prog_execute.c

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

   else {
      return GL_FALSE;
   }
}



/**
 * Store 4 floats into a register.  Observe the instructions saturate and
 * set-condition-code flags.
 */
static void
store_vector4(const struct prog_instruction *inst,
              struct gl_program_machine *machine, const GLfloat value[4])
{
   const struct prog_dst_register *dest = &(inst->DstReg);
   const GLboolean clamp = inst->SaturateMode == SATURATE_ZERO_ONE;
   GLfloat *dstReg;
   GLfloat dummyReg[4];
   GLfloat clampedValue[4];
   GLuint writeMask = dest->WriteMask;

   switch (dest->File) {
   case PROGRAM_OUTPUT:
      ASSERT(dest->Index < MAX_PROGRAM_OUTPUTS);
      dstReg = machine->Outputs[dest->Index];
      break;
   case PROGRAM_TEMPORARY:
      ASSERT(dest->Index < MAX_PROGRAM_TEMPS);
      dstReg = machine->Temporaries[dest->Index];
      break;
   case PROGRAM_WRITE_ONLY:
      dstReg = dummyReg;
      return;
   default:
      _mesa_problem(NULL, "bad register file in store_vector4(fp)");
      return;
   }

#if 0
   if (value[0] > 1.0e10 ||
       IS_INF_OR_NAN(value[0]) ||
       IS_INF_OR_NAN(value[1]) ||
       IS_INF_OR_NAN(value[2]) || IS_INF_OR_NAN(value[3]))
      printf("store %g %g %g %g\n", value[0], value[1], value[2], value[3]);
#endif

   if (clamp) {
      clampedValue[0] = CLAMP(value[0], 0.0F, 1.0F);
      clampedValue[1] = CLAMP(value[1], 0.0F, 1.0F);
      clampedValue[2] = CLAMP(value[2], 0.0F, 1.0F);
      clampedValue[3] = CLAMP(value[3], 0.0F, 1.0F);
      value = clampedValue;
   }

   if (dest->CondMask != COND_TR) {
      /* condition codes may turn off some writes */
      if (writeMask & WRITEMASK_X) {
         if (!test_cc(machine->CondCodes[GET_SWZ(dest->CondSwizzle, 0)],
                      dest->CondMask))
            writeMask &= ~WRITEMASK_X;
      }
      if (writeMask & WRITEMASK_Y) {
         if (!test_cc(machine->CondCodes[GET_SWZ(dest->CondSwizzle, 1)],
                      dest->CondMask))
            writeMask &= ~WRITEMASK_Y;
      }
      if (writeMask & WRITEMASK_Z) {
         if (!test_cc(machine->CondCodes[GET_SWZ(dest->CondSwizzle, 2)],
                      dest->CondMask))
            writeMask &= ~WRITEMASK_Z;
      }
      if (writeMask & WRITEMASK_W) {
         if (!test_cc(machine->CondCodes[GET_SWZ(dest->CondSwizzle, 3)],
                      dest->CondMask))
            writeMask &= ~WRITEMASK_W;
      }
   }

   if (writeMask & WRITEMASK_X)
      dstReg[0] = value[0];
   if (writeMask & WRITEMASK_Y)
      dstReg[1] = value[1];
   if (writeMask & WRITEMASK_Z)
      dstReg[2] = value[2];
   if (writeMask & WRITEMASK_W)
      dstReg[3] = value[3];

   if (inst->CondUpdate) {
      if (writeMask & WRITEMASK_X)
         machine->CondCodes[0] = generate_cc(value[0]);
      if (writeMask & WRITEMASK_Y)
         machine->CondCodes[1] = generate_cc(value[1]);
      if (writeMask & WRITEMASK_Z)
         machine->CondCodes[2] = generate_cc(value[2]);
      if (writeMask & WRITEMASK_W)
         machine->CondCodes[3] = generate_cc(value[3]);
#if DEBUG_PROG
      printf("CondCodes=(%s,%s,%s,%s) for:\n",
             _mesa_condcode_string(machine->CondCodes[0]),
             _mesa_condcode_string(machine->CondCodes[1]),
             _mesa_condcode_string(machine->CondCodes[2]),
             _mesa_condcode_string(machine->CondCodes[3]));
#endif
   }
}


/**
 * Execute the given vertex/fragment program.
 *
 * \param ctx  rendering context
 * \param program  the program to execute
 * \param machine  machine state (must be initialized)
 * \return GL_TRUE if program completed or GL_FALSE if program executed KIL.
 */
GLboolean
_mesa_execute_program(GLcontext * ctx,
                      const struct gl_program *program,
                      struct gl_program_machine *machine)
{
   const GLuint numInst = program->NumInstructions;
   const GLuint maxExec = 10000;
   GLint pc, numExec = 0;

   machine->CurProgram = program;

   if (DEBUG_PROG) {
      printf("execute program %u --------------------\n", program->Id);
   }

