[clang] [llvm] [AArch64] Decouple feature dependency expansion. (PR #94279)
Alexandros Lamprineas via cfe-commits
cfe-commits at lists.llvm.org
Wed Jun 5 06:46:09 PDT 2024
https://github.com/labrinea updated https://github.com/llvm/llvm-project/pull/94279
>From a413428bb0fdcd45b4a251e385b6cc00ff6a0a2c Mon Sep 17 00:00:00 2001
From: Alexandros Lamprineas <alexandros.lamprineas at arm.com>
Date: Mon, 3 Jun 2024 19:57:49 +0100
Subject: [PATCH 1/3] [AArch64] Decouple feature dependency expansion.
The dependency expansion step which was introduced by FMV has been
erroneously used for non-FMV features, for example when parsing the
target attribute. The PR #93695 has rectified most of the tests which
were relying on dependency expansion of target features specified on
the -cc1 command line. In this patch I am decoupling the dependency
expansion of features specified on the target attribute from FMV.
To do that first I am expanding FMV dependencies before passing the
list of target features to initFeatureMap(). Similarly when parsing
the target attribute I am reconstructing an ExtensionSet from the
list of target features which was created during the command line
option parsing. The attribute parsing may toggle bits of that
ExtensionSet and at the end it is converted to a list of target
features. Those are passed to initFeatureMap(), which no longer
requires an override.
A side effect of this refactoring is that features specified on the
target_version attribute now supersede the command line options,
which is what should be happening in the first place.
---
clang/include/clang/AST/ASTContext.h | 3 -
clang/lib/AST/ASTContext.cpp | 53 ++++-----
clang/lib/Basic/Targets/AArch64.cpp | 99 +++++------------
clang/lib/Basic/Targets/AArch64.h | 4 -
.../CodeGen/aarch64-cpu-supports-target.c | 2 +-
.../aarch64-sme-attrs.cpp | 2 +-
clang/test/CodeGen/aarch64-targetattr.c | 41 ++++---
clang/test/CodeGen/attr-target-version.c | 46 ++++----
clang/test/Sema/aarch64-neon-target.c | 4 +-
clang/test/Sema/builtin-cpu-supports.c | 2 +-
.../llvm/TargetParser/AArch64TargetParser.h | 104 +++++++++++-------
llvm/lib/Target/AArch64/AArch64Features.td | 2 +-
llvm/lib/TargetParser/AArch64TargetParser.cpp | 53 ++++++---
13 files changed, 201 insertions(+), 214 deletions(-)
diff --git a/clang/include/clang/AST/ASTContext.h b/clang/include/clang/AST/ASTContext.h
index a1d1d1c51cd41..8bce4812f0d48 100644
--- a/clang/include/clang/AST/ASTContext.h
+++ b/clang/include/clang/AST/ASTContext.h
@@ -3203,9 +3203,6 @@ class ASTContext : public RefCountedBase<ASTContext> {
/// valid feature names.
ParsedTargetAttr filterFunctionTargetAttrs(const TargetAttr *TD) const;
- std::vector<std::string>
- filterFunctionTargetVersionAttrs(const TargetVersionAttr *TV) const;
-
void getFunctionFeatureMap(llvm::StringMap<bool> &FeatureMap,
const FunctionDecl *) const;
void getFunctionFeatureMap(llvm::StringMap<bool> &FeatureMap,
diff --git a/clang/lib/AST/ASTContext.cpp b/clang/lib/AST/ASTContext.cpp
index 73d3b152c49f1..6b7cfbe48796a 100644
--- a/clang/lib/AST/ASTContext.cpp
+++ b/clang/lib/AST/ASTContext.cpp
@@ -87,6 +87,7 @@
#include "llvm/Support/MD5.h"
#include "llvm/Support/MathExtras.h"
#include "llvm/Support/raw_ostream.h"
+#include "llvm/TargetParser/AArch64TargetParser.h"
#include "llvm/TargetParser/Triple.h"
#include <algorithm>
#include <cassert>
@@ -13663,17 +13664,18 @@ QualType ASTContext::getCorrespondingSignedFixedPointType(QualType Ty) const {
}
}
-std::vector<std::string> ASTContext::filterFunctionTargetVersionAttrs(
- const TargetVersionAttr *TV) const {
- assert(TV != nullptr);
- llvm::SmallVector<StringRef, 8> Feats;
- std::vector<std::string> ResFeats;
- TV->getFeatures(Feats);
- for (auto &Feature : Feats)
- if (Target->validateCpuSupports(Feature.str()))
- // Use '?' to mark features that came from TargetVersion.
- ResFeats.push_back("?" + Feature.str());
- return ResFeats;
+// Given a list of FMV features, add each of their backend features to the list.
+static void
+getFMVBackendFeaturesFor(const llvm::SmallVectorImpl<StringRef> &FMVFeatStrings,
+ std::vector<std::string> &BackendFeats) {
+ for (StringRef F : FMVFeatStrings) {
+ if (auto FMVExt = llvm::AArch64::parseArchExtension(F)) {
+ SmallVector<StringRef, 8> Feats;
+ FMVExt->DependentFeatures.split(Feats, ',', -1, false);
+ for (StringRef F : Feats)
+ BackendFeats.push_back(F.str());
+ }
+ }
}
ParsedTargetAttr
@@ -13708,10 +13710,12 @@ void ASTContext::getFunctionFeatureMap(llvm::StringMap<bool> &FeatureMap,
// Make a copy of the features as passed on the command line into the
// beginning of the additional features from the function to override.
- ParsedAttr.Features.insert(
- ParsedAttr.Features.begin(),
- Target->getTargetOpts().FeaturesAsWritten.begin(),
- Target->getTargetOpts().FeaturesAsWritten.end());
+ // AArch64 handles command line option features in parseTargetAttr().
+ if (!Target->getTriple().isAArch64())
+ ParsedAttr.Features.insert(
+ ParsedAttr.Features.begin(),
+ Target->getTargetOpts().FeaturesAsWritten.begin(),
+ Target->getTargetOpts().FeaturesAsWritten.end());
if (ParsedAttr.CPU != "" && Target->isValidCPUName(ParsedAttr.CPU))
TargetCPU = ParsedAttr.CPU;
@@ -13734,13 +13738,9 @@ void ASTContext::getFunctionFeatureMap(llvm::StringMap<bool> &FeatureMap,
} else if (const auto *TC = FD->getAttr<TargetClonesAttr>()) {
std::vector<std::string> Features;
if (Target->getTriple().isAArch64()) {
- // TargetClones for AArch64
llvm::SmallVector<StringRef, 8> Feats;
TC->getFeatures(Feats, GD.getMultiVersionIndex());
- for (StringRef Feat : Feats)
- if (Target->validateCpuSupports(Feat.str()))
- // Use '?' to mark features that came from AArch64 TargetClones.
- Features.push_back("?" + Feat.str());
+ getFMVBackendFeaturesFor(Feats, Features);
Features.insert(Features.begin(),
Target->getTargetOpts().FeaturesAsWritten.begin(),
Target->getTargetOpts().FeaturesAsWritten.end());
@@ -13753,11 +13753,14 @@ void ASTContext::getFunctionFeatureMap(llvm::StringMap<bool> &FeatureMap,
}
Target->initFeatureMap(FeatureMap, getDiagnostics(), TargetCPU, Features);
} else if (const auto *TV = FD->getAttr<TargetVersionAttr>()) {
- std::vector<std::string> Feats = filterFunctionTargetVersionAttrs(TV);
- Feats.insert(Feats.begin(),
- Target->getTargetOpts().FeaturesAsWritten.begin(),
- Target->getTargetOpts().FeaturesAsWritten.end());
- Target->initFeatureMap(FeatureMap, getDiagnostics(), TargetCPU, Feats);
+ llvm::SmallVector<StringRef, 8> Feats;
+ TV->getFeatures(Feats);
+ std::vector<std::string> Features;
+ getFMVBackendFeaturesFor(Feats, Features);
+ Features.insert(Features.begin(),
+ Target->getTargetOpts().FeaturesAsWritten.begin(),
+ Target->getTargetOpts().FeaturesAsWritten.end());
+ Target->initFeatureMap(FeatureMap, getDiagnostics(), TargetCPU, Features);
} else {
FeatureMap = Target->getTargetOpts().FeatureMap;
}
diff --git a/clang/lib/Basic/Targets/AArch64.cpp b/clang/lib/Basic/Targets/AArch64.cpp
index 5db1ce78c657f..d8bb4fa3a3b49 100644
--- a/clang/lib/Basic/Targets/AArch64.cpp
+++ b/clang/lib/Basic/Targets/AArch64.cpp
@@ -1050,51 +1050,6 @@ bool AArch64TargetInfo::handleTargetFeatures(std::vector<std::string> &Features,
return true;
}
-bool AArch64TargetInfo::initFeatureMap(
- llvm::StringMap<bool> &Features, DiagnosticsEngine &Diags, StringRef CPU,
- const std::vector<std::string> &FeaturesVec) const {
- std::vector<std::string> UpdatedFeaturesVec;
- // Parse the CPU and add any implied features.
