[DirectX] ForwardHandle needs to check if globals were stored on allocas #151751
Conversation
This file contains hidden or 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
fixes llvm#140819 SROA pass is making it so that some globals get loaded into stack allocations. This means we find an alloca where we use to expect a load and now need to walk an alloca -> store -> maybe load chain before we find the global. Doing so fixes All but two instances of llvm#137715 And fixes every instance of `Load of "8.sroa.0" is not a global resource handle we are currently seeing in the DML shaders.
|
@llvm/pr-subscribers-backend-directx Author: Farzon Lotfi (farzonl) Changesfixes #140819 SROA pass is making it so that some globals get loaded into stack allocations. This means we find an alloca where we use to expect a load and now need to walk an alloca -> store -> maybe load chain before we find the global. Doing so fixes All but two instances of #137715 And fixes every instance of `Load of "8.sroa.0" is not a global resource handle we are currently seeing in the DML shaders. Full diff: https://github.com/llvm/llvm-project/pull/151751.diff 2 Files Affected:
diff --git a/llvm/lib/Target/DirectX/DXILForwardHandleAccesses.cpp b/llvm/lib/Target/DirectX/DXILForwardHandleAccesses.cpp
index 73abfe7c48584..747472b1a4bc1 100644
--- a/llvm/lib/Target/DirectX/DXILForwardHandleAccesses.cpp
+++ b/llvm/lib/Target/DirectX/DXILForwardHandleAccesses.cpp
@@ -87,17 +87,48 @@ static bool forwardHandleAccesses(Function &F, DominatorTree &DT) {
for (LoadInst *LI : LoadsToProcess) {
Value *V = LI->getPointerOperand();
- auto *GV = dyn_cast<GlobalVariable>(LI->getPointerOperand());
+ auto *GV = dyn_cast<GlobalVariable>(V);
// If we didn't find the global, we may need to walk through a level of
// indirection. This generally happens at -O0.
- if (!GV)
+ if (!GV) {
if (auto *NestedLI = dyn_cast<LoadInst>(V)) {
BasicBlock::iterator BBI(NestedLI);
Value *Loaded = FindAvailableLoadedValue(
NestedLI, NestedLI->getParent(), BBI, 0, nullptr, nullptr);
GV = dyn_cast_or_null<GlobalVariable>(Loaded);
+ } else if (auto *NestedAlloca = dyn_cast<AllocaInst>(V)) {
+ for (auto &Use : NestedAlloca->uses()) {
+ auto *Store = dyn_cast<StoreInst>(Use.getUser());
+ if (!Store)
+ continue;
+
+ Value *StoredVal = Store->getValueOperand();
+ if (!StoredVal)
+ continue;
+
+ // Try direct global match
+ GV = dyn_cast<GlobalVariable>(StoredVal);
+ if (GV)
+ break;
+
+ // If it's a load, check its source
+ if (auto *Load = dyn_cast<LoadInst>(StoredVal)) {
+ GV = dyn_cast<GlobalVariable>(Load->getPointerOperand());
+ if (GV)
+ break;
+
+ // Try to find available loaded value
+ BasicBlock::iterator BBI(Load);
+ Value *Loaded = FindAvailableLoadedValue(Load, Load->getParent(),
+ BBI, 0, nullptr, nullptr);
+ GV = dyn_cast<GlobalVariable>(Loaded);
+ if (GV)
+ break;
+ }
+ }
}
+ }
auto It = HandleMap.find(GV);
if (It == HandleMap.end()) {
diff --git a/llvm/test/CodeGen/DirectX/issue-140819_allow_forward_handle_on_alloca.ll b/llvm/test/CodeGen/DirectX/issue-140819_allow_forward_handle_on_alloca.ll
new file mode 100644
index 0000000000000..0c1ccb50caee7
--- /dev/null
+++ b/llvm/test/CodeGen/DirectX/issue-140819_allow_forward_handle_on_alloca.ll
@@ -0,0 +1,39 @@
+; NOTE: Assertions have been autogenerated by utils/update_test_checks.py UTC_ARGS: --version 5
+; RUN: opt -S -dxil-forward-handle-accesses %s | FileCheck %s
+
+%"class.hlsl::RWStructuredBuffer" = type { target("dx.RawBuffer", i32, 1, 0) }
+@_ZL4dest = internal unnamed_addr global %"class.hlsl::RWStructuredBuffer" poison, align 4
+@.str = private unnamed_addr constant [5 x i8] c"dest\00", align 1