#if FEATURE_MESA_program_debug
   CurrentMachine = machine;
#endif

   if (program->Target == GL_VERTEX_PROGRAM_ARB) {
      machine->EnvParams = ctx->VertexProgram.Parameters;
   }
   else {
      machine->EnvParams = ctx->FragmentProgram.Parameters;
   }

   for (pc = 0; pc < numInst; pc++) {
      const struct prog_instruction *inst = program->Instructions + pc;

#if FEATURE_MESA_program_debug
      if (ctx->FragmentProgram.CallbackEnabled &&
          ctx->FragmentProgram.Callback) {
         ctx->FragmentProgram.CurrentPosition = inst->StringPos;
         ctx->FragmentProgram.Callback(program->Target,
                                       ctx->FragmentProgram.CallbackData);
      }
#endif

      if (DEBUG_PROG) {
         _mesa_print_instruction(inst);
      }

      switch (inst->Opcode) {
      case OPCODE_ABS:
         {
            GLfloat a[4], result[4];
            fetch_vector4(&inst->SrcReg[0], machine, a);
            result[0] = FABSF(a[0]);
            result[1] = FABSF(a[1]);
            result[2] = FABSF(a[2]);
            result[3] = FABSF(a[3]);
            store_vector4(inst, machine, result);
         }
         break;
      case OPCODE_ADD:
         {
            GLfloat a[4], b[4], result[4];
            fetch_vector4(&inst->SrcReg[0], machine, a);
            fetch_vector4(&inst->SrcReg[1], machine, b);
            result[0] = a[0] + b[0];
            result[1] = a[1] + b[1];
            result[2] = a[2] + b[2];
            result[3] = a[3] + b[3];
            store_vector4(inst, machine, result);
            if (DEBUG_PROG) {
               printf("ADD (%g %g %g %g) = (%g %g %g %g) + (%g %g %g %g)\n",
                      result[0], result[1], result[2], result[3],
                      a[0], a[1], a[2], a[3], b[0], b[1], b[2], b[3]);
            }
         }
         break;
      case OPCODE_ARL:
         {
            GLfloat t[4];
            fetch_vector4(&inst->SrcReg[0], machine, t);
            machine->AddressReg[0][0] = (GLint) FLOORF(t[0]);
         }
         break;
      case OPCODE_BGNLOOP:
         /* no-op */
         break;
      case OPCODE_ENDLOOP:
         /* subtract 1 here since pc is incremented by for(pc) loop */
         pc = inst->BranchTarget - 1;   /* go to matching BNGLOOP */
         break;
      case OPCODE_BGNSUB:      /* begin subroutine */
         break;
      case OPCODE_ENDSUB:      /* end subroutine */
         break;
      case OPCODE_BRA:         /* branch (conditional) */
         /* fall-through */
      case OPCODE_BRK:         /* break out of loop (conditional) */
         /* fall-through */
      case OPCODE_CONT:        /* continue loop (conditional) */
         if (eval_condition(machine, inst)) {
            /* take branch */
            /* Subtract 1 here since we'll do pc++ at end of for-loop */
            pc = inst->BranchTarget - 1;
         }
         break;
      case OPCODE_CAL:         /* Call subroutine (conditional) */
         if (eval_condition(machine, inst)) {
            /* call the subroutine */
            if (machine->StackDepth >= MAX_PROGRAM_CALL_DEPTH) {
               return GL_TRUE;  /* Per GL_NV_vertex_program2 spec */
            }
            machine->CallStack[machine->StackDepth++] = pc + 1; /* next inst */
            /* Subtract 1 here since we'll do pc++ at end of for-loop */
            pc = inst->BranchTarget - 1;
         }
         break;
      case OPCODE_CMP:
         {
            GLfloat a[4], b[4], c[4], result[4];
            fetch_vector4(&inst->SrcReg[0], machine, a);
            fetch_vector4(&inst->SrcReg[1], machine, b);
            fetch_vector4(&inst->SrcReg[2], machine, c);
            result[0] = a[0] < 0.0F ? b[0] : c[0];
            result[1] = a[1] < 0.0F ? b[1] : c[1];
            result[2] = a[2] < 0.0F ? b[2] : c[2];
            result[3] = a[3] < 0.0F ? b[3] : c[3];
            store_vector4(inst, machine, result);
         }
         break;
      case OPCODE_COS:
         {
            GLfloat a[4], result[4];
            fetch_vector1(&inst->SrcReg[0], machine, a);
            result[0] = result[1] = result[2] = result[3]
               = (GLfloat) _mesa_cos(a[0]);
            store_vector4(inst, machine, result);
         }
         break;
      case OPCODE_DDX:         /* Partial derivative with respect to X */
         {
            GLfloat result[4];
            fetch_vector4_deriv(ctx, &inst->SrcReg[0], machine,
                                'X', result);
