[llvm] [AMDGPU] Fix AAAMDWavesPerEU to union caller waves-per-eu ranges (PR #207903)
Arseniy Obolenskiy via llvm-commits
llvm-commits at lists.llvm.org
Wed Jul 15 04:49:11 PDT 2026
https://github.com/aobolensk updated https://github.com/llvm/llvm-project/pull/207903
>From 67ecff476555135f3c4bfbe218e2571a85f6a69a Mon Sep 17 00:00:00 2001
From: Arseniy Obolenskiy <arseniy.obolenskiy at amd.com>
Date: Tue, 7 Jul 2026 08:09:03 +0200
Subject: [PATCH 1/3] [AMDGPU] Fix AAAMDWavesPerEU to union caller waves-per-eu
ranges
updateImpl used max() on both bounds, so a callee shared by kernels with [1,1] and [8,8] got [8,8] instead of [1,8]
Remove custom implementation that leads to miscompilation and reuse `updateImplImpl<AAAMDWavesPerEU>()`
---
llvm/lib/Target/AMDGPU/AMDGPUAttributor.cpp | 34 +-------------
.../CodeGen/AMDGPU/propagate-waves-per-eu.ll | 44 +++++++++----------
2 files changed, 23 insertions(+), 55 deletions(-)
diff --git a/llvm/lib/Target/AMDGPU/AMDGPUAttributor.cpp b/llvm/lib/Target/AMDGPU/AMDGPUAttributor.cpp
index 630ffad96e451..e4b40a7e7be7b 100644
--- a/llvm/lib/Target/AMDGPU/AMDGPUAttributor.cpp
+++ b/llvm/lib/Target/AMDGPU/AMDGPUAttributor.cpp
@@ -1142,39 +1142,7 @@ struct AAAMDWavesPerEU : public AAAMDSizeRangeAttribute {
}
ChangeStatus updateImpl(Attributor &A) override {
- ChangeStatus Change = ChangeStatus::UNCHANGED;
-
- auto CheckCallSite = [&](AbstractCallSite CS) {
- Function *Caller = CS.getInstruction()->getFunction();
- Function *Func = getAssociatedFunction();
- LLVM_DEBUG(dbgs() << '[' << getName() << "] Call " << Caller->getName()
- << "->" << Func->getName() << '\n');
- (void)Func;
-
- const auto *CallerAA = A.getAAFor<AAAMDWavesPerEU>(
- *this, IRPosition::function(*Caller), DepClassTy::REQUIRED);
- if (!CallerAA || !CallerAA->isValidState())
- return false;
-
- ConstantRange Assumed = getAssumed();
- unsigned Min = std::max(Assumed.getLower().getZExtValue(),
- CallerAA->getAssumed().getLower().getZExtValue());
- unsigned Max = std::max(Assumed.getUpper().getZExtValue(),
- CallerAA->getAssumed().getUpper().getZExtValue());
- ConstantRange Range(APInt(32, Min), APInt(32, Max));
- IntegerRangeState RangeState(Range);
- getState() = RangeState;
- Change |= getState() == Assumed ? ChangeStatus::UNCHANGED
- : ChangeStatus::CHANGED;
-
- return true;
- };
-
- bool AllCallSitesKnown = true;
- if (!A.checkForAllCallSites(CheckCallSite, *this, true, AllCallSitesKnown))
- return indicatePessimisticFixpoint();
-
- return Change;
+ return updateImplImpl<AAAMDWavesPerEU>(A);
}
/// Create an abstract attribute view for the position \p IRP.
diff --git a/llvm/test/CodeGen/AMDGPU/propagate-waves-per-eu.ll b/llvm/test/CodeGen/AMDGPU/propagate-waves-per-eu.ll
index a19fe155d1c0e..9d19a66433d72 100644
--- a/llvm/test/CodeGen/AMDGPU/propagate-waves-per-eu.ll
+++ b/llvm/test/CodeGen/AMDGPU/propagate-waves-per-eu.ll
@@ -14,7 +14,7 @@ define internal void @default_to_1_8_a() {
define amdgpu_kernel void @kernel_1_8() #0 {
; CHECK-LABEL: define amdgpu_kernel void @kernel_1_8
-; CHECK-SAME: () #[[ATTR1:[0-9]+]] {
+; CHECK-SAME: () #[[ATTR0]] {
; CHECK-NEXT: call void @default_to_1_8_a()
; CHECK-NEXT: ret void
;
@@ -25,7 +25,7 @@ define amdgpu_kernel void @kernel_1_8() #0 {
; Called from a single kernel with 1,2
define internal void @default_to_1_2() {
; CHECK-LABEL: define internal void @default_to_1_2
-; CHECK-SAME: () #[[ATTR2:[0-9]+]] {
+; CHECK-SAME: () #[[ATTR1:[0-9]+]] {
; CHECK-NEXT: ret void
;
ret void
@@ -33,7 +33,7 @@ define internal void @default_to_1_2() {
define amdgpu_kernel void @kernel_1_2() #1 {
; CHECK-LABEL: define amdgpu_kernel void @kernel_1_2
-; CHECK-SAME: () #[[ATTR2]] {
+; CHECK-SAME: () #[[ATTR1]] {
; CHECK-NEXT: call void @default_to_1_2()
; CHECK-NEXT: call void @flat_group_1_1()
; CHECK-NEXT: call void @default_to_1_8_b()
@@ -50,7 +50,7 @@ define amdgpu_kernel void @kernel_1_2() #1 {
; Called from a single kernel with 1,4
define internal void @default_to_1_4() {
; CHECK-LABEL: define internal void @default_to_1_4
-; CHECK-SAME: () #[[ATTR3:[0-9]+]] {
+; CHECK-SAME: () #[[ATTR2:[0-9]+]] {
; CHECK-NEXT: ret void
;
ret void
@@ -58,7 +58,7 @@ define internal void @default_to_1_4() {
define amdgpu_kernel void @kernel_1_4() #2 {
; CHECK-LABEL: define amdgpu_kernel void @kernel_1_4
-; CHECK-SAME: () #[[ATTR3]] {
+; CHECK-SAME: () #[[ATTR2]] {
; CHECK-NEXT: call void @default_to_1_4()
; CHECK-NEXT: ret void
;
@@ -69,7 +69,7 @@ define amdgpu_kernel void @kernel_1_4() #2 {
; Called from kernels with 2,9 and 9,9
define internal void @default_to_2_9() {
; CHECK-LABEL: define internal void @default_to_2_9
-; CHECK-SAME: () #[[ATTR4:[0-9]+]] {
+; CHECK-SAME: () #[[ATTR3:[0-9]+]] {
; CHECK-NEXT: ret void
;
ret void
@@ -79,7 +79,7 @@ define internal void @default_to_2_9() {
; bounds, and should not be changed.
define internal void @flat_group_1_1() #3 {
; CHECK-LABEL: define internal void @flat_group_1_1
-; CHECK-SAME: () #[[ATTR5:[0-9]+]] {
+; CHECK-SAME: () #[[ATTR4:[0-9]+]] {
; CHECK-NEXT: ret void
;
ret void
@@ -88,7 +88,7 @@ define internal void @flat_group_1_1() #3 {
; 2,8 -> 2,2
define internal void @flat_group_2_8() #4 {
; CHECK-LABEL: define internal void @flat_group_2_8
-; CHECK-SAME: () #[[ATTR0]] {
+; CHECK-SAME: () #[[ATTR5:[0-9]+]] {
; CHECK-NEXT: ret void
;
ret void
@@ -105,7 +105,7 @@ define internal void @flat_group_9_10() #5 {
define amdgpu_kernel void @kernel_2_9() #6 {
; CHECK-LABEL: define amdgpu_kernel void @kernel_2_9
-; CHECK-SAME: () #[[ATTR7:[0-9]+]] {
+; CHECK-SAME: () #[[ATTR3]] {
; CHECK-NEXT: call void @default_to_2_9()
; CHECK-NEXT: call void @flat_group_1_1()
; CHECK-NEXT: ret void
@@ -117,7 +117,7 @@ define amdgpu_kernel void @kernel_2_9() #6 {
define amdgpu_kernel void @kernel_9_9() #7 {
; CHECK-LABEL: define amdgpu_kernel void @kernel_9_9
-; CHECK-SAME: () #[[ATTR4]] {
+; CHECK-SAME: () #[[ATTR7:[0-9]+]] {
; CHECK-NEXT: call void @default_to_2_9()
; CHECK-NEXT: call void @flat_group_9_10()
; CHECK-NEXT: ret void
@@ -140,7 +140,7 @@ define internal void @default_to_1_8_b() {
; this should probably be illegal.
define amdgpu_kernel void @kernel_2_8() #4 {
; CHECK-LABEL: define amdgpu_kernel void @kernel_2_8
-; CHECK-SAME: () #[[ATTR0]] {
+; CHECK-SAME: () #[[ATTR5]] {
; CHECK-NEXT: call void @default_to_1_8_a()
; CHECK-NEXT: call void @default_to_1_8_b()
; CHECK-NEXT: ret void
@@ -153,7 +153,7 @@ define amdgpu_kernel void @kernel_2_8() #4 {
; 1,2 -> 2,2
define internal void @merge_cycle_0() #1 {
; CHECK-LABEL: define internal void @merge_cycle_0
-; CHECK-SAME: () #[[ATTR2]] {
+; CHECK-SAME: () #[[ATTR1]] {
; CHECK-NEXT: call void @merge_cycle_1()
; CHECK-NEXT: ret void
;
@@ -165,7 +165,7 @@ define internal void @merge_cycle_0() #1 {
; 2,8 -> 2,8
define internal void @merge_cycle_1() #4 {
; CHECK-LABEL: define internal void @merge_cycle_1
-; CHECK-SAME: () #[[ATTR0]] {
+; CHECK-SAME: () #[[ATTR5]] {
; CHECK-NEXT: call void @merge_cycle_0()
; CHECK-NEXT: ret void
;
@@ -218,7 +218,7 @@ define internal i32 @bitcasted_function() {
define internal void @called_from_invalid_bounds_0() {
; CHECK-LABEL: define internal void @called_from_invalid_bounds_0
-; CHECK-SAME: () #[[ATTR1]] {
+; CHECK-SAME: () #[[ATTR0]] {
; CHECK-NEXT: ret void
;
ret void
@@ -235,7 +235,7 @@ define internal void @called_from_invalid_bounds_1() {
; Invalid range for amdgpu-waves-per-eu
define amdgpu_kernel void @kernel_invalid_bounds_0_8() #9 {
; CHECK-LABEL: define amdgpu_kernel void @kernel_invalid_bounds_0_8
-; CHECK-SAME: () #[[ATTR1]] {
+; CHECK-SAME: () #[[ATTR0]] {
; CHECK-NEXT: call void @called_from_invalid_bounds_0()
; CHECK-NEXT: ret void
;
@@ -399,14 +399,14 @@ attributes #17 = { "amdgpu-waves-per-eu"="5,8" }
attributes #18 = { "amdgpu-waves-per-eu"="9,10" }
attributes #19 = { "amdgpu-waves-per-eu"="8,9" }
;.
