fdo-mirrors/mesa
1
0
Fork 0
mirror of https://gitlab.freedesktop.org/mesa/mesa.git synced 2026-05-08 19:58:09 +02:00
Code Issues Projects Releases Packages Wiki Activity Actions 1

intel/brw: Pull emit_dummy_mov_instruction out of fs_visitor

Browse source 
Remove references to the workaround number from the callsites.  Instead
the function has "workaround" as part of the name and the number is
in its definition.  Return bool for consistency with other passes.

Reviewed-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Reviewed-by: Ian Romanick <ian.d.romanick@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/26887>
This commit is contained in:
Caio Oliveira 2024-01-04 16:24:21 -08:00 • committed by Marge Bot
parent 755d6e8c44
commit 5a5a242210
2 changed files with 20 additions and 37 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

55
src/intel/compiler/brw_fs.cpp
View file

@ -5894,28 +5894,29 @@ needs_dummy_fence(const intel_device_info *devinfo, fs_inst *inst)
* The first instruction of any kernel should have non-zero emask.
* Make sure this happens by introducing a dummy mov instruction.
*/
void
fs_visitor::emit_dummy_mov_instruction()
bool
brw_fs_workaround_emit_dummy_mov_instruction(fs_visitor &s)
{
if (!intel_needs_workaround(devinfo, 14015360517))
return;
if (!intel_needs_workaround(s.devinfo, 14015360517))
return false;
struct backend_instruction *first_inst =
cfg->first_block()->start();
s.cfg->first_block()->start();
/* We can skip the WA if first instruction is marked with
* force_writemask_all or exec_size equals dispatch_width.
*/
if (first_inst->force_writemask_all ||
first_inst->exec_size == dispatch_width)
return;
first_inst->exec_size == s.dispatch_width)
return false;
/* Insert dummy mov as first instruction. */
const fs_builder ubld =
fs_builder(this, cfg->first_block(), (fs_inst *)first_inst).exec_all().group(8, 0);
fs_builder(&s, s.cfg->first_block(), (fs_inst *)first_inst).exec_all().group(8, 0);
ubld.MOV(ubld.null_reg_ud(), brw_imm_ud(0u));
invalidate_analysis(DEPENDENCY_INSTRUCTIONS | DEPENDENCY_VARIABLES);
s.invalidate_analysis(DEPENDENCY_INSTRUCTIONS | DEPENDENCY_VARIABLES);
return true;
}
/* Wa_22013689345
@ -6352,9 +6353,7 @@ fs_visitor::run_vs()
brw_fs_lower_3src_null_dest(*this);
brw_fs_workaround_memory_fence_before_eot(*this);
/* Wa_14015360517 */
emit_dummy_mov_instruction();
brw_fs_workaround_emit_dummy_mov_instruction(*this);
allocate_registers(true /* allow_spilling */);
@ -6482,9 +6481,7 @@ fs_visitor::run_tcs()
brw_fs_lower_3src_null_dest(*this);
brw_fs_workaround_memory_fence_before_eot(*this);
/* Wa_14015360517 */
emit_dummy_mov_instruction();
brw_fs_workaround_emit_dummy_mov_instruction(*this);
allocate_registers(true /* allow_spilling */);
@ -6514,9 +6511,7 @@ fs_visitor::run_tes()
brw_fs_lower_3src_null_dest(*this);
brw_fs_workaround_memory_fence_before_eot(*this);
/* Wa_14015360517 */
emit_dummy_mov_instruction();
brw_fs_workaround_emit_dummy_mov_instruction(*this);
allocate_registers(true /* allow_spilling */);
@ -6563,9 +6558,7 @@ fs_visitor::run_gs()
brw_fs_lower_3src_null_dest(*this);
brw_fs_workaround_memory_fence_before_eot(*this);
/* Wa_14015360517 */
emit_dummy_mov_instruction();
brw_fs_workaround_emit_dummy_mov_instruction(*this);
allocate_registers(true /* allow_spilling */);
@ -6671,9 +6664,7 @@ fs_visitor::run_fs(bool allow_spilling, bool do_rep_send)
brw_fs_lower_3src_null_dest(*this);
brw_fs_workaround_memory_fence_before_eot(*this);
/* Wa_14015360517 */
emit_dummy_mov_instruction();
brw_fs_workaround_emit_dummy_mov_instruction(*this);
allocate_registers(allow_spilling);
}
@ -6712,9 +6703,7 @@ fs_visitor::run_cs(bool allow_spilling)
brw_fs_lower_3src_null_dest(*this);
brw_fs_workaround_memory_fence_before_eot(*this);
/* Wa_14015360517 */
emit_dummy_mov_instruction();
brw_fs_workaround_emit_dummy_mov_instruction(*this);
allocate_registers(allow_spilling);
@ -6744,9 +6733,7 @@ fs_visitor::run_bs(bool allow_spilling)
brw_fs_lower_3src_null_dest(*this);
brw_fs_workaround_memory_fence_before_eot(*this);
/* Wa_14015360517 */
emit_dummy_mov_instruction();
brw_fs_workaround_emit_dummy_mov_instruction(*this);
allocate_registers(allow_spilling);
@ -6777,9 +6764,7 @@ fs_visitor::run_task(bool allow_spilling)
brw_fs_lower_3src_null_dest(*this);
brw_fs_workaround_memory_fence_before_eot(*this);
/* Wa_14015360517 */
emit_dummy_mov_instruction();
brw_fs_workaround_emit_dummy_mov_instruction(*this);
allocate_registers(allow_spilling);
@ -6810,9 +6795,7 @@ fs_visitor::run_mesh(bool allow_spilling)
brw_fs_lower_3src_null_dest(*this);
brw_fs_workaround_memory_fence_before_eot(*this);
/* Wa_14015360517 */
emit_dummy_mov_instruction();
brw_fs_workaround_emit_dummy_mov_instruction(*this);
allocate_registers(allow_spilling);

2
src/intel/compiler/brw_fs.h
View file

@ -237,7 +237,6 @@ public:
void optimize();
void allocate_registers(bool allow_spilling);
uint32_t compute_max_register_pressure();
void emit_dummy_mov_instruction();
void assign_curb_setup();
void assign_urb_setup();
void convert_attr_sources_to_hw_regs(fs_inst *inst);
@ -619,6 +618,7 @@ bool brw_fs_opt_split_sends(fs_visitor &s);
bool brw_fs_opt_split_virtual_grfs(fs_visitor &s);
bool brw_fs_opt_zero_samples(fs_visitor &s);
bool brw_fs_workaround_emit_dummy_mov_instruction(fs_visitor &s);
bool brw_fs_workaround_memory_fence_before_eot(fs_visitor &s);
bool brw_fs_workaround_nomask_control_flow(fs_visitor &s);