[clang][AArch64][FMV] Stop emitting alias to ifunc. #96221
Merged
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
Long story short the interaction of two optimizations happening in GlobalOpt results in a crash. For more details look at the issue llvm#96197. I will be fixing this in GlobalOpt but it is a conservative solution since it won't allow us to optimize resolvers which return a pointer to a function whose definition is in another TU when compiling without LTO: __attribute__((target_version("simd"))) void bar(void); __attribute__((target_version("default"))) void bar(void); int foo() { bar(); }
llvmbot
added
clang
|
@llvm/pr-subscribers-clang-codegen @llvm/pr-subscribers-clang Author: Alexandros Lamprineas (labrinea) ChangesLong story short the interaction of two optimizations happening in GlobalOpt results in a crash. For more details look at the issue #96197. I will be fixing this in GlobalOpt but it is a conservative solution since it won't allow us to optimize resolvers which return a pointer to a function whose definition is in another TU when compiling without LTO: attribute((target_version("simd"))) void bar(void); attribute((target_version("default"))) void bar(void); int foo() { bar(); } Full diff: https://github.com/llvm/llvm-project/pull/96221.diff 7 Files Affected:
diff --git a/clang/lib/CodeGen/CodeGenModule.cpp b/clang/lib/CodeGen/CodeGenModule.cpp
index dd4a665ebc78b..76534475e88f7 100644
--- a/clang/lib/CodeGen/CodeGenModule.cpp
+++ b/clang/lib/CodeGen/CodeGenModule.cpp
@@ -4259,8 +4259,8 @@ void CodeGenModule::emitMultiVersionFunctions() {
llvm::Constant *ResolverConstant = GetOrCreateMultiVersionResolver(GD);
if (auto *IFunc = dyn_cast<llvm::GlobalIFunc>(ResolverConstant)) {
ResolverConstant = IFunc->getResolver();
- if (FD->isTargetClonesMultiVersion() ||
- FD->isTargetVersionMultiVersion()) {
+ if (FD->isTargetClonesMultiVersion() &&
+ !getTarget().getTriple().isAArch64()) {
std::string MangledName = getMangledNameImpl(
*this, GD, FD, /*OmitMultiVersionMangling=*/true);
if (!GetGlobalValue(MangledName + ".ifunc")) {
diff --git a/clang/test/CodeGen/aarch64-mixed-target-attributes.c b/clang/test/CodeGen/aarch64-mixed-target-attributes.c
index 6aa747d4cb461..3c047fec6ceed 100644
--- a/clang/test/CodeGen/aarch64-mixed-target-attributes.c
+++ b/clang/test/CodeGen/aarch64-mixed-target-attributes.c
@@ -30,9 +30,6 @@ __attribute__((target_version("jscvt"))) int default_def_with_version_decls(void
//.
// CHECK: @__aarch64_cpu_features = external dso_local global { i64 }
-// CHECK: @explicit_default.ifunc = weak_odr alias i32 (), ptr @explicit_default
-// CHECK: @implicit_default.ifunc = weak_odr alias i32 (), ptr @implicit_default
-// CHECK: @default_def_with_version_decls.ifunc = weak_odr alias i32 (), ptr @default_def_with_version_decls
// CHECK: @explicit_default = weak_odr ifunc i32 (), ptr @explicit_default.resolver
// CHECK: @implicit_default = weak_odr ifunc i32 (), ptr @implicit_default.resolver
// CHECK: @default_def_with_version_decls = weak_odr ifunc i32 (), ptr @default_def_with_version_decls.resolver
diff --git a/clang/test/CodeGen/attr-target-clones-aarch64.c b/clang/test/CodeGen/attr-target-clones-aarch64.c
index ad6079a91fcd5..60f9c7f1fc24e 100644
--- a/clang/test/CodeGen/attr-target-clones-aarch64.c
+++ b/clang/test/CodeGen/attr-target-clones-aarch64.c
@@ -27,14 +27,6 @@ inline int __attribute__((target_clones("fp16", "sve2-bitperm+fcma", "default"))
//.
// CHECK: @__aarch64_cpu_features = external dso_local global { i64 }
-// CHECK: @ftc.ifunc = weak_odr alias i32 (), ptr @ftc
-// CHECK: @ftc_def.ifunc = weak_odr alias i32 (), ptr @ftc_def
-// CHECK: @ftc_dup1.ifunc = weak_odr alias i32 (), ptr @ftc_dup1
-// CHECK: @ftc_dup2.ifunc = weak_odr alias i32 (), ptr @ftc_dup2
-// CHECK: @ftc_dup3.ifunc = weak_odr alias i32 (), ptr @ftc_dup3
-// CHECK: @ftc_inline2.ifunc = weak_odr alias i32 (), ptr @ftc_inline2
-// CHECK: @ftc_inline1.ifunc = weak_odr alias i32 (), ptr @ftc_inline1
-// CHECK: @ftc_inline3.ifunc = weak_odr alias i32 (), ptr @ftc_inline3
// CHECK: @ftc = weak_odr ifunc i32 (), ptr @ftc.resolver
// CHECK: @ftc_def = weak_odr ifunc i32 (), ptr @ftc_def.resolver
// CHECK: @ftc_dup1 = weak_odr ifunc i32 (), ptr @ftc_dup1.resolver
@@ -45,14 +37,6 @@ inline int __attribute__((target_clones("fp16", "sve2-bitperm+fcma", "default"))
// CHECK: @ftc_inline3 = weak_odr ifunc i32 (), ptr @ftc_inline3.resolver
//.