- std::optional<llvm::AArch64::CpuInfo> CpuInfo = llvm::AArch64::parseCpu(CPU);
- if (CpuInfo) {
- auto Exts = CpuInfo->getImpliedExtensions();
- std::vector<StringRef> CPUFeats;
- llvm::AArch64::getExtensionFeatures(Exts, CPUFeats);
- for (auto F : CPUFeats) {
- assert((F[0] == '+' || F[0] == '-') && "Expected +/- in target feature!");
- UpdatedFeaturesVec.push_back(F.str());
- }
- }
-
- // Process target and dependent features. This is done in two loops collecting
- // them into UpdatedFeaturesVec: first to add dependent '+'features, second to
- // add target '+/-'features that can later disable some of features added on
- // the first loop. Function Multi Versioning features begin with '?'.
- for (const auto &Feature : FeaturesVec)
- if (((Feature[0] == '?' || Feature[0] == '+')) &&
- AArch64TargetInfo::doesFeatureAffectCodeGen(Feature.substr(1))) {
- StringRef DepFeatures =
- AArch64TargetInfo::getFeatureDependencies(Feature.substr(1));
- SmallVector<StringRef, 1> AttrFeatures;
- DepFeatures.split(AttrFeatures, ",");
- for (auto F : AttrFeatures)
- UpdatedFeaturesVec.push_back(F.str());
- }
- for (const auto &Feature : FeaturesVec)
- if (Feature[0] != '?') {
- std::string UpdatedFeature = Feature;
- if (Feature[0] == '+') {
- std::optional<llvm::AArch64::ExtensionInfo> Extension =
- llvm::AArch64::parseArchExtension(Feature.substr(1));
- if (Extension)
- UpdatedFeature = Extension->Feature.str();
- }
- UpdatedFeaturesVec.push_back(UpdatedFeature);
- }
-
- return TargetInfo::initFeatureMap(Features, Diags, CPU, UpdatedFeaturesVec);
-}
-
// Parse AArch64 Target attributes, which are a comma separated list of:
// "arch=<arch>" - parsed to features as per -march=..
// "cpu=<cpu>" - parsed to features as per -mcpu=.., with CPU set to <cpu>
@@ -1110,23 +1065,26 @@ ParsedTargetAttr AArch64TargetInfo::parseTargetAttr(StringRef Features) const {
bool FoundArch = false;
auto SplitAndAddFeatures = [](StringRef FeatString,
- std::vector<std::string> &Features) {
+ std::vector<std::string> &Features,
+ llvm::AArch64::ExtensionSet &FeatureBits) {
SmallVector<StringRef, 8> SplitFeatures;
FeatString.split(SplitFeatures, StringRef("+"), -1, false);
for (StringRef Feature : SplitFeatures) {
- StringRef FeatureName = llvm::AArch64::getArchExtFeature(Feature);
- if (!FeatureName.empty())
- Features.push_back(FeatureName.str());
+ if (FeatureBits.parseModifier(Feature))
+ continue;
+ // Pushing the original feature string to give a sema error later on
+ // when they get checked.
+ if (Feature.starts_with("no"))
+ Features.push_back("-" + Feature.drop_front(2).str());
else
- // Pushing the original feature string to give a sema error later on
- // when they get checked.
- if (Feature.starts_with("no"))
- Features.push_back("-" + Feature.drop_front(2).str());
- else
- Features.push_back("+" + Feature.str());
+ Features.push_back("+" + Feature.str());
}
};
+ llvm::AArch64::ExtensionSet FeatureBits;
+ // Reconstruct the bitset from the command line option features.
+ FeatureBits.reconstructFromParsedFeatures(getTargetOpts().FeaturesAsWritten);
+
for (auto &Feature : AttrFeatures) {
Feature = Feature.trim();
if (Feature.starts_with("fpmath="))
@@ -1149,9 +1107,9 @@ ParsedTargetAttr AArch64TargetInfo::parseTargetAttr(StringRef Features) const {
// Ret.Features.
if (!AI)
continue;
- Ret.Features.push_back(AI->ArchFeature.str());
+ FeatureBits.addArchDefaults(*AI);
// Add any extra features, after the +
- SplitAndAddFeatures(Split.second, Ret.Features);
+ SplitAndAddFeatures(Split.second, Ret.Features, FeatureBits);
} else if (Feature.starts_with("cpu=")) {
if (!Ret.CPU.empty())
Ret.Duplicate = "cpu=";
@@ -1161,7 +1119,10 @@ ParsedTargetAttr AArch64TargetInfo::parseTargetAttr(StringRef Features) const {
std::pair<StringRef, StringRef> Split =
Feature.split("=").second.trim().split("+");
Ret.CPU = Split.first;
- SplitAndAddFeatures(Split.second, Ret.Features);
+ if (auto CpuInfo = llvm::AArch64::parseCpu(Ret.CPU)) {
+ FeatureBits.addCPUDefaults(*CpuInfo);
+ SplitAndAddFeatures(Split.second, Ret.Features, FeatureBits);
+ }
}
} else if (Feature.starts_with("tune=")) {
if (!Ret.Tune.empty())
@@ -1169,25 +1130,19 @@ ParsedTargetAttr AArch64TargetInfo::parseTargetAttr(StringRef Features) const {
else
Ret.Tune = Feature.split("=").second.trim();
} else if (Feature.starts_with("+")) {
- SplitAndAddFeatures(Feature, Ret.Features);
- } else if (Feature.starts_with("no-")) {
- StringRef FeatureName =
- llvm::AArch64::getArchExtFeature(Feature.split("-").second);
- if (!FeatureName.empty())
- Ret.Features.push_back("-" + FeatureName.drop_front(1).str());
- else
- Ret.Features.push_back("-" + Feature.split("-").second.str());
+ SplitAndAddFeatures(Feature, Ret.Features, FeatureBits);
} else {
- // Try parsing the string to the internal target feature name. If it is
- // invalid, add the original string (which could already be an internal
- // name). These should be checked later by isValidFeatureName.
- StringRef FeatureName = llvm::AArch64::getArchExtFeature(Feature);
- if (!FeatureName.empty())
- Ret.Features.push_back(FeatureName.str());
+ if (FeatureBits.parseModifier(Feature))
+ continue;
+ // Pushing the original feature string to give a sema error later on
+ // when they get checked.
