static void fire_fb_write( struct brw_wm_compile *c,
			   GLuint base_reg,
			   GLuint nr,
			   GLuint target,
			   GLuint eot )
{
   struct brw_compile *p = &c->func;
   
   /* Pass through control information:
    */
/*  mov (8) m1.0<1>:ud   r1.0<8;8,1>:ud   { Align1 NoMask } */
   {
      brw_push_insn_state(p);
      brw_set_mask_control(p, BRW_MASK_DISABLE); /* ? */
      brw_set_compression_control(p, BRW_COMPRESSION_NONE);
      brw_MOV(p, 
	       brw_message_reg(base_reg + 1),
	       brw_vec8_grf(1, 0));
      brw_pop_insn_state(p);
   }

   /* Send framebuffer write message: */
/*  send (16) null.0<1>:uw m0               r0.0<8;8,1>:uw   0x85a04000:ud    { Align1 EOT } */
   brw_fb_WRITE(p,
		retype(vec16(brw_null_reg()), BRW_REGISTER_TYPE_UW),
		base_reg,
		retype(brw_vec8_grf(0, 0), BRW_REGISTER_TYPE_UW),
		target,		
		nr,
		0, 
		eot);
}

static void emit_pixel_xy(struct brw_compile *p,
			  const struct brw_reg *dst,
			  GLuint mask,
			  const struct brw_reg *arg0)
{
   struct brw_reg r1 = brw_vec1_grf(1, 0);
   struct brw_reg r1_uw = retype(r1, BRW_REGISTER_TYPE_UW);

   brw_set_compression_control(p, BRW_COMPRESSION_NONE);

   /* Calculate pixel centers by adding 1 or 0 to each of the
    * micro-tile coordinates passed in r1.
    */
   if (mask & WRITEMASK_X) {
      brw_ADD(p,
	      vec16(retype(dst[0], BRW_REGISTER_TYPE_UW)),
	      stride(suboffset(r1_uw, 4), 2, 4, 0),
	      brw_imm_v(0x10101010));
   }

   if (mask & WRITEMASK_Y) {
      brw_ADD(p,
	      vec16(retype(dst[1], BRW_REGISTER_TYPE_UW)),
	      stride(suboffset(r1_uw,5), 2, 4, 0),
	      brw_imm_v(0x11001100));
   }

   brw_set_compression_control(p, BRW_COMPRESSION_COMPRESSED);
}

static void emit_wpos_xy(struct brw_wm_compile *c,
			 const struct brw_reg *dst,
			 GLuint mask,
			 const struct brw_reg *arg0)
{
   struct brw_compile *p = &c->func;

   /* Calculate the pixel offset from window bottom left into destination
    * X and Y channels.
    */
   if (mask & WRITEMASK_X) {
      /* X' = X - origin */
      brw_ADD(p,
	      dst[0],
	      retype(arg0[0], BRW_REGISTER_TYPE_W),
	      brw_imm_d(0 - c->key.origin_x));
   }

   if (mask & WRITEMASK_Y) {
      /* Y' = height - (Y - origin_y) = height + origin_y - Y */
      brw_ADD(p,
	      dst[1],
	      negate(retype(arg0[1], BRW_REGISTER_TYPE_W)),
	      brw_imm_d(c->key.origin_y + c->key.drawable_height - 1));
   }
}

static void emit_wpos_xy(struct brw_wm_compile *c,
                struct prog_instruction *inst)
{
    struct brw_compile *p = &c->func;
    GLuint mask = inst->DstReg.WriteMask;
    struct brw_reg src0[2], dst[2];

    dst[0] = get_dst_reg(c, inst, 0, 1);
    dst[1] = get_dst_reg(c, inst, 1, 1);

    src0[0] = get_src_reg(c, &inst->SrcReg[0], 0, 1);
    src0[1] = get_src_reg(c, &inst->SrcReg[0], 1, 1);

