[AMDGPU] Fix setting nontemporal in memory legalizer #83815
Merged
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
|
|
|
@llvm/pr-subscribers-backend-amdgpu Author: Mirko Brkušanin (mbrkusanin) ChangesIterator MI can advance in insertWait() but we need original instruction to Patch is 1.84 MiB, truncated to 20.00 KiB below, full version: https://github.com/llvm/llvm-project/pull/83815.diff 24 Files Affected:
diff --git a/llvm/lib/Target/AMDGPU/SIMemoryLegalizer.cpp b/llvm/lib/Target/AMDGPU/SIMemoryLegalizer.cpp
index 4069a368f68719..73c23f0f987c3d 100644
--- a/llvm/lib/Target/AMDGPU/SIMemoryLegalizer.cpp
+++ b/llvm/lib/Target/AMDGPU/SIMemoryLegalizer.cpp
@@ -2392,6 +2392,11 @@ bool SIGfx12CacheControl::enableVolatileAndOrNonTemporal(
bool Changed = false;
+ if (IsNonTemporal) {
+ // Set non-temporal hint for all cache levels.
+ Changed |= setTH(MI, AMDGPU::CPol::TH_NT);
+ }
+
if (IsVolatile) {
Changed |= setScope(MI, AMDGPU::CPol::SCOPE_SYS);
@@ -2407,11 +2412,6 @@ bool SIGfx12CacheControl::enableVolatileAndOrNonTemporal(
Position::AFTER);
}
- if (IsNonTemporal) {
- // Set non-temporal hint for all cache levels.
- Changed |= setTH(MI, AMDGPU::CPol::TH_NT);
- }
-
return Changed;
}
diff --git a/llvm/test/CodeGen/AMDGPU/memory-legalizer-fence.ll b/llvm/test/CodeGen/AMDGPU/memory-legalizer-fence.ll
index 77962fadcacfc6..e13542f61474e2 100644
--- a/llvm/test/CodeGen/AMDGPU/memory-legalizer-fence.ll
+++ b/llvm/test/CodeGen/AMDGPU/memory-legalizer-fence.ll
@@ -10,6 +10,8 @@
; RUN: llc -mtriple=amdgcn-amd-amdhsa -mcpu=gfx940 -mattr=+tgsplit < %s | FileCheck -check-prefixes=GFX940-TGSPLIT %s
; RUN: llc -mtriple=amdgcn-amd-amdhsa -mcpu=gfx1100 < %s | FileCheck --check-prefixes=GFX11-WGP %s
; RUN: llc -mtriple=amdgcn-amd-amdhsa -mcpu=gfx1100 -mattr=+cumode < %s | FileCheck --check-prefixes=GFX11-CU %s
+; RUN: llc -mtriple=amdgcn-amd-amdhsa -mcpu=gfx1200 < %s | FileCheck --check-prefixes=GFX12-WGP %s
+; RUN: llc -mtriple=amdgcn-amd-amdhsa -mcpu=gfx1200 -mattr=+cumode < %s | FileCheck --check-prefixes=GFX12-CU %s
define amdgpu_kernel void @singlethread_acquire_fence() {
; GFX6-LABEL: singlethread_acquire_fence:
@@ -55,6 +57,14 @@ define amdgpu_kernel void @singlethread_acquire_fence() {
; GFX11-CU-LABEL: singlethread_acquire_fence:
; GFX11-CU: ; %bb.0: ; %entry
; GFX11-CU-NEXT: s_endpgm
+;
+; GFX12-WGP-LABEL: singlethread_acquire_fence:
+; GFX12-WGP: ; %bb.0: ; %entry
+; GFX12-WGP-NEXT: s_endpgm
+;
+; GFX12-CU-LABEL: singlethread_acquire_fence:
+; GFX12-CU: ; %bb.0: ; %entry
