[ARM][Clang] Make +nosimd functional for AArch32 Targets (#130623)

`+simd` and `+nosimd` are used to enable or disable NEON Instructions
when compiling for ARM Targets. However, up until now, using these
has not been possible. To enable this, these options are mapped to the
relevant LLVM backend option (`+neon` and `-neon`) so it can be both
enabled and disabled successfully by the user.

Tests have been added to ensure this behaviour is maintained in the
future, along with updates to existing tests as behaviour has now changed
relating to the use of `+simd` and `+nosimd`.

As `simd` has been mapped within the ARMTargetParser.def, support for
this extension is also added for the `--print-support-extensions` command
when the target is AArch32. This will print the `simd` option, along with the
description that relates to the Neon feature. This previously was not
possible as `simd` did not have a related Feature or Negative Feature.

To make this functional as intended, MVE and MVE.FP now rely on their
own Enum identifier, rather than `AEK_SIMD`. While SIMD does refer to
both Neon and Helium technologies, in terms of command line options,
SIMD relates to Neon. Helium relates to MVE and MVE.FP. The Enum
now reflects this too.
This commit is contained in:
Jack Styles
2025-04-15 09:00:14 +01:00
committed by GitHub
parent 0f526498a1
commit 06da00ae2d
10 changed files with 43 additions and 13 deletions

View File

@@ -525,6 +525,9 @@ X86 Support
Arm and AArch64 Support
^^^^^^^^^^^^^^^^^^^^^^^
- For ARM targets, cc1as now considers the FPU's features for the selected CPU or Architecture.
- The ``+nosimd`` attribute is now fully supported for ARM. Previously, this had no effect when being used with
ARM targets, however this will now disable NEON instructions being generated. The ``simd`` option is
also now printed when the ``--print-supported-extensions`` option is used.
Android Support
^^^^^^^^^^^^^^^

View File

@@ -119,4 +119,10 @@
// RUN: %clang_cc1 -triple thumbv9.3a-linux-gnueabihf -emit-llvm -o - %s | FileCheck %s --check-prefix=CHECK-ARCH93
// CHECK-ARCH93: "target-features"="+armv9.3-a,+thumb-mode,+v8.1a,+v8.2a,+v8.3a,+v8.4a,+v8.5a,+v8.6a,+v8.7a,+v8.8a,+v9.1a,+v9.2a,+v9.3a,+v9a"
// RUN: %clang_cc1 -triple arm-none-eabi -target-cpu cortex-r52 -emit-llvm -o - %s | FileCheck %s --check-prefix=CHECK-ARMSIMD-R52
// CHECK-ARMSIMD-R52: "target-features"="+armv8-r,+crc,+d32,+dsp,+fp-armv8,+fp-armv8d16,+fp-armv8d16sp,+fp-armv8sp,+fp16,+fp64,+hwdiv,+hwdiv-arm,+neon,+vfp2,+vfp2sp,+vfp3,+vfp3d16,+vfp3d16sp,+vfp3sp,+vfp4,+vfp4d16,+vfp4d16sp,+vfp4sp,-thumb-mode"
// RUN: %clang_cc1 -triple armv8-none-eabi -target-cpu cortex-a57 -emit-llvm -o - %s | FileCheck %s --check-prefix=CHECK-ARMSIMD-A57
// CHECK-ARMSIMD-A57: "target-features"="+aes,+armv8-a,+crc,+d32,+dsp,+fp-armv8,+fp-armv8d16,+fp-armv8d16sp,+fp-armv8sp,+fp16,+fp64,+hwdiv,+hwdiv-arm,+neon,+sha2,+vfp2,+vfp2sp,+vfp3,+vfp3d16,+vfp3d16sp,+vfp3sp,+vfp4,+vfp4d16,+vfp4d16sp,+vfp4sp,-thumb-mode"
void foo() {}