    /* Calculate the pixel offset from window bottom left into destination
     * X and Y channels.
     */
    if (mask & WRITEMASK_X) {
	/* X' = X - origin_x */
	brw_ADD(p,
		dst[0],
		retype(src0[0], BRW_REGISTER_TYPE_W),
		brw_imm_d(0 - c->key.origin_x));
    }

    if (mask & WRITEMASK_Y) {
	/* Y' = height - (Y - origin_y) = height + origin_y - Y */
	brw_ADD(p,
		dst[1],
		negate(retype(src0[1], BRW_REGISTER_TYPE_W)),
		brw_imm_d(c->key.origin_y + c->key.drawable_height - 1));
    }
}

static void emit_delta_xy(struct brw_compile *p,
			  const struct brw_reg *dst,
			  GLuint mask,
			  const struct brw_reg *arg0,
			  const struct brw_reg *arg1)
{
   struct brw_reg r1 = brw_vec1_grf(1, 0);

   /* Calc delta X,Y by subtracting origin in r1 from the pixel
    * centers.
    */
   if (mask & WRITEMASK_X) {
      brw_ADD(p,
	      dst[0],
	      retype(arg0[0], BRW_REGISTER_TYPE_UW),
	      negate(r1));
   }

   if (mask & WRITEMASK_Y) {
      brw_ADD(p,
	      dst[1],
	      retype(arg0[1], BRW_REGISTER_TYPE_UW),
	      negate(suboffset(r1,1)));

   }
}

/**
 * Allocate registers for GS.
 *
 * If sol_program is true, then:
 *
 * - The thread will be spawned with the "SVBI Payload Enable" bit set, so GRF
 *   1 needs to be set aside to hold the streamed vertex buffer indices.
 *
 * - The thread will need to use the destination_indices register.
 */
static void brw_gs_alloc_regs( struct brw_gs_compile *c,
			       GLuint nr_verts,
                               bool sol_program )
{
   GLuint i = 0,j;

   /* Register usage is static, precompute here:
    */
   c->reg.R0 = retype(brw_vec8_grf(i, 0), BRW_REGISTER_TYPE_UD); i++;

   /* Streamed vertex buffer indices */
   if (sol_program)
      c->reg.SVBI = retype(brw_vec8_grf(i++, 0), BRW_REGISTER_TYPE_UD);

   /* Payload vertices plus space for more generated vertices:
    */
   for (j = 0; j < nr_verts; j++) {
      c->reg.vertex[j] = brw_vec4_grf(i, 0);
      i += c->nr_regs;
   }

   c->reg.header = retype(brw_vec8_grf(i++, 0), BRW_REGISTER_TYPE_UD);
   c->reg.temp = retype(brw_vec8_grf(i++, 0), BRW_REGISTER_TYPE_UD);

   if (sol_program) {
      c->reg.destination_indices =
         retype(brw_vec4_grf(i++, 0), BRW_REGISTER_TYPE_UD);
   }

   c->prog_data.urb_read_length = c->nr_regs; 
   c->prog_data.total_grf = i;
}

static void emit_pixel_xy(struct brw_wm_compile *c,
		struct prog_instruction *inst)
{
    struct brw_reg r1 = brw_vec1_grf(1, 0);
    struct brw_reg r1_uw = retype(r1, BRW_REGISTER_TYPE_UW);

    struct brw_reg dst0, dst1;
    struct brw_compile *p = &c->func;
    GLuint mask = inst->DstReg.WriteMask;

    dst0 = get_dst_reg(c, inst, 0, 1);
    dst1 = get_dst_reg(c, inst, 1, 1);
    /* Calculate pixel centers by adding 1 or 0 to each of the
     * micro-tile coordinates passed in r1.
     */
    if (mask & WRITEMASK_X) {
	brw_ADD(p,
		vec8(retype(dst0, BRW_REGISTER_TYPE_UW)),
		stride(suboffset(r1_uw, 4), 2, 4, 0),
		brw_imm_v(0x10101010));
    }

    if (mask & WRITEMASK_Y) {
	brw_ADD(p,
		vec8(retype(dst1, BRW_REGISTER_TYPE_UW)),
		stride(suboffset(r1_uw, 5), 2, 4, 0),
		brw_imm_v(0x11001100));
    }