-; CHECK: attributes #[[ATTR0]] = { "amdgpu-flat-work-group-size"="1,64" "amdgpu-no-cluster-id-x" "amdgpu-no-cluster-id-y" "amdgpu-no-cluster-id-z" "amdgpu-no-completion-action" "amdgpu-no-default-queue" "amdgpu-no-dispatch-id" "amdgpu-no-dispatch-ptr" "amdgpu-no-flat-scratch-init" "amdgpu-no-heap-ptr" "amdgpu-no-hostcall-ptr" "amdgpu-no-implicitarg-ptr" "amdgpu-no-lds-kernel-id" "amdgpu-no-multigrid-sync-arg" "amdgpu-no-queue-ptr" "amdgpu-no-workgroup-id-x" "amdgpu-no-workgroup-id-y" "amdgpu-no-workgroup-id-z" "amdgpu-no-workitem-id-x" "amdgpu-no-workitem-id-y" "amdgpu-no-workitem-id-z" "amdgpu-no-wwm" "amdgpu-waves-per-eu"="2,8" }
-; CHECK: attributes #[[ATTR1]] = { "amdgpu-flat-work-group-size"="1,64" "amdgpu-no-cluster-id-x" "amdgpu-no-cluster-id-y" "amdgpu-no-cluster-id-z" "amdgpu-no-completion-action" "amdgpu-no-default-queue" "amdgpu-no-dispatch-id" "amdgpu-no-dispatch-ptr" "amdgpu-no-flat-scratch-init" "amdgpu-no-heap-ptr" "amdgpu-no-hostcall-ptr" "amdgpu-no-implicitarg-ptr" "amdgpu-no-lds-kernel-id" "amdgpu-no-multigrid-sync-arg" "amdgpu-no-queue-ptr" "amdgpu-no-workgroup-id-x" "amdgpu-no-workgroup-id-y" "amdgpu-no-workgroup-id-z" "amdgpu-no-workitem-id-x" "amdgpu-no-workitem-id-y" "amdgpu-no-workitem-id-z" "amdgpu-no-wwm" "amdgpu-waves-per-eu"="1,8" }
-; CHECK: attributes #[[ATTR2]] = { "amdgpu-flat-work-group-size"="1,64" "amdgpu-no-cluster-id-x" "amdgpu-no-cluster-id-y" "amdgpu-no-cluster-id-z" "amdgpu-no-completion-action" "amdgpu-no-default-queue" "amdgpu-no-dispatch-id" "amdgpu-no-dispatch-ptr" "amdgpu-no-flat-scratch-init" "amdgpu-no-heap-ptr" "amdgpu-no-hostcall-ptr" "amdgpu-no-implicitarg-ptr" "amdgpu-no-lds-kernel-id" "amdgpu-no-multigrid-sync-arg" "amdgpu-no-queue-ptr" "amdgpu-no-workgroup-id-x" "amdgpu-no-workgroup-id-y" "amdgpu-no-workgroup-id-z" "amdgpu-no-workitem-id-x" "amdgpu-no-workitem-id-y" "amdgpu-no-workitem-id-z" "amdgpu-no-wwm" "amdgpu-waves-per-eu"="1,2" }
-; CHECK: attributes #[[ATTR3]] = { "amdgpu-flat-work-group-size"="1,64" "amdgpu-no-cluster-id-x" "amdgpu-no-cluster-id-y" "amdgpu-no-cluster-id-z" "amdgpu-no-completion-action" "amdgpu-no-default-queue" "amdgpu-no-dispatch-id" "amdgpu-no-dispatch-ptr" "amdgpu-no-flat-scratch-init" "amdgpu-no-heap-ptr" "amdgpu-no-hostcall-ptr" "amdgpu-no-implicitarg-ptr" "amdgpu-no-lds-kernel-id" "amdgpu-no-multigrid-sync-arg" "amdgpu-no-queue-ptr" "amdgpu-no-workgroup-id-x" "amdgpu-no-workgroup-id-y" "amdgpu-no-workgroup-id-z" "amdgpu-no-workitem-id-x" "amdgpu-no-workitem-id-y" "amdgpu-no-workitem-id-z" "amdgpu-no-wwm" "amdgpu-waves-per-eu"="1,4" }
-; CHECK: attributes #[[ATTR4]] = { "amdgpu-flat-work-group-size"="1,64" "amdgpu-no-cluster-id-x" "amdgpu-no-cluster-id-y" "amdgpu-no-cluster-id-z" "amdgpu-no-completion-action" "amdgpu-no-default-queue" "amdgpu-no-dispatch-id" "amdgpu-no-dispatch-ptr" "amdgpu-no-flat-scratch-init" "amdgpu-no-heap-ptr" "amdgpu-no-hostcall-ptr" "amdgpu-no-implicitarg-ptr" "amdgpu-no-lds-kernel-id" "amdgpu-no-multigrid-sync-arg" "amdgpu-no-queue-ptr" "amdgpu-no-workgroup-id-x" "amdgpu-no-workgroup-id-y" "amdgpu-no-workgroup-id-z" "amdgpu-no-workitem-id-x" "amdgpu-no-workitem-id-y" "amdgpu-no-workitem-id-z" "amdgpu-no-wwm" "amdgpu-waves-per-eu"="9,9" }
-; CHECK: attributes #[[ATTR5]] = { "amdgpu-flat-work-group-size"="1,64" "amdgpu-no-cluster-id-x" "amdgpu-no-cluster-id-y" "amdgpu-no-cluster-id-z" "amdgpu-no-completion-action" "amdgpu-no-default-queue" "amdgpu-no-dispatch-id" "amdgpu-no-dispatch-ptr" "amdgpu-no-flat-scratch-init" "amdgpu-no-heap-ptr" "amdgpu-no-hostcall-ptr" "amdgpu-no-implicitarg-ptr" "amdgpu-no-lds-kernel-id" "amdgpu-no-multigrid-sync-arg" "amdgpu-no-queue-ptr" "amdgpu-no-workgroup-id-x" "amdgpu-no-workgroup-id-y" "amdgpu-no-workgroup-id-z" "amdgpu-no-workitem-id-x" "amdgpu-no-workitem-id-y" "amdgpu-no-workitem-id-z" "amdgpu-no-wwm" "amdgpu-waves-per-eu"="1,1" }
+; CHECK: attributes #[[ATTR0]] = { "amdgpu-flat-work-group-size"="1,64" "amdgpu-no-cluster-id-x" "amdgpu-no-cluster-id-y" "amdgpu-no-cluster-id-z" "amdgpu-no-completion-action" "amdgpu-no-default-queue" "amdgpu-no-dispatch-id" "amdgpu-no-dispatch-ptr" "amdgpu-no-flat-scratch-init" "amdgpu-no-heap-ptr" "amdgpu-no-hostcall-ptr" "amdgpu-no-implicitarg-ptr" "amdgpu-no-lds-kernel-id" "amdgpu-no-multigrid-sync-arg" "amdgpu-no-queue-ptr" "amdgpu-no-workgroup-id-x" "amdgpu-no-workgroup-id-y" "amdgpu-no-workgroup-id-z" "amdgpu-no-workitem-id-x" "amdgpu-no-workitem-id-y" "amdgpu-no-workitem-id-z" "amdgpu-no-wwm" "amdgpu-waves-per-eu"="1,8" }
+; CHECK: attributes #[[ATTR1]] = { "amdgpu-flat-work-group-size"="1,64" "amdgpu-no-cluster-id-x" "amdgpu-no-cluster-id-y" "amdgpu-no-cluster-id-z" "amdgpu-no-completion-action" "amdgpu-no-default-queue" "amdgpu-no-dispatch-id" "amdgpu-no-dispatch-ptr" "amdgpu-no-flat-scratch-init" "amdgpu-no-heap-ptr" "amdgpu-no-hostcall-ptr" "amdgpu-no-implicitarg-ptr" "amdgpu-no-lds-kernel-id" "amdgpu-no-multigrid-sync-arg" "amdgpu-no-queue-ptr" "amdgpu-no-workgroup-id-x" "amdgpu-no-workgroup-id-y" "amdgpu-no-workgroup-id-z" "amdgpu-no-workitem-id-x" "amdgpu-no-workitem-id-y" "amdgpu-no-workitem-id-z" "amdgpu-no-wwm" "amdgpu-waves-per-eu"="1,2" }
+; CHECK: attributes #[[ATTR2]] = { "amdgpu-flat-work-group-size"="1,64" "amdgpu-no-cluster-id-x" "amdgpu-no-cluster-id-y" "amdgpu-no-cluster-id-z" "amdgpu-no-completion-action" "amdgpu-no-default-queue" "amdgpu-no-dispatch-id" "amdgpu-no-dispatch-ptr" "amdgpu-no-flat-scratch-init" "amdgpu-no-heap-ptr" "amdgpu-no-hostcall-ptr" "amdgpu-no-implicitarg-ptr" "amdgpu-no-lds-kernel-id" "amdgpu-no-multigrid-sync-arg" "amdgpu-no-queue-ptr" "amdgpu-no-workgroup-id-x" "amdgpu-no-workgroup-id-y" "amdgpu-no-workgroup-id-z" "amdgpu-no-workitem-id-x" "amdgpu-no-workitem-id-y" "amdgpu-no-workitem-id-z" "amdgpu-no-wwm" "amdgpu-waves-per-eu"="1,4" }
+; CHECK: attributes #[[ATTR3]] = { "amdgpu-flat-work-group-size"="1,64" "amdgpu-no-cluster-id-x" "amdgpu-no-cluster-id-y" "amdgpu-no-cluster-id-z" "amdgpu-no-completion-action" "amdgpu-no-default-queue" "amdgpu-no-dispatch-id" "amdgpu-no-dispatch-ptr" "amdgpu-no-flat-scratch-init" "amdgpu-no-heap-ptr" "amdgpu-no-hostcall-ptr" "amdgpu-no-implicitarg-ptr" "amdgpu-no-lds-kernel-id" "amdgpu-no-multigrid-sync-arg" "amdgpu-no-queue-ptr" "amdgpu-no-workgroup-id-x" "amdgpu-no-workgroup-id-y" "amdgpu-no-workgroup-id-z" "amdgpu-no-workitem-id-x" "amdgpu-no-workitem-id-y" "amdgpu-no-workitem-id-z" "amdgpu-no-wwm" "amdgpu-waves-per-eu"="2,9" }