// CHECK-MTE-BTI: @__aarch64_cpu_features = external dso_local global { i64 }
-// CHECK-MTE-BTI: @ftc.ifunc = weak_odr alias i32 (), ptr @ftc
-// CHECK-MTE-BTI: @ftc_def.ifunc = weak_odr alias i32 (), ptr @ftc_def
-// CHECK-MTE-BTI: @ftc_dup1.ifunc = weak_odr alias i32 (), ptr @ftc_dup1
-// CHECK-MTE-BTI: @ftc_dup2.ifunc = weak_odr alias i32 (), ptr @ftc_dup2
-// CHECK-MTE-BTI: @ftc_dup3.ifunc = weak_odr alias i32 (), ptr @ftc_dup3
-// CHECK-MTE-BTI: @ftc_inline2.ifunc = weak_odr alias i32 (), ptr @ftc_inline2
-// CHECK-MTE-BTI: @ftc_inline1.ifunc = weak_odr alias i32 (), ptr @ftc_inline1
-// CHECK-MTE-BTI: @ftc_inline3.ifunc = weak_odr alias i32 (), ptr @ftc_inline3
// CHECK-MTE-BTI: @ftc = weak_odr ifunc i32 (), ptr @ftc.resolver
// CHECK-MTE-BTI: @ftc_def = weak_odr ifunc i32 (), ptr @ftc_def.resolver
// CHECK-MTE-BTI: @ftc_dup1 = weak_odr ifunc i32 (), ptr @ftc_dup1.resolver
diff --git a/clang/test/CodeGen/attr-target-version.c b/clang/test/CodeGen/attr-target-version.c
index 75f8734e5aaf3..024aafffca629 100644
--- a/clang/test/CodeGen/attr-target-version.c
+++ b/clang/test/CodeGen/attr-target-version.c
@@ -137,19 +137,6 @@ int caller(void) { return used_def_without_default_decl() + used_decl_without_de
//.
// CHECK: @__aarch64_cpu_features = external dso_local global { i64 }
-// CHECK: @fmv.ifunc = weak_odr alias i32 (), ptr @fmv
-// CHECK: @fmv_one.ifunc = weak_odr alias i32 (), ptr @fmv_one
-// CHECK: @fmv_two.ifunc = weak_odr alias i32 (), ptr @fmv_two
-// CHECK: @fmv_e.ifunc = weak_odr alias i32 (), ptr @fmv_e
-// CHECK: @fmv_d.ifunc = internal alias i32 (), ptr @fmv_d
-// CHECK: @fmv_c.ifunc = weak_odr alias void (), ptr @fmv_c
-// CHECK: @fmv_inline.ifunc = weak_odr alias i32 (), ptr @fmv_inline
-// CHECK: @unused_with_default_def.ifunc = weak_odr alias i32 (), ptr @unused_with_default_def
-// CHECK: @unused_with_implicit_default_def.ifunc = weak_odr alias i32 (), ptr @unused_with_implicit_default_def
-// CHECK: @unused_with_implicit_forward_default_def.ifunc = weak_odr alias i32 (), ptr @unused_with_implicit_forward_default_def
-// CHECK: @default_def_with_version_decls.ifunc = weak_odr alias i32 (), ptr @default_def_with_version_decls
-// CHECK: @used_def_without_default_decl.ifunc = weak_odr alias i32 (), ptr @used_def_without_default_decl
-// CHECK: @used_decl_without_default_decl.ifunc = weak_odr alias i32 (), ptr @used_decl_without_default_decl
// CHECK: @fmv = weak_odr ifunc i32 (), ptr @fmv.resolver
// CHECK: @fmv_one = weak_odr ifunc i32 (), ptr @fmv_one.resolver
// CHECK: @fmv_two = weak_odr ifunc i32 (), ptr @fmv_two.resolver
diff --git a/clang/test/CodeGenCXX/attr-target-clones-aarch64.cpp b/clang/test/CodeGenCXX/attr-target-clones-aarch64.cpp
index 7953f902bf09b..29ae6b6856500 100644
--- a/clang/test/CodeGenCXX/attr-target-clones-aarch64.cpp
+++ b/clang/test/CodeGenCXX/attr-target-clones-aarch64.cpp
@@ -40,10 +40,6 @@ void run_foo_tml() {
//.
