[HIP][HIPSTDPAR] Re-work allocation interposition for hipstdpar
#138790
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
|
@llvm/pr-subscribers-clang @llvm/pr-subscribers-llvm-transforms Author: Alex Voicu (AlexVlx) ChangesThe allocation interposition mode had a number of issues, which are primarily addressed in the library component via <ROCm/rocThrust#543>. However, it is necessary to interpose some additional symbols, which this patch does. Furthermore, to implement this in a compatible way, we guard the new implementation under a V1 macro, which is defined in addition to the existing Full diff: https://github.com/llvm/llvm-project/pull/138790.diff 5 Files Affected:
diff --git a/clang/docs/HIPSupport.rst b/clang/docs/HIPSupport.rst
index b2ac53843aeed..dfc27089f9d4d 100644
--- a/clang/docs/HIPSupport.rst
+++ b/clang/docs/HIPSupport.rst
@@ -518,7 +518,7 @@ Predefined Macros
* - ``__HIPSTDPAR__``
- Defined when Clang is compiling code in algorithm offload mode, enabled
with the ``--hipstdpar`` compiler option.
- * - ``__HIPSTDPAR_INTERPOSE_ALLOC__``
+ * - ``__HIPSTDPAR_INTERPOSE_ALLOC__`` / ``__HIPSTDPAR_INTERPOSE_ALLOC_V1__``
- Defined only when compiling in algorithm offload mode, when the user
enables interposition mode with the ``--hipstdpar-interpose-alloc``
compiler option, indicating that all dynamic memory allocation /
diff --git a/clang/lib/Frontend/InitPreprocessor.cpp b/clang/lib/Frontend/InitPreprocessor.cpp
index 906b0faf44067..96d6fb64a6319 100644
--- a/clang/lib/Frontend/InitPreprocessor.cpp
+++ b/clang/lib/Frontend/InitPreprocessor.cpp
@@ -618,8 +618,10 @@ static void InitializeStandardPredefinedMacros(const TargetInfo &TI,
Builder.defineMacro("__HIP_MEMORY_SCOPE_SYSTEM", "5");
if (LangOpts.HIPStdPar) {
Builder.defineMacro("__HIPSTDPAR__");
- if (LangOpts.HIPStdParInterposeAlloc)
+ if (LangOpts.HIPStdParInterposeAlloc) {
Builder.defineMacro("__HIPSTDPAR_INTERPOSE_ALLOC__");
+ Builder.defineMacro("__HIPSTDPAR_INTERPOSE_ALLOC_V1__");
+ }
}
if (LangOpts.CUDAIsDevice) {
Builder.defineMacro("__HIP_DEVICE_COMPILE__");
diff --git a/clang/test/Preprocessor/predefined-macros.c b/clang/test/Preprocessor/predefined-macros.c
index 633ba4681ac52..b7765bfa2fb14 100644
--- a/clang/test/Preprocessor/predefined-macros.c
+++ b/clang/test/Preprocessor/predefined-macros.c
@@ -304,11 +304,13 @@
// RUN: %clang_cc1 %s -E -dM -o - -x hip --hipstdpar -triple x86_64-unknown-linux-gnu \
// RUN: | FileCheck -match-full-lines %s --check-prefix=CHECK-HIPSTDPAR
// CHECK-HIPSTDPAR: #define __HIPSTDPAR__ 1
+// CHECK-HIPSTDPAR-NOT: #define __HIPSTDPAR_INTERPOSE_ALLOC_V1__ 1
// CHECK-HIPSTDPAR-NOT: #define __HIPSTDPAR_INTERPOSE_ALLOC__ 1
// RUN: %clang_cc1 %s -E -dM -o - -x hip --hipstdpar --hipstdpar-interpose-alloc \
// RUN: -triple x86_64-unknown-linux-gnu | FileCheck -match-full-lines %s \
// RUN: --check-prefix=CHECK-HIPSTDPAR-INTERPOSE
