AlexIndustrial/mesa
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

aco/gfx12: don't assume memory operations complete in order

Browse Source 
Reviewed-by: Rhys Perry <pendingchaos02@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/32569>
This commit is contained in:
Georg Lehmann
2024-12-09 13:40:49 +01:00
committed by Marge Bot
parent c3becade15
commit 576a2e798c
1 changed files with 2 additions and 2 deletions
Download Patch File Download Diff File Expand all files Collapse all files
src/amd/compiler/aco_insert_waitcnt.cpp
+2 -2
View File
@@ -319,9 +319,9 @@ perform_barrier(wait_ctx& ctx, wait_imm& imm, memory_sync_info sync, unsigned se
if (bar_scope_lds <= subgroup_scope)
events &= ~event_lds;
/* in non-WGP, the L1 (L0 on GFX10+) cache keeps all memory operations
/* Until GFX12, in non-WGP, the L1 (L0 on GFX10+) cache keeps all memory operations
* in-order for the same workgroup */
if (!ctx.program->wgp_mode && sync.scope <= scope_workgroup)
if (ctx.gfx_level < GFX12 && !ctx.program->wgp_mode && sync.scope <= scope_workgroup)
events &= ~(event_vmem | event_vmem_store | event_smem);
if (events)