View File

@@ -12,6 +12,7 @@
// CHECK-NEXT: dsp Supports DSP instructions in ARM and/or Thumb2
// CHECK-NEXT: mve Support M-Class Vector Extension with integer ops
// CHECK-NEXT: mve.fp Support M-Class Vector Extension with integer and floating ops
// CHECK-NEXT: simd Enable NEON instructions
// CHECK-NEXT: fp16 Enable half-precision floating point
// CHECK-NEXT: ras Enable Reliability, Availability and Serviceability extensions
// CHECK-NEXT: fp16fml Enable full half-precision floating point fml instructions

View File

@@ -1027,3 +1027,19 @@
// CHECK-R52-NEXT: #define __ARM_VFPV4__ 1
// CHECK-R52-NOT: #define __ARM_NEON 1
// CHECK-R52-NOT: #define __ARM_NEON__
// Check that on AArch32, Neon is correctly activated when the target supports the feature
// RUN: %clang -target arm-none-eabi -march=armv8-a -mfloat-abi=hard -x c -E -dM -o - %s | FileCheck -check-prefix=CHECK-SIMD %s
// RUN: %clang -target arm-none-eabi -mcpu=cortex-r52 -mfloat-abi=hard -x c -E -dM -o - %s | FileCheck -check-prefix=CHECK-SIMD %s
// RUN: %clang -target arm-none-eabi -mcpu=cortex-a57 -mfloat-abi=hard -x c -E -dM -o - %s | FileCheck -check-prefix=CHECK-SIMD %s
// CHECK-SIMD: #define __ARM_NEON 1
// CHECK-SIMD: #define __ARM_NEON_FP 0x6
// CHECK-SIMD: #define __ARM_NEON__ 1
// Check that on AArch32 appropriate targets, +nosimd correctly disables NEON instructions.
// RUN: %clang -target arm-none-eabi -march=armv8-a+nosimd -mfloat-abi=hard -x c -E -dM -o - %s | FileCheck -check-prefix=CHECK-NOSIMD %s
// RUN: %clang -target arm-none-eabi -mcpu=cortex-r52+nosimd -mfloat-abi=hard -x c -E -dM -o - %s | FileCheck -check-prefix=CHECK-NOSIMD %s
// RUN: %clang -target arm-none-eabi -mcpu=cortex-a57+nosimd -mfloat-abi=hard -x c -E -dM -o - %s | FileCheck -check-prefix=CHECK-NOSIMD %s
// CHECK-NOSIMD-NOT: #define __ARM_NEON 1
// CHECK-NOSIMD-NOT: #define __ARM_NEON_FP 0x6
// CHECK-NOSIMD-NOT: #define __ARM_NEON__ 1