}

/* Kill pixel - set execution mask to zero for those pixels which
 * fail.
 */
static void emit_kil( struct brw_wm_compile *c,
		      struct brw_reg *arg0)
{
   struct brw_compile *p = &c->func;
   struct intel_context *intel = &p->brw->intel;
   struct brw_reg pixelmask;
   GLuint i, j;

   if (intel->gen >= 6)
      pixelmask = retype(brw_vec1_grf(1, 7), BRW_REGISTER_TYPE_UW);
   else
      pixelmask = retype(brw_vec1_grf(0, 0), BRW_REGISTER_TYPE_UW);

   for (i = 0; i < 4; i++) {
      /* Check if we've already done the comparison for this reg
       * -- common when someone does KIL TEMP.wwww.
       */
      for (j = 0; j < i; j++) {
	 if (memcmp(&arg0[j], &arg0[i], sizeof(arg0[0])) == 0)
	    break;
      }
      if (j != i)
	 continue;

      brw_push_insn_state(p);
      brw_CMP(p, brw_null_reg(), BRW_CONDITIONAL_GE, arg0[i], brw_imm_f(0));   
      brw_set_predicate_control_flag_value(p, 0xff);
      brw_set_compression_control(p, BRW_COMPRESSION_NONE);
      brw_AND(p, pixelmask, brw_flag_reg(), pixelmask);
      brw_pop_insn_state(p);
   }
}

void brw_NOP(struct brw_compile *p)
{
   struct brw_instruction *insn = next_insn(p, BRW_OPCODE_NOP);   
   brw_set_dest(insn, retype(brw_vec4_grf(0,0), BRW_REGISTER_TYPE_UD));
   brw_set_src0(insn, retype(brw_vec4_grf(0,0), BRW_REGISTER_TYPE_UD));
   brw_set_src1(insn, brw_imm_ud(0x0));
}

void
gen8_vec4_generator::generate_scratch_write(vec4_instruction *ir,
                                            struct brw_reg dst,
                                            struct brw_reg src,
                                            struct brw_reg index)
{
   struct brw_reg header = brw_vec8_grf(GEN7_MRF_HACK_START + ir->base_mrf, 0);

   MOV_RAW(header, brw_vec8_grf(0, 0));

   generate_oword_dual_block_offsets(brw_message_reg(ir->base_mrf + 1), index);

   MOV(retype(brw_message_reg(ir->base_mrf + 2), BRW_REGISTER_TYPE_D),
       retype(src, BRW_REGISTER_TYPE_D));

   /* Each of the 8 channel enables is considered for whether each
    * dword is written.
    */
   gen8_instruction *send = next_inst(BRW_OPCODE_SEND);
   gen8_set_dst(brw, send, dst);
   gen8_set_src0(brw, send, header);
   gen8_set_pred_control(send, ir->predicate);
   gen8_set_dp_message(brw, send, GEN7_SFID_DATAPORT_DATA_CACHE,
                       255, /* binding table index: stateless access */
                       GEN7_DATAPORT_WRITE_MESSAGE_OWORD_DUAL_BLOCK_WRITE,
                       BRW_DATAPORT_OWORD_DUAL_BLOCK_1OWORD,
                       3,      /* mlen */
                       0,      /* rlen */
                       true,   /* header present */
                       false); /* EOT */
}

static void brw_gs_ff_sync(struct brw_gs_compile *c, int num_prim)
{
	struct brw_compile *p = &c->func;
	struct intel_context *intel = &c->func.brw->intel;

