[Clang]Throw frontend error for target feature mismatch when using `flatten` attribute #150044

abhishek-kaushik22 · 2025-07-22T15:35:15Z

Fixes #149866

llvmbot · 2025-07-22T15:35:49Z

@llvm/pr-subscribers-clang-codegen

@llvm/pr-subscribers-clang

Author: Abhishek Kaushik (abhishek-kaushik22)

Changes

Fixes #149866

Full diff: https://github.com/llvm/llvm-project/pull/150044.diff

2 Files Affected:

(modified) clang/lib/CodeGen/CGCall.cpp (+5-3)
(added) clang/test/CodeGen/target-features-error-3.c (+12)

diff --git a/clang/lib/CodeGen/CGCall.cpp b/clang/lib/CodeGen/CGCall.cpp
index 0bceecec6e555..589716460b2de 100644
--- a/clang/lib/CodeGen/CGCall.cpp
+++ b/clang/lib/CodeGen/CGCall.cpp
@@ -5232,9 +5232,11 @@ RValue CodeGenFunction::EmitCall(const CGFunctionInfo &CallInfo,
     // since otherwise we could be making a conditional call after a check for
     // the proper cpu features (and it won't cause code generation issues due to
     // function based code generation).
-    if (TargetDecl->hasAttr<AlwaysInlineAttr>() &&
-        (TargetDecl->hasAttr<TargetAttr>() ||
-         (CurFuncDecl && CurFuncDecl->hasAttr<TargetAttr>())))
+    if ((TargetDecl->hasAttr<AlwaysInlineAttr>() &&
+         (TargetDecl->hasAttr<TargetAttr>() ||
+          (CurFuncDecl && CurFuncDecl->hasAttr<TargetAttr>()))) ||
+        (CurFuncDecl && CurFuncDecl->hasAttr<FlattenAttr>() &&
+         TargetDecl->hasAttr<TargetAttr>()))
       checkTargetFeatures(Loc, FD);
   }
 
diff --git a/clang/test/CodeGen/target-features-error-3.c b/clang/test/CodeGen/target-features-error-3.c
new file mode 100644
index 0000000000000..8900edfa257c7
--- /dev/null
+++ b/clang/test/CodeGen/target-features-error-3.c
@@ -0,0 +1,12 @@
+// RUN: %clang_cc1 %s -triple=x86_64-linux-gnu -S -verify -o -
+
+typedef double __v2df __attribute__((__vector_size__(16)));
+
+__v2df __attribute__((target("sse4.1"))) foo() {
+    __v2df v = {0.0, 0.0};
+    return __builtin_ia32_roundpd(v, 2);
+}
+
+__v2df __attribute__((flatten)) bar() {
+    return foo(); // expected-error {{always_inline function 'foo' requires target feature 'sse4.1', but would be inlined into function 'bar' that is compiled without support for 'sse4.1'}}
+}

abhishek-kaushik22 · 2025-07-22T15:44:50Z

@topperc can you please review since you've reviewed similar commits in the past?

clang/lib/CodeGen/CGCall.cpp

clang/test/CodeGen/target-features-error-3.c

abhishek-kaushik22 · 2025-08-10T13:10:24Z

@efriedma-quic can you please review again?

abhishek-kaushik22 · 2025-08-17T10:24:07Z

Ping!

…llvm#153217) I was looking over this pass and noticed it was using shared pointers for CompositeNodes. However, all nodes are owned by the deinterleaving graph and are not released until the graph is destroyed. This means a bump allocator and raw pointers can be used, which have a simpler ownership model and less overhead than shared pointers. The changes in this PR are to: - Add a `SpecificBumpPtrAllocator<CompositeNode>` to the `ComplexDeinterleavingGraph` - This allocates new nodes and will deallocate them when the graph is destroyed - Replace `NodePtr` and `RawNodePtr` with `CompositeNode *`

In the AGPR case this will be an 8 byte instruction, which is part of why this case is a pain to deal with in the first place.

…#137299) Using early returns tends to make the code easier to read, without any changes to the generated code.

This change adds support for MemberExpr with VarDecl ComplexType Issue: llvm#141365

…m#154484) Fixes llvm#154312

This patch handle struct of fixed vector and struct of array of fixed vector correctly for VLS calling convention in EmitFunctionProlog, EmitFunctionEpilog and EmitCall. stack on: llvm#147173

The previous version of the code would prevent the use of the compiler builtins.

…de (llvm#154305) In streaming mode, both the @llvm.aarch64.sme.cnts and @llvm.aarch64.sve.cnt intrinsics are equivalent. For SVE, cnt* is lowered in instCombineIntrinsic to @llvm.sme.vscale(). This patch lowers the SME intrinsic similarly when in streaming-mode.