+
+
+; NOTE: intent of this test is to confirm load target("dx.RawBuffer", i32, 1, 0)
+; is replaced with call @llvm.dx.resource.getpointer
+define void @CSMain() local_unnamed_addr {
+; CHECK-LABEL: define void @CSMain() local_unnamed_addr {
+; CHECK-NEXT: [[ENTRY:.*:]]
+; CHECK-NEXT: [[AGG_TMP_I1_SROA_0:%.*]] = alloca target("dx.RawBuffer", i32, 1, 0), align 8
+; CHECK-NEXT: [[TMP0:%.*]] = tail call target("dx.RawBuffer", i32, 1, 0) @llvm.dx.resource.handlefrombinding.tdx.RawBuffer_i32_1_0t(i32 0, i32 3, i32 1, i32 0, i1 false, ptr nonnull @.str)
+; CHECK-NEXT: store target("dx.RawBuffer", i32, 1, 0) [[TMP0]], ptr @_ZL4dest, align 4
+; CHECK-NEXT: [[TMP1:%.*]] = tail call i32 @llvm.dx.thread.id(i32 0)
+; CHECK-NEXT: [[TMP2:%.*]] = load i32, ptr @_ZL4dest, align 4
+; CHECK-NEXT: call void @llvm.lifetime.start.p0(i64 4, ptr nonnull [[AGG_TMP_I1_SROA_0]])
+; CHECK-NEXT: store i32 [[TMP2]], ptr [[AGG_TMP_I1_SROA_0]], align 8
+; CHECK-NEXT: [[TMP3:%.*]] = tail call noundef nonnull align 4 dereferenceable(4) ptr @llvm.dx.resource.getpointer.p0.tdx.RawBuffer_i32_1_0t(target("dx.RawBuffer", i32, 1, 0) [[TMP0]], i32 [[TMP1]])
+; CHECK-NEXT: store i32 0, ptr [[TMP3]], align 4
+; CHECK-NEXT: call void @llvm.lifetime.end.p0(i64 4, ptr nonnull [[AGG_TMP_I1_SROA_0]])
+; CHECK-NEXT: ret void
+;
+entry:
+ %agg.tmp.i1.sroa.0 = alloca target("dx.RawBuffer", i32, 1, 0), align 8
+ %0 = tail call target("dx.RawBuffer", i32, 1, 0) @llvm.dx.resource.handlefrombinding.tdx.RawBuffer_i32_1_0t(i32 0, i32 3, i32 1, i32 0, i1 false, ptr nonnull @.str)
+ store target("dx.RawBuffer", i32, 1, 0) %0, ptr @_ZL4dest, align 4
+ %1 = tail call i32 @llvm.dx.thread.id(i32 0)
+ %2 = load i32, ptr @_ZL4dest, align 4
+ call void @llvm.lifetime.start.p0(i64 4, ptr nonnull %agg.tmp.i1.sroa.0)
+ store i32 %2, ptr %agg.tmp.i1.sroa.0, align 8
+ %agg.tmp.i1.sroa.0.0.agg.tmp.i1.sroa.0.0.agg.tmp.i1.sroa.0.0.agg.tmp.i1.sroa.0.0. = load target("dx.RawBuffer", i32, 1, 0), ptr %agg.tmp.i1.sroa.0, align 8
+ %3 = tail call noundef nonnull align 4 dereferenceable(4) ptr @llvm.dx.resource.getpointer.p0.tdx.RawBuffer_i32_1_0t(target("dx.RawBuffer", i32, 1, 0) %agg.tmp.i1.sroa.0.0.agg.tmp.i1.sroa.0.0.agg.tmp.i1.sroa.0.0.agg.tmp.i1.sroa.0.0., i32 %1)
+ store i32 0, ptr %3, align 4
+ call void @llvm.lifetime.end.p0(i64 4, ptr nonnull %agg.tmp.i1.sroa.0)
+ ret void
+}
|
|
LGTM. |
bogner
reviewed
Aug 1, 2025
llvm/test/CodeGen/DirectX/issue-140819_allow_forward_handle_on_alloca.ll
Outdated
Show resolved
Hide resolved
bogner
reviewed
Aug 1, 2025
llvm/test/CodeGen/DirectX/issue-140819_allow_forward_handle_on_alloca.ll
Outdated
Show resolved
Hide resolved
bogner
approved these changes
Aug 6, 2025
|
Failure is in compilerRT not any of our code, going to bypass merge rule. |
|
LLVM Buildbot has detected a new failure on builder Full details are available at: https://lab.llvm.org/buildbot/#/builders/190/builds/24994 Here is the relevant piece of the build log for the reference |
|
LLVM Buildbot has detected a new failure on builder Full details are available at: https://lab.llvm.org/buildbot/#/builders/141/builds/10645 Here is the relevant piece of the build log for the reference |
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.
fixes #140819
SROA pass is making it so that some globals get loaded into stack allocations. This means we find an alloca where we use to expect a load and now need to walk an alloca -> store -> maybe load chain before we find the global. Doing so fixes All but two instances of #137715 And fixes every instance of `Load of "8.sroa.0" is not a global resource handle we are currently seeing in the DML shaders.