+ if (Feature.starts_with("no-"))
+ Ret.Features.push_back("-" + Feature.drop_front(3).str());
else
Ret.Features.push_back("+" + Feature.str());
}
}
+ FeatureBits.toLLVMFeatureList(Ret.Features);
return Ret;
}
diff --git a/clang/lib/Basic/Targets/AArch64.h b/clang/lib/Basic/Targets/AArch64.h
index 12fb50286f751..696553ef8038a 100644
--- a/clang/lib/Basic/Targets/AArch64.h
+++ b/clang/lib/Basic/Targets/AArch64.h
@@ -107,10 +107,6 @@ class LLVM_LIBRARY_VISIBILITY AArch64TargetInfo : public TargetInfo {
unsigned multiVersionSortPriority(StringRef Name) const override;
unsigned multiVersionFeatureCost() const override;
- bool
- initFeatureMap(llvm::StringMap<bool> &Features, DiagnosticsEngine &Diags,
- StringRef CPU,
- const std::vector<std::string> &FeaturesVec) const override;
bool useFP16ConversionIntrinsics() const override {
return false;
}
diff --git a/clang/test/CodeGen/aarch64-cpu-supports-target.c b/clang/test/CodeGen/aarch64-cpu-supports-target.c
index e023944b24e53..4a8f7cf8595ca 100644
--- a/clang/test/CodeGen/aarch64-cpu-supports-target.c
+++ b/clang/test/CodeGen/aarch64-cpu-supports-target.c
@@ -48,5 +48,5 @@ int test_versions() {
return code();
}
// CHECK: attributes #0 = { noinline nounwind optnone "no-trapping-math"="true" "stack-protector-buffer-size"="8" }
-// CHECK: attributes #1 = { noinline nounwind optnone "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-features"="+neon" }
+// CHECK: attributes #1 = { noinline nounwind optnone "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-features"="+fp-armv8,+neon" }
// CHECK: attributes #2 = { noinline nounwind optnone "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-features"="+fp-armv8,+fullfp16,+neon,+sve" }
diff --git a/clang/test/CodeGen/aarch64-sme-intrinsics/aarch64-sme-attrs.cpp b/clang/test/CodeGen/aarch64-sme-intrinsics/aarch64-sme-attrs.cpp
index af8933d93d6cb..9885ac45e6a0e 100644
--- a/clang/test/CodeGen/aarch64-sme-intrinsics/aarch64-sme-attrs.cpp
+++ b/clang/test/CodeGen/aarch64-sme-intrinsics/aarch64-sme-attrs.cpp
@@ -1,4 +1,4 @@
-// RUN: %clang_cc1 -triple aarch64-none-linux-gnu -target-feature +sme \
+// RUN: %clang_cc1 -triple aarch64-none-linux-gnu -target-feature +sme -target-feature +bf16 \
// RUN: -disable-O0-optnone -Werror -emit-llvm -o - %s \
// RUN: | opt -S -passes=mem2reg \
// RUN: | opt -S -passes=inline \
diff --git a/clang/test/CodeGen/aarch64-targetattr.c b/clang/test/CodeGen/aarch64-targetattr.c
index 3e7a209245607..f84e72bdac6b7 100644
--- a/clang/test/CodeGen/aarch64-targetattr.c
+++ b/clang/test/CodeGen/aarch64-targetattr.c
@@ -58,58 +58,57 @@ void v1msve() {}
// CHECK-LABEL: @plussve() #12
__attribute__((target("+sve")))
void plussve() {}
-// CHECK-LABEL: @plussveplussve2() #13
+// CHECK-LABEL: @plussveplussve2() #12
__attribute__((target("+sve+nosve2")))
void plussveplussve2() {}
-// CHECK-LABEL: @plussveminusnosve2() #13
+// CHECK-LABEL: @plussveminusnosve2() #12
__attribute__((target("sve,no-sve2")))
void plussveminusnosve2() {}
-// CHECK-LABEL: @plusfp16() #14
+// CHECK-LABEL: @plusfp16() #13
__attribute__((target("+fp16")))
void plusfp16() {}
-// CHECK-LABEL: @all() #15
+// CHECK-LABEL: @all() #14
__attribute__((target("cpu=neoverse-n1,tune=cortex-a710,arch=armv8.6-a+sve2")))
void all() {}
-// CHECK-LABEL: @allplusbranchprotection() #16
+// CHECK-LABEL: @allplusbranchprotection() #15
__attribute__((target("cpu=neoverse-n1,tune=cortex-a710,arch=armv8.6-a+sve2,branch-protection=standard")))
void allplusbranchprotection() {}
// These tests check that the user facing and internal llvm name are both accepted.
-// CHECK-LABEL: @plusnoneon() #17
+// CHECK-LABEL: @plusnoneon() #16
__attribute__((target("+noneon")))
void plusnoneon() {}
-// CHECK-LABEL: @plusnosimd() #17
+// CHECK-LABEL: @plusnosimd() #16
__attribute__((target("+nosimd")))
void plusnosimd() {}
-// CHECK-LABEL: @noneon() #17
+// CHECK-LABEL: @noneon() #16
__attribute__((target("no-neon")))
void noneon() {}
-// CHECK-LABEL: @nosimd() #17
+// CHECK-LABEL: @nosimd() #16
__attribute__((target("no-simd")))
void nosimd() {}
// This isn't part of the standard interface, but test that -arch features should not apply anything else.
-// CHECK-LABEL: @minusarch() #18
+// CHECK-LABEL: @minusarch() #17
__attribute__((target("no-v9.3a")))
void minusarch() {}
// CHECK: attributes #0 = { {{.*}} "target-features"="+crc,+fp-armv8,+lse,+neon,+ras,+rdm,+v8.1a,+v8.2a,+v8a" }
// CHECK: attributes #1 = { {{.*}} "target-features"="+crc,+fp-armv8,+fullfp16,+lse,+neon,+ras,+rdm,+sve,+v8.1a,+v8.2a,+v8a" }
// CHECK: attributes #2 = { {{.*}} "target-features"="+crc,+fp-armv8,+fullfp16,+lse,+neon,+ras,+rdm,+sve,+sve2,+v8.1a,+v8.2a,+v8a" }
-// CHECK: attributes #3 = { {{.*}} "target-features"="+bf16,+complxnum,+crc,+dotprod,+fp-armv8,+fullfp16,+i8mm,+jsconv,+lse,+neon,+pauth,+ras,+rcpc,+rdm,+sve,+sve2,+v8.1a,+v8.2a,+v8.3a,+v8.4a,+v8.5a,+v8.6a,+v8a" }
-// CHECK: attributes #4 = { {{.*}} "target-cpu"="cortex-a710" "target-features"="+bf16,+complxnum,+crc,+dotprod,+flagm,+fp-armv8,+fp16fml,+fullfp16,+i8mm,+jsconv,+lse,+mte,+neon,+pauth,+ras,+rcpc,+rdm,+sb,+sve,+sve2,+sve2-bitperm" }
+// CHECK: attributes #3 = { {{.*}} "target-features"="+bf16,+complxnum,+crc,+dotprod,+fp-armv8,+fp16fml,+fullfp16,+i8mm,+jsconv,+lse,+neon,+pauth,+ras,+rcpc,+rdm,+sve,+sve2,+v8.1a,+v8.2a,+v8.3a,+v8.4a,+v8.5a,+v8.6a,+v8a" }
+// CHECK: attributes #4 = { {{.*}} "target-cpu"="cortex-a710" "target-features"="+bf16,+complxnum,+crc,+dotprod,+flagm,+fp-armv8,+fp16fml,+fullfp16,+i8mm,+jsconv,+lse,+mte,+neon,+pauth,+ras,+rcpc,+rdm,+sb,+sve,+sve2,+sve2-bitperm,+v8.1a,+v8.2a,+v8.3a,+v8.4a,+v8.5a,+v8a,+v9a" }
// CHECK: attributes #5 = { {{.*}} "tune-cpu"="cortex-a710" }
// CHECK: attributes #6 = { {{.*}} "target-cpu"="generic" }
// CHECK: attributes #7 = { {{.*}} "tune-cpu"="generic" }
-// CHECK: attributes #8 = { {{.*}} "target-cpu"="neoverse-n1" "target-features"="+aes,+crc,+dotprod,+fp-armv8,+fullfp16,+lse,+neon,+ras,+rcpc,+rdm,+sha2,+spe,+ssbs" "tune-cpu"="cortex-a710" }
+// CHECK: attributes #8 = { {{.*}} "target-cpu"="neoverse-n1" "target-features"="+aes,+crc,+dotprod,+fp-armv8,+fullfp16,+lse,+neon,+ras,+rcpc,+rdm,+sha2,+spe,+ssbs,+v8.1a,+v8.2a,+v8a" "tune-cpu"="cortex-a710" }
// CHECK: attributes #9 = { {{.*}} "target-features"="+fp-armv8,+fullfp16,+neon,+sve" "tune-cpu"="cortex-a710" }
-// CHECK: attributes #10 = { {{.*}} "target-cpu"="neoverse-v1" "target-features"="+aes,+bf16,+complxnum,+crc,+dotprod,+fp-armv8,+fp16fml,+fullfp16,+i8mm,+jsconv,+lse,+neon,+pauth,+rand,+ras,+rcpc,+rdm,+sha2,+sha3,+sm4,+spe,+ssbs,+sve,+sve2" }
-// CHECK: attributes #11 = { {{.*}} "target-cpu"="neoverse-v1" "target-features"="+aes,+bf16,+complxnum,+crc,+dotprod,+fp-armv8,+fp16fml,+fullfp16,+i8mm,+jsconv,+lse,+neon,+pauth,+rand,+ras,+rcpc,+rdm,+sha2,+sha3,+sm4,+spe,+ssbs,-sve" }
+// CHECK: attributes #10 = { {{.*}} "target-cpu"="neoverse-v1" "target-features"="+aes,+bf16,+complxnum,+crc,+dotprod,+fp-armv8,+fp16fml,+fullfp16,+i8mm,+jsconv,+lse,+neon,+pauth,+rand,+ras,+rcpc,+rdm,+sha2,+sha3,+sm4,+spe,+ssbs,+sve,+sve2,+v8.1a,+v8.2a,+v8.3a,+v8.4a,+v8a" }
+// CHECK: attributes #11 = { {{.*}} "target-cpu"="neoverse-v1" "target-features"="+aes,+bf16,+complxnum,+crc,+dotprod,+fp-armv8,+fp16fml,+fullfp16,+i8mm,+jsconv,+lse,+neon,+pauth,+rand,+ras,+rcpc,+rdm,+sha2,+sha3,+sm4,+spe,+ssbs,+v8.1a,+v8.2a,+v8.3a,+v8.4a,+v8a,-sve" }
// CHECK: attributes #12 = { {{.*}} "target-features"="+fp-armv8,+fullfp16,+neon,+sve" }
