https://github.com/Ko496-glitch updated https://github.com/llvm/llvm-project/pull/199779
>From 670097190c10eff9e253f0b1030b6817426af01c Mon Sep 17 00:00:00 2001 From: Kartik Ohlan <[email protected]> Date: Tue, 26 May 2026 17:53:48 -0400 Subject: [PATCH 1/3] IEEE754 --- .../lib/CIR/CodeGen/CIRGenBuiltinAArch64.cpp | 3 ++ clang/test/CodeGen/AArch64/neon-across.c | 18 ---------- clang/test/CodeGen/AArch64/neon/intrinsics.c | 34 +++++++++++++++++++ 3 files changed, 37 insertions(+), 18 deletions(-) delete mode 100644 clang/test/CodeGen/AArch64/neon-across.c diff --git a/clang/lib/CIR/CodeGen/CIRGenBuiltinAArch64.cpp b/clang/lib/CIR/CodeGen/CIRGenBuiltinAArch64.cpp index ca5a4680ee79f..a956a4dddb57f 100644 --- a/clang/lib/CIR/CodeGen/CIRGenBuiltinAArch64.cpp +++ b/clang/lib/CIR/CodeGen/CIRGenBuiltinAArch64.cpp @@ -372,6 +372,9 @@ static mlir::Value emitCommonNeonSISDBuiltinExpr( case NEON::BI__builtin_neon_vqrshrund_n_s64: case NEON::BI__builtin_neon_vqrshrnd_n_s64: case NEON::BI__builtin_neon_vqrshrnd_n_u64: + case NEON::BI__builtin_neon_vmaxnmv_f32: + case NEON::BI__builtin_neon_vmaxnmvq_f32: + case NEON::BI__builtin_neon_vmaxnmvq_f64: return emitNeonCall(cgf.cgm, cgf.getBuilder(), {cgf.convertType(expr->getArg(0)->getType())}, ops, llvmIntrName, cgf.convertType(expr->getType()), loc); diff --git a/clang/test/CodeGen/AArch64/neon-across.c b/clang/test/CodeGen/AArch64/neon-across.c deleted file mode 100644 index f248e31b374ba..0000000000000 --- a/clang/test/CodeGen/AArch64/neon-across.c +++ /dev/null @@ -1,18 +0,0 @@ -// NOTE: Assertions have been autogenerated by utils/update_cc_test_checks.py UTC_ARGS: --function-signature -// RUN: %clang_cc1 -triple arm64-none-linux-gnu -target-feature +neon \ -// RUN: -disable-O0-optnone -emit-llvm -o - %s | opt -S -passes=mem2reg | FileCheck %s - -// REQUIRES: aarch64-registered-target || arm-registered-target - -#include <arm_neon.h> - - -// CHECK-LABEL: define {{[^@]+}}@test_vmaxnmvq_f32 -// CHECK-SAME: (<4 x float> noundef [[A:%.*]]) #[[ATTR0:[0-9]+]] { -// CHECK-NEXT: entry: -// CHECK-NEXT: [[VMAXNMVQ_F32_I:%.*]] = call float @llvm.aarch64.neon.fmaxnmv.f32.v4f32(<4 x float> [[A]]) -// CHECK-NEXT: ret float [[VMAXNMVQ_F32_I]] -// -float32_t test_vmaxnmvq_f32(float32x4_t a) { - return vmaxnmvq_f32(a); -} diff --git a/clang/test/CodeGen/AArch64/neon/intrinsics.c b/clang/test/CodeGen/AArch64/neon/intrinsics.c index b07f61a712f90..bf92b05ddc239 100644 --- a/clang/test/CodeGen/AArch64/neon/intrinsics.c +++ b/clang/test/CodeGen/AArch64/neon/intrinsics.c @@ -24,6 +24,40 @@ #include <arm_neon.h> +//===------------------------------------------------------===// +// 2.1.1.13.5 Maximum across vector (IEEE754) +//===------------------------------------------------------===// + +//ALL-LABEL: @test_vmaxnmv_f32( +float32_t test_vmaxnmv_f32(float32x2_t a) { + //CIR: cir.call @vmaxnmv_f32 + // + // LLVM-SAME: <2 x float> {{.*}} [[A:%.*]]) + // LLVM: [[VMAXNMV_F32_I:%.*]] = call float @llvm.aarch64.neon.fmaxnmv.f32.v2f32(<2 x float> [[A]]) + // LLVM: ret float [[VMAXNMV_F32_I]] + return vmaxnmv_f32(a); +} + +// ALL-LABEL: @test_vmaxnmvq_f32( +float32_t test_vmaxnmvq_f32(float32x4_t a) { + //CIR: cir.call @vmaxnmvq_f32 + + // LLVM-SAME: <4 x float> {{.*}}[[A:%.*]]) + // LLVM: [[VMAXNMVQ_F32_I:%.*]] = call float @llvm.aarch64.neon.fmaxnmv.f32.v4f32(<4 x float> [[A]]) + // LLVM: ret float [[VMAXNMVQ_F32_I]] + return vmaxnmvq_f32(a); +} + +// ALL-LABEL: @test_vmaxnmvq_f64( +float64_t test_vmaxnmvq_f64(float64x2_t a) { + //CIR: cir.call @vmaxnmvq_f64 + + // LLVM-SAME: <2 x double> {{.*}}[[A:%.*]]) + // LLVM: [[VMAXNMVQ_F64_I:%.