+; CHECK: attributes #[[ATTR4]] = { "amdgpu-flat-work-group-size"="1,64" "amdgpu-no-cluster-id-x" "amdgpu-no-cluster-id-y" "amdgpu-no-cluster-id-z" "amdgpu-no-completion-action" "amdgpu-no-default-queue" "amdgpu-no-dispatch-id" "amdgpu-no-dispatch-ptr" "amdgpu-no-flat-scratch-init" "amdgpu-no-heap-ptr" "amdgpu-no-hostcall-ptr" "amdgpu-no-implicitarg-ptr" "amdgpu-no-lds-kernel-id" "amdgpu-no-multigrid-sync-arg" "amdgpu-no-queue-ptr" "amdgpu-no-workgroup-id-x" "amdgpu-no-workgroup-id-y" "amdgpu-no-workgroup-id-z" "amdgpu-no-workitem-id-x" "amdgpu-no-workitem-id-y" "amdgpu-no-workitem-id-z" "amdgpu-no-wwm" "amdgpu-waves-per-eu"="1,1" }
+; CHECK: attributes #[[ATTR5]] = { "amdgpu-flat-work-group-size"="1,64" "amdgpu-no-cluster-id-x" "amdgpu-no-cluster-id-y" "amdgpu-no-cluster-id-z" "amdgpu-no-completion-action" "amdgpu-no-default-queue" "amdgpu-no-dispatch-id" "amdgpu-no-dispatch-ptr" "amdgpu-no-flat-scratch-init" "amdgpu-no-heap-ptr" "amdgpu-no-hostcall-ptr" "amdgpu-no-implicitarg-ptr" "amdgpu-no-lds-kernel-id" "amdgpu-no-multigrid-sync-arg" "amdgpu-no-queue-ptr" "amdgpu-no-workgroup-id-x" "amdgpu-no-workgroup-id-y" "amdgpu-no-workgroup-id-z" "amdgpu-no-workitem-id-x" "amdgpu-no-workitem-id-y" "amdgpu-no-workitem-id-z" "amdgpu-no-wwm" "amdgpu-waves-per-eu"="2,8" }
; CHECK: attributes #[[ATTR6]] = { "amdgpu-flat-work-group-size"="1,64" "amdgpu-no-cluster-id-x" "amdgpu-no-cluster-id-y" "amdgpu-no-cluster-id-z" "amdgpu-no-completion-action" "amdgpu-no-default-queue" "amdgpu-no-dispatch-id" "amdgpu-no-dispatch-ptr" "amdgpu-no-flat-scratch-init" "amdgpu-no-heap-ptr" "amdgpu-no-hostcall-ptr" "amdgpu-no-implicitarg-ptr" "amdgpu-no-lds-kernel-id" "amdgpu-no-multigrid-sync-arg" "amdgpu-no-queue-ptr" "amdgpu-no-workgroup-id-x" "amdgpu-no-workgroup-id-y" "amdgpu-no-workgroup-id-z" "amdgpu-no-workitem-id-x" "amdgpu-no-workitem-id-y" "amdgpu-no-workitem-id-z" "amdgpu-no-wwm" "amdgpu-waves-per-eu"="9,10" }
-; CHECK: attributes #[[ATTR7]] = { "amdgpu-flat-work-group-size"="1,64" "amdgpu-no-cluster-id-x" "amdgpu-no-cluster-id-y" "amdgpu-no-cluster-id-z" "amdgpu-no-completion-action" "amdgpu-no-default-queue" "amdgpu-no-dispatch-id" "amdgpu-no-dispatch-ptr" "amdgpu-no-flat-scratch-init" "amdgpu-no-heap-ptr" "amdgpu-no-hostcall-ptr" "amdgpu-no-implicitarg-ptr" "amdgpu-no-lds-kernel-id" "amdgpu-no-multigrid-sync-arg" "amdgpu-no-queue-ptr" "amdgpu-no-workgroup-id-x" "amdgpu-no-workgroup-id-y" "amdgpu-no-workgroup-id-z" "amdgpu-no-workitem-id-x" "amdgpu-no-workitem-id-y" "amdgpu-no-workitem-id-z" "amdgpu-no-wwm" "amdgpu-waves-per-eu"="2,9" }
+; CHECK: attributes #[[ATTR7]] = { "amdgpu-flat-work-group-size"="1,64" "amdgpu-no-cluster-id-x" "amdgpu-no-cluster-id-y" "amdgpu-no-cluster-id-z" "amdgpu-no-completion-action" "amdgpu-no-default-queue" "amdgpu-no-dispatch-id" "amdgpu-no-dispatch-ptr" "amdgpu-no-flat-scratch-init" "amdgpu-no-heap-ptr" "amdgpu-no-hostcall-ptr" "amdgpu-no-implicitarg-ptr" "amdgpu-no-lds-kernel-id" "amdgpu-no-multigrid-sync-arg" "amdgpu-no-queue-ptr" "amdgpu-no-workgroup-id-x" "amdgpu-no-workgroup-id-y" "amdgpu-no-workgroup-id-z" "amdgpu-no-workitem-id-x" "amdgpu-no-workitem-id-y" "amdgpu-no-workitem-id-z" "amdgpu-no-wwm" "amdgpu-waves-per-eu"="9,9" }
; CHECK: attributes #[[ATTR8]] = { "amdgpu-flat-work-group-size"="1,64" "amdgpu-no-cluster-id-x" "amdgpu-no-cluster-id-y" "amdgpu-no-cluster-id-z" "amdgpu-no-completion-action" "amdgpu-no-default-queue" "amdgpu-no-dispatch-id" "amdgpu-no-dispatch-ptr" "amdgpu-no-flat-scratch-init" "amdgpu-no-heap-ptr" "amdgpu-no-hostcall-ptr" "amdgpu-no-implicitarg-ptr" "amdgpu-no-lds-kernel-id" "amdgpu-no-multigrid-sync-arg" "amdgpu-no-queue-ptr" "amdgpu-no-workgroup-id-x" "amdgpu-no-workgroup-id-y" "amdgpu-no-workgroup-id-z" "amdgpu-no-workitem-id-x" "amdgpu-no-workitem-id-y" "amdgpu-no-workitem-id-z" "amdgpu-no-wwm" "amdgpu-waves-per-eu"="3,8" }
; CHECK: attributes #[[ATTR9]] = { "amdgpu-no-cluster-id-x" "amdgpu-no-cluster-id-y" "amdgpu-no-cluster-id-z" "amdgpu-no-completion-action" "amdgpu-no-default-queue" "amdgpu-no-dispatch-id" "amdgpu-no-dispatch-ptr" "amdgpu-no-flat-scratch-init" "amdgpu-no-heap-ptr" "amdgpu-no-hostcall-ptr" "amdgpu-no-implicitarg-ptr" "amdgpu-no-lds-kernel-id" "amdgpu-no-multigrid-sync-arg" "amdgpu-no-queue-ptr" "amdgpu-no-workgroup-id-x" "amdgpu-no-workgroup-id-y" "amdgpu-no-workgroup-id-z" "amdgpu-no-workitem-id-x" "amdgpu-no-workitem-id-y" "amdgpu-no-workitem-id-z" "amdgpu-no-wwm" }
; CHECK: attributes #[[ATTR10]] = { "amdgpu-flat-work-group-size"="1,64" "amdgpu-no-cluster-id-x" "amdgpu-no-cluster-id-y" "amdgpu-no-cluster-id-z" "amdgpu-no-completion-action" "amdgpu-no-default-queue" "amdgpu-no-dispatch-id" "amdgpu-no-dispatch-ptr" "amdgpu-no-flat-scratch-init" "amdgpu-no-heap-ptr" "amdgpu-no-hostcall-ptr" "amdgpu-no-implicitarg-ptr" "amdgpu-no-lds-kernel-id" "amdgpu-no-multigrid-sync-arg" "amdgpu-no-queue-ptr" "amdgpu-no-workgroup-id-x" "amdgpu-no-workgroup-id-y" "amdgpu-no-workgroup-id-z" "amdgpu-no-workitem-id-x" "amdgpu-no-workitem-id-y" "amdgpu-no-workitem-id-z" "amdgpu-no-wwm" }
>From ffac73c600039aded702e7472a61910ee3fe4f4b Mon Sep 17 00:00:00 2001
From: Arseniy Obolenskiy <arseniy.obolenskiy at amd.com>
Date: Tue, 7 Jul 2026 14:00:28 +0200
Subject: [PATCH 2/3] Clean up amdgpu-no-*
---
.../CodeGen/AMDGPU/propagate-waves-per-eu.ll | 148 +++++++++++++-----
1 file changed, 109 insertions(+), 39 deletions(-)
diff --git a/llvm/test/CodeGen/AMDGPU/propagate-waves-per-eu.ll b/llvm/test/CodeGen/AMDGPU/propagate-waves-per-eu.ll
index 9d19a66433d72..67caeb760ce44 100644
--- a/llvm/test/CodeGen/AMDGPU/propagate-waves-per-eu.ll
+++ b/llvm/test/CodeGen/AMDGPU/propagate-waves-per-eu.ll
@@ -3,12 +3,18 @@
; Check propagation of amdgpu-flat-work-group-size attribute.
+; External call to avoid inferring argument attributes. This makes the
+; final attribute groups easier to read.
