 + inst->TexSrcUnit) {
      /* simple texture fetch for which we should have derivatives */
      GLuint attr = inst->SrcReg[0].Index;
      machine->FetchTexelDeriv(ctx, texcoord,
                               machine->DerivX[attr],
                               machine->DerivY[attr],
                               lodBias, unit, color);
   }
   else {
      machine->FetchTexelLod(ctx, texcoord, lodBias, unit, color);
   }
}


/**
 * Test value against zero and return GT, LT, EQ or UN if NaN.
 */
static INLINE GLuint
generate_cc(float value)
{
   if (value != value)
      return COND_UN;           /* NaN */
   if (value > 0.0F)
      return COND_GT;
   if (value < 0.0F)
      return COND_LT;
   return COND_EQ;
}


/**
 * Test if the ccMaskRule is satisfied by the given condition code.
 * Used to mask destination writes according to the current condition code.
 */
static INLINE GLboolean
test_cc(GLuint condCode, GLuint ccMaskRule)
{
   switch (ccMaskRule) {
   case COND_EQ: return (condCode == COND_EQ);
   case COND_NE: return (condCode != COND_EQ);
   case COND_LT: return (condCode == COND_LT);
   case COND_GE: return (condCode == COND_GT || condCode == COND_EQ);
   case COND_LE: return (condCode == COND_LT || condCode == COND_EQ);
   case COND_GT: return (condCode == COND_GT);
   case COND_TR: return GL_TRUE;
   case COND_FL: return GL_FALSE;
   default:      return GL_TRUE;
   }
}


/**
 * Evaluate the 4 condition codes against a predicate and return GL_TRUE
 * or GL_FALSE to indicate result.
 */
static INLINE GLboolean
eval_condition(const struct gl_program_machine *machine,
               const struct prog_instruction *inst)
{
   const GLuint swizzle = inst->DstReg.CondSwizzle;
   const GLuint condMask = inst->DstReg.CondMask;
   if (test_cc(machine->CondCodes[GET_SWZ(swizzle, 0)], condMask) ||
       test_cc(machine->CondCodes[GET_SWZ(swizzle, 1)], condMask) ||
       test_cc(machine->CondCodes[GET_SWZ(swizzle, 2)], condMask) ||
       test_cc(machine->CondCodes[GET_SWZ(swizzle, 3)], condMask)) {
      return GL_TRUE;
   }