Skip to content

[AArch64][Clang] Limit variadic onstack args to 8 bytes on Arm64EC. #154578

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Closed
wants to merge 1 commit into from
Closed

[AArch64][Clang] Limit variadic onstack args to 8 bytes on Arm64EC. #154578

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
14 changes: 10 additions & 4 deletions clang/lib/CodeGen/Targets/AArch64.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -474,8 +474,11 @@ ABIArgInfo AArch64ABIInfo::classifyArgumentType(QualType Ty, bool IsVariadicFn,
Ty, IsNamedArg, NVec, NPred, UnpaddedCoerceToSeq, NSRN, NPRN);
}

// Aggregates <= 16 bytes are passed directly in registers or on the stack.
if (Size <= 128) {
// Aggregates <= 16 bytes (8 bytes for variadic Arm64EC) are passed directly
// in registers or on the stack.
uint64_t MaxDirectSize =
(IsVariadicFn && getTarget().getTriple().isWindowsArm64EC()) ? 64 : 128;
if (Size <= MaxDirectSize) {
unsigned Alignment;
if (Kind == AArch64ABIKind::AAPCS) {
Alignment = getContext().getTypeUnadjustedAlign(Ty);
Expand Down Expand Up @@ -1152,8 +1155,11 @@ RValue AArch64ABIInfo::EmitMSVAArg(CodeGenFunction &CGF, Address VAListAddr,
QualType Ty, AggValueSlot Slot) const {
bool IsIndirect = false;

// Composites larger than 16 bytes are passed by reference.
if (isAggregateTypeForABI(Ty) && getContext().getTypeSize(Ty) > 128)
// Composites larger than 16 bytes (8 bytes on Arm64EC) are passed by
// reference.
uint64_t MaxDirectSize =
getTarget().getTriple().isWindowsArm64EC() ? 64 : 128;
if (isAggregateTypeForABI(Ty) && getContext().getTypeSize(Ty) > MaxDirectSize)
IsIndirect = true;

return emitVoidPtrVAArg(CGF, VAListAddr, Ty, IsIndirect,
Expand Down
18 changes: 17 additions & 1 deletion clang/test/CodeGen/AArch64/varargs-ms.c
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -1,4 +1,5 @@
// RUN: %clang_cc1 -triple arm64-windows-msvc -emit-llvm -o - %s | FileCheck %s
// RUN: %clang_cc1 -triple arm64-windows-msvc -emit-llvm -o - %s | FileCheck %s --check-prefix=CHECK --check-prefix=CHECK-A64
// RUN: %clang_cc1 -triple arm64ec-windows-msvc -emit-llvm -o - %s | FileCheck %s --check-prefix=CHECK --check-prefix=CHECK-EC

#include <stdarg.h>

Expand All @@ -8,3 +9,18 @@ int simple_int(va_list ap) {
// CHECK: [[RESULT:%[a-z_0-9]+]] = load i32, ptr %argp.cur
// CHECK: ret i32 [[RESULT]]
}

struct bigstruct {
int item[4];
};
struct bigstruct big_struct(va_list ap) {
// CHECK-LABEL: define dso_local [2 x i64] @big_struct
return va_arg(ap, struct bigstruct);
// CHECK-EC: %argp.next = getelementptr inbounds i8, ptr %argp.cur, i64 8
// CHECK-EC: [[PTR:%[0-9]+]] = load ptr, ptr %argp.cur, align 8
// CHECK-EC: call void @llvm.memcpy.p0.p0.i64(ptr align 4 %retval, ptr align 4 [[PTR]], i64 16, i1 false)
// CHECK-A64: %argp.next = getelementptr inbounds i8, ptr %argp.cur, i64 16
// CHECK-A64: call void @llvm.memcpy.p0.p0.i64(ptr align 4 %retval, ptr align 8 %argp.cur, i64 16, i1 false)
// CHECK: [[RESULT:%[0-9]+]] = load [2 x i64], ptr %coerce.dive
// CHECK: ret [2 x i64] [[RESULT]]
}