-// CHECK: attributes #13 = { {{.*}} "target-features"="+fp-armv8,+fullfp16,+neon,+sve,-sve2" }
-// CHECK: attributes #14 = { {{.*}} "target-features"="+fullfp16" }
-// CHECK: attributes #15 = { {{.*}} "target-cpu"="neoverse-n1" "target-features"="+aes,+bf16,+complxnum,+crc,+dotprod,+fp-armv8,+fullfp16,+i8mm,+jsconv,+lse,+neon,+pauth,+ras,+rcpc,+rdm,+sha2,+spe,+ssbs,+sve,+sve2,+v8.1a,+v8.2a,+v8.3a,+v8.4a,+v8.5a,+v8.6a,+v8a" "tune-cpu"="cortex-a710" }
-// CHECK: attributes #16 = { {{.*}} "branch-target-enforcement"="true" "guarded-control-stack"="true" {{.*}} "target-features"="+aes,+bf16,+complxnum,+crc,+dotprod,+fp-armv8,+fullfp16,+i8mm,+jsconv,+lse,+neon,+pauth,+ras,+rcpc,+rdm,+sha2,+spe,+ssbs,+sve,+sve2,+v8.1a,+v8.2a,+v8.3a,+v8.4a,+v8.5a,+v8.6a,+v8a" "tune-cpu"="cortex-a710" }
-// CHECK: attributes #17 = { {{.*}} "target-features"="-neon" }
-// CHECK: attributes #18 = { {{.*}} "target-features"="-v9.3a" }
+// CHECK: attributes #13 = { {{.*}} "target-features"="+fp-armv8,+fullfp16,+neon" }
+// CHECK: attributes #14 = { {{.*}} "target-cpu"="neoverse-n1" "target-features"="+aes,+bf16,+complxnum,+crc,+dotprod,+fp-armv8,+fullfp16,+i8mm,+jsconv,+lse,+neon,+pauth,+ras,+rcpc,+rdm,+sha2,+spe,+ssbs,+sve,+sve2,+v8.1a,+v8.2a,+v8.3a,+v8.4a,+v8.5a,+v8.6a,+v8a" "tune-cpu"="cortex-a710" }
+// CHECK: attributes #15 = { {{.*}} "branch-target-enforcement"="true" "guarded-control-stack"="true" {{.*}} "target-features"="+aes,+bf16,+complxnum,+crc,+dotprod,+fp-armv8,+fullfp16,+i8mm,+jsconv,+lse,+neon,+pauth,+ras,+rcpc,+rdm,+sha2,+spe,+ssbs,+sve,+sve2,+v8.1a,+v8.2a,+v8.3a,+v8.4a,+v8.5a,+v8.6a,+v8a" "tune-cpu"="cortex-a710" }
+// CHECK-NOT: attributes #16 = {{.*}} "target-features"
+// CHECK: attributes #17 = { {{.*}} "target-features"="-v9.3a" }
diff --git a/clang/test/CodeGen/attr-target-version.c b/clang/test/CodeGen/attr-target-version.c
index 3597711333d34..75f8734e5aaf3 100644
--- a/clang/test/CodeGen/attr-target-version.c
+++ b/clang/test/CodeGen/attr-target-version.c
@@ -1129,42 +1129,42 @@ int caller(void) { return used_def_without_default_decl() + used_decl_without_de
// CHECK-NOFMV-NEXT: ret i32 0
//
//.
-// CHECK: attributes #[[ATTR0]] = { noinline nounwind optnone "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-features"="+flagm,+fp16fml,+fullfp16,+neon,+rand,-fp-armv8,-v9.5a" }
+// CHECK: attributes #[[ATTR0]] = { noinline nounwind optnone "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-features"="+flagm,+fp-armv8,+fp16fml,+fullfp16,+neon,+rand,-v9.5a" }
// CHECK: attributes #[[ATTR1]] = { noinline nounwind optnone "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-features"="+altnzcv,+bf16,+flagm,+sme,+sme-i16i64,-fp-armv8,-v9.5a" }
-// CHECK: attributes #[[ATTR2]] = { noinline nounwind optnone "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-features"="+lse,+neon,+sha2,-fp-armv8,-v9.5a" }
-// CHECK: attributes #[[ATTR3]] = { noinline nounwind optnone "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-features"="+dotprod,+ls64,+neon,-fp-armv8,-v9.5a" }
-// CHECK: attributes #[[ATTR4]] = { noinline nounwind optnone "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-features"="+fp16fml,+fullfp16,+neon,-fp-armv8,-v9.5a" }
-// CHECK: attributes #[[ATTR5]] = { noinline nounwind optnone "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-features"="+neon,-fp-armv8,-v9.5a" }
+// CHECK: attributes #[[ATTR2]] = { noinline nounwind optnone "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-features"="+fp-armv8,+lse,+neon,+sha2,-v9.5a" }
+// CHECK: attributes #[[ATTR3]] = { noinline nounwind optnone "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-features"="+dotprod,+fp-armv8,+ls64,+neon,-v9.5a" }
+// CHECK: attributes #[[ATTR4]] = { noinline nounwind optnone "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-features"="+fp-armv8,+fp16fml,+fullfp16,+neon,-v9.5a" }
+// CHECK: attributes #[[ATTR5]] = { noinline nounwind optnone "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-features"="+fp-armv8,+neon,-v9.5a" }
// CHECK: attributes #[[ATTR6]] = { noinline nounwind optnone "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-features"="+crc,-fp-armv8,-v9.5a" }
// CHECK: attributes #[[ATTR7]] = { noinline nounwind optnone "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-features"="+bti,-fp-armv8,-v9.5a" }
// CHECK: attributes #[[ATTR8]] = { noinline nounwind optnone "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-features"="+bf16,+sme,+sme2,-fp-armv8,-v9.5a" }
// CHECK: attributes #[[ATTR9:[0-9]+]] = { "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-features"="-fp-armv8,-v9.5a" }
// CHECK: attributes #[[ATTR10]] = { noinline nounwind optnone "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-features"="+ccpp,-fp-armv8,-v9.5a" }
// CHECK: attributes #[[ATTR11]] = { noinline nounwind optnone "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-features"="-fp-armv8,-v9.5a" }
-// CHECK: attributes #[[ATTR12]] = { noinline nounwind optnone "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-features"="+fullfp16,+neon,-fp-armv8,-v9.5a" }
+// CHECK: attributes #[[ATTR12]] = { noinline nounwind optnone "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-features"="+fp-armv8,+fullfp16,+neon,-v9.5a" }
// CHECK: attributes #[[ATTR13]] = { noinline nounwind optnone "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-features"="+sb,-fp-armv8,-v9.5a" }
// CHECK: attributes #[[ATTR14]] = { noinline nounwind optnone "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-features"="+mops,-fp-armv8,-v9.5a" }
-// CHECK: attributes #[[ATTR15]] = { noinline nounwind optnone "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-features"="+dotprod,+neon,-fp-armv8,-v9.5a" }
-// CHECK: attributes #[[ATTR16]] = { noinline nounwind optnone "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-features"="+fullfp16,+neon,+sve,-fp-armv8,-v9.5a" }
+// CHECK: attributes #[[ATTR15]] = { noinline nounwind optnone "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-features"="+dotprod,+fp-armv8,+neon,-v9.5a" }
+// CHECK: attributes #[[ATTR16]] = { noinline nounwind optnone "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-features"="+fp-armv8,+fullfp16,+neon,+sve,-v9.5a" }
// CHECK: attributes #[[ATTR17]] = { noinline nounwind optnone "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-features"="+lse,-fp-armv8,-v9.5a" }
-// CHECK: attributes #[[ATTR18]] = { noinline nounwind optnone "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-features"="+neon,+rdm,-fp-armv8,-v9.5a" }
-// CHECK: attributes #[[ATTR19:[0-9]+]] = { "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-features"="+jsconv,+neon,-fp-armv8,-v9.5a" }
-// CHECK: attributes #[[ATTR20:[0-9]+]] = { "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-features"="+neon,+rdm,-fp-armv8,-v9.5a" }
-// CHECK: attributes #[[ATTR21]] = { noinline nounwind optnone "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-features"="+jsconv,+neon,-fp-armv8,-v9.5a" }
-// CHECK: attributes #[[ATTR22]] = { noinline nounwind optnone "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-features"="+aes,+f64mm,+fullfp16,+neon,+sve,-fp-armv8,-v9.5a" }
-// CHECK: attributes #[[ATTR23]] = { noinline nounwind optnone "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-features"="+bf16,+complxnum,+fullfp16,+neon,+rdm,+sme,-fp-armv8,-v9.5a" }