…154452) This is a cherry pick of llvm#154053 with a fix for bad handling of endianess when loading float and double litteral from the binary. --------- Co-authored-by: Ferdinand Lemaire <ferdinand.lemaire@woven-planet.global> Co-authored-by: Jessica Paquette <jessica.paquette@woven-planet.global> Co-authored-by: Luc Forget <luc.forget@woven.toyota>

… -mattr argument. NFC. The unnecessary quotes prevents DOS scripts from executing the RUN line.

/llvm-project/clang/lib/CodeGen/Targets/RISCV.cpp:865:9: error: unused variable 'FixedSrcTy' [-Werror,-Wunused-variable] auto *FixedSrcTy = cast<llvm::FixedVectorType>(SrcTy); ^ 1 error generated.

…. NFC

…vm#154326) A couple of minor readability changes now that we're not supporting both intrinsics and records.

Add trailing newlines to the following files to comply with POSIX standards: - llvm/lib/Target/RISCV/RISCVInstrInfoXSpacemiT.td - llvm/test/MC/RISCV/xsmtvdot-invalid.s - llvm/test/MC/RISCV/xsmtvdot-valid.s Closes llvm#151706

…context" (llvm#154458) The immediate evaluation context needs the lambda scope info to propagate some flags, however that LSI was removed in ActOnFinishFunctionBody which happened before rebuilding a lambda expression. The last attempt destroyed LSI at the end of the block scope, after which we still need it in DiagnoseShadowingLambdaDecls. This also converts the wrapper function to default arguments as a drive-by fix, as well as does some cleanup. Fixes llvm#145776

…lvm#154340) Adds legalization support for `G_TRUNC_SSAT_S`, `G_TRUNC_SSAT_S`, `G_TRUNC_USAT_U` instructions for GlobalISel.

Whether the argument is fixed is now available via ArgFlags, so make use of it. The previous implementation was quite problematic, because it stored the number of fixed arguments in a global variable, which is not thread safe.

…4289) If FunctionAttrs infers additional attributes on a function, it also invalidates analysis on callers of that function. The way it does this right now limits this to calls with matching signature. However, the function attributes will also be used when the signatures do not match. Use getCalledOperand() to avoid a signature check. This is not a correctness fix, just improves analysis quality. I noticed this due to llvm#144497 (comment), where LICM ends up with a stale MemoryDef that could be a MemoryUse (which is a bug in LICM, but still non-optimal).

…ection by name only (llvm#154159) The AArch64 build attribute specification now allows switching to an already-defined subsection using its name alone, without repeating the optionality and type parameters. This patch updates the parser to support that behavior. Spec reference: https://github.com/ARM-software/abi-aa/pull/230/files

…ilability. (llvm#154143) The FEAT_SVE_B16B16 arithmetic instructions are only available to streaming mode functions when SME2 is available.

Co-authored-by: Aaron Ballman <aaron@aaronballman.com>

…roject into 149866

abhishek-kaushik22 · 2025-08-21T17:20:35Z

Sorry for this I don't know what happened here, I just did a rebase from main :(

llvmbot added clang Clang issues not falling into any other category clang:codegen IR generation bugs: mangling, exceptions, etc. labels Jul 22, 2025

abhishek-kaushik22 requested a review from topperc July 22, 2025 15:44

efriedma-quic reviewed Jul 22, 2025

View reviewed changes

clang/lib/CodeGen/CGCall.cpp Outdated Show resolved Hide resolved

clang/test/CodeGen/target-features-error-3.c Outdated Show resolved Hide resolved

abhishek-kaushik22 requested a review from efriedma-quic July 23, 2025 20:35

llvmbot added the clang:frontend Language frontend issues, e.g. anything involving "Sema" label Aug 10, 2025

MacDue and others added 20 commits August 20, 2025 08:59

AMDGPU: Correct inst size for av_mov_b32_imm_pseudo (llvm#154459)

0fa6fdf

In the AGPR case this will be an 8 byte instruction, which is part of why this case is a pain to deal with in the first place.

[libc++][NFC] Use early returns in a few basic_string functions (llvm…

2dc0a5f

…#137299) Using early returns tends to make the code easier to read, without any changes to the generated code.

[CIR] Implement MemberExpr with VarDecl for ComplexType (llvm#154307)

018c5ba

This change adds support for MemberExpr with VarDecl ComplexType Issue: llvm#141365

[clang][bytecode] Diagnose one-past-end reads from global arrays (llv…

e16ced3

…m#154484) Fixes llvm#154312

[clang][RISCV] Fix crash on VLS calling convention (llvm#145489)

52a2e68

This patch handle struct of fixed vector and struct of array of fixed vector correctly for VLS calling convention in EmitFunctionProlog, EmitFunctionEpilog and EmitCall. stack on: llvm#147173

[libc] Fix constexpr add_with_carry/sub_with_borrow (llvm#154282)

4dd9e99

The previous version of the code would prevent the use of the compiler builtins.