+declare void @dummy()
+
; Called from a single kernel with 1,8
define internal void @default_to_1_8_a() {
; CHECK-LABEL: define internal void @default_to_1_8_a
; CHECK-SAME: () #[[ATTR0:[0-9]+]] {
+; CHECK-NEXT: call void @dummy()
; CHECK-NEXT: ret void
;
+ call void @dummy()
ret void
}
@@ -16,9 +22,11 @@ define amdgpu_kernel void @kernel_1_8() #0 {
; CHECK-LABEL: define amdgpu_kernel void @kernel_1_8
; CHECK-SAME: () #[[ATTR0]] {
; CHECK-NEXT: call void @default_to_1_8_a()
+; CHECK-NEXT: call void @dummy()
; CHECK-NEXT: ret void
;
call void @default_to_1_8_a()
+ call void @dummy()
ret void
}
@@ -26,8 +34,10 @@ define amdgpu_kernel void @kernel_1_8() #0 {
define internal void @default_to_1_2() {
; CHECK-LABEL: define internal void @default_to_1_2
; CHECK-SAME: () #[[ATTR1:[0-9]+]] {
+; CHECK-NEXT: call void @dummy()
; CHECK-NEXT: ret void
;
+ call void @dummy()
ret void
}
@@ -38,12 +48,14 @@ define amdgpu_kernel void @kernel_1_2() #1 {
; CHECK-NEXT: call void @flat_group_1_1()
; CHECK-NEXT: call void @default_to_1_8_b()
; CHECK-NEXT: call void @flat_group_2_8()
+; CHECK-NEXT: call void @dummy()
; CHECK-NEXT: ret void
;
call void @default_to_1_2()
call void @flat_group_1_1()
call void @default_to_1_8_b()
call void @flat_group_2_8()
+ call void @dummy()
ret void
}
@@ -51,8 +63,10 @@ define amdgpu_kernel void @kernel_1_2() #1 {
define internal void @default_to_1_4() {
; CHECK-LABEL: define internal void @default_to_1_4
; CHECK-SAME: () #[[ATTR2:[0-9]+]] {
+; CHECK-NEXT: call void @dummy()
; CHECK-NEXT: ret void
;
+ call void @dummy()
ret void
}
@@ -60,9 +74,11 @@ define amdgpu_kernel void @kernel_1_4() #2 {
; CHECK-LABEL: define amdgpu_kernel void @kernel_1_4
; CHECK-SAME: () #[[ATTR2]] {
; CHECK-NEXT: call void @default_to_1_4()
+; CHECK-NEXT: call void @dummy()
; CHECK-NEXT: ret void
;
call void @default_to_1_4()
+ call void @dummy()
ret void
}
@@ -70,8 +86,10 @@ define amdgpu_kernel void @kernel_1_4() #2 {
define internal void @default_to_2_9() {
; CHECK-LABEL: define internal void @default_to_2_9
; CHECK-SAME: () #[[ATTR3:[0-9]+]] {
+; CHECK-NEXT: call void @dummy()
; CHECK-NEXT: ret void
;
+ call void @dummy()
ret void
}
@@ -80,8 +98,10 @@ define internal void @default_to_2_9() {
define internal void @flat_group_1_1() #3 {
; CHECK-LABEL: define internal void @flat_group_1_1
; CHECK-SAME: () #[[ATTR4:[0-9]+]] {
+; CHECK-NEXT: call void @dummy()
; CHECK-NEXT: ret void
;
+ call void @dummy()
ret void
}
@@ -89,8 +109,10 @@ define internal void @flat_group_1_1() #3 {
define internal void @flat_group_2_8() #4 {
; CHECK-LABEL: define internal void @flat_group_2_8
; CHECK-SAME: () #[[ATTR5:[0-9]+]] {
+; CHECK-NEXT: call void @dummy()
; CHECK-NEXT: ret void
;
+ call void @dummy()
ret void
}
@@ -98,8 +120,10 @@ define internal void @flat_group_2_8() #4 {
define internal void @flat_group_9_10() #5 {
; CHECK-LABEL: define internal void @flat_group_9_10
; CHECK-SAME: () #[[ATTR6:[0-9]+]] {
+; CHECK-NEXT: call void @dummy()
; CHECK-NEXT: ret void
;
+ call void @dummy()
ret void
}
@@ -108,10 +132,12 @@ define amdgpu_kernel void @kernel_2_9() #6 {
; CHECK-SAME: () #[[ATTR3]] {
; CHECK-NEXT: call void @default_to_2_9()
; CHECK-NEXT: call void @flat_group_1_1()
+; CHECK-NEXT: call void @dummy()
; CHECK-NEXT: ret void
;
call void @default_to_2_9()
call void @flat_group_1_1()
+ call void @dummy()
ret void
}
@@ -120,10 +146,12 @@ define amdgpu_kernel void @kernel_9_9() #7 {
; CHECK-SAME: () #[[ATTR7:[0-9]+]] {
; CHECK-NEXT: call void @default_to_2_9()
; CHECK-NEXT: call void @flat_group_9_10()
+; CHECK-NEXT: call void @dummy()
; CHECK-NEXT: ret void
;
call void @default_to_2_9()
call void @flat_group_9_10()
+ call void @dummy()
ret void
}
@@ -131,8 +159,10 @@ define amdgpu_kernel void @kernel_9_9() #7 {
define internal void @default_to_1_8_b() {
; CHECK-LABEL: define internal void @default_to_1_8_b
; CHECK-SAME: () #[[ATTR0]] {
+; CHECK-NEXT: call void @dummy()
; CHECK-NEXT: ret void
;
+ call void @dummy()
ret void
}
@@ -143,10 +173,12 @@ define amdgpu_kernel void @kernel_2_8() #4 {
; CHECK-SAME: () #[[ATTR5]] {
; CHECK-NEXT: call void @default_to_1_8_a()
; CHECK-NEXT: call void @default_to_1_8_b()
+; CHECK-NEXT: call void @dummy()
; CHECK-NEXT: ret void
;
call void @default_to_1_8_a()
call void @default_to_1_8_b()
+ call void @dummy()
ret void
}
@@ -155,9 +187,11 @@ define internal void @merge_cycle_0() #1 {
; CHECK-LABEL: define internal void @merge_cycle_0
; CHECK-SAME: () #[[ATTR1]] {
; CHECK-NEXT: call void @merge_cycle_1()
+; CHECK-NEXT: call void @dummy()
; CHECK-NEXT: ret void
;
call void @merge_cycle_1()
+ call void @dummy()
ret void
}
@@ -167,9 +201,11 @@ define internal void @merge_cycle_1() #4 {
; CHECK-LABEL: define internal void @merge_cycle_1
; CHECK-SAME: () #[[ATTR5]] {
; CHECK-NEXT: call void @merge_cycle_0()
+; CHECK-NEXT: call void @dummy()
; CHECK-NEXT: ret void
;
call void @merge_cycle_0()
+ call void @dummy()
ret void
}
@@ -180,30 +216,34 @@ define amdgpu_kernel void @kernel_3_8() #8 {
; CHECK-NEXT: call void @default_captured_address()
; CHECK-NEXT: call void @externally_visible_default()
; CHECK-NEXT: [[F32:%.*]] = call float @bitcasted_function()
+; CHECK-NEXT: call void @dummy()
; CHECK-NEXT: ret void
;
call void @merge_cycle_0()
call void @default_captured_address()
call void @externally_visible_default()
%f32 = call float @bitcasted_function()
+ call void @dummy()
ret void
}
define internal void @default_captured_address() {
-; CHECK-LABEL: define internal void @default_captured_address
-; CHECK-SAME: () #[[ATTR9:[0-9]+]] {
+; CHECK-LABEL: define internal void @default_captured_address() {
; CHECK-NEXT: store volatile ptr @default_captured_address, ptr poison, align 8
+; CHECK-NEXT: call void @dummy()
; CHECK-NEXT: ret void
;
store volatile ptr @default_captured_address, ptr poison, align 8
+ call void @dummy()
ret void
}
define void @externally_visible_default() {
-; CHECK-LABEL: define void @externally_visible_default
-; CHECK-SAME: () #[[ATTR9]] {
+; CHECK-LABEL: define void @externally_visible_default() {
+; CHECK-NEXT: call void @dummy()
; CHECK-NEXT: ret void
;
+ call void @dummy()
ret void
}
@@ -211,24 +251,30 @@ define void @externally_visible_default() {
define internal i32 @bitcasted_function() {
; CHECK-LABEL: define internal i32 @bitcasted_function
; CHECK-SAME: () #[[ATTR8]] {
+; CHECK-NEXT: call void @dummy()
; CHECK-NEXT: ret i32 0
;
+ call void @dummy()
ret i32 0
}
define internal void @called_from_invalid_bounds_0() {
; CHECK-LABEL: define internal void @called_from_invalid_bounds_0
; CHECK-SAME: () #[[ATTR0]] {
+; CHECK-NEXT: call void @dummy()
; CHECK-NEXT: ret void
;
+ call void @dummy()
ret void
}
define internal void @called_from_invalid_bounds_1() {
; CHECK-LABEL: define internal void @called_from_invalid_bounds_1
-; CHECK-SAME: () #[[ATTR10:[0-9]+]] {
+; CHECK-SAME: () #[[ATTR9:[0-9]+]] {
+; CHECK-NEXT: call void @dummy()
; CHECK-NEXT: ret void
;
+ call void @dummy()
ret void
}
@@ -237,20 +283,24 @@ define amdgpu_kernel void @kernel_invalid_bounds_0_8() #9 {
; CHECK-LABEL: define amdgpu_kernel void @kernel_invalid_bounds_0_8
; CHECK-SAME: () #[[ATTR0]] {
