nvvertexec.c

mesa-6.5-minigui源码

/*
 * Mesa 3-D graphics library
 * Version:  6.5
 *
 * Copyright (C) 1999-2005  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 nvvertexec.c
 * Code to execute vertex programs.
 * \author Brian Paul
 */

#include "glheader.h"
#include "context.h"
#include "imports.h"
#include "macros.h"
#include "mtypes.h"
#include "nvvertexec.h"
#include "program_instruction.h"
#include "program.h"
#include "math/m_matrix.h"


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


/**
 * Load/initialize the vertex program registers which need to be set
 * per-vertex.
 */
void
_mesa_init_vp_per_vertex_registers(GLcontext *ctx)
{
   /* Input registers get initialized from the current vertex attribs */
   MEMCPY(ctx->VertexProgram.Inputs, ctx->Current.Attrib,
          VERT_ATTRIB_MAX * 4 * sizeof(GLfloat));

   if (ctx->VertexProgram.Current->IsNVProgram) {
      GLuint i;
      /* Output/result regs are initialized to [0,0,0,1] */
      for (i = 0; i < MAX_NV_VERTEX_PROGRAM_OUTPUTS; i++) {
         ASSIGN_4V(ctx->VertexProgram.Outputs[i], 0.0F, 0.0F, 0.0F, 1.0F);
      }
      /* Temp regs are initialized to [0,0,0,0] */
      for (i = 0; i < MAX_NV_VERTEX_PROGRAM_TEMPS; i++) {
         ASSIGN_4V(ctx->VertexProgram.Temporaries[i], 0.0F, 0.0F, 0.0F, 0.0F);
      }
      ASSIGN_4V(ctx->VertexProgram.AddressReg, 0, 0, 0, 0);
   }
}



/**
 * Copy the 16 elements of a matrix into four consecutive program
 * registers starting at 'pos'.
 */
static void
load_matrix(GLfloat registers[][4], GLuint pos, const GLfloat mat[16])
{
   GLuint i;
   for (i = 0; i < 4; i++) {
      registers[pos + i][0] = mat[0 + i];
      registers[pos + i][1] = mat[4 + i];
      registers[pos + i][2] = mat[8 + i];
      registers[pos + i][3] = mat[12 + i];
   }
}


/**
 * As above, but transpose the matrix.
 */
static void
load_transpose_matrix(GLfloat registers[][4], GLuint pos,
                      const GLfloat mat[16])
{
   MEMCPY(registers[pos], mat, 16 * sizeof(GLfloat));
}


/**
 * Load program parameter registers with tracked matrices (if NV program)
 * or GL state values (if ARB program).
 * This needs to be done per glBegin/glEnd, not per-vertex.
 */
void
_mesa_init_vp_per_primitive_registers(GLcontext *ctx)
{
   if (ctx->VertexProgram.Current->IsNVProgram) {
      GLuint i;

      for (i = 0; i < MAX_NV_VERTEX_PROGRAM_PARAMS / 4; i++) {
         /* point 'mat' at source matrix */
         GLmatrix *mat;
         if (ctx->VertexProgram.TrackMatrix[i] == GL_MODELVIEW) {
            mat = ctx->ModelviewMatrixStack.Top;
         }
         else if (ctx->VertexProgram.TrackMatrix[i] == GL_PROJECTION) {
            mat = ctx->ProjectionMatrixStack.Top;
         }
         else if (ctx->VertexProgram.TrackMatrix[i] == GL_TEXTURE) {
            mat = ctx->TextureMatrixStack[ctx->Texture.CurrentUnit].Top;
         }
         else if (ctx->VertexProgram.TrackMatrix[i] == GL_COLOR) {
            mat = ctx->ColorMatrixStack.Top;
         }
         else if (ctx->VertexProgram.TrackMatrix[i]==GL_MODELVIEW_PROJECTION_NV) {
            /* XXX verify the combined matrix is up to date */
            mat = &ctx->_ModelProjectMatrix;
         }
         else if (ctx->VertexProgram.TrackMatrix[i] >= GL_MATRIX0_NV &&
                  ctx->VertexProgram.TrackMatrix[i] <= GL_MATRIX7_NV) {
            GLuint n = ctx->VertexProgram.TrackMatrix[i] - GL_MATRIX0_NV;
            ASSERT(n < MAX_PROGRAM_MATRICES);
            mat = ctx->ProgramMatrixStack[n].Top;
         }
         else {
            /* no matrix is tracked, but we leave the register values as-is */
            assert(ctx->VertexProgram.TrackMatrix[i] == GL_NONE);
            continue;
         }