	if (intel->gen < 6) {
	    brw_MOV(p, get_element_ud(c->reg.R0, 1), brw_imm_ud(num_prim));
	    brw_ff_sync(p,
		        c->reg.R0,
			0,
			c->reg.R0,
			1, /* allocate */
			1, /* response length */
			0 /* eot */);
	} else {
	    brw_MOV(p, retype(c->reg.temp, BRW_REGISTER_TYPE_UD),
		    retype(c->reg.R0, BRW_REGISTER_TYPE_UD));
	    brw_MOV(p, get_element_ud(c->reg.temp, 1), brw_imm_ud(num_prim));
	    brw_ff_sync(p,
		        c->reg.temp,
			0,
			c->reg.temp,
			1, /* allocate */
			1, /* response length */
			0 /* eot */);
	    brw_MOV(p, get_element_ud(c->reg.R0, 0),
		    get_element_ud(c->reg.temp, 0));
	}
}

static void emit_delta_xy(struct brw_wm_compile *c,
		struct prog_instruction *inst)
{
    struct brw_reg r1 = brw_vec1_grf(1, 0);
    struct brw_reg dst0, dst1, src0, src1;
    struct brw_compile *p = &c->func;
    GLuint mask = inst->DstReg.WriteMask;

    dst0 = get_dst_reg(c, inst, 0, 1);
    dst1 = get_dst_reg(c, inst, 1, 1);
    src0 = get_src_reg(c, &inst->SrcReg[0], 0, 1);
    src1 = get_src_reg(c, &inst->SrcReg[0], 1, 1);
    /* Calc delta X,Y by subtracting origin in r1 from the pixel
     * centers.
     */
    if (mask & WRITEMASK_X) {
	brw_ADD(p,
		dst0,
		retype(src0, BRW_REGISTER_TYPE_UW),
		negate(r1));
    }

    if (mask & WRITEMASK_Y) {
	brw_ADD(p,
		dst1,
		retype(src1, BRW_REGISTER_TYPE_UW),
		negate(suboffset(r1,1)));

    }

}

void
vec4_generator::generate_gs_set_vertex_count(struct brw_reg dst,
        struct brw_reg src)
{
    brw_push_insn_state(p);
    brw_set_access_mode(p, BRW_ALIGN_1);
    brw_set_mask_control(p, BRW_MASK_DISABLE);

    /* If we think of the src and dst registers as composed of 8 DWORDs each,
     * we want to pick up the contents of DWORDs 0 and 4 from src, truncate
     * them to WORDs, and then pack them into DWORD 2 of dst.
     *
     * It's easier to get the EU to do this if we think of the src and dst
     * registers as composed of 16 WORDS each; then, we want to pick up the
     * contents of WORDs 0 and 8 from src, and pack them into WORDs 4 and 5 of
     * dst.
     *
     * We can do that by the following EU instruction:
     *
     *     mov (2) dst.4<1>:uw src<8;1,0>:uw   { Align1, Q1, NoMask }
     */
    brw_MOV(p, suboffset(stride(retype(dst, BRW_REGISTER_TYPE_UW), 2, 2, 1), 4),
            stride(retype(src, BRW_REGISTER_TYPE_UW), 8, 1, 0));
    brw_set_access_mode(p, BRW_ALIGN_16);
    brw_pop_insn_state(p);
}

static void
gen_ADD_GRF_GRF_IMM_d(struct brw_codegen *p)
{
   struct brw_reg g0 = retype(brw_vec8_grf(0, 0), BRW_REGISTER_TYPE_D);
   struct brw_reg g2 = retype(brw_vec8_grf(2, 0), BRW_REGISTER_TYPE_D);

   brw_ADD(p, g0, g2, brw_imm_d(1));
}

void ObjectEditor::typeChanged(int type){
    switch(type){
    case 0: retype(EOBJECT); break;
    case 1: retype(E_PARALAX);break;
    case 2: retype(E_GAME_RESURS);break;
    default: break;
    }
}

void brw_ENDIF(struct brw_compile *p, 
	       struct brw_instruction *patch_insn)
{
   GLuint br = 1;