+; GFX12-CU-NEXT: s_endpgm
entry:
fence syncscope("singlethread") acquire
ret void
@@ -104,6 +114,14 @@ define amdgpu_kernel void @singlethread_release_fence() {
; GFX11-CU-LABEL: singlethread_release_fence:
; GFX11-CU: ; %bb.0: ; %entry
; GFX11-CU-NEXT: s_endpgm
+;
+; GFX12-WGP-LABEL: singlethread_release_fence:
+; GFX12-WGP: ; %bb.0: ; %entry
+; GFX12-WGP-NEXT: s_endpgm
+;
+; GFX12-CU-LABEL: singlethread_release_fence:
+; GFX12-CU: ; %bb.0: ; %entry
+; GFX12-CU-NEXT: s_endpgm
entry:
fence syncscope("singlethread") release
ret void
@@ -153,6 +171,14 @@ define amdgpu_kernel void @singlethread_acq_rel_fence() {
; GFX11-CU-LABEL: singlethread_acq_rel_fence:
; GFX11-CU: ; %bb.0: ; %entry
; GFX11-CU-NEXT: s_endpgm
+;
+; GFX12-WGP-LABEL: singlethread_acq_rel_fence:
+; GFX12-WGP: ; %bb.0: ; %entry
+; GFX12-WGP-NEXT: s_endpgm
+;
+; GFX12-CU-LABEL: singlethread_acq_rel_fence:
+; GFX12-CU: ; %bb.0: ; %entry
+; GFX12-CU-NEXT: s_endpgm
entry:
fence syncscope("singlethread") acq_rel
ret void
@@ -202,6 +228,14 @@ define amdgpu_kernel void @singlethread_seq_cst_fence() {
; GFX11-CU-LABEL: singlethread_seq_cst_fence:
; GFX11-CU: ; %bb.0: ; %entry
; GFX11-CU-NEXT: s_endpgm
+;
+; GFX12-WGP-LABEL: singlethread_seq_cst_fence:
+; GFX12-WGP: ; %bb.0: ; %entry
+; GFX12-WGP-NEXT: s_endpgm
+;
+; GFX12-CU-LABEL: singlethread_seq_cst_fence:
+; GFX12-CU: ; %bb.0: ; %entry
+; GFX12-CU-NEXT: s_endpgm
entry:
fence syncscope("singlethread") seq_cst
ret void
@@ -251,6 +285,14 @@ define amdgpu_kernel void @singlethread_one_as_acquire_fence() {
; GFX11-CU-LABEL: singlethread_one_as_acquire_fence:
; GFX11-CU: ; %bb.0: ; %entry
; GFX11-CU-NEXT: s_endpgm
+;
+; GFX12-WGP-LABEL: singlethread_one_as_acquire_fence:
+; GFX12-WGP: ; %bb.0: ; %entry
+; GFX12-WGP-NEXT: s_endpgm
+;
+; GFX12-CU-LABEL: singlethread_one_as_acquire_fence:
+; GFX12-CU: ; %bb.0: ; %entry
+; GFX12-CU-NEXT: s_endpgm
entry:
fence syncscope("singlethread-one-as") acquire
ret void
@@ -300,6 +342,14 @@ define amdgpu_kernel void @singlethread_one_as_release_fence() {
; GFX11-CU-LABEL: singlethread_one_as_release_fence:
; GFX11-CU: ; %bb.0: ; %entry
; GFX11-CU-NEXT: s_endpgm
+;
+; GFX12-WGP-LABEL: singlethread_one_as_release_fence:
+; GFX12-WGP: ; %bb.0: ; %entry
+; GFX12-WGP-NEXT: s_endpgm
+;
+; GFX12-CU-LABEL: singlethread_one_as_release_fence:
+; GFX12-CU: ; %bb.0: ; %entry
+; GFX12-CU-NEXT: s_endpgm
entry:
fence syncscope("singlethread-one-as") release
ret void
@@ -349,6 +399,14 @@ define amdgpu_kernel void @singlethread_one_as_acq_rel_fence() {