         /* load the matrix */
         if (ctx->VertexProgram.TrackMatrixTransform[i] == GL_IDENTITY_NV) {
            load_matrix(ctx->VertexProgram.Parameters, i*4, mat->m);
         }
         else if (ctx->VertexProgram.TrackMatrixTransform[i] == GL_INVERSE_NV) {
            _math_matrix_analyse(mat); /* update the inverse */
            ASSERT(!_math_matrix_is_dirty(mat));
            load_matrix(ctx->VertexProgram.Parameters, i*4, mat->inv);
         }
         else if (ctx->VertexProgram.TrackMatrixTransform[i] == GL_TRANSPOSE_NV) {
            load_transpose_matrix(ctx->VertexProgram.Parameters, i*4, mat->m);
         }
         else {
            assert(ctx->VertexProgram.TrackMatrixTransform[i]
                   == GL_INVERSE_TRANSPOSE_NV);
            _math_matrix_analyse(mat); /* update the inverse */
            ASSERT(!_math_matrix_is_dirty(mat));
            load_transpose_matrix(ctx->VertexProgram.Parameters, i*4, mat->inv);
         }
      }
   }
   else {
      /* Using and ARB vertex program */
      if (ctx->VertexProgram.Current->Base.Parameters) {
         /* Grab the state GL state and put into registers */
         _mesa_load_state_parameters(ctx,
                                 ctx->VertexProgram.Current->Base.Parameters);
      }
   }
}



/**
 * For debugging.  Dump the current vertex program machine registers.
 */
void
_mesa_dump_vp_state( const struct gl_vertex_program_state *state )
{
   int i;
   _mesa_printf("VertexIn:\n");
   for (i = 0; i < MAX_NV_VERTEX_PROGRAM_INPUTS; i++) {
      _mesa_printf("%d: %f %f %f %f   ", i,
                   state->Inputs[i][0],
                   state->Inputs[i][1],
                   state->Inputs[i][2],
                   state->Inputs[i][3]);
   }
   _mesa_printf("\n");

   _mesa_printf("VertexOut:\n");
   for (i = 0; i < MAX_NV_VERTEX_PROGRAM_OUTPUTS; i++) {
      _mesa_printf("%d: %f %f %f %f   ", i,
                  state->Outputs[i][0],
                  state->Outputs[i][1],
                  state->Outputs[i][2],
                  state->Outputs[i][3]);
   }
   _mesa_printf("\n");

   _mesa_printf("Registers:\n");
   for (i = 0; i < MAX_NV_VERTEX_PROGRAM_TEMPS; i++) {
      _mesa_printf("%d: %f %f %f %f   ", i,
                  state->Temporaries[i][0],
                  state->Temporaries[i][1],
                  state->Temporaries[i][2],
                  state->Temporaries[i][3]);
   }
   _mesa_printf("\n");

   _mesa_printf("Parameters:\n");
   for (i = 0; i < MAX_NV_VERTEX_PROGRAM_PARAMS; i++) {
      _mesa_printf("%d: %f %f %f %f   ", i,
                  state->Parameters[i][0],
                  state->Parameters[i][1],
                  state->Parameters[i][2],
                  state->Parameters[i][3]);
   }
   _mesa_printf("\n");
}



/**
 * 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_vertex_program_state *state )
{
   if (source->RelAddr) {
      const GLint reg = source->Index + state->AddressReg[0];
      ASSERT( (source->File == PROGRAM_ENV_PARAM) || 
        (source->File == PROGRAM_STATE_VAR) );
      if (reg < 0 || reg > MAX_NV_VERTEX_PROGRAM_PARAMS)
         return ZeroVec;
      else if (source->File == PROGRAM_ENV_PARAM)
         return state->Parameters[reg];
      else
         return state->Current->Base.Parameters->ParameterValues[reg];
   }
   else {
      switch (source->File) {
         case PROGRAM_TEMPORARY:
            ASSERT(source->Index < MAX_NV_VERTEX_PROGRAM_TEMPS);
            return state->Temporaries[source->Index];
         case PROGRAM_INPUT:
            ASSERT(source->Index < MAX_NV_VERTEX_PROGRAM_INPUTS);
            return state->Inputs[source->Index];
         case PROGRAM_OUTPUT:
            /* This is only needed for the PRINT instruction */
            ASSERT(source->Index < MAX_NV_VERTEX_PROGRAM_OUTPUTS);
            return state->Outputs[source->Index];
         case PROGRAM_LOCAL_PARAM:
            ASSERT(source->Index < MAX_PROGRAM_LOCAL_PARAMS);
            return state->Current->Base.LocalParams[source->Index];
         case PROGRAM_ENV_PARAM:
            ASSERT(source->Index < MAX_NV_VERTEX_PROGRAM_PARAMS);
            return state->Parameters[source->Index];
         case PROGRAM_STATE_VAR:
            ASSERT(source->Index < state->Current->Base.Parameters->NumParameters);
            return state->Current->Base.Parameters->ParameterValues[source->Index];
         default:
            _mesa_problem(NULL,
                          "Bad source register file in get_register_pointer");
            return NULL;
      }
   }
   return NULL;
}


