i915_fragprog.c

mesa-6.5-minigui源码

	 i915_emit_arith( p,
			 A0_MUL,
			 tmp, A0_DEST_CHANNEL_X, 0,
			 tmp, src1, 0);


	 i915_emit_arith( p,
			 A0_EXP,
			 get_result_vector( p, inst ), 
			 flags, 0,
			 swizzle(tmp,X,X,X,X), 0, 0);

	 break;

      case OPCODE_RCP: 
	 src0 = src_vector( p, &inst->SrcReg[0], program);

	 i915_emit_arith( p, 
			 A0_RCP,
			 get_result_vector( p, inst ), 
			 get_result_flags( inst ), 0,
			 swizzle(src0,X,X,X,X), 0, 0);
	 break;

      case OPCODE_RSQ: 

	 src0 = src_vector( p, &inst->SrcReg[0], program);

	 i915_emit_arith( p, 
			 A0_RSQ,
			 get_result_vector( p, inst ), 
			 get_result_flags( inst ), 0,
			 swizzle(src0,X,X,X,X), 0, 0);
	 break;
	 
      case OPCODE_SCS:
	 src0 = src_vector( p, &inst->SrcReg[0], program);
	 tmp = i915_get_utemp( p );

	 /* 
	  * t0.xy = MUL x.xx11, x.x1111  ; x^2, x, 1, 1
	  * t0 = MUL t0.xyxy t0.xx11 ; x^4, x^3, x^2, x
	  * t1 = MUL t0.xyyw t0.yz11    ; x^7 x^5 x^3 x
	  * scs.x = DP4 t1, sin_constants
	  * t1 = MUL t0.xxz1 t0.z111    ; x^6 x^4 x^2 1
	  * scs.y = DP4 t1, cos_constants
	  */
	 i915_emit_arith( p, 
			 A0_MUL,
			 tmp, A0_DEST_CHANNEL_XY, 0,
			 swizzle(src0, X,X,ONE,ONE), 
			 swizzle(src0, X,ONE,ONE,ONE), 0);

	 i915_emit_arith( p, 
			 A0_MUL,
			 tmp, A0_DEST_CHANNEL_ALL, 0,
			 swizzle(tmp, X,Y,X,Y), 
			 swizzle(tmp, X,X,ONE,ONE), 0);

	 if (inst->DstReg.WriteMask & WRITEMASK_Y) {
	    GLuint tmp1;
	    
	    if (inst->DstReg.WriteMask & WRITEMASK_X)
	       tmp1 = i915_get_utemp( p );
	    else
	       tmp1 = tmp;

	    i915_emit_arith( p, 
			    A0_MUL,
			    tmp1, A0_DEST_CHANNEL_ALL, 0,
			    swizzle(tmp, X,Y,Y,W), 
			    swizzle(tmp, X,Z,ONE,ONE), 0);
	    
	    i915_emit_arith( p, 
			    A0_DP4,
			    get_result_vector( p, inst ), 
			    A0_DEST_CHANNEL_Y, 0,
			    swizzle(tmp1, W,Z,Y,X),
			    i915_emit_const4fv( p, sin_constants ), 0);
	 }

	 if (inst->DstReg.WriteMask & WRITEMASK_X) {
	    i915_emit_arith( p, 
			    A0_MUL,
			    tmp, A0_DEST_CHANNEL_XYZ, 0,
			    swizzle(tmp, X,X,Z,ONE), 
			    swizzle(tmp, Z,ONE,ONE,ONE), 0);
	    
	    i915_emit_arith( p, 
			    A0_DP4,
			    get_result_vector( p, inst ), 
			    A0_DEST_CHANNEL_X, 0,
			    swizzle(tmp, ONE,Z,Y,X), 
			    i915_emit_const4fv( p, cos_constants ), 0);
	 }
	 break;

      case OPCODE_SGE: 
	 EMIT_2ARG_ARITH( A0_SGE );
	 break;

      case OPCODE_SIN:
	 src0 = src_vector( p, &inst->SrcReg[0], program);
	 tmp = i915_get_utemp( p );

	 i915_emit_arith( p, 
			 A0_MUL,
			 tmp, A0_DEST_CHANNEL_X, 0,
			 src0, 
			 i915_emit_const1f(p, 1.0/(M_PI * 2)),
			 0);

	 i915_emit_arith( p, 
			 A0_MOD,
			 tmp, A0_DEST_CHANNEL_X, 0,
			 tmp, 
			 0, 0 );

	 /* By choosing different taylor constants, could get rid of this mul:
	  */
	 i915_emit_arith( p, 
			 A0_MUL,
			 tmp, A0_DEST_CHANNEL_X, 0,
			 tmp, 
			 i915_emit_const1f(p, (M_PI * 2)),
			 0);