; GFX11-CU-LABEL: singlethread_one_as_acq_rel_fence:
; GFX11-CU: ; %bb.0: ; %entry
; GFX11-CU-NEXT: s_endpgm
+;
+; GFX12-WGP-LABEL: singlethread_one_as_acq_rel_fence:
+; GFX12-WGP: ; %bb.0: ; %entry
+; GFX12-WGP-NEXT: s_endpgm
+;
+; GFX12-CU-LABEL: singlethread_one_as_acq_rel_fence:
+; GFX12-CU: ; %bb.0: ; %entry
+; GFX12-CU-NEXT: s_endpgm
entry:
fence syncscope("singlethread-one-as") acq_rel
ret void
@@ -398,6 +456,14 @@ define amdgpu_kernel void @singlethread_one_as_seq_cst_fence() {
; GFX11-CU-LABEL: singlethread_one_as_seq_cst_fence:
; GFX11-CU: ; %bb.0: ; %entry
; GFX11-CU-NEXT: s_endpgm
+;
+; GFX12-WGP-LABEL: singlethread_one_as_seq_cst_fence:
+; GFX12-WGP: ; %bb.0: ; %entry
+; GFX12-WGP-NEXT: s_endpgm
+;
+; GFX12-CU-LABEL: singlethread_one_as_seq_cst_fence:
+; GFX12-CU: ; %bb.0: ; %entry
+; GFX12-CU-NEXT: s_endpgm
entry:
fence syncscope("singlethread-one-as") seq_cst
ret void
@@ -447,6 +513,14 @@ define amdgpu_kernel void @wavefront_acquire_fence() {
; GFX11-CU-LABEL: wavefront_acquire_fence:
; GFX11-CU: ; %bb.0: ; %entry
; GFX11-CU-NEXT: s_endpgm
+;
+; GFX12-WGP-LABEL: wavefront_acquire_fence:
+; GFX12-WGP: ; %bb.0: ; %entry
+; GFX12-WGP-NEXT: s_endpgm
+;
+; GFX12-CU-LABEL: wavefront_acquire_fence:
+; GFX12-CU: ; %bb.0: ; %entry
+; GFX12-CU-NEXT: s_endpgm
entry:
fence syncscope("wavefront") acquire
ret void
@@ -496,6 +570,14 @@ define amdgpu_kernel void @wavefront_release_fence() {
; GFX11-CU-LABEL: wavefront_release_fence:
; GFX11-CU: ; %bb.0: ; %entry
; GFX11-CU-NEXT: s_endpgm
+;
+; GFX12-WGP-LABEL: wavefront_release_fence:
+; GFX12-WGP: ; %bb.0: ; %entry
+; GFX12-WGP-NEXT: s_endpgm
+;
+; GFX12-CU-LABEL: wavefront_release_fence:
+; GFX12-CU: ; %bb.0: ; %entry
+; GFX12-CU-NEXT: s_endpgm
entry:
fence syncscope("wavefront") release
ret void
@@ -545,6 +627,14 @@ define amdgpu_kernel void @wavefront_acq_rel_fence() {
; GFX11-CU-LABEL: wavefront_acq_rel_fence:
; GFX11-CU: ; %bb.0: ; %entry
; GFX11-CU-NEXT: s_endpgm
+;
+; GFX12-WGP-LABEL: wavefront_acq_rel_fence:
+; GFX12-WGP: ; %bb.0: ; %entry
+; GFX12-WGP-NEXT: s_endpgm
+;
+; GFX12-CU-LABEL: wavefront_acq_rel_fence:
+; GFX12-CU: ; %bb.0: ; %entry
+; GFX12-CU-NEXT: s_endpgm
entry:
fence syncscope("wavefront") acq_rel
ret void
@@ -594,6 +684,14 @@ define amdgpu_kernel void @wavefront_seq_cst_fence() {
; GFX11-CU-LABEL: wavefront_seq_cst_fence:
; GFX11-CU: ; %bb.0: ; %entry
; GFX11-CU-NEXT: s_endpgm
+;
+; GFX12-WGP-LABEL: wavefront_seq_cst_fence:
+; GFX12-WGP: ; %bb.0: ; %entry
+; GFX12-WGP-NEXT: s_endpgm
+;
+; GFX12-CU-LABEL: wavefront_seq_cst_fence:
+; GFX12-CU: ; %bb.0: ; %entry
+; GFX12-CU-NEXT: s_endpgm
entry:
fence syncscope("wavefront") seq_cst
ret void
@@ -643,6 +741,14 @@ define amdgpu_kernel void @wavefront_one_as_acquire_fence() {
; GFX11-CU-LABEL: wavefront_one_as_acquire_fence:
; GFX11-CU: ; %bb.0: ; %entry
; GFX11-CU-NEXT: s_endpgm
+;
+; GFX12-WGP-LABEL: wavefront_one_as_acquire_fence:
+; GFX12-WGP: ; %bb.0: ; %entry
+; GFX12-WGP-NEXT: s_endpgm
+;
+; GFX12-CU-LABEL: wavefront_one_as_acquire_fence:
+; GFX12-CU: ; %bb.0: ; %entry
+; GFX12-CU-NEXT: s_endpgm
entry:
fence syncscope("wavefront-one-as") acquire
ret void
@@ -692,6 +798,14 @@ define amdgpu_kernel void @wavefront_one_as_release_fence() {
; GFX11-CU-LABEL: wavefront_one_as_release_fence:
; GFX11-CU: ; %bb.0: ; %entry
; GFX11-CU-NEXT: s_endpgm
+;
+; GFX12-WGP-LABEL: wavefront_one_as_release_fence:
+; GFX12-WGP: ; %bb.0: ; %entry
+; GFX12-WGP-NEXT: s_endpgm
+;
+; GFX12-CU-LABEL: wavefront_one_as_release_fence:
+; GFX12-CU: ; %bb.0: ; %entry
+; GFX12-CU-NEXT: s_endpgm
entry:
fence syncscope("wavefront-one-as") release
ret void
@@ -741,6 +855,14 @@ define amdgpu_kernel void @wavefront_one_as_acq_rel_fence() {
; GFX11-CU-LABEL: wavefront_one_as_acq_rel_fence:
; GFX11-CU: ; %bb.0: ; %entry
; GFX11-CU-NEXT: s_endpgm
+;
+; GFX12-WGP-LABEL: wavefront_one_as_acq_rel_fence:
+; GFX12-WGP: ; %bb.0: ; %entry
+; GFX12-WGP-NEXT: s_endpgm
+;
+; GFX12-CU-LABEL: wavefront_one_as_acq_rel_fence:
+; GFX12-CU: ; %bb.0: ; %entry
+; GFX12-CU-NEXT: s_endpgm
entry:
fence syncscope("wavefront-one-as") acq_rel
ret void
@@ -790,6 +912,14 @@ define amdgpu_kernel void @wavefront_one_as_seq_cst_fence() {
; GFX11-CU-LABEL: wavefront_one_as_seq_cst_fence:
; GFX11-CU: ; %bb.0: ; %entry
; GFX11-CU-NEXT: s_endpgm
+;
+; GFX12-WGP-LABEL: wavefront_one_as_seq_cst_fence:
+; GFX12-WGP: ; %bb.0: ; %entry
+; GFX12-WGP-NEXT: s_endpgm
+;
+; GFX12-CU-LABEL: wavefront_one_as_seq_cst_fence:
+; GFX12-CU: ; %bb.0: ; %entry
+; GFX12-CU-NEXT: s_endpgm
entry:
fence syncscope("wavefront-one-as") seq_cst
ret void
@@ -843,6 +973,15 @@ define amdgpu_kernel void @workgroup_acquire_fence() {
; GFX11-CU-LABEL: workgroup_acquire_fence:
; GFX11-CU: ; %bb.0: ; %entry
; GFX11-CU-NEXT: s_endpgm
+;
+; GFX12-WGP-LABEL: workgroup_acquire_fence:
+; GFX12-WGP: ; %bb.0: ; %entry
+; GFX12-WGP-NEXT: global_inv scope:SCOPE_SE
+; GFX12-WGP-NEXT: s_endpgm
+;
+; GFX12-CU-LABEL: workgroup_acquire_fence:
+; GFX12-CU: ; %bb.0: ; %entry
+; GFX12-CU-NEXT: s_endpgm
entry:
fence syncscope("workgroup") acquire
ret void
