[llvm-branch-commits] [Clang][MemProf] Add end-to-end test for PGHO rewriting (PR #169243)

via llvm-branch-commits llvm-branch-commits at lists.llvm.org
Sun Nov 23 13:21:03 PST 2025


llvmbot wrote:


<!--LLVM PR SUMMARY COMMENT-->

@llvm/pr-subscribers-clang

Author: Marco Elver (melver)

<details>
<summary>Changes</summary>

Adds an end-to-end (non-LTO) test verifying that the optimization
pipeline is set up correctly for Profile Guided Heap Optimization (PGHO)
transforms. Ensure that both PGHO and AllocToken can stack, and the
AllocToken pass does not interfere with PGHO and vice versa.


---
Full diff: https://github.com/llvm/llvm-project/pull/169243.diff


1 Files Affected:

- (added) clang/test/CodeGen/memprof-pgho.cpp (+61) 


``````````diff
diff --git a/clang/test/CodeGen/memprof-pgho.cpp b/clang/test/CodeGen/memprof-pgho.cpp
new file mode 100644
index 0000000000000..73bec2b303f13
--- /dev/null
+++ b/clang/test/CodeGen/memprof-pgho.cpp
@@ -0,0 +1,61 @@
+// Test end-to-end optimization pipeline with PGHO, that it does not interfere
+// with other allocation instrumentation features.
+//
+// RUN: split-file %s %t
+// RUN: llvm-profdata merge %t/memprof.yaml -o %t/use.profdata
+// RUN: %clang_cc1 -O2 -debug-info-kind=limited -fmemory-profile-use=%t/use.profdata -mllvm -optimize-hot-cold-new \
+// RUN:            %t/src.cpp -triple x86_64-linux-gnu -emit-llvm -o - | FileCheck %s --check-prefixes=CHECK,DEFAULT
+// RUN: %clang_cc1 -O2 -fsanitize=alloc-token -debug-info-kind=limited -fmemory-profile-use=%t/use.profdata -mllvm -optimize-hot-cold-new \
+// RUN:             %t/src.cpp -triple x86_64-linux-gnu -emit-llvm -o - | FileCheck %s --check-prefixes=CHECK,ALLOCTOKEN
+
+//--- memprof.yaml
+---
+HeapProfileRecords:
+  - GUID: 0x7f8d88fcc70a347b
+    AllocSites:
+    - Callstack:
+      - { Function: 0x7f8d88fcc70a347b, LineOffset: 1, Column: 10, IsInlineFrame: false }
+      - { Function: 0xdb956436e78dd5fa, LineOffset: 1, Column: 13, IsInlineFrame: false }
+      MemInfoBlock:
+        AllocCount: 1
+        TotalAccessCount: 0
+        MinAccessCount: 0
+        MaxAccessCount: 0
+        TotalSize: 10
+        MinSize: 10
+        MaxSize: 10
+        AllocTimestamp: 100
+        DeallocTimestamp: 100
+        TotalLifetime: 100000
+        MinLifetime: 100000
+        MaxLifetime: 100000
+        AllocCpuId: 0
+        DeallocCpuId: 0
+        NumMigratedCpu: 0
+        NumLifetimeOverlaps: 0
+        NumSameAllocCpu: 0
+        NumSameDeallocCpu: 0
+        DataTypeId: 0
+        TotalAccessDensity: 0
+        MinAccessDensity: 0
+        MaxAccessDensity: 0
+        TotalLifetimeAccessDensity: 0
+        MinLifetimeAccessDensity: 0
+        MaxLifetimeAccessDensity: 0
+        AccessHistogramSize: 0
+        AccessHistogram: 0
+...
+
+//--- src.cpp
+// CHECK-LABEL: define{{.*}} ptr @_Z3foov()
+// DEFAULT:    call {{.*}} ptr @_Znam12__hot_cold_t(i64 10, i8 -128)
+// ALLOCTOKEN: call {{.*}} ptr @__alloc_token__Znam12__hot_cold_t(i64 10, i8 -128, i64 1538840549748785101){{.*}} !alloc_token
+char *foo() {
+  return new char[10];
+}
+
+int main() {
+  char *a = foo();
+  delete[] a;
+  return 0;
+}

``````````

</details>


https://github.com/llvm/llvm-project/pull/169243


More information about the llvm-branch-commits mailing list