[clang] [ARM64EC] Add support for parsing __vectorcall (PR #87725)

via cfe-commits cfe-commits at lists.llvm.org
Thu Apr 4 16:19:56 PDT 2024


llvmbot wrote:


<!--LLVM PR SUMMARY COMMENT-->

@llvm/pr-subscribers-backend-aarch64

Author: Eli Friedman (efriedma-quic)

<details>
<summary>Changes</summary>

MSVC doesn't support generating __vectorcall calls in Arm64EC mode, but it does treat it as a distinct type.  The Microsoft STL depends on this functionality.  (Not sure if this is intentional.)  Add support for parsing the same way as MSVC, and add some checks to ensure we don't try to actually generate code.

The error handling in CodeGen is ugly, but I can't think of a better way to do it.

---
Full diff: https://github.com/llvm/llvm-project/pull/87725.diff


4 Files Affected:

- (modified) clang/lib/Basic/Targets/AArch64.cpp (+4-1) 
- (modified) clang/lib/CodeGen/CGCall.cpp (+6) 
- (modified) clang/lib/CodeGen/CodeGenModule.cpp (+8) 
- (added) clang/test/CodeGenCXX/arm64ec-vectorcall.cpp (+14) 


``````````diff
diff --git a/clang/lib/Basic/Targets/AArch64.cpp b/clang/lib/Basic/Targets/AArch64.cpp
index 1569b5e04b770a..c8d243a8fb7aea 100644
--- a/clang/lib/Basic/Targets/AArch64.cpp
+++ b/clang/lib/Basic/Targets/AArch64.cpp
@@ -1543,10 +1543,13 @@ WindowsARM64TargetInfo::getBuiltinVaListKind() const {
 TargetInfo::CallingConvCheckResult
 WindowsARM64TargetInfo::checkCallingConvention(CallingConv CC) const {
   switch (CC) {
+  case CC_X86VectorCall:
+    if (getTriple().isWindowsArm64EC())
+      return CCCR_OK;
+    return CCCR_Ignore;
   case CC_X86StdCall:
   case CC_X86ThisCall:
   case CC_X86FastCall:
-  case CC_X86VectorCall:
     return CCCR_Ignore;
   case CC_C:
   case CC_OpenCLKernel:
diff --git a/clang/lib/CodeGen/CGCall.cpp b/clang/lib/CodeGen/CGCall.cpp
index f12765b826935b..3f5463a9a70e9d 100644
--- a/clang/lib/CodeGen/CGCall.cpp
+++ b/clang/lib/CodeGen/CGCall.cpp
@@ -5591,6 +5591,12 @@ RValue CodeGenFunction::EmitCall(const CGFunctionInfo &CallInfo,
                              /*AttrOnCallSite=*/true,
                              /*IsThunk=*/false);
 
+  if (CallingConv == llvm::CallingConv::X86_VectorCall &&
+      getTarget().getTriple().isWindowsArm64EC()) {
+    CGM.Error(Loc, "__vectorcall calling convention is not currently "
+                   "supported");
+  }
+
   if (const FunctionDecl *FD = dyn_cast_or_null<FunctionDecl>(CurFuncDecl)) {
     if (FD->hasAttr<StrictFPAttr>())
       // All calls within a strictfp function are marked strictfp
diff --git a/clang/lib/CodeGen/CodeGenModule.cpp b/clang/lib/CodeGen/CodeGenModule.cpp
index 00b3bfcaa0bc25..75519be8bba052 100644
--- a/clang/lib/CodeGen/CodeGenModule.cpp
+++ b/clang/lib/CodeGen/CodeGenModule.cpp
@@ -2087,6 +2087,14 @@ void CodeGenModule::SetLLVMFunctionAttributes(GlobalDecl GD,
   llvm::AttributeList PAL;
   ConstructAttributeList(F->getName(), Info, GD, PAL, CallingConv,
                          /*AttrOnCallSite=*/false, IsThunk);
+  if (CallingConv == llvm::CallingConv::X86_VectorCall &&
+      getTarget().getTriple().isWindowsArm64EC()) {
+    SourceLocation Loc;
+    if (const Decl *D = GD.getDecl())
+      Loc = D->getLocation();
+
+    Error(Loc, "__vectorcall calling convention is not currently supported");
+  }
   F->setAttributes(PAL);
   F->setCallingConv(static_cast<llvm::CallingConv::ID>(CallingConv));
 }
diff --git a/clang/test/CodeGenCXX/arm64ec-vectorcall.cpp b/clang/test/CodeGenCXX/arm64ec-vectorcall.cpp
new file mode 100644
index 00000000000000..73d2d63835917c
--- /dev/null
+++ b/clang/test/CodeGenCXX/arm64ec-vectorcall.cpp
@@ -0,0 +1,14 @@
+// RUN: %clang_cc1 -triple arm64ec-windows-msvc -emit-llvm -o - %s -verify
+
+// ARM64EC doesn't support generating __vectorcall calls... but __vectorcall
+// function types need to be distinct from __cdecl function types to support
+// compiling the STL. Make sure we only diagnose constructs that actually
+// require generating code.
+void __vectorcall f1();
+void f2(void __vectorcall p()) {}
+void f2(void p()) {}
+void __vectorcall (*f3)();
+void __vectorcall f4(); // expected-error {{__vectorcall}}
+void __vectorcall f5() { // expected-error {{__vectorcall}}
+  f4(); // expected-error{{__vectorcall}}
+}

``````````

</details>


https://github.com/llvm/llvm-project/pull/87725


More information about the cfe-commits mailing list