-// CHECK: attributes #[[ATTR24]] = { noinline nounwind optnone "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-features"="+f32mm,+fullfp16,+i8mm,+neon,+sha2,+sha3,+sve,-fp-armv8,-v9.5a" }
-// CHECK: attributes #[[ATTR25]] = { noinline nounwind optnone "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-features"="+bf16,+dit,+fullfp16,+neon,+sve,-fp-armv8,-v9.5a" }
+// CHECK: attributes #[[ATTR18]] = { noinline nounwind optnone "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-features"="+fp-armv8,+neon,+rdm,-v9.5a" }
+// CHECK: attributes #[[ATTR19:[0-9]+]] = { "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-features"="+fp-armv8,+jsconv,+neon,-v9.5a" }
+// CHECK: attributes #[[ATTR20:[0-9]+]] = { "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-features"="+fp-armv8,+neon,+rdm,-v9.5a" }
+// CHECK: attributes #[[ATTR21]] = { noinline nounwind optnone "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-features"="+fp-armv8,+jsconv,+neon,-v9.5a" }
+// CHECK: attributes #[[ATTR22]] = { noinline nounwind optnone "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-features"="+aes,+f64mm,+fp-armv8,+fullfp16,+neon,+sve,-v9.5a" }
+// CHECK: attributes #[[ATTR23]] = { noinline nounwind optnone "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-features"="+bf16,+complxnum,+fp-armv8,+fullfp16,+neon,+rdm,+sme,-v9.5a" }
+// CHECK: attributes #[[ATTR24]] = { noinline nounwind optnone "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-features"="+f32mm,+fp-armv8,+fullfp16,+i8mm,+neon,+sha2,+sha3,+sve,-v9.5a" }
+// CHECK: attributes #[[ATTR25]] = { noinline nounwind optnone "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-features"="+bf16,+dit,+fp-armv8,+fullfp16,+neon,+sve,-v9.5a" }
// CHECK: attributes #[[ATTR26]] = { noinline nounwind optnone "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-features"="+ccpp,+rcpc,-fp-armv8,-v9.5a" }
-// CHECK: attributes #[[ATTR27]] = { noinline nounwind optnone "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-features"="+ccdp,+ccpp,+jsconv,+neon,-fp-armv8,-v9.5a" }
+// CHECK: attributes #[[ATTR27]] = { noinline nounwind optnone "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-features"="+ccdp,+ccpp,+fp-armv8,+jsconv,+neon,-v9.5a" }
// CHECK: attributes #[[ATTR28]] = { noinline nounwind optnone "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-features"="+fptoint,+rcpc,-fp-armv8,-v9.5a" }
-// CHECK: attributes #[[ATTR29]] = { noinline nounwind optnone "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-features"="+bf16,+fullfp16,+neon,+sve,-fp-armv8,-v9.5a" }
-// CHECK: attributes #[[ATTR30]] = { noinline nounwind optnone "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-features"="+fullfp16,+neon,+sve,+sve2,+sve2-aes,+sve2-sha3,-fp-armv8,-v9.5a" }
-// CHECK: attributes #[[ATTR31]] = { noinline nounwind optnone "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-features"="+fullfp16,+neon,+sve,+sve2,+sve2-aes,+sve2-bitperm,-fp-armv8,-v9.5a" }
-// CHECK: attributes #[[ATTR32]] = { noinline nounwind optnone "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-features"="+fullfp16,+mte,+neon,+sve,+sve2,+sve2-sm4,-fp-armv8,-v9.5a" }
+// CHECK: attributes #[[ATTR29]] = { noinline nounwind optnone "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-features"="+bf16,+fp-armv8,+fullfp16,+neon,+sve,-v9.5a" }
+// CHECK: attributes #[[ATTR30]] = { noinline nounwind optnone "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-features"="+fp-armv8,+fullfp16,+neon,+sve,+sve2,+sve2-aes,+sve2-sha3,-v9.5a" }
+// CHECK: attributes #[[ATTR31]] = { noinline nounwind optnone "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-features"="+fp-armv8,+fullfp16,+neon,+sve,+sve2,+sve2-aes,+sve2-bitperm,-v9.5a" }
+// CHECK: attributes #[[ATTR32]] = { noinline nounwind optnone "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-features"="+fp-armv8,+fullfp16,+mte,+neon,+sve,+sve2,+sve2-sm4,-v9.5a" }
// CHECK: attributes #[[ATTR33]] = { noinline nounwind optnone "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-features"="+mops,+mte,+rcpc,+rcpc3,-fp-armv8,-v9.5a" }
-// CHECK: attributes #[[ATTR34]] = { noinline nounwind optnone "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-features"="+neon,+sm4,-fp-armv8,-v9.5a" }
-// CHECK: attributes #[[ATTR35]] = { noinline nounwind optnone "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-features"="+lse,+neon,+rdm,-fp-armv8,-v9.5a" }
+// CHECK: attributes #[[ATTR34]] = { noinline nounwind optnone "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-features"="+fp-armv8,+neon,+sm4,-v9.5a" }
+// CHECK: attributes #[[ATTR35]] = { noinline nounwind optnone "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-features"="+fp-armv8,+lse,+neon,+rdm,-v9.5a" }
//.
// CHECK-NOFMV: attributes #[[ATTR0]] = { noinline nounwind optnone "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-features"="-fmv" }
// CHECK-NOFMV: attributes #[[ATTR1:[0-9]+]] = { "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-features"="-fmv" }
diff --git a/clang/test/Sema/aarch64-neon-target.c b/clang/test/Sema/aarch64-neon-target.c
index fa45fff1d183d..642afddd88c15 100644
--- a/clang/test/Sema/aarch64-neon-target.c
+++ b/clang/test/Sema/aarch64-neon-target.c
@@ -69,8 +69,8 @@ void undefined(uint32x2_t v2i32, uint32x4_t v4i32, uint16x8_t v8i16, uint8x16_t
vrnd_f16(v4f16); // expected-error {{always_inline function 'vrnd_f16' requires target feature 'fullfp16'}}
vmaxnm_f16(v4f16, v4f16); // expected-error {{always_inline function 'vmaxnm_f16' requires target feature 'fullfp16'}}
vrndi_f16(v4f16); // expected-error {{always_inline function 'vrndi_f16' requires target feature 'fullfp16'}}
- // fp16fml
- vfmlal_low_f16(v2f32, v4f16, v4f16); // expected-error {{always_inline function 'vfmlal_low_f16' requires target feature 'fp16fml'}}
+ // fp16fml depends on fp-armv8
+ vfmlal_low_f16(v2f32, v4f16, v4f16); // expected-error {{always_inline function 'vfmlal_low_f16' requires target feature 'fp-armv8'}}
// i8mm
vmmlaq_s32(v4i32, v8i16, v8i16); // expected-error {{always_inline function 'vmmlaq_s32' requires target feature 'i8mm'}}
vusdot_laneq_s32(v2i32, v8i8, v8i16, 0); // expected-error {{always_inline function 'vusdot_s32' requires target feature 'i8mm'}}
diff --git a/clang/test/Sema/builtin-cpu-supports.c b/clang/test/Sema/builtin-cpu-supports.c
index 51ee9661807f8..21a405c5a6bd7 100644
--- a/clang/test/Sema/builtin-cpu-supports.c
+++ b/clang/test/Sema/builtin-cpu-supports.c
@@ -27,7 +27,7 @@ int main(void) {
(void)__builtin_cpu_supports("x86-64-v4");
(void)__builtin_cpu_supports("x86-64-v5"); // expected-warning {{invalid cpu feature string for builtin}}
#else
- if (__builtin_cpu_supports("neon")) // expected-warning {{invalid cpu feature string for builtin}}
+ if (__builtin_cpu_supports("neon"))
a("vsx");
if (__builtin_cpu_is("cortex-x3")) // expected-error {{builtin is not supported on this target}}
diff --git a/llvm/include/llvm/TargetParser/AArch64TargetParser.h b/llvm/include/llvm/TargetParser/AArch64TargetParser.h
index 5025ab2491de8..0a3a3dd695c45 100644
--- a/llvm/include/llvm/TargetParser/AArch64TargetParser.h
+++ b/llvm/include/llvm/TargetParser/AArch64TargetParser.h
@@ -132,48 +132,6 @@ struct ExtensionInfo {
#define EMIT_EXTENSIONS
#include "llvm/TargetParser/AArch64TargetParserDef.inc"
-struct ExtensionSet {
- // Set of extensions which are currently enabled.