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

   if (source->NegateBase) {
      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)];
   }
   else {
      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)];
   }
}



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

   if (source->NegateBase) {
      result[0] = -src[GET_SWZ(source->Swizzle, 0)];
   }
   else {
      result[0] = src[GET_SWZ(source->Swizzle, 0)];
   }
}


/**
 * Store 4 floats into a register.
 */
static void
store_vector4( const struct prog_dst_register *dest,
               struct gl_vertex_program_state *state,
               const GLfloat value[4] )
{
   GLfloat *dst;
   switch (dest->File) {
      case PROGRAM_TEMPORARY:
         dst = state->Temporaries[dest->Index];
         break;
      case PROGRAM_OUTPUT:
         dst = state->Outputs[dest->Index];
         break;
      case PROGRAM_ENV_PARAM:
         {
            /* a slight hack */
            GET_CURRENT_CONTEXT(ctx);
            dst = ctx->VertexProgram.Parameters[dest->Index];
         }
         break;
      default:
         _mesa_problem(NULL, "Invalid register file in store_vector4(file=%d)",
                       dest->File);
         return;
   }

   if (dest->WriteMask & WRITEMASK_X)
      dst[0] = value[0];
   if (dest->WriteMask & WRITEMASK_Y)
      dst[1] = value[1];
   if (dest->WriteMask & WRITEMASK_Z)
      dst[2] = value[2];
   if (dest->WriteMask & WRITEMASK_W)
      dst[3] = value[3];
}


/**
 * 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


/**
 * Execute the given vertex program
 */
void
_mesa_exec_vertex_program(GLcontext *ctx, const struct vertex_program *program)
{
   struct gl_vertex_program_state *state = &ctx->VertexProgram;
   const struct prog_instruction *inst;

   ctx->_CurrentProgram = GL_VERTEX_PROGRAM_ARB; /* or NV, doesn't matter */

   /* If the program is position invariant, multiply the input position
    * by the MVP matrix and store in the vertex position result register.
    */
   if (ctx->VertexProgram.Current->IsPositionInvariant) {
      TRANSFORM_POINT( ctx->VertexProgram.Outputs[VERT_RESULT_HPOS], 
                       ctx->_ModelProjectMatrix.m, 
                       ctx->VertexProgram.Inputs[VERT_ATTRIB_POS]);

      /* XXX: This could go elsewhere */
      ctx->VertexProgram.Current->Base.OutputsWritten |= VERT_BIT_POS;
   }

   for (inst = program->Base.Instructions; ; inst++) {

      if (ctx->VertexProgram.CallbackEnabled &&
          ctx->VertexProgram.Callback) {
         ctx->VertexProgram.CurrentPosition = inst->StringPos;
         ctx->VertexProgram.Callback(program->Base.Target,
                                     ctx->VertexProgram.CallbackData);
      }

      switch (inst->Opcode) {
         case OPCODE_MOV:
            {
               GLfloat t[4];
               fetch_vector4( &inst->SrcReg[0], state, t );
               store_vector4( &inst->DstReg, state, t );
            }
            break;
         case OPCODE_LIT:
            {
               const GLfloat epsilon = 1.0F / 256.0F; /* per NV spec */
               GLfloat t[4], lit[4];
               fetch_vector4( &inst->SrcReg[0], state, t );
               t[0] = MAX2(t[0], 0.0F);
               t[1] = MAX2(t[1], 0.0F);
               t[3] = CLAMP(t[3], -(128.0F - epsilon), (128.0F - epsilon));
               lit[0] = 1.0;
               lit[1] = t[0];