            store_vector4(inst, machine, result);
         }
         break;
      case OPCODE_DDY:         /* Partial derivative with respect to Y */
         {
            GLfloat result[4];
            fetch_vector4_deriv(ctx, &inst->SrcReg[0], machine,
                                'Y', result);
            store_vector4(inst, machine, result);
         }
         break;
      case OPCODE_DP3:
         {
            GLfloat a[4], b[4], result[4];
            fetch_vector4(&inst->SrcReg[0], machine, a);
            fetch_vector4(&inst->SrcReg[1], machine, b);
            result[0] = result[1] = result[2] = result[3] = DOT3(a, b);
            store_vector4(inst, machine, result);
            if (DEBUG_PROG) {
               printf("DP3 %g = (%g %g %g) . (%g %g %g)\n",
                      result[0], a[0], a[1], a[2], b[0], b[1], b[2]);
            }
         }
         break;
      case OPCODE_DP4:
         {
            GLfloat a[4], b[4], result[4];
            fetch_vector4(&inst->SrcReg[0], machine, a);
            fetch_vector4(&inst->SrcReg[1], machine, b);
            result[0] = result[1] = result[2] = result[3] = DOT4(a, b);
            store_vector4(inst, machine, result);
            if (DEBUG_PROG) {
               printf("DP4 %g = (%g, %g %g %g) . (%g, %g %g %g)\n",
                      result[0], a[0], a[1], a[2], a[3],
                      b[0], b[1], b[2], b[3]);
            }
         }
         break;
      case OPCODE_DPH:
         {
            GLfloat a[4], b[4], result[4];
            fetch_vector4(&inst->SrcReg[0], machine, a);
            fetch_vector4(&inst->SrcReg[1], machine, b);
            result[0] = result[1] = result[2] = result[3] =
               a[0] * b[0] + a[1] * b[1] + a[2] * b[2] + b[3];
            store_vector4(inst, machine, result);
         }
         break;
      case OPCODE_DST:         /* Distance vector */
         {
            GLfloat a[4], b[4], result[4];
            fetch_vector4(&inst->SrcReg[0], machine, a);
            fetch_vector4(&inst->SrcReg[1], machine, b);
            result[0] = 1.0F;
            result[1] = a[1] * b[1];
            result[2] = a[2];
            result[3] = b[3];
            store_vector4(inst, machine, result);
         }
         break;
      case OPCODE_EXP:
         {
            GLfloat t[4], q[4], floor_t0;
            fetch_vector1(&inst->SrcReg[0], machine, t);
            floor_t0 = FLOORF(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 {
               q[0] = LDEXPF(1.0, (int) floor_t0);
               /* Note: GL_NV_vertex_program expects 
                * result.z = result.x * APPX(result.y)
                * We do what the ARB extension says.
                */
               q[2] = (GLfloat) pow(2.0, t[0]);
            }
            q[1] = t[0] - floor_t0;
            q[3] = 1.0F;
            store_vector4( inst, machine, q );
         }
         break;
      case OPCODE_EX2:         /* Exponential base 2 */
         {
            GLfloat a[4], result[4];
            fetch_vector1(&inst->SrcReg[0], machine, a);
            result[0] = result[1] = result[2] = result[3] =
               (GLfloat) _mesa_pow(2.0, a[0]);
            store_vector4(inst, machine, result);
         }
         break;
      case OPCODE_FLR:
         {
            GLfloat a[4], result[4];
            fetch_vector4(&inst->SrcReg[0], machine, a);
            result[0] = FLOORF(a[0]);
            result[1] = FLOORF(a[1]);
            result[2] = FLOORF(a[2]);
            result[3] = FLOORF(a[3]);
            store_vector4(inst, machine, result);
         }
         break;
      case OPCODE_FRC:
         {
            GLfloat a[4], result[4];
            fetch_vector4(&inst->SrcReg[0], machine, a);
            result[0] = a[0] - FLOORF(a[0]);
            result[1] = a[1] - FLOORF(a[1]);
            result[2] = a[2] - FLOORF(a[2]);
            result[3] = a[3] - FLOORF(a[3]);
            store_vector4(inst, machine, result);
         }
         break;
      case OPCODE_IF:
         {
            GLboolean cond;
            /* eval condition */
            if (inst->SrcReg[0].File != PROGRAM_UNDEFINED) {
               GLfloat a[4];
               fetch_vector1(&inst->SrcReg[0], machine, a);
               cond = (a[0] != 0.0);
            }
            else {
               cond = eval_condition(machine, inst);
            }
            if (DEBUG_PROG) {
               printf("IF: %d\n", cond);
            }