*]] = call double @llvm.aarch64.neon.fmaxnmv.f64.v2f64(<2 x double> [[A]]) + // LLVM: ret double [[VMAXNMVQ_F64_I]] + return vmaxnmvq_f64(a); +} + //===------------------------------------------------------===// // 2.1.3.2.7 Vector saturating rounding shift right and narrow // TODO: Implement SISD variants >From ec6acd8b74976282e2056a3c687f9a798e15fc52 Mon Sep 17 00:00:00 2001 From: Kartik Ohlan <[email protected]> Date: Tue, 2 Jun 2026 17:57:20 -0700 Subject: [PATCH 2/3] added revisions --- clang/test/CodeGen/AArch64/neon-intrinsics.c | 20 -------------------- clang/test/CodeGen/AArch64/neon/intrinsics.c | 10 +++++----- 2 files changed, 5 insertions(+), 25 deletions(-) diff --git a/clang/test/CodeGen/AArch64/neon-intrinsics.c b/clang/test/CodeGen/AArch64/neon-intrinsics.c index 35b886c53a5ec..d9377d4420188 100644 --- a/clang/test/CodeGen/AArch64/neon-intrinsics.c +++ b/clang/test/CodeGen/AArch64/neon-intrinsics.c @@ -19257,26 +19257,6 @@ int64x1_t test_vneg_s64(int64x1_t a) { return vneg_s64(a); } -// CHECK-LABEL: define dso_local double @test_vmaxnmvq_f64( -// CHECK-SAME: <2 x double> noundef [[A:%.*]]) #[[ATTR0]] { -// CHECK-NEXT: [[ENTRY:.*:]] -// CHECK-NEXT: [[VMAXNMVQ_F64_I:%.*]] = call double @llvm.aarch64.neon.fmaxnmv.f64.v2f64(<2 x double> [[A]]) -// CHECK-NEXT: ret double [[VMAXNMVQ_F64_I]] -// -float64_t test_vmaxnmvq_f64(float64x2_t a) { - return vmaxnmvq_f64(a); -} - -// CHECK-LABEL: define dso_local float @test_vmaxnmv_f32( -// CHECK-SAME: <2 x float> noundef [[A:%.*]]) #[[ATTR0]] { -// CHECK-NEXT: [[ENTRY:.*:]] -// CHECK-NEXT: [[VMAXNMV_F32_I:%.*]] = call float @llvm.aarch64.neon.fmaxnmv.f32.v2f32(<2 x float> [[A]]) -// CHECK-NEXT: ret float [[VMAXNMV_F32_I]] -// -float32_t test_vmaxnmv_f32(float32x2_t a) { - return vmaxnmv_f32(a); -} - // CHECK-LABEL: define dso_local <2 x i64> @test_vpaddq_s64( // CHECK-SAME: <2 x i64> noundef [[A:%.*]], <2 x i64> noundef [[B:%.*]]) #[[ATTR0]] { // CHECK-NEXT: [[ENTRY:.*:]] diff --git a/clang/test/CodeGen/AArch64/neon/intrinsics.c b/clang/test/CodeGen/AArch64/neon/intrinsics.c index bf92b05ddc239..a70ac833ef6bd 100644 --- a/clang/test/CodeGen/AArch64/neon/intrinsics.c +++ b/clang/test/CodeGen/AArch64/neon/intrinsics.c @@ -30,11 +30,11 @@ //ALL-LABEL: @test_vmaxnmv_f32( float32_t test_vmaxnmv_f32(float32x2_t a) { - //CIR: cir.call @vmaxnmv_f32 - // - // LLVM-SAME: <2 x float> {{.*}} [[A:%.*]]) - // LLVM: [[VMAXNMV_F32_I:%.*]] = call float @llvm.aarch64.neon.fmaxnmv.f32.v2f32(<2 x float> [[A]]) - // LLVM: ret float [[VMAXNMV_F32_I]] + //CIR: cir.call @vmaxnmv_f32 + + // LLVM-SAME: <2 x float> {{.*}} [[A:%.*]]) + // LLVM: [[VMAXNMV_F32_I:%.*]] = call float @llvm.aarch64.neon.fmaxnmv.f32.v2f32(<2 x float> [[A]]) + // LLVM: ret float [[VMAXNMV_F32_I]] return vmaxnmv_f32(a); } >From 0271dfa892ea05c4a85a4f2bcb015e51326deeae Mon Sep 17 00:00:00 2001 From: Kartik Ohlan <[email protected]> Date: Wed, 3 Jun 2026 20:00:25 -0700 Subject: [PATCH 3/3] Remove emitNeonCall for certain NEON builtins Removed a call to emitNeonCall for specific NEON builtins. --- clang/lib/CIR/CodeGen/CIRGenBuiltinAArch64.cpp | 3 --- 1 file changed, 3 deletions(-) diff --git a/clang/lib/CIR/CodeGen/CIRGenBuiltinAArch64.cpp b/clang/lib/CIR/CodeGen/CIRGenBuiltinAArch64.cpp index 04264e415340d..a88288a49ffd5 100644 --- a/clang/lib/CIR/CodeGen/CIRGenBuiltinAArch64.cpp +++ b/clang/lib/CIR/CodeGen/CIRGenBuiltinAArch64.cpp @@ -437,9 +437,6 @@ static mlir::Value emitCommonNeonSISDBuiltinExpr( case NEON::BI__builtin_neon_vmaxnmv_f32: case NEON::BI__builtin_neon_vmaxnmvq_f32: case NEON::BI__builtin_neon_vmaxnmvq_f64: - return emitNeonCall(cgf.cgm, cgf.getBuilder(), - {cgf.convertType(expr->getArg(0)->getType())}, ops, - llvmIntrName, cgf.convertType(expr->getType()), loc); case NEON::BI__builtin_neon_vsrid_n_s64: case NEON::BI__builtin_neon_vsrid_n_u64: break; _______________________________________________ cfe-commits mailing list [email protected] https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