+// CHECK-HIPSTDPAR-INTERPOSE: #define __HIPSTDPAR_INTERPOSE_ALLOC_V1__ 1
// CHECK-HIPSTDPAR-INTERPOSE: #define __HIPSTDPAR_INTERPOSE_ALLOC__ 1
// CHECK-HIPSTDPAR-INTERPOSE: #define __HIPSTDPAR__ 1
@@ -316,4 +318,5 @@
// RUN: -triple amdgcn-amd-amdhsa -fcuda-is-device | FileCheck -match-full-lines \
// RUN: %s --check-prefix=CHECK-HIPSTDPAR-INTERPOSE-DEV-NEG
// CHECK-HIPSTDPAR-INTERPOSE-DEV-NEG: #define __HIPSTDPAR__ 1
+// CHECK-HIPSTDPAR-INTERPOSE-DEV-NEG-NOT: #define __HIPSTDPAR_INTERPOSE_ALLOC_V1__ 1
// CHECK-HIPSTDPAR-INTERPOSE-DEV-NEG-NOT: #define __HIPSTDPAR_INTERPOSE_ALLOC__ 1
diff --git a/llvm/lib/Transforms/HipStdPar/HipStdPar.cpp b/llvm/lib/Transforms/HipStdPar/HipStdPar.cpp
index 40164a34f08ac..03a65152e9833 100644
--- a/llvm/lib/Transforms/HipStdPar/HipStdPar.cpp
+++ b/llvm/lib/Transforms/HipStdPar/HipStdPar.cpp
@@ -236,6 +236,8 @@ static constexpr std::pair<StringLiteral, StringLiteral> ReplaceMap[]{
{"free", "__hipstdpar_free"},
{"malloc", "__hipstdpar_malloc"},
{"memalign", "__hipstdpar_aligned_alloc"},
+ {"mmap", "__hipstdpar_mmap"},
+ {"munmap", "__hipstdpar_munmap"},
{"posix_memalign", "__hipstdpar_posix_aligned_alloc"},
{"realloc", "__hipstdpar_realloc"},
{"reallocarray", "__hipstdpar_realloc_array"},
@@ -271,6 +273,16 @@ static constexpr std::pair<StringLiteral, StringLiteral> ReplaceMap[]{
{"__libc_realloc", "__hipstdpar_realloc"}
};
+static constexpr std::pair<StringLiteral, StringLiteral> HiddenMap[]{
+ // hidden_malloc and hidden_free are only kept for backwards compatibility /
+ // legacy purposes, and we should remove them in the future
+ {"__hipstdpar_hidden_malloc", "__libc_malloc"},
+ {"__hipstdpar_hidden_free", "__libc_free"},
+ {"__hipstdpar_hidden_memalign", "__libc_memalign"},
+ {"__hipstdpar_hidden_mmap", "mmap"},
+ {"__hipstdpar_hidden_munmap", "munmap"}
+};
+
PreservedAnalyses
HipStdParAllocationInterpositionPass::run(Module &M, ModuleAnalysisManager&) {
SmallDenseMap<StringRef, StringRef> AllocReplacements(std::cbegin(ReplaceMap),
@@ -299,19 +311,14 @@ HipStdParAllocationInterpositionPass::run(Module &M, ModuleAnalysisManager&) {
}
}
- if (auto F = M.getFunction("__hipstdpar_hidden_malloc")) {
- auto LibcMalloc = M.getOrInsertFunction(
- "__libc_malloc", F->getFunctionType(), F->getAttributes());
- F->replaceAllUsesWith(LibcMalloc.getCallee());
+ for (auto &&HR : HiddenMap) {
+ if (auto F = M.getFunction(HR.first)) {
+ auto R = M.getOrInsertFunction(HR.second, F->getFunctionType(),
+ F->getAttributes());
+ F->replaceAllUsesWith(R.getCallee());
- eraseFromModule(*F);
- }
- if (auto F = M.getFunction("__hipstdpar_hidden_free")) {
- auto LibcFree = M.getOrInsertFunction("__libc_free", F->getFunctionType(),
- F->getAttributes());
- F->replaceAllUsesWith(LibcFree.getCallee());
-
- eraseFromModule(*F);
+ eraseFromModule(*F);
+ }
}
return PreservedAnalyses::none();