   if (BRW_IS_IGDNG(p->brw))
      br = 2; 
 
   if (p->single_program_flow) {
      /* In single program flow mode, there's no need to execute an ENDIF,
       * since we don't need to do any stack operations, and if we're executing
       * currently, we want to just continue executing.
       */
      struct brw_instruction *next = &p->store[p->nr_insn];

      assert(patch_insn->header.opcode == BRW_OPCODE_ADD);

      patch_insn->bits3.ud = (next - patch_insn) * 16;
   } else {
      struct brw_instruction *insn = next_insn(p, BRW_OPCODE_ENDIF);

      brw_set_dest(insn, retype(brw_vec4_grf(0,0), BRW_REGISTER_TYPE_UD));
      brw_set_src0(insn, retype(brw_vec4_grf(0,0), BRW_REGISTER_TYPE_UD));
      brw_set_src1(insn, brw_imm_d(0x0));

      insn->header.compression_control = BRW_COMPRESSION_NONE;
      insn->header.execution_size = patch_insn->header.execution_size;
      insn->header.mask_control = BRW_MASK_ENABLE;
      insn->header.thread_control = BRW_THREAD_SWITCH;

      assert(patch_insn->bits3.if_else.jump_count == 0);

      /* Patch the if or else instructions to point at this or the next
       * instruction respectively.
       */
      if (patch_insn->header.opcode == BRW_OPCODE_IF) {
	 /* Automagically turn it into an IFF:
	  */
	 patch_insn->header.opcode = BRW_OPCODE_IFF;
	 patch_insn->bits3.if_else.jump_count = br * (insn - patch_insn + 1);
	 patch_insn->bits3.if_else.pop_count = 0;
	 patch_insn->bits3.if_else.pad0 = 0;
      } else if (patch_insn->header.opcode == BRW_OPCODE_ELSE) {
	 patch_insn->bits3.if_else.jump_count = br * (insn - patch_insn + 1);
	 patch_insn->bits3.if_else.pop_count = 1;
	 patch_insn->bits3.if_else.pad0 = 0;
      } else {
	 assert(0);
      }

      /* Also pop item off the stack in the endif instruction:
       */
      insn->bits3.if_else.jump_count = 0;
      insn->bits3.if_else.pop_count = 1;
      insn->bits3.if_else.pad0 = 0;
   }
}

static void brw_clip_line_alloc_regs( struct brw_clip_compile *c )
{
   const struct gen_device_info *devinfo = c->func.devinfo;
   GLuint i = 0,j;

   /* Register usage is static, precompute here:
    */
   c->reg.R0 = retype(brw_vec8_grf(i, 0), BRW_REGISTER_TYPE_UD); i++;

   if (c->key.nr_userclip) {
      c->reg.fixed_planes = brw_vec4_grf(i, 0);
      i += (6 + c->key.nr_userclip + 1) / 2;

      c->prog_data.curb_read_length = (6 + c->key.nr_userclip + 1) / 2;
   }
   else
      c->prog_data.curb_read_length = 0;


   /* Payload vertices plus space for more generated vertices:
    */
   for (j = 0; j < 4; j++) {
      c->reg.vertex[j] = brw_vec4_grf(i, 0);
      i += c->nr_regs;
   }

   c->reg.t           = brw_vec1_grf(i, 0);
   c->reg.t0          = brw_vec1_grf(i, 1);
   c->reg.t1          = brw_vec1_grf(i, 2);
   c->reg.planemask   = retype(brw_vec1_grf(i, 3), BRW_REGISTER_TYPE_UD);
   c->reg.plane_equation = brw_vec4_grf(i, 4);
   i++;

   c->reg.dp0         = brw_vec1_grf(i, 0); /* fixme - dp4 will clobber r.1,2,3 */
   c->reg.dp1         = brw_vec1_grf(i, 4);
   i++;

   if (!c->key.nr_userclip) {
      c->reg.fixed_planes = brw_vec8_grf(i, 0);
      i++;
   }

