We set header_present but then pass it some random garbage. Give it g0
instead. I'm not actually sure this does anything but g0 is the usual
header data and this is what the windows driver does so it seems like a
good idea.
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
void
brw_memory_fence(struct brw_codegen *p,
struct brw_reg dst,
void
brw_memory_fence(struct brw_codegen *p,
struct brw_reg dst,
enum opcode send_op);
void
enum opcode send_op);
void
void
brw_memory_fence(struct brw_codegen *p,
struct brw_reg dst,
void
brw_memory_fence(struct brw_codegen *p,
struct brw_reg dst,
enum opcode send_op)
{
const struct gen_device_info *devinfo = p->devinfo;
enum opcode send_op)
{
const struct gen_device_info *devinfo = p->devinfo;
brw_push_insn_state(p);
brw_set_default_mask_control(p, BRW_MASK_DISABLE);
brw_set_default_exec_size(p, BRW_EXECUTE_1);
brw_push_insn_state(p);
brw_set_default_mask_control(p, BRW_MASK_DISABLE);
brw_set_default_exec_size(p, BRW_EXECUTE_1);
+ dst = retype(vec1(dst), BRW_REGISTER_TYPE_UW);
+ src = retype(vec1(src), BRW_REGISTER_TYPE_UD);
/* Set dst as destination for dependency tracking, the MEMORY_FENCE
* message doesn't write anything back.
*/
insn = next_insn(p, send_op);
/* Set dst as destination for dependency tracking, the MEMORY_FENCE
* message doesn't write anything back.
*/
insn = next_insn(p, send_op);
- dst = retype(dst, BRW_REGISTER_TYPE_UW);
brw_set_dest(p, insn, dst);
brw_set_dest(p, insn, dst);
- brw_set_src0(p, insn, dst);
+ brw_set_src0(p, insn, src);
brw_set_memory_fence_message(p, insn, GEN7_SFID_DATAPORT_DATA_CACHE,
commit_enable);
brw_set_memory_fence_message(p, insn, GEN7_SFID_DATAPORT_DATA_CACHE,
commit_enable);
*/
insn = next_insn(p, send_op);
brw_set_dest(p, insn, offset(dst, 1));
*/
insn = next_insn(p, send_op);
brw_set_dest(p, insn, offset(dst, 1));
- brw_set_src0(p, insn, offset(dst, 1));
+ brw_set_src0(p, insn, src);
brw_set_memory_fence_message(p, insn, GEN6_SFID_DATAPORT_RENDER_CACHE,
commit_enable);
brw_set_memory_fence_message(p, insn, GEN6_SFID_DATAPORT_RENDER_CACHE,
commit_enable);
break;
case SHADER_OPCODE_MEMORY_FENCE:
break;
case SHADER_OPCODE_MEMORY_FENCE:
- brw_memory_fence(p, dst, BRW_OPCODE_SEND);
+ brw_memory_fence(p, dst, src[0], BRW_OPCODE_SEND);
break;
case SHADER_OPCODE_INTERLOCK:
assert(devinfo->gen >= 9);
/* The interlock is basically a memory fence issued via sendc */
break;
case SHADER_OPCODE_INTERLOCK:
assert(devinfo->gen >= 9);
/* The interlock is basically a memory fence issued via sendc */
- brw_memory_fence(p, dst, BRW_OPCODE_SENDC);
+ brw_memory_fence(p, dst, src[0], BRW_OPCODE_SENDC);
break;
case SHADER_OPCODE_FIND_LIVE_CHANNEL: {
break;
case SHADER_OPCODE_FIND_LIVE_CHANNEL: {
case nir_intrinsic_memory_barrier: {
const fs_builder ubld = bld.group(8, 0);
const fs_reg tmp = ubld.vgrf(BRW_REGISTER_TYPE_UD, 2);
case nir_intrinsic_memory_barrier: {
const fs_builder ubld = bld.group(8, 0);
const fs_reg tmp = ubld.vgrf(BRW_REGISTER_TYPE_UD, 2);
- ubld.emit(SHADER_OPCODE_MEMORY_FENCE, tmp)
+ ubld.emit(SHADER_OPCODE_MEMORY_FENCE, tmp, brw_vec8_grf(0, 0))
->size_written = 2 * REG_SIZE;
break;
}
->size_written = 2 * REG_SIZE;
break;
}
const fs_builder ubld = bld.group(8, 0);
const fs_reg tmp = ubld.vgrf(BRW_REGISTER_TYPE_UD, 2);
const fs_builder ubld = bld.group(8, 0);
const fs_reg tmp = ubld.vgrf(BRW_REGISTER_TYPE_UD, 2);
- ubld.emit(SHADER_OPCODE_INTERLOCK, tmp)->size_written = 2 *
- REG_SIZE;
-
+ ubld.emit(SHADER_OPCODE_INTERLOCK, tmp, brw_vec8_grf(0, 0))
+ ->size_written = 2 * REG_SIZE;
break;
case SHADER_OPCODE_MEMORY_FENCE:
break;
case SHADER_OPCODE_MEMORY_FENCE:
- brw_memory_fence(p, dst, BRW_OPCODE_SEND);
+ brw_memory_fence(p, dst, src[0], BRW_OPCODE_SEND);
break;
case SHADER_OPCODE_FIND_LIVE_CHANNEL: {
break;
case SHADER_OPCODE_FIND_LIVE_CHANNEL: {
const vec4_builder bld =
vec4_builder(this).at_end().annotate(current_annotation, base_ir);
const dst_reg tmp = bld.vgrf(BRW_REGISTER_TYPE_UD, 2);
const vec4_builder bld =
vec4_builder(this).at_end().annotate(current_annotation, base_ir);
const dst_reg tmp = bld.vgrf(BRW_REGISTER_TYPE_UD, 2);
- bld.emit(SHADER_OPCODE_MEMORY_FENCE, tmp)
+ bld.emit(SHADER_OPCODE_MEMORY_FENCE, tmp, brw_vec8_grf(0, 0))
->size_written = 2 * REG_SIZE;
break;
}
->size_written = 2 * REG_SIZE;
break;
}