; CHECK-NEXT: call void @called_from_invalid_bounds_0()
+; CHECK-NEXT: call void @dummy()
; CHECK-NEXT: ret void
;
call void @called_from_invalid_bounds_0()
+ call void @dummy()
ret void
}
; Invalid range for amdgpu-waves-per-eu
define amdgpu_kernel void @kernel_invalid_bounds_1_123() #10 {
; CHECK-LABEL: define amdgpu_kernel void @kernel_invalid_bounds_1_123
-; CHECK-SAME: () #[[ATTR11:[0-9]+]] {
+; CHECK-SAME: () #[[ATTR10:[0-9]+]] {
; CHECK-NEXT: call void @called_from_invalid_bounds_1()
+; CHECK-NEXT: call void @dummy()
; CHECK-NEXT: ret void
;
call void @called_from_invalid_bounds_1()
+ call void @dummy()
ret void
}
@@ -260,100 +310,119 @@ define amdgpu_kernel void @kernel_invalid_bounds_1_123() #10 {
; -> 2,10
define void @larger_group_size_implies_lower_minimum() #11 {
; CHECK-LABEL: define void @larger_group_size_implies_lower_minimum
-; CHECK-SAME: () #[[ATTR12:[0-9]+]] {
+; CHECK-SAME: () #[[ATTR11:[0-9]+]] {
+; CHECK-NEXT: call void @dummy()
; CHECK-NEXT: ret void
;
+ call void @dummy()
ret void
}
define amdgpu_kernel void @kernel_3_6() #12 {
; CHECK-LABEL: define amdgpu_kernel void @kernel_3_6
-; CHECK-SAME: () #[[ATTR13:[0-9]+]] {
+; CHECK-SAME: () #[[ATTR12:[0-9]+]] {
; CHECK-NEXT: call void @larger_group_size_implies_lower_minimum()
+; CHECK-NEXT: call void @dummy()
; CHECK-NEXT: ret void
;
call void @larger_group_size_implies_lower_minimum()
+ call void @dummy()
ret void
}
; 3,6 -> 6,9
define internal void @refine_upper_func_3_6() #13 {
; CHECK-LABEL: define internal void @refine_upper_func_3_6
-; CHECK-SAME: () #[[ATTR14:[0-9]+]] {
+; CHECK-SAME: () #[[ATTR13:[0-9]+]] {
+; CHECK-NEXT: call void @dummy()
; CHECK-NEXT: ret void
;
+ call void @dummy()
ret void
}
; 4,8 -> 6,8
define internal void @refine_lower_func_4_8() #14 {
; CHECK-LABEL: define internal void @refine_lower_func_4_8
-; CHECK-SAME: () #[[ATTR15:[0-9]+]] {
+; CHECK-SAME: () #[[ATTR14:[0-9]+]] {
; CHECK-NEXT: call void @refine_upper_func_3_6()
+; CHECK-NEXT: call void @dummy()
; CHECK-NEXT: ret void
;
call void @refine_upper_func_3_6()
+ call void @dummy()
ret void
}
define amdgpu_kernel void @kernel_foo_6_8() #15 {
; CHECK-LABEL: define amdgpu_kernel void @kernel_foo_6_8
-; CHECK-SAME: () #[[ATTR16:[0-9]+]] {
+; CHECK-SAME: () #[[ATTR15:[0-9]+]] {
; CHECK-NEXT: call void @refine_upper_func_3_6()
; CHECK-NEXT: call void @refine_lower_func_4_8()
; CHECK-NEXT: call void @func_9_10_a()
+; CHECK-NEXT: call void @dummy()
; CHECK-NEXT: ret void
;
call void @refine_upper_func_3_6()
call void @refine_lower_func_4_8()
call void @func_9_10_a()
+ call void @dummy()
ret void
}
; 5,5 -> 5,5
define internal void @func_5_5() #16 {
; CHECK-LABEL: define internal void @func_5_5
-; CHECK-SAME: () #[[ATTR17:[0-9]+]] {
+; CHECK-SAME: () #[[ATTR16:[0-9]+]] {
+; CHECK-NEXT: call void @dummy()
; CHECK-NEXT: ret void
;
+ call void @dummy()
ret void
}
; 5,8 -> 8,8
define internal void @func_5_8() #17 {
; CHECK-LABEL: define internal void @func_5_8
-; CHECK-SAME: () #[[ATTR18:[0-9]+]] {
+; CHECK-SAME: () #[[ATTR17:[0-9]+]] {
+; CHECK-NEXT: call void @dummy()
; CHECK-NEXT: ret void
;
+ call void @dummy()
ret void
}
; 9,10 -> 9,10
define internal void @func_9_10_a() #18 {
; CHECK-LABEL: define internal void @func_9_10_a
-; CHECK-SAME: () #[[ATTR19:[0-9]+]] {
+; CHECK-SAME: () #[[ATTR18:[0-9]+]] {
+; CHECK-NEXT: call void @dummy()
; CHECK-NEXT: ret void
;
+ call void @dummy()
ret void
}
; 9,10 -> 9,9
define internal void @func_9_10_b() #18 {
; CHECK-LABEL: define internal void @func_9_10_b
-; CHECK-SAME: () #[[ATTR19]] {
+; CHECK-SAME: () #[[ATTR18]] {
+; CHECK-NEXT: call void @dummy()
; CHECK-NEXT: ret void
;
+ call void @dummy()
ret void
}
define amdgpu_kernel void @kernel_bar_8_9() #19 {
; CHECK-LABEL: define amdgpu_kernel void @kernel_bar_8_9
-; CHECK-SAME: () #[[ATTR20:[0-9]+]] {
+; CHECK-SAME: () #[[ATTR19:[0-9]+]] {
; CHECK-NEXT: call void @refine_upper_func_3_6()
; CHECK-NEXT: call void @func_5_5()
; CHECK-NEXT: call void @func_9_10_b()
; CHECK-NEXT: call void @func_5_8()
; CHECK-NEXT: call void @externally_visible()
+; CHECK-NEXT: call void @dummy()
; CHECK-NEXT: ret void
;
call void @refine_upper_func_3_6()
@@ -361,16 +430,18 @@ define amdgpu_kernel void @kernel_bar_8_9() #19 {
call void @func_9_10_b()
call void @func_5_8()
call void @externally_visible()
+ call void @dummy()
ret void
}
; This is an optimization hint based on users, so it's not strictly
; required that all callers be visible.
define void @externally_visible() {
-; CHECK-LABEL: define void @externally_visible
-; CHECK-SAME: () #[[ATTR9]] {
+; CHECK-LABEL: define void @externally_visible() {
+; CHECK-NEXT: call void @dummy()
; CHECK-NEXT: ret void
;
+ call void @dummy()
ret void
}
@@ -399,25 +470,24 @@ attributes #17 = { "amdgpu-waves-per-eu"="5,8" }
attributes #18 = { "amdgpu-waves-per-eu"="9,10" }
attributes #19 = { "amdgpu-waves-per-eu"="8,9" }
;.
-; CHECK: attributes #[[ATTR0]] = { "amdgpu-flat-work-group-size"="1,64" "amdgpu-no-cluster-id-x" "amdgpu-no-cluster-id-y" "amdgpu-no-cluster-id-z" "amdgpu-no-completion-action" "amdgpu-no-default-queue" "amdgpu-no-dispatch-id" "amdgpu-no-dispatch-ptr" "amdgpu-no-flat-scratch-init" "amdgpu-no-heap-ptr" "amdgpu-no-hostcall-ptr" "amdgpu-no-implicitarg-ptr" "amdgpu-no-lds-kernel-id" "amdgpu-no-multigrid-sync-arg" "amdgpu-no-queue-ptr" "amdgpu-no-workgroup-id-x" "amdgpu-no-workgroup-id-y" "amdgpu-no-workgroup-id-z" "amdgpu-no-workitem-id-x" "amdgpu-no-workitem-id-y" "amdgpu-no-workitem-id-z" "amdgpu-no-wwm" "amdgpu-waves-per-eu"="1,8" }
-; CHECK: attributes #[[ATTR1]] = { "amdgpu-flat-work-group-size"="1,64" "amdgpu-no-cluster-id-x" "amdgpu-no-cluster-id-y" "amdgpu-no-cluster-id-z" "amdgpu-no-completion-action" "amdgpu-no-default-queue" "amdgpu-no-dispatch-id" "amdgpu-no-dispatch-ptr" "amdgpu-no-flat-scratch-init" "amdgpu-no-heap-ptr" "amdgpu-no-hostcall-ptr" "amdgpu-no-implicitarg-ptr" "amdgpu-no-lds-kernel-id" "amdgpu-no-multigrid-sync-arg" "amdgpu-no-queue-ptr" "amdgpu-no-workgroup-id-x" "amdgpu-no-workgroup-id-y" "amdgpu-no-workgroup-id-z" "amdgpu-no-workitem-id-x" "amdgpu-no-workitem-id-y" "amdgpu-no-workitem-id-z" "amdgpu-no-wwm" "amdgpu-waves-per-eu"="1,2" }
-; CHECK: attributes #[[ATTR2]] = { "amdgpu-flat-work-group-size"="1,64" "amdgpu-no-cluster-id-x" "amdgpu-no-cluster-id-y" "amdgpu-no-cluster-id-z" "amdgpu-no-completion-action" "amdgpu-no-default-queue" "amdgpu-no-dispatch-id" "amdgpu-no-dispatch-ptr" "amdgpu-no-flat-scratch-init" "amdgpu-no-heap-ptr" "amdgpu-no-hostcall-ptr" "amdgpu-no-implicitarg-ptr" "amdgpu-no-lds-kernel-id" "amdgpu-no-multigrid-sync-arg" "amdgpu-no-queue-ptr" "amdgpu-no-workgroup-id-x" "amdgpu-no-workgroup-id-y" "amdgpu-no-workgroup-id-z" "amdgpu-no-workitem-id-x" "amdgpu-no-workitem-id-y" "amdgpu-no-workitem-id-z" "amdgpu-no-wwm" "amdgpu-waves-per-eu"="1,4" }
-; CHECK: attributes #[[ATTR3]] = { "amdgpu-flat-work-group-size"="1,64" "amdgpu-no-cluster-id-x" "amdgpu-no-cluster-id-y" "amdgpu-no-cluster-id-z" "amdgpu-no-completion-action" "amdgpu-no-default-queue" "amdgpu-no-dispatch-id" "amdgpu-no-dispatch-ptr" "amdgpu-no-flat-scratch-init" "amdgpu-no-heap-ptr" "amdgpu-no-hostcall-ptr" "amdgpu-no-implicitarg-ptr" "amdgpu-no-lds-kernel-id" "amdgpu-no-multigrid-sync-arg" "amdgpu-no-queue-ptr" "amdgpu-no-workgroup-id-x" "amdgpu-no-workgroup-id-y" "amdgpu-no-workgroup-id-z" "amdgpu-no-workitem-id-x" "amdgpu-no-workitem-id-y" "amdgpu-no-workitem-id-z" "amdgpu-no-wwm" "amdgpu-waves-per-eu"="2,9" }
