[clang] Emit convergence tokens for loop in global array init#140120
Merged
Conversation
When initializing a global array, a loop is generated, but no convergence is emitted for the loop. This fixes that up.
Member
|
@llvm/pr-subscribers-clang-codegen @llvm/pr-subscribers-hlsl Author: Steven Perron (s-perron) ChangesWhen initializing a global array, a loop is generated, but no Full diff: https://github.com/llvm/llvm-project/pull/140120.diff 2 Files Affected:
diff --git a/clang/lib/CodeGen/CGClass.cpp b/clang/lib/CodeGen/CGClass.cpp
index befbfc64a680c..44062739d5dd4 100644
--- a/clang/lib/CodeGen/CGClass.cpp
+++ b/clang/lib/CodeGen/CGClass.cpp
@@ -2042,6 +2042,8 @@ void CodeGenFunction::EmitCXXAggrConstructorCall(const CXXConstructorDecl *ctor,
cur->addIncoming(arrayBegin, entryBB);
// Inside the loop body, emit the constructor call on the array element.
+ if (CGM.shouldEmitConvergenceTokens())
+ ConvergenceTokenStack.push_back(emitConvergenceLoopToken(loopBB));
// The alignment of the base, adjusted by the size of a single element,
// provides a conservative estimate of the alignment of every element.
@@ -2101,6 +2103,9 @@ void CodeGenFunction::EmitCXXAggrConstructorCall(const CXXConstructorDecl *ctor,
// Patch the earlier check to skip over the loop.
if (zeroCheckBranch) zeroCheckBranch->setSuccessor(0, contBB);
+ if (CGM.shouldEmitConvergenceTokens())
+ ConvergenceTokenStack.pop_back();
+
EmitBlock(contBB);
}
diff --git a/clang/test/CodeGenHLSL/convergence/global_array.hlsl b/clang/test/CodeGenHLSL/convergence/global_array.hlsl
new file mode 100644
index 0000000000000..e11de82c3ded0
--- /dev/null
+++ b/clang/test/CodeGenHLSL/convergence/global_array.hlsl
@@ -0,0 +1,16 @@
+// RUN: %clang_cc1 -Wno-hlsl-implicit-binding -finclude-default-header -triple spirv-unknown-vulkan-compute -emit-llvm -disable-llvm-passes -o - %s | FileCheck %s
+
+// CHECK: define internal spir_func void @__cxx_global_var_init()
+// CHECK: [[entry_token:%.*]] = call token @llvm.experimental.convergence.entry()
+// CHECK: br label %[[loop_entry:.*]]
+
+// CHECK: [[loop_entry]]:
+// CHECK: [[loop_token:%.*]] = call token @llvm.experimental.convergence.loop() [ "convergencectrl"(token [[entry_token]]) ]
+// CHECK: call void {{.*}} [ "convergencectrl"(token [[loop_token]]) ]
+// CHECK: br i1 {{%.*}} label {{%.*}} label %[[loop_entry]]
+RWBuffer<float> e[2];
+
+[numthreads(4,1,1)]
+void main() {
+}
+
|
Member
|
@llvm/pr-subscribers-clang Author: Steven Perron (s-perron) ChangesWhen initializing a global array, a loop is generated, but no Full diff: https://github.com/llvm/llvm-project/pull/140120.diff 2 Files Affected:
diff --git a/clang/lib/CodeGen/CGClass.cpp b/clang/lib/CodeGen/CGClass.cpp
index befbfc64a680c..44062739d5dd4 100644
--- a/clang/lib/CodeGen/CGClass.cpp
+++ b/clang/lib/CodeGen/CGClass.cpp
@@ -2042,6 +2042,8 @@ void CodeGenFunction::EmitCXXAggrConstructorCall(const CXXConstructorDecl *ctor,
cur->addIncoming(arrayBegin, entryBB);
// Inside the loop body, emit the constructor call on the array element.
+ if (CGM.shouldEmitConvergenceTokens())
+ ConvergenceTokenStack.push_back(emitConvergenceLoopToken(loopBB));
// The alignment of the base, adjusted by the size of a single element,
// provides a conservative estimate of the alignment of every element.
@@ -2101,6 +2103,9 @@ void CodeGenFunction::EmitCXXAggrConstructorCall(const CXXConstructorDecl *ctor,
// Patch the earlier check to skip over the loop.
if (zeroCheckBranch) zeroCheckBranch->setSuccessor(0, contBB);
+ if (CGM.shouldEmitConvergenceTokens())
+ ConvergenceTokenStack.pop_back();
+
EmitBlock(contBB);
}
diff --git a/clang/test/CodeGenHLSL/convergence/global_array.hlsl b/clang/test/CodeGenHLSL/convergence/global_array.hlsl
new file mode 100644
index 0000000000000..e11de82c3ded0
--- /dev/null
+++ b/clang/test/CodeGenHLSL/convergence/global_array.hlsl
@@ -0,0 +1,16 @@
+// RUN: %clang_cc1 -Wno-hlsl-implicit-binding -finclude-default-header -triple spirv-unknown-vulkan-compute -emit-llvm -disable-llvm-passes -o - %s | FileCheck %s
+
+// CHECK: define internal spir_func void @__cxx_global_var_init()
+// CHECK: [[entry_token:%.*]] = call token @llvm.experimental.convergence.entry()
+// CHECK: br label %[[loop_entry:.*]]
+
+// CHECK: [[loop_entry]]:
+// CHECK: [[loop_token:%.*]] = call token @llvm.experimental.convergence.loop() [ "convergencectrl"(token [[entry_token]]) ]
+// CHECK: call void {{.*}} [ "convergencectrl"(token [[loop_token]]) ]
+// CHECK: br i1 {{%.*}} label {{%.*}} label %[[loop_entry]]
+RWBuffer<float> e[2];
+
+[numthreads(4,1,1)]
+void main() {
+}
+
|
cassiebeckley
approved these changes
May 15, 2025
bogner
approved these changes
May 15, 2025
This was referenced Aug 6, 2025
inbelic
added a commit
to inbelic/llvm-project
that referenced
this pull request
Mar 26, 2026
This follows the previous correction made in llvm#140120
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
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
4 participants
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.
When initializing a global array, a loop is generated, but no
convergence is emitted for the loop. This fixes that up.