Skip to content

[clang][bytecode] Implement ia32_select* builtins #154758

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

Merged
merged 1 commit into from
Aug 21, 2025
Merged

[clang][bytecode] Implement ia32_select* builtins #154758

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
61 changes: 61 additions & 0 deletions clang/lib/AST/ByteCode/InterpBuiltin.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -2778,6 +2778,40 @@ static bool interp__builtin_elementwise_fma(InterpState &S, CodePtr OpPC,
return true;
}

/// AVX512 predicated move: "Result = Mask[] ? LHS[] : RHS[]".
static bool interp__builtin_select(InterpState &S, CodePtr OpPC,
const CallExpr *Call) {
const Pointer &RHS = S.Stk.pop<Pointer>();
Copy link
Collaborator

@shafik shafik Aug 25, 2025

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

assert(Call->getNumArgs() == 1);?

const Pointer &LHS = S.Stk.pop<Pointer>();
PrimType MaskT = *S.getContext().classify(Call->getArg(0));
APSInt Mask = popToAPSInt(S.Stk, MaskT);
const Pointer &Dst = S.Stk.peek<Pointer>();

assert(LHS.getNumElems() == RHS.getNumElems());
assert(LHS.getNumElems() == Dst.getNumElems());
unsigned NumElems = LHS.getNumElems();
PrimType ElemT = LHS.getFieldDesc()->getPrimType();
PrimType DstElemT = Dst.getFieldDesc()->getPrimType();

for (unsigned I = 0; I != NumElems; ++I) {
if (ElemT == PT_Float) {
assert(DstElemT == PT_Float);
Dst.elem<Floating>(I) =
Mask[I] ? LHS.elem<Floating>(I) : RHS.elem<Floating>(I);
} else {
APSInt Elem;
INT_TYPE_SWITCH(ElemT, {
Copy link
Contributor

@brandonxin brandonxin Sep 11, 2025

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Hello, I am new to LLVM/Clang. I want to ask why there is an INT_TYPE_SWITCH and then another INT_TYPE_SWITCH_NO_BOOL. Is it possible that ElemT is PT_Bool? I checked BuiltinsX86.td and all these select buitins require vectors of integers as their parameters.

Copy link
Contributor Author

@tbaederr tbaederr Sep 12, 2025

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

I don't remember anymore, but looking at this now, I'd probably use one only one of those switches. If it can't be bool, there should be an assert somewhere.

Elem = Mask[I] ? LHS.elem<T>(I).toAPSInt() : RHS.elem<T>(I).toAPSInt();
});
INT_TYPE_SWITCH_NO_BOOL(DstElemT,
{ Dst.elem<T>(I) = static_cast<T>(Elem); });
}
}
Dst.initializeAllElements();

return true;
}

bool InterpretBuiltin(InterpState &S, CodePtr OpPC, const CallExpr *Call,
uint32_t BuiltinID) {
if (!S.getASTContext().BuiltinInfo.isConstantEvaluated(BuiltinID))
Expand Down Expand Up @@ -3210,9 +3244,36 @@ bool InterpretBuiltin(InterpState &S, CodePtr OpPC, const CallExpr *Call,
case clang::X86::BI__builtin_ia32_pmuludq256:
case clang::X86::BI__builtin_ia32_pmuludq512:
return interp__builtin_ia32_pmul(S, OpPC, Call, BuiltinID);

case Builtin::BI__builtin_elementwise_fma:
return interp__builtin_elementwise_fma(S, OpPC, Call);

case X86::BI__builtin_ia32_selectb_128:
case X86::BI__builtin_ia32_selectb_256:
case X86::BI__builtin_ia32_selectb_512:
case X86::BI__builtin_ia32_selectw_128:
case X86::BI__builtin_ia32_selectw_256:
case X86::BI__builtin_ia32_selectw_512:
case X86::BI__builtin_ia32_selectd_128:
case X86::BI__builtin_ia32_selectd_256:
case X86::BI__builtin_ia32_selectd_512:
case X86::BI__builtin_ia32_selectq_128:
case X86::BI__builtin_ia32_selectq_256:
case X86::BI__builtin_ia32_selectq_512:
case X86::BI__builtin_ia32_selectph_128:
case X86::BI__builtin_ia32_selectph_256:
case X86::BI__builtin_ia32_selectph_512:
case X86::BI__builtin_ia32_selectpbf_128:
case X86::BI__builtin_ia32_selectpbf_256:
case X86::BI__builtin_ia32_selectpbf_512:
case X86::BI__builtin_ia32_selectps_128:
case X86::BI__builtin_ia32_selectps_256:
case X86::BI__builtin_ia32_selectps_512:
case X86::BI__builtin_ia32_selectpd_128:
case X86::BI__builtin_ia32_selectpd_256:
case X86::BI__builtin_ia32_selectpd_512:
return interp__builtin_select(S, OpPC, Call);

default:
S.FFDiag(S.Current->getLocation(OpPC),
diag::note_invalid_subexpr_in_const_expr)
Expand Down
5 changes: 5 additions & 0 deletions clang/test/CodeGen/X86/avx512f-builtins.c
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -3,6 +3,11 @@
// RUN: %clang_cc1 -x c++ -flax-vector-conversions=none -ffreestanding %s -triple=x86_64-apple-darwin -target-feature +avx512f -emit-llvm -o - -Wall -Werror -Wsign-conversion | FileCheck %s
// RUN: %clang_cc1 -x c++ -flax-vector-conversions=none -fms-extensions -fms-compatibility -ffreestanding %s -triple=x86_64-windows-msvc -target-feature +avx512f -emit-llvm -o - -Wall -Werror -Wsign-conversion | FileCheck %s

// RUN: %clang_cc1 -x c -flax-vector-conversions=none -ffreestanding %s -triple=x86_64-apple-darwin -target-feature +avx512f -emit-llvm -o - -Wall -Werror -Wsign-conversion -fexperimental-new-constant-interpreter | FileCheck %s
// RUN: %clang_cc1 -x c -flax-vector-conversions=none -fms-extensions -fms-compatibility -ffreestanding %s -triple=x86_64-windows-msvc -target-feature +avx512f -emit-llvm -o - -Wall -Werror -Wsign-conversion -fexperimental-new-constant-interpreter | FileCheck %s
// RUN: %clang_cc1 -x c++ -flax-vector-conversions=none -ffreestanding %s -triple=x86_64-apple-darwin -target-feature +avx512f -emit-llvm -o - -Wall -Werror -Wsign-conversion -fexperimental-new-constant-interpreter | FileCheck %s
// RUN: %clang_cc1 -x c++ -flax-vector-conversions=none -fms-extensions -fms-compatibility -ffreestanding %s -triple=x86_64-windows-msvc -target-feature +avx512f -emit-llvm -o - -Wall -Werror -Wsign-conversion -fexperimental-new-constant-interpreter | FileCheck %s

#include <immintrin.h>
#include "builtin_test_helpers.h"

Expand Down