-; CHECK: attributes #[[ATTR4]] = { "amdgpu-flat-work-group-size"="1,64" "amdgpu-no-cluster-id-x" "amdgpu-no-cluster-id-y" "amdgpu-no-cluster-id-z" "amdgpu-no-completion-action" "amdgpu-no-default-queue" "amdgpu-no-dispatch-id" "amdgpu-no-dispatch-ptr" "amdgpu-no-flat-scratch-init" "amdgpu-no-heap-ptr" "amdgpu-no-hostcall-ptr" "amdgpu-no-implicitarg-ptr" "amdgpu-no-lds-kernel-id" "amdgpu-no-multigrid-sync-arg" "amdgpu-no-queue-ptr" "amdgpu-no-workgroup-id-x" "amdgpu-no-workgroup-id-y" "amdgpu-no-workgroup-id-z" "amdgpu-no-workitem-id-x" "amdgpu-no-workitem-id-y" "amdgpu-no-workitem-id-z" "amdgpu-no-wwm" "amdgpu-waves-per-eu"="1,1" }
-; CHECK: attributes #[[ATTR5]] = { "amdgpu-flat-work-group-size"="1,64" "amdgpu-no-cluster-id-x" "amdgpu-no-cluster-id-y" "amdgpu-no-cluster-id-z" "amdgpu-no-completion-action" "amdgpu-no-default-queue" "amdgpu-no-dispatch-id" "amdgpu-no-dispatch-ptr" "amdgpu-no-flat-scratch-init" "amdgpu-no-heap-ptr" "amdgpu-no-hostcall-ptr" "amdgpu-no-implicitarg-ptr" "amdgpu-no-lds-kernel-id" "amdgpu-no-multigrid-sync-arg" "amdgpu-no-queue-ptr" "amdgpu-no-workgroup-id-x" "amdgpu-no-workgroup-id-y" "amdgpu-no-workgroup-id-z" "amdgpu-no-workitem-id-x" "amdgpu-no-workitem-id-y" "amdgpu-no-workitem-id-z" "amdgpu-no-wwm" "amdgpu-waves-per-eu"="2,8" }
-; CHECK: attributes #[[ATTR6]] = { "amdgpu-flat-work-group-size"="1,64" "amdgpu-no-cluster-id-x" "amdgpu-no-cluster-id-y" "amdgpu-no-cluster-id-z" "amdgpu-no-completion-action" "amdgpu-no-default-queue" "amdgpu-no-dispatch-id" "amdgpu-no-dispatch-ptr" "amdgpu-no-flat-scratch-init" "amdgpu-no-heap-ptr" "amdgpu-no-hostcall-ptr" "amdgpu-no-implicitarg-ptr" "amdgpu-no-lds-kernel-id" "amdgpu-no-multigrid-sync-arg" "amdgpu-no-queue-ptr" "amdgpu-no-workgroup-id-x" "amdgpu-no-workgroup-id-y" "amdgpu-no-workgroup-id-z" "amdgpu-no-workitem-id-x" "amdgpu-no-workitem-id-y" "amdgpu-no-workitem-id-z" "amdgpu-no-wwm" "amdgpu-waves-per-eu"="9,10" }
-; CHECK: attributes #[[ATTR7]] = { "amdgpu-flat-work-group-size"="1,64" "amdgpu-no-cluster-id-x" "amdgpu-no-cluster-id-y" "amdgpu-no-cluster-id-z" "amdgpu-no-completion-action" "amdgpu-no-default-queue" "amdgpu-no-dispatch-id" "amdgpu-no-dispatch-ptr" "amdgpu-no-flat-scratch-init" "amdgpu-no-heap-ptr" "amdgpu-no-hostcall-ptr" "amdgpu-no-implicitarg-ptr" "amdgpu-no-lds-kernel-id" "amdgpu-no-multigrid-sync-arg" "amdgpu-no-queue-ptr" "amdgpu-no-workgroup-id-x" "amdgpu-no-workgroup-id-y" "amdgpu-no-workgroup-id-z" "amdgpu-no-workitem-id-x" "amdgpu-no-workitem-id-y" "amdgpu-no-workitem-id-z" "amdgpu-no-wwm" "amdgpu-waves-per-eu"="9,9" }
-; CHECK: attributes #[[ATTR8]] = { "amdgpu-flat-work-group-size"="1,64" "amdgpu-no-cluster-id-x" "amdgpu-no-cluster-id-y" "amdgpu-no-cluster-id-z" "amdgpu-no-completion-action" "amdgpu-no-default-queue" "amdgpu-no-dispatch-id" "amdgpu-no-dispatch-ptr" "amdgpu-no-flat-scratch-init" "amdgpu-no-heap-ptr" "amdgpu-no-hostcall-ptr" "amdgpu-no-implicitarg-ptr" "amdgpu-no-lds-kernel-id" "amdgpu-no-multigrid-sync-arg" "amdgpu-no-queue-ptr" "amdgpu-no-workgroup-id-x" "amdgpu-no-workgroup-id-y" "amdgpu-no-workgroup-id-z" "amdgpu-no-workitem-id-x" "amdgpu-no-workitem-id-y" "amdgpu-no-workitem-id-z" "amdgpu-no-wwm" "amdgpu-waves-per-eu"="3,8" }
-; CHECK: attributes #[[ATTR9]] = { "amdgpu-no-cluster-id-x" "amdgpu-no-cluster-id-y" "amdgpu-no-cluster-id-z" "amdgpu-no-completion-action" "amdgpu-no-default-queue" "amdgpu-no-dispatch-id" "amdgpu-no-dispatch-ptr" "amdgpu-no-flat-scratch-init" "amdgpu-no-heap-ptr" "amdgpu-no-hostcall-ptr" "amdgpu-no-implicitarg-ptr" "amdgpu-no-lds-kernel-id" "amdgpu-no-multigrid-sync-arg" "amdgpu-no-queue-ptr" "amdgpu-no-workgroup-id-x" "amdgpu-no-workgroup-id-y" "amdgpu-no-workgroup-id-z" "amdgpu-no-workitem-id-x" "amdgpu-no-workitem-id-y" "amdgpu-no-workitem-id-z" "amdgpu-no-wwm" }
-; CHECK: attributes #[[ATTR10]] = { "amdgpu-flat-work-group-size"="1,64" "amdgpu-no-cluster-id-x" "amdgpu-no-cluster-id-y" "amdgpu-no-cluster-id-z" "amdgpu-no-completion-action" "amdgpu-no-default-queue" "amdgpu-no-dispatch-id" "amdgpu-no-dispatch-ptr" "amdgpu-no-flat-scratch-init" "amdgpu-no-heap-ptr" "amdgpu-no-hostcall-ptr" "amdgpu-no-implicitarg-ptr" "amdgpu-no-lds-kernel-id" "amdgpu-no-multigrid-sync-arg" "amdgpu-no-queue-ptr" "amdgpu-no-workgroup-id-x" "amdgpu-no-workgroup-id-y" "amdgpu-no-workgroup-id-z" "amdgpu-no-workitem-id-x" "amdgpu-no-workitem-id-y" "amdgpu-no-workitem-id-z" "amdgpu-no-wwm" }
-; CHECK: attributes #[[ATTR11]] = { "amdgpu-flat-work-group-size"="1,64" "amdgpu-no-cluster-id-x" "amdgpu-no-cluster-id-y" "amdgpu-no-cluster-id-z" "amdgpu-no-completion-action" "amdgpu-no-default-queue" "amdgpu-no-dispatch-id" "amdgpu-no-dispatch-ptr" "amdgpu-no-flat-scratch-init" "amdgpu-no-heap-ptr" "amdgpu-no-hostcall-ptr" "amdgpu-no-implicitarg-ptr" "amdgpu-no-lds-kernel-id" "amdgpu-no-multigrid-sync-arg" "amdgpu-no-queue-ptr" "amdgpu-no-workgroup-id-x" "amdgpu-no-workgroup-id-y" "amdgpu-no-workgroup-id-z" "amdgpu-no-workitem-id-x" "amdgpu-no-workitem-id-y" "amdgpu-no-workitem-id-z" "amdgpu-no-wwm" "amdgpu-waves-per-eu"="1,123" }
-; CHECK: attributes #[[ATTR12]] = { "amdgpu-flat-work-group-size"="1,512" "amdgpu-no-cluster-id-x" "amdgpu-no-cluster-id-y" "amdgpu-no-cluster-id-z" "amdgpu-no-completion-action" "amdgpu-no-default-queue" "amdgpu-no-dispatch-id" "amdgpu-no-dispatch-ptr" "amdgpu-no-flat-scratch-init" "amdgpu-no-heap-ptr" "amdgpu-no-hostcall-ptr" "amdgpu-no-implicitarg-ptr" "amdgpu-no-lds-kernel-id" "amdgpu-no-multigrid-sync-arg" "amdgpu-no-queue-ptr" "amdgpu-no-workgroup-id-x" "amdgpu-no-workgroup-id-y" "amdgpu-no-workgroup-id-z" "amdgpu-no-workitem-id-x" "amdgpu-no-workitem-id-y" "amdgpu-no-workitem-id-z" "amdgpu-no-wwm" }
-; CHECK: attributes #[[ATTR13]] = { "amdgpu-flat-work-group-size"="1,512" "amdgpu-no-cluster-id-x" "amdgpu-no-cluster-id-y" "amdgpu-no-cluster-id-z" "amdgpu-no-completion-action" "amdgpu-no-default-queue" "amdgpu-no-dispatch-id" "amdgpu-no-dispatch-ptr" "amdgpu-no-flat-scratch-init" "amdgpu-no-heap-ptr" "amdgpu-no-hostcall-ptr" "amdgpu-no-implicitarg-ptr" "amdgpu-no-lds-kernel-id" "amdgpu-no-multigrid-sync-arg" "amdgpu-no-queue-ptr" "amdgpu-no-workgroup-id-x" "amdgpu-no-workgroup-id-y" "amdgpu-no-workgroup-id-z" "amdgpu-no-workitem-id-x" "amdgpu-no-workitem-id-y" "amdgpu-no-workitem-id-z" "amdgpu-no-wwm" "amdgpu-waves-per-eu"="3,6" }
-; CHECK: attributes #[[ATTR14]] = { "amdgpu-no-cluster-id-x" "amdgpu-no-cluster-id-y" "amdgpu-no-cluster-id-z" "amdgpu-no-completion-action" "amdgpu-no-default-queue" "amdgpu-no-dispatch-id" "amdgpu-no-dispatch-ptr" "amdgpu-no-flat-scratch-init" "amdgpu-no-heap-ptr" "amdgpu-no-hostcall-ptr" "amdgpu-no-implicitarg-ptr" "amdgpu-no-lds-kernel-id" "amdgpu-no-multigrid-sync-arg" "amdgpu-no-queue-ptr" "amdgpu-no-workgroup-id-x" "amdgpu-no-workgroup-id-y" "amdgpu-no-workgroup-id-z" "amdgpu-no-workitem-id-x" "amdgpu-no-workitem-id-y" "amdgpu-no-workitem-id-z" "amdgpu-no-wwm" "amdgpu-waves-per-eu"="3,6" }