- ExtensionBitset Enabled;
- // Set of extensions which have been enabled or disabled at any point. Used
- // to avoid cluttering the cc1 command-line with lots of unneeded features.
- ExtensionBitset Touched;
- // Base architecture version, which we need to know because some feature
- // dependencies change depending on this.
- const ArchInfo *BaseArch;
-
- ExtensionSet() : Enabled(), Touched(), BaseArch(nullptr) {}
-
- // Enable the given architecture extension, and any other extensions it
- // depends on. Does not change the base architecture, or follow dependencies
- // between features which are only related by required arcitecture versions.
- void enable(ArchExtKind E);
-
- // Disable the given architecture extension, and any other extensions which
- // depend on it. Does not change the base architecture, or follow
- // dependencies between features which are only related by required
- // arcitecture versions.
- void disable(ArchExtKind E);
-
- // Add default extensions for the given CPU. Records the base architecture,
- // to later resolve dependencies which depend on it.
- void addCPUDefaults(const CpuInfo &CPU);
-
- // Add default extensions for the given architecture version. Records the
- // base architecture, to later resolve dependencies which depend on it.
- void addArchDefaults(const ArchInfo &Arch);
-
- // Add or remove a feature based on a modifier string. The string must be of
- // the form "<name>" to enable a feature or "no<name>" to disable it. This
- // will also enable or disable any features as required by the dependencies
- // between them.
- bool parseModifier(StringRef Modifier);
-
- // Convert the set of enabled extension to an LLVM feature list, appending
- // them to Features.
- void toLLVMFeatureList(std::vector<StringRef> &Features) const;
-};
-
// Represents a dependency between two architecture extensions. Later is the
// feature which was added to the architecture after Earlier, and expands the
// functionality provided by it. If Later is enabled, then Earlier will also be
@@ -579,6 +537,62 @@ inline constexpr CpuInfo CpuInfos[] = {
AArch64::AEK_SSBS, AArch64::AEK_CSSC})},
};
+struct ExtensionSet {
+ // Set of extensions which are currently enabled.
+ ExtensionBitset Enabled;
+ // Set of extensions which have been enabled or disabled at any point. Used
+ // to avoid cluttering the cc1 command-line with lots of unneeded features.
+ ExtensionBitset Touched;
+ // Base architecture version, which we need to know because some feature
+ // dependencies change depending on this.
+ const ArchInfo *BaseArch;
+
+ ExtensionSet() : Enabled(), Touched(), BaseArch(nullptr) {}
+
+ // Enable the given architecture extension, and any other extensions it
+ // depends on. Does not change the base architecture, or follow dependencies
+ // between features which are only related by required arcitecture versions.
+ void enable(ArchExtKind E);
+
+ // Disable the given architecture extension, and any other extensions which
+ // depend on it. Does not change the base architecture, or follow
+ // dependencies between features which are only related by required
+ // arcitecture versions.
+ void disable(ArchExtKind E);
+
+ // Add default extensions for the given CPU. Records the base architecture,
+ // to later resolve dependencies which depend on it.
+ void addCPUDefaults(const CpuInfo &CPU);
+
+ // Add default extensions for the given architecture version. Records the
+ // base architecture, to later resolve dependencies which depend on it.
+ void addArchDefaults(const ArchInfo &Arch);
+
+ // Add or remove a feature based on a modifier string. The string must be of
+ // the form "<name>" to enable a feature or "no<name>" to disable it. This
+ // will also enable or disable any features as required by the dependencies
+ // between them.
+ bool parseModifier(StringRef Modifier);
+
+ void reconstructFromParsedFeatures(std::vector<std::string> &Features);
+
+ // Convert the set of enabled extension to an LLVM feature list, appending
+ // them to Features.
+ template <typename T> void toLLVMFeatureList(std::vector<T> &Features) const {
+ if (BaseArch && !BaseArch->ArchFeature.empty())
+ Features.emplace_back(T(BaseArch->ArchFeature));
+
+ for (const auto &E : Extensions) {
+ if (E.Feature.empty() || !Touched.test(E.ID))
+ continue;
+ if (Enabled.test(E.ID))
+ Features.emplace_back(T(E.Feature));
+ else
+ Features.emplace_back(T(E.NegFeature));
+ }
+ }
+};
+
// Name alias.
struct Alias {
StringRef AltName;
@@ -602,7 +616,13 @@ const ArchInfo *getArchForCpu(StringRef CPU);
// Parser
const ArchInfo *parseArch(StringRef Arch);
+
+// Return the extension which has the given -target-feature name.
+std::optional<ExtensionInfo> targetFeatureToExtension(StringRef TargetFeature);
+
+// Parse a name as defined by the Extension class in tablegen.
std::optional<ExtensionInfo> parseArchExtension(StringRef Extension);
+
// Given the name of a CPU or alias, return the correponding CpuInfo.
std::optional<CpuInfo> parseCpu(StringRef Name);
// Used by target parser tests
diff --git a/llvm/lib/Target/AArch64/AArch64Features.td b/llvm/lib/Target/AArch64/AArch64Features.td
index ffb899a301459..8269cf19a210f 100644
--- a/llvm/lib/Target/AArch64/AArch64Features.td
+++ b/llvm/lib/Target/AArch64/AArch64Features.td
@@ -106,7 +106,7 @@ def FeatureFPARMv8 : Extension<"fp-armv8", "FPARMv8",
"Enable ARMv8 (FEAT_FP)", [],
"FEAT_FP", "+fp-armv8,+neon", 90>;
-let ArchExtKindSpelling = "AEK_SIMD", MArchName = "simd" in
+let ArchExtKindSpelling = "AEK_SIMD", MArchName = "simd", MArchAlias = "neon" in
def FeatureNEON : Extension<"neon", "NEON",
"Enable Advanced SIMD instructions (FEAT_AdvSIMD)", [FeatureFPARMv8],
"FEAT_SIMD", "+fp-armv8,+neon", 100>;
diff --git a/llvm/lib/TargetParser/AArch64TargetParser.cpp b/llvm/lib/TargetParser/AArch64TargetParser.cpp
index ca356ec82bf1f..b448682cb6c01 100644
--- a/llvm/lib/TargetParser/AArch64TargetParser.cpp
+++ b/llvm/lib/TargetParser/AArch64TargetParser.cpp
@@ -122,6 +122,14 @@ AArch64::parseArchExtension(StringRef ArchExt) {
return {};
}
+std::optional<AArch64::ExtensionInfo>
+AArch64::targetFeatureToExtension(StringRef TargetFeature) {
+ for (const auto &E : Extensions)
+ if (TargetFeature == E.Feature)
+ return E;
+ return {};
+}
+
std::optional<AArch64::CpuInfo> AArch64::parseCpu(StringRef Name) {
// Resolve aliases first.
Name = resolveCPUAlias(Name);
@@ -166,6 +174,11 @@ void AArch64::ExtensionSet::enable(ArchExtKind E) {
Touched.set(E);
Enabled.set(E);
+ // These depend on each other, meaning you can't have one without the other.
+ // We don't want this to be in ExtensionDependencies to avoid infinite loops.
+ if (E == AEK_FP)
+ enable(AEK_SIMD);
+
// Recursively enable all features that this one depends on. This handles all
// of the simple cases, where the behaviour doesn't depend on the base
// architecture version.