@@ -892,6 +1031,14 @@ define amdgpu_kernel void @workgroup_release_fence() {
; GFX11-CU-LABEL: workgroup_release_fence:
; GFX11-CU: ; %bb.0: ; %entry
; GFX11-CU-NEXT: s_endpgm
+;
+; GFX12-WGP-LABEL: workgroup_release_fence:
+; GFX12-WGP: ; %bb.0: ; %entry
+; GFX12-WGP-NEXT: s_endpgm
+;
+; GFX12-CU-LABEL: workgroup_release_fence:
+; GFX12-CU: ; %bb.0: ; %entry
+; GFX12-CU-NEXT: s_endpgm
entry:
fence syncscope("workgroup") release
ret void
@@ -945,6 +1092,15 @@ define amdgpu_kernel void @workgroup_acq_rel_fence() {
; GFX11-CU-LABEL: workgroup_acq_rel_fence:
; GFX11-CU: ; %bb.0: ; %entry
; GFX11-CU-NEXT: s_endpgm
+;
+; GFX12-WGP-LABEL: workgroup_acq_rel_fence:
+; GFX12-WGP: ; %bb.0: ; %entry
+; GFX12-WGP-NEXT: global_inv scope:SCOPE_SE
+; GFX12-WGP-NEXT: s_endpgm
+;
+; GFX12-CU-LABEL: workgroup_acq_rel_fence:
+; GFX12-CU: ; %bb.0: ; %entry
+; GFX12-CU-NEXT: s_endpgm
entry:
fence syncscope("workgroup") acq_rel
ret void
@@ -998,6 +1154,15 @@ define amdgpu_kernel void @workgroup_seq_cst_fence() {
; GFX11-CU-LABEL: workgroup_seq_cst_fence:
; GFX11-CU: ; %bb.0: ; %entry
; GFX11-CU-NEXT: s_endpgm
+;
+; GFX12-WGP-LABEL: workgroup_seq_cst_fence:
+; GFX12-WGP: ; %bb.0: ; %entry
+; GFX12-WGP-NEXT: global_inv scope:SCOPE_SE
+; GFX12-WGP-NEXT: s_endpgm
+;
+; GFX12-CU-LABEL: workgroup_seq_cst_fence:
+; GFX12-CU: ; %bb.0: ; %entry
+; GFX12-CU-NEXT: s_endpgm
entry:
fence syncscope("workgroup") seq_cst
ret void
@@ -1051,6 +1216,15 @@ define amdgpu_kernel void @workgroup_one_as_acquire_fence() {
; GFX11-CU-LABEL: workgroup_one_as_acquire_fence:
; GFX11-CU: ; %bb.0: ; %entry
; GFX11-CU-NEXT: s_endpgm
+;
+; GFX12-WGP-LABEL: workgroup_one_as_acquire_fence:
+; GFX12-WGP: ; %bb.0: ; %entry
+; GFX12-WGP-NEXT: global_inv scope:SCOPE_SE
+; GFX12-WGP-NEXT: s_endpgm
+;
+; GFX12-CU-LABEL: workgroup_one_as_acquire_fence:
+; GFX12-CU: ; %bb.0: ; %entry
+; GFX12-CU-NEXT: s_endpgm
entry:
fence syncscope("workgroup-one-as") acquire
ret void
@@ -1100,6 +1274,14 @@ define amdgpu_kernel void @workgroup_one_as_release_fence() {
; GFX11-CU-LABEL: workgroup_one_as_release_fence:
; GFX11-CU: ; %bb.0: ; %entry
; GFX11-CU-NEXT: s_endpgm
+;
+; GFX12-WGP-LABEL: workgroup_one_as_release_fence:
+; GFX12-WGP: ; %bb.0: ; %entry
+; GFX12-WGP-NEXT: s_endpgm
+;
+; GFX12-CU-LABEL: workgroup_one_as_release_fence:
+; GFX12-CU: ; %bb.0: ; %entry
+; GFX12-CU-NEXT: s_endpgm
entry:
fence syncscope("workgroup-one-as") release
ret void
@@ -1153,6 +1335,15 @@ define amdgpu_kernel void @workgroup_one_as_acq_rel_fence() {
; GFX11-CU-LABEL: workgroup_one_as_acq_rel_fence:
; GFX11-CU: ; %bb.0: ; %entry
; GFX11-CU-NEXT: s_endpgm
+;
+; GFX12-WGP-LABEL: workgroup_one_as_acq_rel_fence:
+; GFX12-WGP: ; %bb.0: ; %entry
+; GFX12-WGP-NEXT: global_inv scope:SCOPE_SE
+; GFX12-WGP-NEXT: s_endpgm
+;
+; GFX12-CU-LABEL: workgroup_one_as_acq_rel_fence:
+; GFX12-CU: ; %bb.0: ; %entry
+; GFX12-CU-NEXT: s_endpgm
entry:
fence syncscope("workgroup-one-as") acq_rel
ret void
@@ -1206,6 +1397,15 @@ define amdgpu_kernel void @workgroup_one_as_seq_cst_fence() {
; GFX11-CU-LABEL: workgroup_one_as_seq_cst_fence:
; GFX11-CU: ; %bb.0: ; %entry
; GFX11-CU-NEXT: s_endpgm
+;
+; GFX12-WGP-LABEL: workgroup_one_as_seq_cst_fence:
+; GFX12-WGP: ; %bb.0: ; %entry
+; GFX12-WGP-NEXT: global_inv scope:SCOPE_SE
+; GFX12-WGP-NEXT: s_endpgm
+;
+; GFX12-CU-LABEL: workgroup_one_as_seq_cst_fence:
+; GFX12-CU: ; %bb.0: ; %entry
+; GFX12-CU-NEXT: s_endpgm
entry:
fence syncscope("workgroup-one-as") seq_cst
ret void
@@ -1269,6 +1469,16 @@ define amdgpu_kernel void @agent_acquire_fence() {
; GFX11-CU-NEXT: buffer_gl1_inv
; GFX11-CU-NEXT: buffer_gl0_inv
; GFX11-CU-NEXT: s_endpgm
+;
+; GFX12-WGP-LABEL: agent_acquire_fence:
+; GFX12-WGP: ; %bb.0: ; %entry
+; GFX12-WGP-NEXT: global_inv scope:SCOPE_DEV
+; GFX12-WGP-NEXT: s_endpgm
+;
+; GFX12-CU-LABEL: agent_acquire_fence:
+; GFX12-CU: ; %bb.0: ; %entry
+; GFX12-CU-NEXT: global_inv scope:SCOPE_DEV
+; GFX12-CU-NEXT: s_endpgm
entry:
fence syncscope("agent") acquire
ret void
@@ -1320,6 +1530,14 @@ define amdgpu_kernel void @agent_release_fence() {
; GFX11-CU-LABEL: agent_release_fence:
; GFX11-CU: ; %bb.0: ; %entry
; GFX11-CU-NEXT: s_endpgm
+;
+; GFX12-WGP-LABEL: agent_release_fence:
+; GFX12-WGP: ; %bb.0: ; %entry
+; GFX12-WGP-NEXT: s_endpgm
+;
+; GFX12-CU-LABEL: agent_release_fence:
+; GFX12-CU: ; %bb.0: ; %entry
+; GFX12-CU-NEXT: s_endpgm
entry:
fence syncscope("agent") release
ret void
@@ -1385,6 +1603,16 @@ define amdgpu_kernel void @agent_acq_rel_fence() {
; GFX11-CU-NEXT: buffer_gl1_inv
; GFX11-CU-NEXT: buffer_gl0_inv
; GFX11-CU-NEXT: s_endpgm
+;
+; GFX12-WGP-LABEL: agent_acq_rel_fence:
+; GFX12-WGP: ; %bb.0: ; %entry
+; GFX12-WGP-NEXT: global_inv scope:SCOPE_DEV
+; GFX12-WGP-NEXT: s_endpgm
+;
+; GFX12-CU-LABEL: agent_acq_rel_fence:
+; GFX12-CU: ; %bb.0: ; %entry
+; GFX12-CU-NEXT: global_inv scope:SCOPE_DEV
+; GFX12-CU-NEXT: s_endpgm
entry:
fence syncscope("agent") acq_rel
ret void
@@ -1450,6 +1678,16 @@ define amdgpu_kernel void @agent_seq_cst_fence() {
; GFX11-CU-NEXT: buffer_gl1_inv
; GFX11-CU-NEXT: buffer_gl0_inv
; GFX11-CU-NEXT: s_endpgm