-; CHECK: attributes #[[ATTR15]] = { "amdgpu-no-cluster-id-x" "amdgpu-no-cluster-id-y" "amdgpu-no-cluster-id-z" "amdgpu-no-completion-action" "amdgpu-no-default-queue" "amdgpu-no-dispatch-id" "amdgpu-no-dispatch-ptr" "amdgpu-no-flat-scratch-init" "amdgpu-no-heap-ptr" "amdgpu-no-hostcall-ptr" "amdgpu-no-implicitarg-ptr" "amdgpu-no-lds-kernel-id" "amdgpu-no-multigrid-sync-arg" "amdgpu-no-queue-ptr" "amdgpu-no-workgroup-id-x" "amdgpu-no-workgroup-id-y" "amdgpu-no-workgroup-id-z" "amdgpu-no-workitem-id-x" "amdgpu-no-workitem-id-y" "amdgpu-no-workitem-id-z" "amdgpu-no-wwm" "amdgpu-waves-per-eu"="4,8" }
-; CHECK: attributes #[[ATTR16]] = { "amdgpu-no-cluster-id-x" "amdgpu-no-cluster-id-y" "amdgpu-no-cluster-id-z" "amdgpu-no-completion-action" "amdgpu-no-default-queue" "amdgpu-no-dispatch-id" "amdgpu-no-dispatch-ptr" "amdgpu-no-flat-scratch-init" "amdgpu-no-heap-ptr" "amdgpu-no-hostcall-ptr" "amdgpu-no-implicitarg-ptr" "amdgpu-no-lds-kernel-id" "amdgpu-no-multigrid-sync-arg" "amdgpu-no-queue-ptr" "amdgpu-no-workgroup-id-x" "amdgpu-no-workgroup-id-y" "amdgpu-no-workgroup-id-z" "amdgpu-no-workitem-id-x" "amdgpu-no-workitem-id-y" "amdgpu-no-workitem-id-z" "amdgpu-no-wwm" "amdgpu-waves-per-eu"="6,8" }
-; CHECK: attributes #[[ATTR17]] = { "amdgpu-no-cluster-id-x" "amdgpu-no-cluster-id-y" "amdgpu-no-cluster-id-z" "amdgpu-no-completion-action" "amdgpu-no-default-queue" "amdgpu-no-dispatch-id" "amdgpu-no-dispatch-ptr" "amdgpu-no-flat-scratch-init" "amdgpu-no-heap-ptr" "amdgpu-no-hostcall-ptr" "amdgpu-no-implicitarg-ptr" "amdgpu-no-lds-kernel-id" "amdgpu-no-multigrid-sync-arg" "amdgpu-no-queue-ptr" "amdgpu-no-workgroup-id-x" "amdgpu-no-workgroup-id-y" "amdgpu-no-workgroup-id-z" "amdgpu-no-workitem-id-x" "amdgpu-no-workitem-id-y" "amdgpu-no-workitem-id-z" "amdgpu-no-wwm" "amdgpu-waves-per-eu"="5,5" }
-; CHECK: attributes #[[ATTR18]] = { "amdgpu-no-cluster-id-x" "amdgpu-no-cluster-id-y" "amdgpu-no-cluster-id-z" "amdgpu-no-completion-action" "amdgpu-no-default-queue" "amdgpu-no-dispatch-id" "amdgpu-no-dispatch-ptr" "amdgpu-no-flat-scratch-init" "amdgpu-no-heap-ptr" "amdgpu-no-hostcall-ptr" "amdgpu-no-implicitarg-ptr" "amdgpu-no-lds-kernel-id" "amdgpu-no-multigrid-sync-arg" "amdgpu-no-queue-ptr" "amdgpu-no-workgroup-id-x" "amdgpu-no-workgroup-id-y" "amdgpu-no-workgroup-id-z" "amdgpu-no-workitem-id-x" "amdgpu-no-workitem-id-y" "amdgpu-no-workitem-id-z" "amdgpu-no-wwm" "amdgpu-waves-per-eu"="5,8" }
-; CHECK: attributes #[[ATTR19]] = { "amdgpu-no-cluster-id-x" "amdgpu-no-cluster-id-y" "amdgpu-no-cluster-id-z" "amdgpu-no-completion-action" "amdgpu-no-default-queue" "amdgpu-no-dispatch-id" "amdgpu-no-dispatch-ptr" "amdgpu-no-flat-scratch-init" "amdgpu-no-heap-ptr" "amdgpu-no-hostcall-ptr" "amdgpu-no-implicitarg-ptr" "amdgpu-no-lds-kernel-id" "amdgpu-no-multigrid-sync-arg" "amdgpu-no-queue-ptr" "amdgpu-no-workgroup-id-x" "amdgpu-no-workgroup-id-y" "amdgpu-no-workgroup-id-z" "amdgpu-no-workitem-id-x" "amdgpu-no-workitem-id-y" "amdgpu-no-workitem-id-z" "amdgpu-no-wwm" "amdgpu-waves-per-eu"="9,10" }
-; CHECK: attributes #[[ATTR20]] = { "amdgpu-no-cluster-id-x" "amdgpu-no-cluster-id-y" "amdgpu-no-cluster-id-z" "amdgpu-no-completion-action" "amdgpu-no-default-queue" "amdgpu-no-dispatch-id" "amdgpu-no-dispatch-ptr" "amdgpu-no-flat-scratch-init" "amdgpu-no-heap-ptr" "amdgpu-no-hostcall-ptr" "amdgpu-no-implicitarg-ptr" "amdgpu-no-lds-kernel-id" "amdgpu-no-multigrid-sync-arg" "amdgpu-no-queue-ptr" "amdgpu-no-workgroup-id-x" "amdgpu-no-workgroup-id-y" "amdgpu-no-workgroup-id-z" "amdgpu-no-workitem-id-x" "amdgpu-no-workitem-id-y" "amdgpu-no-workitem-id-z" "amdgpu-no-wwm" "amdgpu-waves-per-eu"="8,9" }
+; CHECK: attributes #[[ATTR0]] = { "amdgpu-flat-work-group-size"="1,64" "amdgpu-waves-per-eu"="1,8" }
+; CHECK: attributes #[[ATTR1]] = { "amdgpu-flat-work-group-size"="1,64" "amdgpu-waves-per-eu"="1,2" }
+; CHECK: attributes #[[ATTR2]] = { "amdgpu-flat-work-group-size"="1,64" "amdgpu-waves-per-eu"="1,4" }
+; CHECK: attributes #[[ATTR3]] = { "amdgpu-flat-work-group-size"="1,64" "amdgpu-waves-per-eu"="2,9" }
+; CHECK: attributes #[[ATTR4]] = { "amdgpu-flat-work-group-size"="1,64" "amdgpu-waves-per-eu"="1,1" }
+; CHECK: attributes #[[ATTR5]] = { "amdgpu-flat-work-group-size"="1,64" "amdgpu-waves-per-eu"="2,8" }
+; CHECK: attributes #[[ATTR6]] = { "amdgpu-flat-work-group-size"="1,64" "amdgpu-waves-per-eu"="9,10" }
+; CHECK: attributes #[[ATTR7]] = { "amdgpu-flat-work-group-size"="1,64" "amdgpu-waves-per-eu"="9,9" }
+; CHECK: attributes #[[ATTR8]] = { "amdgpu-flat-work-group-size"="1,64" "amdgpu-waves-per-eu"="3,8" }
+; CHECK: attributes #[[ATTR9]] = { "amdgpu-flat-work-group-size"="1,64" }
+; CHECK: attributes #[[ATTR10]] = { "amdgpu-flat-work-group-size"="1,64" "amdgpu-waves-per-eu"="1,123" }
+; CHECK: attributes #[[ATTR11]] = { "amdgpu-flat-work-group-size"="1,512" }
+; CHECK: attributes #[[ATTR12]] = { "amdgpu-flat-work-group-size"="1,512" "amdgpu-waves-per-eu"="3,6" }
+; CHECK: attributes #[[ATTR13]] = { "amdgpu-waves-per-eu"="3,6" }
+; CHECK: attributes #[[ATTR14]] = { "amdgpu-waves-per-eu"="4,8" }
+; CHECK: attributes #[[ATTR15]] = { "amdgpu-waves-per-eu"="6,8" }
+; CHECK: attributes #[[ATTR16]] = { "amdgpu-waves-per-eu"="5,5" }
+; CHECK: attributes #[[ATTR17]] = { "amdgpu-waves-per-eu"="5,8" }
+; CHECK: attributes #[[ATTR18]] = { "amdgpu-waves-per-eu"="9,10" }
+; CHECK: attributes #[[ATTR19]] = { "amdgpu-waves-per-eu"="8,9" }
;.
>From 412d9ef6ef4a9367c594a4c3cc4b153e167470b6 Mon Sep 17 00:00:00 2001
From: Arseniy Obolenskiy <arseniy.obolenskiy at amd.com>
Date: Wed, 15 Jul 2026 13:48:56 +0200
Subject: [PATCH 3/3] Address comments
---
llvm/lib/Target/AMDGPU/AMDGPUAttributor.cpp | 39 +++++++++++++++-
.../CodeGen/AMDGPU/propagate-waves-per-eu.ll | 44 +++++++++----------
2 files changed, 60 insertions(+), 23 deletions(-)
diff --git a/llvm/lib/Target/AMDGPU/AMDGPUAttributor.cpp b/llvm/lib/Target/AMDGPU/AMDGPUAttributor.cpp
index e4b40a7e7be7b..14795a48174df 100644
--- a/llvm/lib/Target/AMDGPU/AMDGPUAttributor.cpp
+++ b/llvm/lib/Target/AMDGPU/AMDGPUAttributor.cpp
@@ -1142,7 +1142,44 @@ struct AAAMDWavesPerEU : public AAAMDSizeRangeAttribute {
}
ChangeStatus updateImpl(Attributor &A) override {
- return updateImplImpl<AAAMDWavesPerEU>(A);
+ ChangeStatus Change = ChangeStatus::UNCHANGED;
+
+ auto CheckCallSite = [&](AbstractCallSite CS) {
+ Function *Caller = CS.getInstruction()->getFunction();
+ LLVM_DEBUG(dbgs() << '[' << getName() << "] Call " << Caller->getName()
+ << "->" << getAssociatedFunction()->getName() << '\n');
+
+ const auto *CallerAA = A.getAAFor<AAAMDWavesPerEU>(
+ *this, IRPosition::function(*Caller), DepClassTy::REQUIRED);
+ if (!CallerAA || !CallerAA->isValidState())
+ return false;
+
+ ConstantRange Assumed = getAssumed();
+ ConstantRange CallerRange = CallerAA->getAssumed();
+ // Take the conservative (larger) lower bound: a callee reachable from
+ // several kernels must run enough waves per EU for the most demanding
+ // one, otherwise its register budget would be too optimistic. Widen the
+ // upper bound to cover every caller.