@@ -213,21 +226,6 @@ void AArch64::ExtensionSet::disable(ArchExtKind E) {
disable(Dep.Later);
}
-void AArch64::ExtensionSet::toLLVMFeatureList(
- std::vector<StringRef> &Features) const {
- if (BaseArch && !BaseArch->ArchFeature.empty())
- Features.push_back(BaseArch->ArchFeature);
-
- for (const auto &E : Extensions) {
- if (E.Feature.empty() || !Touched.test(E.ID))
- continue;
- if (Enabled.test(E.ID))
- Features.push_back(E.Feature);
- else
- Features.push_back(E.NegFeature);
- }
-}
-
void AArch64::ExtensionSet::addCPUDefaults(const CpuInfo &CPU) {
LLVM_DEBUG(llvm::dbgs() << "addCPUDefaults(" << CPU.Name << ")\n");
BaseArch = &CPU.Arch;
@@ -250,13 +248,17 @@ void AArch64::ExtensionSet::addArchDefaults(const ArchInfo &Arch) {
bool AArch64::ExtensionSet::parseModifier(StringRef Modifier) {
LLVM_DEBUG(llvm::dbgs() << "parseModifier(" << Modifier << ")\n");
- bool IsNegated = Modifier.starts_with("no");
- StringRef ArchExt = IsNegated ? Modifier.drop_front(2) : Modifier;
+ size_t NChars = 0;
+ if (Modifier.starts_with("no-"))
+ NChars = 3;
+ else if (Modifier.starts_with("no"))
+ NChars = 2;
+ StringRef ArchExt = Modifier.drop_front(NChars);
if (auto AE = parseArchExtension(ArchExt)) {
if (AE->Feature.empty() || AE->NegFeature.empty())
return false;
- if (IsNegated)
+ if (NChars)
disable(AE->ID);
else
enable(AE->ID);
@@ -265,6 +267,21 @@ bool AArch64::ExtensionSet::parseModifier(StringRef Modifier) {
return false;
}
+void AArch64::ExtensionSet::reconstructFromParsedFeatures(
+ std::vector<std::string> &Features) {
+ assert(Touched.none() && "Bitset already initialized");
+ for (auto &F : Features) {
+ bool IsNegated = F[0] == '-';
+ if (auto AE = targetFeatureToExtension(F)) {
+ Touched.set(AE->ID);
+ if (IsNegated)
+ Enabled.reset(AE->ID);
+ else
+ Enabled.set(AE->ID);
+ }
+ }
+}
+
const AArch64::ExtensionInfo &
AArch64::getExtensionByID(AArch64::ArchExtKind ExtID) {
return lookupExtensionByID(ExtID);
>From 95c4244f2fc3452361cc14b1054392fb186731f0 Mon Sep 17 00:00:00 2001
From: Alexandros Lamprineas <alexandros.lamprineas at arm.com>
Date: Wed, 5 Jun 2024 12:06:41 +0100
Subject: [PATCH 2/3] Changes from last revision:
* return getFMVBackendFeaturesFor by value
* add comments in AArch64TargetInfo::parseTargetAttr()
* reject 'neon' in target attribute
* add constness to the argument of reconstructFromParsedFeatures()
* split ExtensionSet::parseModifier in two (cmdline vs attribute)
* disable AEK_FP when disabling AEK_SIMD and adjust tests
---
clang/lib/AST/ASTContext.cpp | 17 ++++++-----
clang/lib/Basic/Targets/AArch64.cpp | 18 +++++++----
clang/lib/Driver/ToolChains/Arch/AArch64.cpp | 2 +-
clang/test/CodeGen/aarch64-targetattr.c | 7 -----
.../Preprocessor/aarch64-target-features.c | 16 ++--------
clang/test/Sema/builtin-cpu-supports.c | 2 +-
.../llvm/TargetParser/AArch64TargetParser.h | 16 ++++++++--
llvm/lib/Target/AArch64/AArch64Features.td | 2 +-
llvm/lib/TargetParser/AArch64TargetParser.cpp | 30 ++++++++++++++++---
.../TargetParser/TargetParserTest.cpp | 4 +--
10 files changed, 67 insertions(+), 47 deletions(-)
diff --git a/clang/lib/AST/ASTContext.cpp b/clang/lib/AST/ASTContext.cpp
index 6b7cfbe48796a..d452f5fec0991 100644
--- a/clang/lib/AST/ASTContext.cpp
+++ b/clang/lib/AST/ASTContext.cpp
@@ -13665,9 +13665,9 @@ QualType ASTContext::getCorrespondingSignedFixedPointType(QualType Ty) const {
}
// Given a list of FMV features, add each of their backend features to the list.
-static void
-getFMVBackendFeaturesFor(const llvm::SmallVectorImpl<StringRef> &FMVFeatStrings,
- std::vector<std::string> &BackendFeats) {
+static std::vector<std::string>
+getFMVBackendFeaturesFor(const llvm::SmallVectorImpl<StringRef> &FMVFeatStrings) {
+ std::vector<std::string> BackendFeats;
for (StringRef F : FMVFeatStrings) {
if (auto FMVExt = llvm::AArch64::parseArchExtension(F)) {
SmallVector<StringRef, 8> Feats;
@@ -13676,6 +13676,7 @@ getFMVBackendFeaturesFor(const llvm::SmallVectorImpl<StringRef> &FMVFeatStrings,
BackendFeats.push_back(F.str());
}
}
+ return BackendFeats;
}
ParsedTargetAttr
@@ -13736,27 +13737,27 @@ void ASTContext::getFunctionFeatureMap(llvm::StringMap<bool> &FeatureMap,
Target->getTargetOpts().FeaturesAsWritten.end());
Target->initFeatureMap(FeatureMap, getDiagnostics(), TargetCPU, Features);
} else if (const auto *TC = FD->getAttr<TargetClonesAttr>()) {
- std::vector<std::string> Features;
if (Target->getTriple().isAArch64()) {
llvm::SmallVector<StringRef, 8> Feats;
TC->getFeatures(Feats, GD.getMultiVersionIndex());
- getFMVBackendFeaturesFor(Feats, Features);
+ std::vector<std::string> Features = getFMVBackendFeaturesFor(Feats);
Features.insert(Features.begin(),
Target->getTargetOpts().FeaturesAsWritten.begin(),
Target->getTargetOpts().FeaturesAsWritten.end());
+ Target->initFeatureMap(FeatureMap, getDiagnostics(), TargetCPU, Features);
} else {
+ std::vector<std::string> Features;
StringRef VersionStr = TC->getFeatureStr(GD.getMultiVersionIndex());
if (VersionStr.starts_with("arch="))
TargetCPU = VersionStr.drop_front(sizeof("arch=") - 1);
else if (VersionStr != "default")
Features.push_back((StringRef{"+"} + VersionStr).str());
+ Target->initFeatureMap(FeatureMap, getDiagnostics(), TargetCPU, Features);
}
- Target->initFeatureMap(FeatureMap, getDiagnostics(), TargetCPU, Features);
} else if (const auto *TV = FD->getAttr<TargetVersionAttr>()) {
llvm::SmallVector<StringRef, 8> Feats;
TV->getFeatures(Feats);
- std::vector<std::string> Features;
- getFMVBackendFeaturesFor(Feats, Features);
+ std::vector<std::string> Features = getFMVBackendFeaturesFor(Feats);
Features.insert(Features.begin(),
Target->getTargetOpts().FeaturesAsWritten.begin(),
Target->getTargetOpts().FeaturesAsWritten.end());
diff --git a/clang/lib/Basic/Targets/AArch64.cpp b/clang/lib/Basic/Targets/AArch64.cpp
index d8bb4fa3a3b49..3127da425ce81 100644
--- a/clang/lib/Basic/Targets/AArch64.cpp
+++ b/clang/lib/Basic/Targets/AArch64.cpp
@@ -1056,6 +1056,12 @@ bool AArch64TargetInfo::handleTargetFeatures(std::vector<std::string> &Features,
// "tune=<cpu>" - TuneCPU set to <cpu>
// "feature", "no-feature" - Add (or remove) feature.
// "+feature", "+nofeature" - Add (or remove) feature.
+//
+// A feature may correspond to an Extension (anything with a corresponding
+// AEK_), in which case an ExtensionSet is used to parse it and expand its
+// dependencies. Otherwise the feature is passed through (e.g. +v8.1a,
+// +outline-atomics, -fmv, etc). Features coming from the command line are
+// already parsed, therefore their dependencies do not need expansion.
