nvvertexec.c

winNT技术操作系统,国外开放的原代码和LIUX一样

   if (source->Negate) {
      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 vp_src_register *source,
               const struct gl_vertex_program_state *state,
               GLfloat result[4] )
{
   const GLfloat *src = get_register_pointer(source, state);

   if (source->Negate) {
      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 vp_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 vp_instruction *inst;

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

   /* If the program is position invariant, multiply the input
    * position and the MVP matrix and stick it into the output pos slot
    */
   if (ctx->VertexProgram.Current->IsPositionInvariant) {
      TRANSFORM_POINT( ctx->VertexProgram.Outputs[0], 
                       ctx->_ModelProjectMatrix.m, 
                       ctx->VertexProgram.Inputs[0]);

      /* XXX: This could go elsewhere */
      ctx->VertexProgram.Current->OutputsWritten |= 0x1;
   }
   for (inst = program->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 VP_OPCODE_MOV:
            {
               GLfloat t[4];
               fetch_vector4( &inst->SrcReg[0], state, t );
               store_vector4( &inst->DstReg, state, t );
            }
            break;
         case VP_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];
               lit[2] = (t[0] > 0.0) ? (GLfloat) _mesa_pow(t[1], t[3]) : 0.0F;
               lit[3] = 1.0;
               store_vector4( &inst->DstReg, state, lit );
            }
            break;
         case VP_OPCODE_RCP:
            {
               GLfloat t[4];
               fetch_vector1( &inst->SrcReg[0], state, t );
               if (t[0] != 1.0F)
                  t[0] = 1.0F / t[0];  /* div by zero is infinity! */
               t[1] = t[2] = t[3] = t[0];
               store_vector4( &inst->DstReg, state, t );
            }
            break;
         case VP_OPCODE_RSQ:
            {
               GLfloat t[4];
               fetch_vector1( &inst->SrcReg[0], state, t );
               t[0] = INV_SQRTF(FABSF(t[0]));
               t[1] = t[2] = t[3] = t[0];
               store_vector4( &inst->DstReg, state, t );
            }
            break;
         case VP_OPCODE_EXP:
            {
               GLfloat t[4], q[4], floor_t0;
               fetch_vector1( &inst->SrcReg[0], state, t );
               floor_t0 = (float) floor(t[0]);
               if (floor_t0 > FLT_MAX_EXP) {
                  SET_POS_INFINITY(q[0]);
                  SET_POS_INFINITY(q[2]);
               }
               else if (floor_t0 < FLT_MIN_EXP) {
                  q[0] = 0.0F;
                  q[2] = 0.0F;
               }
               else {
#ifdef USE_IEEE
                  GLint ii = (GLint) floor_t0;
                  ii = (ii < 23) + 0x3f800000;
                  SET_FLOAT_BITS(q[0], ii);
                  q[0] = *((GLfloat *) (void *)&ii);
#else
                  q[0] = (GLfloat) pow(2.0, floor_t0);
#endif
                  q[2] = (GLfloat) (q[0] * LOG2(q[1]));
               }
               q[1] = t[0] - floor_t0;
               q[3] = 1.0F;
               store_vector4( &inst->DstReg, state, q );
            }
            break;
         case VP_OPCODE_LOG:
            {
               GLfloat t[4], q[4], abs_t0;
               fetch_vector1( &inst->SrcReg[0], state, t );
               abs_t0 = (GLfloat) fabs(t[0]);
               if (abs_t0 != 0.0F) {
                  /* Since we really can't handle infinite values on VMS
                   * like other OSes we'll use __MAXFLOAT to represent
                   * infinity.  This may need some tweaking.
                   */
#ifdef VMS
                  if (abs_t0 == __MAXFLOAT)
#else
                  if (IS_INF_OR_NAN(abs_t0))
#endif
                  {
                     SET_POS_INFINITY(q[0]);
                     q[1] = 1.0F;
                     SET_POS_INFINITY(q[2]);
                  }
                  else {
                     int exponent;
                     double mantissa = frexp(t[0], &exponent);
                     q[0] = (GLfloat) (exponent - 1);
                     q[1] = (GLfloat) (2.0 * mantissa); /* map [.5, 1) -> [1, 2) */
                     q[2] = (GLfloat) (q[0] + LOG2(q[1]));
                  }
                  }
               else {
                  SET_NEG_INFINITY(q[0]);
                  q[1] = 1.0F;
                  SET_NEG_INFINITY(q[2]);
               }
               q[3] = 1.0;
               store_vector4( &inst->DstReg, state, q );
            }
            break;
         case VP_OPCODE_MUL:
            {
               GLfloat t[4], u[4], prod[4];
               fetch_vector4( &inst->SrcReg[0], state, t );
               fetch_vector4( &inst->SrcReg[1], state, u );
               prod[0] = t[0] * u[0];
               prod[1] = t[1] * u[1];
               prod[2] = t[2] * u[2];
               prod[3] = t[3] * u[3];
               store_vector4( &inst->DstReg, state, prod );
            }
            break;
         case VP_OPCODE_ADD:
            {
               GLfloat t[4], u[4], sum[4];
               fetch_vector4( &inst->SrcReg[0], state, t );
               fetch_vector4( &inst->SrcReg[1], state, u );
               sum[0] = t[0] + u[0];
               sum[1] = t[1] + u[1];
               sum[2] = t[2] + u[2];
               sum[3] = t[3] + u[3];
               store_vector4( &inst->DstReg, state, sum );
            }
            break;
         case VP_OPCODE_DP3:
            {
               GLfloat t[4], u[4], dot[4];
               fetch_vector4( &inst->SrcReg[0], state, t );
               fetch_vector4( &inst->SrcReg[1], state, u );
               dot[0] = t[0] * u[0] + t[1] * u[1] + t[2] * u[2];
               dot[1] = dot[2] = dot[3] = dot[0];
               store_vector4( &inst->DstReg, state, dot );
            }
            break;
         case VP_OPCODE_DP4:
            {
               GLfloat t[4], u[4], dot[4];
               fetch_vector4( &inst->SrcReg[0], state, t );
               fetch_vector4( &inst->SrcReg[1], state, u );
               dot[0] = t[0] * u[0] + t[1] * u[1] + t[2] * u[2] + t[3] * u[3];
               dot[1] = dot[2] = dot[3] = dot[0];
               store_vector4( &inst->DstReg, state, dot );
            }
            break;
         case VP_OPCODE_DST:
            {
               GLfloat t[4], u[4], dst[4];
               fetch_vector4( &inst->SrcReg[0], state, t );
               fetch_vector4( &inst->SrcReg[1], state, u );
               dst[0] = 1.0F;
               dst[1] = t[1] * u[1];
               dst[2] = t[2];