+ unsigned Min = std::max(Assumed.getLower().getZExtValue(),
+ CallerRange.getLower().getZExtValue());
+ unsigned Max = std::max(Assumed.getUpper().getZExtValue(),
+ CallerRange.getUpper().getZExtValue());
+ IntegerRangeState RangeState(
+ ConstantRange(APInt(32, Min), APInt(32, Max)));
+ getState() = RangeState;
+ Change |= getState() == Assumed ? ChangeStatus::UNCHANGED
+ : ChangeStatus::CHANGED;
+
+ return true;
+ };
+
+ bool AllCallSitesKnown = true;
+ if (!A.checkForAllCallSites(CheckCallSite, *this,
+ /*RequireAllCallSites=*/true,
+ AllCallSitesKnown))
+ return indicatePessimisticFixpoint();
+
+ return Change;
}
/// Create an abstract attribute view for the position \p IRP.
diff --git a/llvm/test/CodeGen/AMDGPU/propagate-waves-per-eu.ll b/llvm/test/CodeGen/AMDGPU/propagate-waves-per-eu.ll
index 67caeb760ce44..0dca1fec39ddc 100644
--- a/llvm/test/CodeGen/AMDGPU/propagate-waves-per-eu.ll
+++ b/llvm/test/CodeGen/AMDGPU/propagate-waves-per-eu.ll
@@ -20,7 +20,7 @@ define internal void @default_to_1_8_a() {
define amdgpu_kernel void @kernel_1_8() #0 {
; CHECK-LABEL: define amdgpu_kernel void @kernel_1_8
-; CHECK-SAME: () #[[ATTR0]] {
+; CHECK-SAME: () #[[ATTR1:[0-9]+]] {
; CHECK-NEXT: call void @default_to_1_8_a()
; CHECK-NEXT: call void @dummy()
; CHECK-NEXT: ret void
@@ -33,7 +33,7 @@ define amdgpu_kernel void @kernel_1_8() #0 {
; Called from a single kernel with 1,2
define internal void @default_to_1_2() {
; CHECK-LABEL: define internal void @default_to_1_2
-; CHECK-SAME: () #[[ATTR1:[0-9]+]] {
+; CHECK-SAME: () #[[ATTR2:[0-9]+]] {
; CHECK-NEXT: call void @dummy()
; CHECK-NEXT: ret void
;
@@ -43,7 +43,7 @@ define internal void @default_to_1_2() {
define amdgpu_kernel void @kernel_1_2() #1 {
; CHECK-LABEL: define amdgpu_kernel void @kernel_1_2
-; CHECK-SAME: () #[[ATTR1]] {
+; CHECK-SAME: () #[[ATTR2]] {
; CHECK-NEXT: call void @default_to_1_2()
; CHECK-NEXT: call void @flat_group_1_1()
; CHECK-NEXT: call void @default_to_1_8_b()
@@ -62,7 +62,7 @@ define amdgpu_kernel void @kernel_1_2() #1 {
; Called from a single kernel with 1,4
define internal void @default_to_1_4() {
; CHECK-LABEL: define internal void @default_to_1_4
-; CHECK-SAME: () #[[ATTR2:[0-9]+]] {
+; CHECK-SAME: () #[[ATTR3:[0-9]+]] {
; CHECK-NEXT: call void @dummy()
; CHECK-NEXT: ret void
;
@@ -72,7 +72,7 @@ define internal void @default_to_1_4() {
define amdgpu_kernel void @kernel_1_4() #2 {
; CHECK-LABEL: define amdgpu_kernel void @kernel_1_4
-; CHECK-SAME: () #[[ATTR2]] {
+; CHECK-SAME: () #[[ATTR3]] {
; CHECK-NEXT: call void @default_to_1_4()
; CHECK-NEXT: call void @dummy()
; CHECK-NEXT: ret void
@@ -85,7 +85,7 @@ define amdgpu_kernel void @kernel_1_4() #2 {
; Called from kernels with 2,9 and 9,9
define internal void @default_to_2_9() {
; CHECK-LABEL: define internal void @default_to_2_9
-; CHECK-SAME: () #[[ATTR3:[0-9]+]] {
+; CHECK-SAME: () #[[ATTR4:[0-9]+]] {
; CHECK-NEXT: call void @dummy()
; CHECK-NEXT: ret void
;
@@ -97,7 +97,7 @@ define internal void @default_to_2_9() {
; bounds, and should not be changed.
define internal void @flat_group_1_1() #3 {
; CHECK-LABEL: define internal void @flat_group_1_1
-; CHECK-SAME: () #[[ATTR4:[0-9]+]] {
+; CHECK-SAME: () #[[ATTR5:[0-9]+]] {
; CHECK-NEXT: call void @dummy()
; CHECK-NEXT: ret void
;
@@ -108,7 +108,7 @@ define internal void @flat_group_1_1() #3 {
; 2,8 -> 2,2
define internal void @flat_group_2_8() #4 {
; CHECK-LABEL: define internal void @flat_group_2_8
-; CHECK-SAME: () #[[ATTR5:[0-9]+]] {
+; CHECK-SAME: () #[[ATTR0]] {
; CHECK-NEXT: call void @dummy()
; CHECK-NEXT: ret void
;
@@ -129,7 +129,7 @@ define internal void @flat_group_9_10() #5 {
define amdgpu_kernel void @kernel_2_9() #6 {
; CHECK-LABEL: define amdgpu_kernel void @kernel_2_9
-; CHECK-SAME: () #[[ATTR3]] {
+; CHECK-SAME: () #[[ATTR7:[0-9]+]] {
; CHECK-NEXT: call void @default_to_2_9()
; CHECK-NEXT: call void @flat_group_1_1()
; CHECK-NEXT: call void @dummy()
@@ -143,7 +143,7 @@ define amdgpu_kernel void @kernel_2_9() #6 {
define amdgpu_kernel void @kernel_9_9() #7 {
; CHECK-LABEL: define amdgpu_kernel void @kernel_9_9
-; CHECK-SAME: () #[[ATTR7:[0-9]+]] {
+; CHECK-SAME: () #[[ATTR4]] {
; CHECK-NEXT: call void @default_to_2_9()
; CHECK-NEXT: call void @flat_group_9_10()
; CHECK-NEXT: call void @dummy()
@@ -170,7 +170,7 @@ define internal void @default_to_1_8_b() {
; this should probably be illegal.
define amdgpu_kernel void @kernel_2_8() #4 {
; CHECK-LABEL: define amdgpu_kernel void @kernel_2_8
-; CHECK-SAME: () #[[ATTR5]] {
+; CHECK-SAME: () #[[ATTR0]] {
; CHECK-NEXT: call void @default_to_1_8_a()
; CHECK-NEXT: call void @default_to_1_8_b()
; CHECK-NEXT: call void @dummy()
@@ -185,7 +185,7 @@ define amdgpu_kernel void @kernel_2_8() #4 {
; 1,2 -> 2,2
define internal void @merge_cycle_0() #1 {
; CHECK-LABEL: define internal void @merge_cycle_0
-; CHECK-SAME: () #[[ATTR1]] {
+; CHECK-SAME: () #[[ATTR2]] {
; CHECK-NEXT: call void @merge_cycle_1()
; CHECK-NEXT: call void @dummy()
; CHECK-NEXT: ret void
@@ -199,7 +199,7 @@ define internal void @merge_cycle_0() #1 {
; 2,8 -> 2,8
define internal void @merge_cycle_1() #4 {
; CHECK-LABEL: define internal void @merge_cycle_1
-; CHECK-SAME: () #[[ATTR5]] {
+; CHECK-SAME: () #[[ATTR0]] {
; CHECK-NEXT: call void @merge_cycle_0()
; CHECK-NEXT: call void @dummy()
; CHECK-NEXT: ret void
@@ -260,7 +260,7 @@ define internal i32 @bitcasted_function() {
define internal void @called_from_invalid_bounds_0() {
; CHECK-LABEL: define internal void @called_from_invalid_bounds_0
-; CHECK-SAME: () #[[ATTR0]] {
+; CHECK-SAME: () #[[ATTR1]] {
; CHECK-NEXT: call void @dummy()
; CHECK-NEXT: ret void
;
@@ -281,7 +281,7 @@ define internal void @called_from_invalid_bounds_1() {
; Invalid range for amdgpu-waves-per-eu
define amdgpu_kernel void @kernel_invalid_bounds_0_8() #9 {
; CHECK-LABEL: define amdgpu_kernel void @kernel_invalid_bounds_0_8
-; CHECK-SAME: () #[[ATTR0]] {
+; CHECK-SAME: () #[[ATTR1]] {
; CHECK-NEXT: call void @called_from_invalid_bounds_0()
; CHECK-NEXT: call void @dummy()
; CHECK-NEXT: ret void
@@ -470,14 +470,14 @@ attributes #17 = { "amdgpu-waves-per-eu"="5,8" }
attributes #18 = { "amdgpu-waves-per-eu"="9,10" }
attributes #19 = { "amdgpu-waves-per-eu"="8,9" }
;.
-; CHECK: attributes #[[ATTR0]] = { "amdgpu-flat-work-group-size"="1,64" "amdgpu-waves-per-eu"="1,8" }
-; CHECK: attributes #[[ATTR1]] = { "amdgpu-flat-work-group-size"="1,64" "amdgpu-waves-per-eu"="1,2" }
-; CHECK: attributes #[[ATTR2]] = { "amdgpu-flat-work-group-size"="1,64" "amdgpu-waves-per-eu"="1,4" }
-; CHECK: attributes #[[ATTR3]] = { "amdgpu-flat-work-group-size"="1,64" "amdgpu-waves-per-eu"="2,9" }
-; CHECK: attributes #[[ATTR4]] = { "amdgpu-flat-work-group-size"="1,64" "amdgpu-waves-per-eu"="1,1" }
-; CHECK: attributes #[[ATTR5]] = { "amdgpu-flat-work-group-size"="1,64" "amdgpu-waves-per-eu"="2,8" }
+; CHECK: attributes #[[ATTR0]] = { "amdgpu-flat-work-group-size"="1,64" "amdgpu-waves-per-eu"="2,8" }
+; CHECK: attributes #[[ATTR1]] = { "amdgpu-flat-work-group-size"="1,64" "amdgpu-waves-per-eu"="1,8" }
+; CHECK: attributes #[[ATTR2]] = { "amdgpu-flat-work-group-size"="1,64" "amdgpu-waves-per-eu"="1,2" }
+; CHECK: attributes #[[ATTR3]] = { "amdgpu-flat-work-group-size"="1,64" "amdgpu-waves-per-eu"="1,4" }
+; CHECK: attributes #[[ATTR4]] = { "amdgpu-flat-work-group-size"="1,64" "amdgpu-waves-per-eu"="9,9" }
+; CHECK: attributes #[[ATTR5]] = { "amdgpu-flat-work-group-size"="1,64" "amdgpu-waves-per-eu"="1,1" }
; CHECK: attributes #[[ATTR6]] = { "amdgpu-flat-work-group-size"="1,64" "amdgpu-waves-per-eu"="9,10" }
-; CHECK: attributes #[[ATTR7]] = { "amdgpu-flat-work-group-size"="1,64" "amdgpu-waves-per-eu"="9,9" }
+; CHECK: attributes #[[ATTR7]] = { "amdgpu-flat-work-group-size"="1,64" "amdgpu-waves-per-eu"="2,9" }
; CHECK: attributes #[[ATTR8]] = { "amdgpu-flat-work-group-size"="1,64" "amdgpu-waves-per-eu"="3,8" }
; CHECK: attributes #[[ATTR9]] = { "amdgpu-flat-work-group-size"="1,64" }
; CHECK: attributes #[[ATTR10]] = { "amdgpu-flat-work-group-size"="1,64" "amdgpu-waves-per-eu"="1,123" }
More information about the llvm-commits
mailing list