+;
+; GFX12-WGP-LABEL: agent_seq_cst_fence:
+; GFX12-WGP: ; %bb.0: ; %entry
+; GFX12-WGP-NEXT: global_inv scope:SCOPE_DEV
+; GFX12-WGP-NEXT: s_endpgm
+;
+; GFX12-CU-LABEL: agent_seq_cst_fence:
+; GFX12-CU: ; %bb.0: ; %entry
+; GFX12-CU-NEXT: global_inv scope:SCOPE_DEV
+; GFX12-CU-NEXT: s_endpgm
entry:
fence syncscope("agent") seq_cst
ret void
@@ -1513,6 +1751,16 @@ define amdgpu_kernel void @agent_one_as_acquire_fence() {
; GFX11-CU-NEXT: buffer_gl1_inv
; GFX11-CU-NEXT: buffer_gl0_inv
; GFX11-CU-NEXT: s_endpgm
+;
+; GFX12-WGP-LABEL: agent_one_as_acquire_fence:
+; GFX12-WGP: ; %bb.0: ; %entry
+; GFX12-WGP-NEXT: global_inv scope:SCOPE_DEV
+; GFX12-WGP-NEXT: s_endpgm
+;
+; GFX12-CU-LABEL: agent_one_as_acquire_fence:
+; GFX12-CU: ; %bb.0: ; %entry
+; GFX12-CU-NEXT: global_inv scope:SCOPE_DEV
+; GFX12-CU-NEXT: s_endpgm
entry:
fence syncscope("agent-one-as") acquire
ret void
@@ -1564,6 +1812,14 @@ define amdgpu_kernel void @agent_one_as_release_fence() {
; GFX11-CU-LABEL: agent_one_as_release_fence:
; GFX11-CU: ; %bb.0: ; %entry
; GFX11-CU-NEXT: s_endpgm
+;
+; GFX12-WGP-LABEL: agent_one_as_release_fence:
+; GFX12-WGP: ; %bb.0: ; %entry
+; GFX12-WGP-NEXT: s_endpgm
+;
+; GFX12-CU-LABEL: agent_one_as_release_fence:
+; GFX12-CU: ; %bb.0: ; %entry
+; GFX12-CU-NEXT: s_endpgm
entry:
fence syncscope("agent-one-as") release
ret void
@@ -1629,6 +1885,16 @@ define amdgpu_kernel void @agent_one_as_acq_rel_fence() {
; GFX11-CU-NEXT: buffer_gl1_inv
; GFX11-CU-NEXT: buffer_gl0_inv
; GFX11-CU-NEXT: s_endpgm
+;
+; GFX12-WGP-LABEL: agent_one_as_acq_rel_fence:
+; GFX12-WGP: ; %bb.0: ; %entry
+; GFX12-WGP-NEXT: global_inv scope:SCOPE_DEV
+; GFX12-WGP-NEXT: s_endpgm
+;
+; GFX12-CU-LABEL: agent_one_as_acq_rel_fence:
+; GFX12-CU: ; %bb.0: ; %entry
+; GFX12-CU-NEXT: global_inv scope:SCOPE_DEV
+; GFX12-CU-NEXT: s_endpgm
entry:
fence syncscope("agent-one-as") acq_rel
ret void
@@ -1694,6 +1960,16 @@ define amdgpu_kernel void @agent_one_as_seq_cst_fence() {
; GFX11-CU-NEXT: buffer_gl1_inv
; GFX11-CU-NEXT: buffer_gl0_inv
; GFX11-CU-NEXT: s_endpgm
+;
+; GFX12-WGP-LABEL: agent_one_as_seq_cst_fence:
+; GFX12-WGP: ; %bb.0: ; %entry
+; GFX12-WGP-NEXT: global_inv scope:SCOPE_DEV
+; GFX12-WGP-NEXT: s_endpgm
+;
+; GFX12-CU-LABEL: agent_one_as_seq_cst_fence:
+; GFX12-CU: ; %bb.0: ; %entry
+; GFX12-CU-NEXT: global_inv scope:SCOPE_DEV
+; GFX12-CU-NEXT: s_endpgm
entry:
fence syncscope("agent-one-as") seq_cst
ret void
@@ -1759,6 +2035,16 @@ define amdgpu_kernel void @system_acquire_fence() {
; GFX11-CU-NEXT: buffer_gl1_inv