diff --git a/llvm/test/Transforms/HipStdPar/allocation-interposition.ll b/llvm/test/Transforms/HipStdPar/allocation-interposition.ll
index 9ec284b1dedb7..bdc9951d0252c 100644
--- a/llvm/test/Transforms/HipStdPar/allocation-interposition.ll
+++ b/llvm/test/Transforms/HipStdPar/allocation-interposition.ll
@@ -16,6 +16,16 @@ declare void @__hipstdpar_hidden_free(ptr)
declare ptr @__hipstdpar_hidden_malloc(i64)
+declare ptr @__hipstdpar_hidden_memalign(i64, i64)
+
+declare ptr @__hipstdpar_hidden_mmap(ptr, i64, i32, i32, i32, i64)
+
+declare i32 @__hipstdpar_hidden_munmap(ptr, i64)
+
+declare ptr @__hipstdpar_mmap(ptr, i64, i32, i32, i32, i64)
+
+declare i32 @__hipstdpar_munmap(ptr, i64)
+
declare ptr @__hipstdpar_realloc(ptr, i64)
declare ptr @__hipstdpar_realloc_array(ptr, i64, i64)
@@ -171,7 +181,21 @@ define dso_local noundef i32 @allocs() {
; CHECK: call void @__hipstdpar_free(ptr noundef %28)
call void @__libc_free(ptr noundef %28)
- ret i32 0
+ ; CHECK: %29 = call ptr @__libc_malloc(i64 noundef 8)
+ %29 = call ptr @__hipstdpar_hidden_malloc(i64 noundef 8)
+ ; CHECK: call void @__libc_free(ptr noundef %29)
+ call void @__hipstdpar_hidden_free(ptr noundef %29)
+
+ ; CHECK: %30 = call ptr @__libc_memalign(i64 noundef 8, i64 noundef 4)
+ %30 = call ptr @__hipstdpar_hidden_memalign(i64 noundef 8, i64 noundef 4)
+ ; CHECK: %31 = call ptr @mmap(ptr %30, i64 8, i32 0, i32 0, i32 0, i64 0)
+ %31 = call ptr @__hipstdpar_hidden_mmap(ptr %30, i64 8, i32 0, i32 0, i32 0, i64 0)
+ ; CHECK: %32 = call i32 @munmap(ptr %31, i64 8)
+ %32 = call i32 @__hipstdpar_hidden_munmap(ptr %31, i64 8)
+ ; CHECK: call void @__libc_free(ptr noundef %30)
+ call void @__hipstdpar_hidden_free(ptr noundef %30)
+
+ ret i32 %32
}
declare noalias ptr @aligned_alloc(i64 noundef, i64 noundef)
@@ -220,4 +244,8 @@ declare void @__libc_free(ptr noundef)
declare ptr @__libc_malloc(i64 noundef)
-declare ptr @__libc_memalign(i64 noundef, i64 noundef)
\ No newline at end of file
+declare ptr @__libc_memalign(i64 noundef, i64 noundef)
+
+declare ptr @mmap(ptr noundef, i64 noundef, i32 noundef, i32 noundef, i32 noundef, i64 noundef)
+
+declare i32 @munmap(ptr noundef, i64 noundef)
|
|
✅ With the latest revision this PR passed the C/C++ code formatter. |
yxsamliu
approved these changes
May 7, 2025
|
LLVM Buildbot has detected a new failure on builder Full details are available at: https://lab.llvm.org/buildbot/#/builders/162/builds/21807 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.
The allocation interposition mode had a number of issues, which are primarily addressed in the library component via ROCm/rocThrust#543. However, it is necessary to interpose some additional symbols, which this patch does. Furthermore, to implement this in a compatible way, we guard the new implementation under a V1 macro, which is defined in addition to the existing
__HIPSTDPAR_INTERPOSE_ALLOC__one.