View File

@@ -224,12 +224,12 @@ ARM_ARCH_EXT_NAME("dotprod", ARM::AEK_DOTPROD, "+dotprod", "-dotprod")
ARM_ARCH_EXT_NAME("dsp", ARM::AEK_DSP, "+dsp", "-dsp")
ARM_ARCH_EXT_NAME("fp", ARM::AEK_FP, {}, {})
ARM_ARCH_EXT_NAME("fp.dp", ARM::AEK_FP_DP, {}, {})
ARM_ARCH_EXT_NAME("mve", (ARM::AEK_DSP | ARM::AEK_SIMD), "+mve", "-mve")
ARM_ARCH_EXT_NAME("mve.fp", (ARM::AEK_DSP | ARM::AEK_SIMD | ARM::AEK_FP),
ARM_ARCH_EXT_NAME("mve", (ARM::AEK_DSP | ARM::AEK_MVE), "+mve", "-mve")
ARM_ARCH_EXT_NAME("mve.fp", (ARM::AEK_DSP | ARM::AEK_MVE | ARM::AEK_FP),
"+mve.fp", "-mve.fp")
ARM_ARCH_EXT_NAME("idiv", (ARM::AEK_HWDIVARM | ARM::AEK_HWDIVTHUMB), {}, {})
ARM_ARCH_EXT_NAME("mp", ARM::AEK_MP, {}, {})
ARM_ARCH_EXT_NAME("simd", ARM::AEK_SIMD, {}, {})
ARM_ARCH_EXT_NAME("simd", ARM::AEK_SIMD, "+neon", "-neon")
ARM_ARCH_EXT_NAME("sec", ARM::AEK_SEC, {}, {})
ARM_ARCH_EXT_NAME("virt", ARM::AEK_VIRT, {}, {})
ARM_ARCH_EXT_NAME("fp16", ARM::AEK_FP16, "+fullfp16", "-fullfp16")
@@ -345,12 +345,12 @@ ARM_CPU_NAME("cortex-m33", ARMV8MMainline, FK_FPV5_SP_D16, false, ARM::AEK_DSP)
ARM_CPU_NAME("star-mc1", ARMV8MMainline, FK_FPV5_SP_D16, false, ARM::AEK_DSP)
ARM_CPU_NAME("cortex-m35p", ARMV8MMainline, FK_FPV5_SP_D16, false, ARM::AEK_DSP)
ARM_CPU_NAME("cortex-m55", ARMV8_1MMainline, FK_FP_ARMV8_FULLFP16_D16, false,
(ARM::AEK_DSP | ARM::AEK_SIMD | ARM::AEK_FP | ARM::AEK_FP16))
(ARM::AEK_DSP | ARM::AEK_MVE | ARM::AEK_FP | ARM::AEK_FP16))
ARM_CPU_NAME("cortex-m85", ARMV8_1MMainline, FK_FP_ARMV8_FULLFP16_D16, false,
(ARM::AEK_DSP | ARM::AEK_SIMD | ARM::AEK_FP | ARM::AEK_FP16 |
(ARM::AEK_DSP | ARM::AEK_MVE | ARM::AEK_FP | ARM::AEK_FP16 |
ARM::AEK_RAS | ARM::AEK_PACBTI))
ARM_CPU_NAME("cortex-m52", ARMV8_1MMainline, FK_FP_ARMV8_FULLFP16_D16, false,
(ARM::AEK_DSP | ARM::AEK_SIMD | ARM::AEK_FP | ARM::AEK_FP16 |
(ARM::AEK_DSP | ARM::AEK_MVE | ARM::AEK_FP | ARM::AEK_FP16 |
ARM::AEK_RAS | ARM::AEK_PACBTI))
ARM_CPU_NAME("cortex-a32", ARMV8A, FK_CRYPTO_NEON_FP_ARMV8, false, ARM::AEK_CRC)
ARM_CPU_NAME("cortex-a35", ARMV8A, FK_CRYPTO_NEON_FP_ARMV8, false, ARM::AEK_CRC)

View File

@@ -61,6 +61,7 @@ enum ArchExtKind : uint64_t {
AEK_CDECP6 = 1 << 28,
AEK_CDECP7 = 1 << 29,
AEK_PACBTI = 1 << 30,
AEK_MVE = 1ULL << 31,
// Unsupported extensions.
AEK_OS = 1ULL << 59,
AEK_IWMMXT = 1ULL << 60,

View File

@@ -12970,7 +12970,7 @@ bool ARMAsmParser::enableArchExtFeature(StringRef Name, SMLoc &ExtLoc) {
{ARM::AEK_CRYPTO,
{Feature_HasV8Bit},
{ARM::FeatureCrypto, ARM::FeatureNEON, ARM::FeatureFPARMv8}},
{(ARM::AEK_DSP | ARM::AEK_SIMD | ARM::AEK_FP),
{(ARM::AEK_DSP | ARM::AEK_MVE | ARM::AEK_FP),
{Feature_HasV8_1MMainlineBit},
{ARM::HasMVEFloatOps}},
{ARM::AEK_FP,

View File

@@ -248,7 +248,7 @@ void ARMTargetStreamer::emitTargetAttributes(const MCSubtargetInfo &STI) {
emitFPU(STI.hasFeature(ARM::FeatureFP64) ? ARM::FK_FPV5_D16
: ARM::FK_FPV5_SP_D16);
if (STI.hasFeature(ARM::HasMVEFloatOps))
emitArchExtension(ARM::AEK_SIMD | ARM::AEK_DSP | ARM::AEK_FP);
emitArchExtension(ARM::AEK_MVE | ARM::AEK_DSP | ARM::AEK_FP);
}
} else if (STI.hasFeature(ARM::FeatureVFP4_D16_SP))
emitFPU(STI.hasFeature(ARM::FeatureD32)