	 /* 
	  * t0.xy = MUL x.xx11, x.x1111  ; x^2, x, 1, 1
	  * t0 = MUL t0.xyxy t0.xx11 ; x^4, x^3, x^2, x
	  * t1 = MUL t0.xyyw t0.yz11    ; x^7 x^5 x^3 x
	  * result = DP4 t1.wzyx, sin_constants
	  */
	 i915_emit_arith( p, 
			 A0_MUL,
			 tmp, A0_DEST_CHANNEL_XY, 0,
			 swizzle(tmp, X,X,ONE,ONE), 
			 swizzle(tmp, X,ONE,ONE,ONE), 0);

	 i915_emit_arith( p, 
			 A0_MUL,
			 tmp, A0_DEST_CHANNEL_ALL, 0,
			 swizzle(tmp, X,Y,X,Y), 
			 swizzle(tmp, X,X,ONE,ONE), 0);

	 i915_emit_arith( p, 
			 A0_MUL,
			 tmp, A0_DEST_CHANNEL_ALL, 0,
			 swizzle(tmp, X,Y,Y,W), 
			 swizzle(tmp, X,Z,ONE,ONE), 0);
	    
	 i915_emit_arith( p, 
			 A0_DP4,
			 get_result_vector( p, inst ), 
			 get_result_flags( inst ), 0,
			 swizzle(tmp, W, Z, Y, X ),
			 i915_emit_const4fv( p, sin_constants ), 0);
	 break;

      case OPCODE_SLT: 
	 EMIT_2ARG_ARITH( A0_SLT );
	 break;

      case OPCODE_SUB: 
	 src0 = src_vector( p, &inst->SrcReg[0], program);
	 src1 = src_vector( p, &inst->SrcReg[1], program);

	 i915_emit_arith( p, 
			 A0_ADD,
			 get_result_vector( p, inst ), 
			 get_result_flags( inst ), 0,
			 src0, negate(src1, 1,1,1,1), 0);
	 break;

      case OPCODE_SWZ: 
	 EMIT_1ARG_ARITH( A0_MOV ); /* extended swizzle handled natively */
	 break;

      case OPCODE_TEX: 
	 EMIT_TEX( T0_TEXLD );
	 break;

      case OPCODE_TXB:
	 EMIT_TEX( T0_TEXLDB );
	 break;

      case OPCODE_TXP:
	 EMIT_TEX( T0_TEXLDP );
	 break;

      case OPCODE_XPD:
	 /* Cross product:
	  *      result.x = src0.y * src1.z - src0.z * src1.y;
	  *      result.y = src0.z * src1.x - src0.x * src1.z;
	  *      result.z = src0.x * src1.y - src0.y * src1.x;
	  *      result.w = undef;
	  */
	 src0 = src_vector( p, &inst->SrcReg[0], program);
	 src1 = src_vector( p, &inst->SrcReg[1], program);
	 tmp = i915_get_utemp( p );
	 
	 i915_emit_arith( p, 
			 A0_MUL,
			 tmp, A0_DEST_CHANNEL_ALL, 0,
			 swizzle(src0,Z,X,Y,ONE), 
			 swizzle(src1,Y,Z,X,ONE), 0);

	 i915_emit_arith( p, 
			 A0_MAD,
			 get_result_vector( p, inst ), 
			 get_result_flags( inst ), 0,
			 swizzle(src0,Y,Z,X,ONE), 
			 swizzle(src1,Z,X,Y,ONE), 
			 negate(tmp,1,1,1,0));
	 break;

      case OPCODE_END:
	 return;
	 
      default:
	 i915_program_error( p, "bad opcode" );
	 return;
      }

      inst++;
      i915_release_utemps( p ); 
   }
}

/* Rather than trying to intercept and jiggle depth writes during
 * emit, just move the value into its correct position at the end of
 * the program:
 */
static void fixup_depth_write( struct i915_fragment_program *p )
{
   if (p->depth_written) {
      GLuint depth = UREG(REG_TYPE_OD, 0);

      i915_emit_arith( p, 
		      A0_MOV,
		      depth, A0_DEST_CHANNEL_W, 0,
		      swizzle(depth,X,Y,Z,Z), 
		      0, 0);
   }
}