[RISCV] incorrect-extract-subvector-combine.ll - remove quotes around…

035f40e

… -mattr argument. NFC. The unnecessary quotes prevents DOS scripts from executing the RUN line.

[RISCV] Silent a warning (NFC)

80bc38b

/llvm-project/clang/lib/CodeGen/Targets/RISCV.cpp:865:9: error: unused variable 'FixedSrcTy' [-Werror,-Wunused-variable] auto *FixedSrcTy = cast<llvm::FixedVectorType>(SrcTy); ^ 1 error generated.

[OMPIRBuilder] Use CreateNUWMul instead of passing flags to CreateMul…

1741358

…. NFC

[RemoveDIs][NFC] Clean up BasicBlockUtils now intrinsics are gone (ll…

6c93525

…vm#154326) A couple of minor readability changes now that we're not supporting both intrinsics and records.

[RISCV][NFC] Ensure files end with newline. (llvm#154457)

46e77eb

Add trailing newlines to the following files to comply with POSIX standards: - llvm/lib/Target/RISCV/RISCVInstrInfoXSpacemiT.td - llvm/test/MC/RISCV/xsmtvdot-invalid.s - llvm/test/MC/RISCV/xsmtvdot-valid.s Closes llvm#151706

[GlobalISel] Legalize Saturated Truncate instructions and intrinsics (l…

9df7ca1

…lvm#154340) Adds legalization support for `G_TRUNC_SSAT_S`, `G_TRUNC_SSAT_S`, `G_TRUNC_USAT_U` instructions for GlobalISel.

[LLVM][CodeGen][SME] hasB16b16() is not sufficient to prove BFADD ava…

d6a688f

…ilability. (llvm#154143) The FEAT_SVE_B16B16 arithmetic instructions are only available to streaming mode functions when SME2 is available.

abhishek-kaushik22 and others added 4 commits August 21, 2025 22:46

Apply suggestions from code review

98e1459

Co-authored-by: Aaron Ballman <aaron@aaronballman.com>

Add non-failing test

53fae7a

Update ReleaseNotes.rst

6171b73

Merge branch '149866' of https://github.com/abhishek-kaushik22/llvm-p…

49735f1

…roject into 149866

abhishek-kaushik22 requested review from DeinAlptraum, aaupov, maksfb, rafaelauler, ayermolo, yota9, paschalis-mpeis, yozhu, JDevlieghere, lanza, bcardosolopes, xlauko, andykaylor, Endilll, nikic and a team as code owners August 21, 2025 17:18

abhishek-kaushik22 closed this Aug 21, 2025

abhishek-kaushik22 deleted the 149866 branch August 21, 2025 17:21

Endilll removed request for a team, Endilll and yozhu August 21, 2025 18:13

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Uh oh!

[Clang]Throw frontend error for target feature mismatch when using `flatten` attribute #150044

[Clang]Throw frontend error for target feature mismatch when using `flatten` attribute #150044

abhishek-kaushik22 commented Jul 22, 2025

Uh oh!

llvmbot commented Jul 22, 2025 •

edited

Loading

Uh oh!

abhishek-kaushik22 commented Jul 22, 2025 •

edited

Loading

Uh oh!

Uh oh!

Uh oh!

abhishek-kaushik22 commented Aug 10, 2025

Uh oh!

abhishek-kaushik22 commented Aug 17, 2025

Uh oh!

abhishek-kaushik22 commented Aug 21, 2025 •

edited

Loading

Uh oh!

Uh oh!

[Clang]Throw frontend error for target feature mismatch when using flatten attribute #150044

[Clang]Throw frontend error for target feature mismatch when using flatten attribute #150044

Conversation

abhishek-kaushik22 commented Jul 22, 2025

Uh oh!

llvmbot commented Jul 22, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

abhishek-kaushik22 commented Jul 22, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

Uh oh!

Uh oh!

abhishek-kaushik22 commented Aug 10, 2025

Uh oh!

abhishek-kaushik22 commented Aug 17, 2025

Uh oh!

abhishek-kaushik22 commented Aug 21, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

Uh oh!

[Clang]Throw frontend error for target feature mismatch when using `flatten` attribute #150044

[Clang]Throw frontend error for target feature mismatch when using `flatten` attribute #150044

llvmbot commented Jul 22, 2025 •

edited

Loading

abhishek-kaushik22 commented Jul 22, 2025 •

edited

Loading

abhishek-kaushik22 commented Aug 21, 2025 •

edited

Loading