View File

@@ -657,6 +657,10 @@ void ARM::PrintSupportedExtensions(StringMap<StringRef> DescMap) {
// Extensions without a feature cannot be used with -march.
if (!Ext.Feature.empty()) {
std::string Description = DescMap[Ext.Name].str();
// With SIMD, this links to the NEON feature, so the description should be
// taken from here, as SIMD does not exist in TableGen.
if (Ext.Name == "simd")
Description = DescMap["neon"].str();
outs() << " "
<< format(Description.empty() ? "%s\n" : "%-20s%s\n",
Ext.Name.str().c_str(), Description.c_str());

View File

@@ -507,17 +507,17 @@ INSTANTIATE_TEST_SUITE_P(
"8-M.Mainline"),
ARMCPUTestParams<uint64_t>(
"cortex-m55", "armv8.1-m.main", "fp-armv8-fullfp16-d16",
ARM::AEK_HWDIVTHUMB | ARM::AEK_DSP | ARM::AEK_SIMD | ARM::AEK_FP |
ARM::AEK_HWDIVTHUMB | ARM::AEK_DSP | ARM::AEK_MVE | ARM::AEK_FP |
ARM::AEK_RAS | ARM::AEK_LOB | ARM::AEK_FP16,
"8.1-M.Mainline"),
ARMCPUTestParams<uint64_t>(
"cortex-m85", "armv8.1-m.main", "fp-armv8-fullfp16-d16",
ARM::AEK_HWDIVTHUMB | ARM::AEK_DSP | ARM::AEK_SIMD | ARM::AEK_FP |
ARM::AEK_HWDIVTHUMB | ARM::AEK_DSP | ARM::AEK_MVE | ARM::AEK_FP |
ARM::AEK_RAS | ARM::AEK_LOB | ARM::AEK_FP16 | ARM::AEK_PACBTI,
"8.1-M.Mainline"),
ARMCPUTestParams<uint64_t>(
"cortex-m52", "armv8.1-m.main", "fp-armv8-fullfp16-d16",
ARM::AEK_HWDIVTHUMB | ARM::AEK_DSP | ARM::AEK_SIMD | ARM::AEK_FP |
ARM::AEK_HWDIVTHUMB | ARM::AEK_DSP | ARM::AEK_MVE | ARM::AEK_FP |
ARM::AEK_RAS | ARM::AEK_LOB | ARM::AEK_FP16 | ARM::AEK_PACBTI,
"8.1-M.Mainline"),
ARMCPUTestParams<uint64_t>("iwmmxt", "iwmmxt", "none", ARM::AEK_NONE,
@@ -801,7 +801,7 @@ TEST(TargetParserTest, ARMArchExtFeature) {
{"fp", "nofp", nullptr, nullptr},
{"idiv", "noidiv", nullptr, nullptr},
{"mp", "nomp", nullptr, nullptr},
{"simd", "nosimd", nullptr, nullptr},
{"simd", "nosimd", "+neon", "-neon"},
{"sec", "nosec", nullptr, nullptr},
{"virt", "novirt", nullptr, nullptr},
{"fp16", "nofp16", "+fullfp16", "-fullfp16"},
@@ -1046,7 +1046,6 @@ TEST(TargetParserTest, ARMPrintSupportedExtensions) {
EXPECT_EQ(std::string::npos, captured.find("invalid"));
// Should not include anything that lacks a feature name. Checking a few here
// but not all as if one is hidden correctly the rest should be.
EXPECT_EQ(std::string::npos, captured.find("simd"));
EXPECT_EQ(std::string::npos, captured.find("maverick"));
EXPECT_EQ(std::string::npos, captured.find("xscale"));
}