#define FRAG_BIT_TEX(n)  (FRAG_BIT_TEX0 << (n))


static void check_wpos( struct i915_fragment_program *p )
{
   GLuint inputs = p->FragProg.Base.InputsRead;
   GLint i;

   p->wpos_tex = -1;

   for (i = 0; i < p->ctx->Const.MaxTextureCoordUnits; i++) {
      if (inputs & FRAG_BIT_TEX(i)) 
	 continue;
      else if (inputs & FRAG_BIT_WPOS) {
	 p->wpos_tex = i;
	 inputs &= ~FRAG_BIT_WPOS;
      }   
   }

   if (inputs & FRAG_BIT_WPOS) {
      i915_program_error(p, "No free texcoord for wpos value");
   }
}


static void translate_program( struct i915_fragment_program *p )
{
   struct i915_context *i915 = I915_CONTEXT(p->ctx);
   
   i915_init_program( i915, p );
   check_wpos( p ); 
   upload_program( p );
   fixup_depth_write( p );
   i915_fini_program( p ); 
   
   p->translated = 1;
}


static void track_params( struct i915_fragment_program *p )
{
   GLint i;

   if (p->nr_params)
      _mesa_load_state_parameters(p->ctx, p->FragProg.Base.Parameters); 

   for (i = 0; i < p->nr_params; i++) {
      GLint reg = p->param[i].reg;
      COPY_4V( p->constant[reg], p->param[i].values );
   }
   
   p->params_uptodate = 1;
   p->on_hardware = 0;		/* overkill */
}


static void i915BindProgram( GLcontext *ctx,
			    GLenum target, 
			    struct program *prog )
{
   if (target == GL_FRAGMENT_PROGRAM_ARB) {
      struct i915_context *i915 = I915_CONTEXT(ctx);
      struct i915_fragment_program *p = (struct i915_fragment_program *)prog;

      if (i915->current_program == p) 
	 return;
      
      if (i915->current_program) {
	 i915->current_program->on_hardware = 0;
	 i915->current_program->params_uptodate = 0;
      }
      
      i915->current_program = p;

      assert(p->on_hardware == 0);
      assert(p->params_uptodate == 0);

      /* Hack: make sure fog is correctly enabled according to this
       * fragment program's fog options.
       */
      ctx->Driver.Enable( ctx, GL_FRAGMENT_PROGRAM_ARB, 
			  ctx->FragmentProgram.Enabled );
   }
}

static struct program *i915NewProgram( GLcontext *ctx,
				      GLenum target, 
				      GLuint id )
{
   switch (target) {
   case GL_VERTEX_PROGRAM_ARB:
      return _mesa_init_vertex_program( ctx, CALLOC_STRUCT(vertex_program),
					target, id );

   case GL_FRAGMENT_PROGRAM_ARB: {
      struct i915_fragment_program *prog = CALLOC_STRUCT(i915_fragment_program);
      if (prog) {
	 i915_init_program( I915_CONTEXT(ctx), prog );

	 return _mesa_init_fragment_program( ctx, &prog->FragProg,
					     target, id );
      }
      else
	 return NULL;
   }

   default:
      /* Just fallback:
       */
      return _mesa_new_program( ctx, target, id );
   }
}

static void i915DeleteProgram( GLcontext *ctx,
			      struct program *prog )
{
   if (prog->Target == GL_FRAGMENT_PROGRAM_ARB) {
      struct i915_context *i915 = I915_CONTEXT(ctx);
      struct i915_fragment_program *p = (struct i915_fragment_program *)prog;
      
      if (i915->current_program == p) 
	 i915->current_program = 0;
   }

   _mesa_delete_program( ctx, prog );
}


static GLboolean i915IsProgramNative( GLcontext *ctx,
				     GLenum target, 
				     struct program *prog )
{
   if (target == GL_FRAGMENT_PROGRAM_ARB) {
      struct i915_fragment_program *p = (struct i915_fragment_program *)prog;

      if (!p->translated)
	 translate_program( p );
      
      return !p->error;
   }
   else
      return GL_TRUE;
}

static void i915ProgramStringNotify( GLcontext *ctx,
				    GLenum target,
				    struct program *prog )
{
   if (target == GL_FRAGMENT_PROGRAM_ARB) {
      struct i915_fragment_program *p = (struct i915_fragment_program *)prog;
      p->translated = 0;

      /* Hack: make sure fog is correctly enabled according to this
       * fragment program's fog options.
       */
      ctx->Driver.Enable( ctx, GL_FRAGMENT_PROGRAM_ARB, 
			  ctx->FragmentProgram.Enabled );
   }
}


void i915ValidateFragmentProgram( struct i915_context *i915 )
{
   GLcontext *ctx = &i915->intel.ctx;
   struct intel_context *intel = intel_context(ctx);
   TNLcontext *tnl = TNL_CONTEXT(ctx);
   struct vertex_buffer *VB = &tnl->vb;

   struct i915_fragment_program *p = 
      (struct i915_fragment_program *)ctx->FragmentProgram._Current;

   const GLuint inputsRead = p->FragProg.Base.InputsRead;
   GLuint s4 = i915->state.Ctx[I915_CTXREG_LIS4] & ~S4_VFMT_MASK;
   GLuint s2 = S2_TEXCOORD_NONE;
   int i, offset = 0;