; GFX11-CU-NEXT: buffer_gl0_inv
; GFX11-CU-NEXT: s_endpgm
+;
+; GFX12-WGP-LABEL: system_acquire_fence:
+; GFX12-WGP: ; %bb.0: ; %entry
+; GFX12-WGP-NEXT: global_inv scope:SCOPE_SYS
+; GFX12-WGP-NEXT: s_endpgm
+;
+; GFX12-CU-LABEL: system_acquire_fence:
+; GFX12-CU: ; %bb.0: ; %entry
+; GFX12-CU-NEXT: global_inv scope:SCOPE_SYS
+; GFX12-CU-NEXT: s_endpgm
entry:
fence acquire
ret void
@@ -1812,6 +2098,14 @@ define amdgpu_kernel void @system_release_fence() {
; GFX11-CU-LABEL: system_release_fence:
; GFX11-CU: ; %bb.0: ; %entry
; GFX11-CU-NEXT: s_endpgm
+;
+; GFX12-WGP-LABEL: system_release_fence:
+; GFX12-WGP: ; %bb.0: ; %entry
+; GFX12-WGP-NEXT: s_endpgm
+;
+; GFX12-CU-LABEL: system_release_fence:
+; GFX12-CU: ; %bb.0: ; %entry
+; GFX12-CU-NEXT: s_endpgm
entry:
fence release
ret void
@@ -1881,6 +2175,16 @@ define amdgpu_kernel void @system_acq_rel_fence() {
; GFX11-CU-NEXT: buffer_gl1_inv
; GFX11-CU-NEXT: buffer_gl0_inv
; GFX11-CU-NEXT: s_endpgm
+;
+; GFX12-WGP-LABEL: system_acq_rel_fence:
+; GFX12-WGP: ; %bb.0: ; %entry
+; GFX12-WGP-NEXT: global_inv scope:SCOPE_SYS
+; GFX12-WGP-NEXT: s_endpgm
+;
+; GFX12-CU-LABEL: system_acq_rel_fence:
+; GFX12-CU: ; %bb.0: ; %entry
+; GFX12-CU-NEXT: global_inv scope:SCOPE_SYS
+; GFX12-CU-NEXT: s_endpgm
entry:
fence acq_rel
ret void
@@ -1950,6 +2254,16 @@ define amdgpu_kernel void @system_seq_cst_fence() {
; GFX11-CU-NEXT: buffer_gl1_inv
; GFX11-CU-NEXT: buffer_gl0_inv
; GFX11-CU-NEXT: s_endpgm
+;
+; GFX12-WGP-LABEL: system_seq_cst_fence:
+; GFX12-WGP: ; %bb.0: ; %entry
+; GFX12-WGP-NEXT: global_inv scope:SCOPE_SYS
+; GFX12-WGP-NEXT: s_endpgm
+;
+; GFX12-CU-LABEL: system_seq_cst_fence:
+; GFX12-CU: ; %bb.0: ; %entry
+; GFX12-CU-NEXT: global_inv scope:SCOPE_SYS
+; GFX12-CU-NEXT: s_endpgm
e...
[truncated]
|
Iterator MI can advance in insertWait() but we need original instruction to set temporal hint. Just move it before handling volatile.
mbrkusanin
force-pushed
the
memory-legalizer-nontemporal-fix
branch
from
c5f3f7c
to
febe0d9
Compare
|
Rebase |
jayfoad
approved these changes
Mar 4, 2024
|
I wonder if this is worth backporting to the LLVM 18 release branch. |
tstellar
pushed a commit
that referenced
this pull request
Apr 26, 2024
Iterator MI can advance in insertWait() but we need original instruction to set temporal hint. Just move it before handling volatile.
Closed
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
Iterator MI can advance in insertWait() but we need original instruction to
set temporal hint. Just move it before handling volatile.