   c->reg.vertex_src_mask = retype(brw_vec1_grf(i, 0), BRW_REGISTER_TYPE_UD);
   c->reg.clipdistance_offset = retype(brw_vec1_grf(i, 1), BRW_REGISTER_TYPE_W);
   i++;

   if (devinfo->gen == 5) {
      c->reg.ff_sync = retype(brw_vec1_grf(i, 0), BRW_REGISTER_TYPE_UD);
      i++;
   }

   c->first_tmp = i;
   c->last_tmp = i;

   c->prog_data.urb_read_length = c->nr_regs; /* ? */
   c->prog_data.total_grf = i;
}

/**
 * Read a float[4] vector from the data port Data Cache (const buffer).
 * Location (in buffer) should be a multiple of 16.
 * Used for fetching shader constants.
 * If relAddr is true, we'll do an indirect fetch using the address register.
 */
void brw_dp_READ_4( struct brw_compile *p,
                    struct brw_reg dest,
                    GLboolean relAddr,
                    GLuint location,
                    GLuint bind_table_index )
{
   /* XXX: relAddr not implemented */
   GLuint msg_reg_nr = 1;
   {
      struct brw_reg b;
      brw_push_insn_state(p);
      brw_set_predicate_control(p, BRW_PREDICATE_NONE);
      brw_set_compression_control(p, BRW_COMPRESSION_NONE);
      brw_set_mask_control(p, BRW_MASK_DISABLE);

   /* Setup MRF[1] with location/offset into const buffer */
      b = brw_message_reg(msg_reg_nr);
      b = retype(b, BRW_REGISTER_TYPE_UD);
      /* XXX I think we're setting all the dwords of MRF[1] to 'location'.
       * when the docs say only dword[2] should be set.  Hmmm.  But it works.
       */
      brw_MOV(p, b, brw_imm_ud(location));
      brw_pop_insn_state(p);
   }

   {
      struct brw_instruction *insn = next_insn(p, BRW_OPCODE_SEND);
   
      insn->header.predicate_control = BRW_PREDICATE_NONE;
      insn->header.compression_control = BRW_COMPRESSION_NONE; 
      insn->header.destreg__conditionalmod = msg_reg_nr;
      insn->header.mask_control = BRW_MASK_DISABLE;
  
      /* cast dest to a uword[8] vector */
      dest = retype(vec8(dest), BRW_REGISTER_TYPE_UW);

      brw_set_dest(insn, dest);
      brw_set_src0(insn, brw_null_reg());

      brw_set_dp_read_message(p->brw,
			      insn,
			      bind_table_index,
			      0,  /* msg_control (0 means 1 Oword) */
			      BRW_DATAPORT_READ_MESSAGE_OWORD_BLOCK_READ, /* msg_type */
			      0, /* source cache = data cache */
			      1, /* msg_length */
			      1, /* response_length (1 Oword) */
			      0); /* eot */
   }
}

static void brw_gs_emit_vue(struct brw_gs_compile *c, 
			    struct brw_reg vert,
			    GLboolean last,
			    GLuint header)
{
   struct brw_compile *p = &c->func;
   struct intel_context *intel = &c->func.brw->intel;
   GLboolean allocate = !last;
   struct brw_reg temp;

   if (intel->gen < 6)
       temp = c->reg.R0;
   else {
       temp = c->reg.temp;
       brw_MOV(p, retype(temp, BRW_REGISTER_TYPE_UD),
	       retype(c->reg.R0, BRW_REGISTER_TYPE_UD));
   }

   /* Overwrite PrimType and PrimStart in the message header, for
    * each vertex in turn:
    */
   brw_MOV(p, get_element_ud(temp, 2), brw_imm_ud(header));

   /* Copy the vertex from vertn into m1..mN+1:
    */
   brw_copy8(p, brw_message_reg(1), vert, c->nr_regs);