// CHECK: @__aarch64_cpu_features = external dso_local global { i64 }
-// CHECK: @_Z7foo_ovli.ifunc = weak_odr alias i32 (i32), ptr @_Z7foo_ovli
-// CHECK: @_Z7foo_ovlv.ifunc = weak_odr alias i32 (), ptr @_Z7foo_ovlv
-// CHECK: @_ZN7MyClassIssE7foo_tmlEv.ifunc = weak_odr alias i32 (ptr), ptr @_ZN7MyClassIssE7foo_tmlEv
-// CHECK: @_ZN7MyClassIisE7foo_tmlEv.ifunc = weak_odr alias i32 (ptr), ptr @_ZN7MyClassIisE7foo_tmlEv
// CHECK: @_Z7foo_ovli = weak_odr ifunc i32 (i32), ptr @_Z7foo_ovli.resolver
// CHECK: @_Z7foo_ovlv = weak_odr ifunc i32 (), ptr @_Z7foo_ovlv.resolver
// CHECK: @_ZN7MyClassIssE7foo_tmlEv = weak_odr ifunc i32 (ptr), ptr @_ZN7MyClassIssE7foo_tmlEv.resolver
diff --git a/clang/test/CodeGenCXX/attr-target-version.cpp b/clang/test/CodeGenCXX/attr-target-version.cpp
index 8b7273fe3bb51..fd19f4c5a3030 100644
--- a/clang/test/CodeGenCXX/attr-target-version.cpp
+++ b/clang/test/CodeGenCXX/attr-target-version.cpp
@@ -62,12 +62,6 @@ int bar() {
//.
// CHECK: @__aarch64_cpu_features = external dso_local global { i64 }
-// CHECK: @_Z3fooi.ifunc = weak_odr alias i32 (i32), ptr @_Z3fooi
-// CHECK: @_Z3foov.ifunc = weak_odr alias i32 (), ptr @_Z3foov
-// CHECK: @_ZN7MyClass3gooEi.ifunc = weak_odr alias i32 (ptr, i32), ptr @_ZN7MyClass3gooEi
-// CHECK: @_ZN7MyClass23unused_with_default_defEv.ifunc = weak_odr alias i32 (ptr), ptr @_ZN7MyClass23unused_with_default_defEv
-// CHECK: @_ZN7MyClass32unused_with_implicit_default_defEv.ifunc = weak_odr alias i32 (ptr), ptr @_ZN7MyClass32unused_with_implicit_default_defEv
-// CHECK: @_ZN7MyClass40unused_with_implicit_forward_default_defEv.ifunc = weak_odr alias i32 (ptr), ptr @_ZN7MyClass40unused_with_implicit_forward_default_defEv
// CHECK: @_ZN7MyClass3gooEi = weak_odr ifunc i32 (ptr, i32), ptr @_ZN7MyClass3gooEi.resolver
// CHECK: @_Z3fooi = weak_odr ifunc i32 (i32), ptr @_Z3fooi.resolver
// CHECK: @_Z3foov = weak_odr ifunc i32 (), ptr @_Z3foov.resolver
diff --git a/clang/test/CodeGenCXX/fmv-namespace.cpp b/clang/test/CodeGenCXX/fmv-namespace.cpp
index 5bcd0da06eebc..193f01db4c5d3 100644
--- a/clang/test/CodeGenCXX/fmv-namespace.cpp
+++ b/clang/test/CodeGenCXX/fmv-namespace.cpp
@@ -19,8 +19,6 @@ int baz() { return OtherName::foo(); }
//.
// CHECK: @__aarch64_cpu_features = external dso_local global { i64 }
-// CHECK: @_ZN4Name3fooEv.ifunc = weak_odr alias i32 (), ptr @_ZN4Name3fooEv
-// CHECK: @_ZN9OtherName3fooEv.ifunc = weak_odr alias i32 (), ptr @_ZN9OtherName3fooEv
// CHECK: @_ZN4Name3fooEv = weak_odr ifunc i32 (), ptr @_ZN4Name3fooEv.resolver
// CHECK: @_ZN9OtherName3fooEv = weak_odr ifunc i32 (), ptr @_ZN9OtherName3fooEv.resolver
//.
|
|
cc @tahonermann |
|
I think this needs a release note. |
DanielKristofKiss
approved these changes
Jun 20, 2024
We removed the global alias that was pointing to AArch64 Function Multiversioning ifuncs because it interacts badly with GlobalOpt.
AlexisPerry
pushed a commit
to llvm-project-tlp/llvm-project
that referenced
this pull request
Jul 9, 2024
Long story short the interaction of two optimizations happening in GlobalOpt results in a crash. For more details look at the issue llvm#96197. I will be fixing this in GlobalOpt but it is a conservative solution since it won't allow us to optimize resolvers which return a pointer to a function whose definition is in another TU when compiling without LTO: ``` __attribute__((target_version("simd"))) void bar(void); __attribute__((target_version("default"))) void bar(void); int foo() { bar(); } ``` fixes: llvm#96197
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.
Long story short the interaction of two optimizations happening in GlobalOpt results in a crash. For more details look at the issue #96197. I will be fixing this in GlobalOpt but it is a conservative solution since it won't allow us to optimize resolvers which return a pointer to a function whose definition is in another TU when compiling without LTO:
fixes: #96197