   /* Important:
    */
   VB->AttribPtr[VERT_ATTRIB_POS] = VB->NdcPtr;

   if (!p->translated) 
      translate_program( p );

   intel->vertex_attr_count = 0;
   intel->wpos_offset = 0;
   intel->wpos_size = 0;
   intel->coloroffset = 0;
   intel->specoffset = 0;

   if (inputsRead & FRAG_BITS_TEX_ANY) {
      EMIT_ATTR( _TNL_ATTRIB_POS, EMIT_4F_VIEWPORT, S4_VFMT_XYZW, 16 );
   }
   else {
      EMIT_ATTR( _TNL_ATTRIB_POS, EMIT_3F_VIEWPORT, S4_VFMT_XYZ, 12 );
   }

   if (inputsRead & FRAG_BIT_COL0) {
      intel->coloroffset = offset / 4;
      EMIT_ATTR( _TNL_ATTRIB_COLOR0, EMIT_4UB_4F_BGRA, S4_VFMT_COLOR, 4 );
   }
   
   if ((inputsRead & (FRAG_BIT_COL1|FRAG_BIT_FOGC)) || 
       i915->vertex_fog != I915_FOG_NONE) {

      if (inputsRead & FRAG_BIT_COL1) {
	 intel->specoffset = offset / 4;
	 EMIT_ATTR( _TNL_ATTRIB_COLOR1, EMIT_3UB_3F_BGR, S4_VFMT_SPEC_FOG, 3 );
      }
      else
	 EMIT_PAD(3);

      if ((inputsRead & FRAG_BIT_FOGC) || i915->vertex_fog != I915_FOG_NONE) 
	 EMIT_ATTR( _TNL_ATTRIB_FOG, EMIT_1UB_1F, S4_VFMT_SPEC_FOG, 1 );
      else
	 EMIT_PAD( 1 );
   }

#if 0
   if ((inputsRead & FRAG_BIT_FOGC) || i915->vertex_fog != I915_FOG_NONE) {
      EMIT_ATTR( _TNL_ATTRIB_FOG, EMIT_1F, S4_VFMT_FOG_PARAM, 4 );
   }
#endif

   for (i = 0; i < p->ctx->Const.MaxTextureCoordUnits; i++) {
      if (inputsRead & FRAG_BIT_TEX(i)) {
	 int sz = VB->TexCoordPtr[i]->size;
	    
	 s2 &= ~S2_TEXCOORD_FMT(i, S2_TEXCOORD_FMT0_MASK);
	 s2 |= S2_TEXCOORD_FMT(i, SZ_TO_HW(sz));

	 EMIT_ATTR( _TNL_ATTRIB_TEX0+i, EMIT_SZ(sz), 0, sz * 4 );
      }
      else if (i == p->wpos_tex) {
	
	 /* If WPOS is required, duplicate the XYZ position data in an
	  * unused texture coordinate:
	  */
	 s2 &= ~S2_TEXCOORD_FMT(i, S2_TEXCOORD_FMT0_MASK);
	 s2 |= S2_TEXCOORD_FMT(i, SZ_TO_HW(3));

	 intel->wpos_offset = offset;
	 intel->wpos_size = 3 * sizeof(GLuint);

	 EMIT_PAD( intel->wpos_size );
      }   
   }

   if (s2 != i915->state.Ctx[I915_CTXREG_LIS2] ||
       s4 != i915->state.Ctx[I915_CTXREG_LIS4]) {
      int k;
    
      I915_STATECHANGE( i915, I915_UPLOAD_CTX );

      /* Must do this *after* statechange, so as not to affect
       * buffered vertices reliant on the old state:
       */
      intel->vertex_size = _tnl_install_attrs( &intel->ctx, 
					       intel->vertex_attrs, 
					       intel->vertex_attr_count,
					       intel->ViewportMatrix.m, 0 ); 

      intel->vertex_size >>= 2;

      i915->state.Ctx[I915_CTXREG_LIS2] = s2;
      i915->state.Ctx[I915_CTXREG_LIS4] = s4;

      k = intel->vtbl.check_vertex_size( intel, intel->vertex_size );
      assert(k);
   }

   if (!p->params_uptodate) 
      track_params( p );

   if (!p->on_hardware) 
      i915_upload_program( i915, p );
}

void i915InitFragProgFuncs( struct dd_function_table *functions )
{
   functions->BindProgram = i915BindProgram;
   functions->NewProgram = i915NewProgram;
   functions->DeleteProgram = i915DeleteProgram;
   functions->IsProgramNative = i915IsProgramNative;
   functions->ProgramStringNotify = i915ProgramStringNotify;
}