   /* Send each vertex as a seperate write to the urb.  This is
    * different to the concept in brw_sf_emit.c, where subsequent
    * writes are used to build up a single urb entry.  Each of these
    * writes instantiates a seperate urb entry, and a new one must be
    * allocated each time.
    */
   brw_urb_WRITE(p, 
		 allocate ? temp : retype(brw_null_reg(), BRW_REGISTER_TYPE_UD),
		 0,
		 temp,
		 allocate,
		 1,		/* used */
		 c->nr_regs + 1, /* msg length */
		 allocate ? 1 : 0, /* response length */
		 allocate ? 0 : 1, /* eot */
		 1,		/* writes_complete */
		 0,		/* urb offset */
		 BRW_URB_SWIZZLE_NONE);

   if (intel->gen >= 6 && allocate)
       brw_MOV(p, get_element_ud(c->reg.R0, 0), get_element_ud(temp, 0));
}

void brw_emit_anyprim_setup( struct brw_sf_compile *c )
{
   struct brw_compile *p = &c->func;
   struct brw_context *brw = p->brw;
   struct brw_reg payload_prim = brw_uw1_reg(BRW_GENERAL_REGISTER_FILE, 1, 0);
   struct brw_reg payload_attr = get_element_ud(brw_vec1_reg(BRW_GENERAL_REGISTER_FILE, 1, 0), 0);
   struct brw_reg primmask;
   int jmp;
   struct brw_reg v1_null_ud = vec1(retype(brw_null_reg(), BRW_REGISTER_TYPE_UD));

   c->nr_verts = 3;
   alloc_regs(c);

   primmask = retype(get_element(c->tmp, 0), BRW_REGISTER_TYPE_UD);

   brw_MOV(p, primmask, brw_imm_ud(1));
   brw_SHL(p, primmask, primmask, payload_prim);

   brw_AND(p, v1_null_ud, primmask, brw_imm_ud((1<<_3DPRIM_TRILIST) |
					       (1<<_3DPRIM_TRISTRIP) |
					       (1<<_3DPRIM_TRIFAN) |
					       (1<<_3DPRIM_TRISTRIP_REVERSE) |
					       (1<<_3DPRIM_POLYGON) |
					       (1<<_3DPRIM_RECTLIST) |
					       (1<<_3DPRIM_TRIFAN_NOSTIPPLE)));
   brw_inst_set_cond_modifier(brw, brw_last_inst, BRW_CONDITIONAL_Z);
   jmp = brw_JMPI(p, brw_imm_d(0), BRW_PREDICATE_NORMAL) - p->store;
   brw_emit_tri_setup(c, false);
   brw_land_fwd_jump(p, jmp);

   brw_AND(p, v1_null_ud, primmask, brw_imm_ud((1<<_3DPRIM_LINELIST) |
					       (1<<_3DPRIM_LINESTRIP) |
					       (1<<_3DPRIM_LINELOOP) |
					       (1<<_3DPRIM_LINESTRIP_CONT) |
					       (1<<_3DPRIM_LINESTRIP_BF) |
					       (1<<_3DPRIM_LINESTRIP_CONT_BF)));
   brw_inst_set_cond_modifier(brw, brw_last_inst, BRW_CONDITIONAL_Z);
   jmp = brw_JMPI(p, brw_imm_d(0), BRW_PREDICATE_NORMAL) - p->store;
   brw_emit_line_setup(c, false);
   brw_land_fwd_jump(p, jmp);

   brw_AND(p, v1_null_ud, payload_attr, brw_imm_ud(1<<BRW_SPRITE_POINT_ENABLE));
   brw_inst_set_cond_modifier(brw, brw_last_inst, BRW_CONDITIONAL_Z);
   jmp = brw_JMPI(p, brw_imm_d(0), BRW_PREDICATE_NORMAL) - p->store;
   brw_emit_point_sprite_setup(c, false);
   brw_land_fwd_jump(p, jmp);

   brw_emit_point_setup( c, false );
}