ParsedTargetAttr AArch64TargetInfo::parseTargetAttr(StringRef Features) const {
ParsedTargetAttr Ret;
if (Features == "default")
@@ -1070,10 +1076,10 @@ ParsedTargetAttr AArch64TargetInfo::parseTargetAttr(StringRef Features) const {
SmallVector<StringRef, 8> SplitFeatures;
FeatString.split(SplitFeatures, StringRef("+"), -1, false);
for (StringRef Feature : SplitFeatures) {
- if (FeatureBits.parseModifier(Feature))
+ if (FeatureBits.parseAttributeModifier(Feature))
continue;
- // Pushing the original feature string to give a sema error later on
- // when they get checked.
+ // Pass through features that are not extensions, e.g. +v8.1a,
+ // +outline-atomics, -fmv, etc.
if (Feature.starts_with("no"))
Features.push_back("-" + Feature.drop_front(2).str());
else
@@ -1132,10 +1138,10 @@ ParsedTargetAttr AArch64TargetInfo::parseTargetAttr(StringRef Features) const {
} else if (Feature.starts_with("+")) {
SplitAndAddFeatures(Feature, Ret.Features, FeatureBits);
} else {
- if (FeatureBits.parseModifier(Feature))
+ if (FeatureBits.parseAttributeModifier(Feature))
continue;
- // Pushing the original feature string to give a sema error later on
- // when they get checked.
+ // Pass through features that are not extensions, e.g. +v8.1a,
+ // +outline-atomics, -fmv, etc.
if (Feature.starts_with("no-"))
Ret.Features.push_back("-" + Feature.drop_front(3).str());
else
diff --git a/clang/lib/Driver/ToolChains/Arch/AArch64.cpp b/clang/lib/Driver/ToolChains/Arch/AArch64.cpp
index 2cd2b35ee51bc..fa21ffcf64b3b 100644
--- a/clang/lib/Driver/ToolChains/Arch/AArch64.cpp
+++ b/clang/lib/Driver/ToolChains/Arch/AArch64.cpp
@@ -81,7 +81,7 @@ static bool DecodeAArch64Features(const Driver &D, StringRef text,
D.Diag(clang::diag::err_drv_no_neon_modifier);
continue;
}
- if (!Extensions.parseModifier(Feature))
+ if (!Extensions.parseCmdLineOptModifier(Feature))
return false;
}
diff --git a/clang/test/CodeGen/aarch64-targetattr.c b/clang/test/CodeGen/aarch64-targetattr.c
index f84e72bdac6b7..c9272c235d4f6 100644
--- a/clang/test/CodeGen/aarch64-targetattr.c
+++ b/clang/test/CodeGen/aarch64-targetattr.c
@@ -75,16 +75,9 @@ void all() {}
__attribute__((target("cpu=neoverse-n1,tune=cortex-a710,arch=armv8.6-a+sve2,branch-protection=standard")))
void allplusbranchprotection() {}
-// These tests check that the user facing and internal llvm name are both accepted.
-// CHECK-LABEL: @plusnoneon() #16
-__attribute__((target("+noneon")))
-void plusnoneon() {}
// CHECK-LABEL: @plusnosimd() #16
__attribute__((target("+nosimd")))
void plusnosimd() {}
-// CHECK-LABEL: @noneon() #16
-__attribute__((target("no-neon")))
-void noneon() {}
// CHECK-LABEL: @nosimd() #16
__attribute__((target("no-simd")))
void nosimd() {}
diff --git a/clang/test/Preprocessor/aarch64-target-features.c b/clang/test/Preprocessor/aarch64-target-features.c
index 82304a15a04a3..0d5dd35e7632f 100644
--- a/clang/test/Preprocessor/aarch64-target-features.c
+++ b/clang/test/Preprocessor/aarch64-target-features.c
@@ -261,18 +261,6 @@
// CHECK-FULLFP16-VECTOR-SCALAR: #define __ARM_FP 0xE
// CHECK-FULLFP16-VECTOR-SCALAR: #define __ARM_FP16_FORMAT_IEEE 1
-// +fp16fml+nosimd doesn't make sense as the fp16fml instructions all require SIMD.
-// However, as +fp16fml implies +fp16 there is a set of defines that we would expect.
-// RUN: %clang --target=aarch64 -march=armv8-a+fp16fml+nosimd -x c -E -dM %s -o - | FileCheck -match-full-lines --check-prefix=CHECK-FULLFP16-SCALAR %s
-// RUN: %clang --target=aarch64 -march=armv8-a+fp16+nosimd -x c -E -dM %s -o - | FileCheck -match-full-lines --check-prefix=CHECK-FULLFP16-SCALAR %s
-// RUN: %clang --target=aarch64 -march=armv8.4-a+fp16fml+nosimd -x c -E -dM %s -o - | FileCheck -match-full-lines --check-prefix=CHECK-FULLFP16-SCALAR %s
-// RUN: %clang --target=aarch64 -march=armv8.4-a+fp16+nosimd -x c -E -dM %s -o - | FileCheck -match-full-lines --check-prefix=CHECK-FULLFP16-SCALAR %s
-// CHECK-FULLFP16-SCALAR-NOT: #define __ARM_FEATURE_FP16_FML 1
-// CHECK-FULLFP16-SCALAR: #define __ARM_FEATURE_FP16_SCALAR_ARITHMETIC 1
-// CHECK-FULLFP16-SCALAR-NOT: #define __ARM_FEATURE_FP16_VECTOR_ARITHMETIC 1
-// CHECK-FULLFP16-SCALAR: #define __ARM_FP 0xE
-// CHECK-FULLFP16-SCALAR: #define __ARM_FP16_FORMAT_IEEE 1
-
// RUN: %clang --target=aarch64 -march=armv8.2-a -x c -E -dM %s -o - | FileCheck -match-full-lines --check-prefix=CHECK-FULLFP16-NOFML-VECTOR-SCALAR %s
// RUN: %clang --target=aarch64 -march=armv8.2-a+nofp16 -x c -E -dM %s -o - | FileCheck -match-full-lines --check-prefix=CHECK-FULLFP16-NOFML-VECTOR-SCALAR %s
// RUN: %clang --target=aarch64 -march=armv8.2-a+nofp16fml -x c -E -dM %s -o - | FileCheck -match-full-lines --check-prefix=CHECK-FULLFP16-NOFML-VECTOR-SCALAR %s
@@ -464,7 +452,7 @@
// RUN: %clang -target aarch64 -mcpu=cortex-a53+noSIMD -### -c %s 2>&1 | FileCheck -check-prefix=CHECK-MCPU-3 %s
// CHECK-MCPU-1: "-cc1"{{.*}} "-triple" "aarch64{{.*}}" "-target-feature" "-aes"{{.*}} "-target-feature" "-sha2"
// CHECK-MCPU-2: "-cc1"{{.*}} "-triple" "aarch64{{.*}}" "-target-feature" "+crc" "-target-feature" "+fp-armv8" "-target-feature" "+neon"
-// CHECK-MCPU-3: "-cc1"{{.*}} "-triple" "aarch64{{.*}}" "-target-feature" "+v8a" "-target-feature" "-aes" "-target-feature" "+crc" "-target-feature" "+fp-armv8" "-target-feature" "-sha2" "-target-feature" "-neon"
+// CHECK-MCPU-3: "-cc1"{{.*}} "-triple" "aarch64{{.*}}" "-target-feature" "+v8a" "-target-feature" "-aes" "-target-feature" "+crc" "-target-feature" "-fp-armv8" "-target-feature" "-sha2" "-target-feature" "-neon"
// RUN: %clang -target aarch64 -mcpu=cyclone+nocrc+nocrypto -march=armv8-a -### -c %s 2>&1 | FileCheck -check-prefix=CHECK-MCPU-MARCH %s
// RUN: %clang -target aarch64 -march=armv8-a -mcpu=cyclone+nocrc+nocrypto -### -c %s 2>&1 | FileCheck -check-prefix=CHECK-MCPU-MARCH %s
@@ -494,7 +482,7 @@
// RUN: %clang -target aarch64 -march=armv8.1a+noSIMD -### -c %s 2>&1 | FileCheck -check-prefix=CHECK-V81A-FEATURE-3 %s
// CHECK-V81A-FEATURE-1: "-cc1"{{.*}} "-triple" "aarch64{{.*}}" "-target-feature" "+v8.1a" "-target-feature" "+aes" "-target-feature" "+crc" "-target-feature" "+crypto" "-target-feature" "+fp-armv8" "-target-feature" "+lse" "-target-feature" "+rdm" "-target-feature" "+sha2" "-target-feature" "+neon"
// CHECK-V81A-FEATURE-2: "-cc1"{{.*}} "-triple" "aarch64{{.*}}" "-target-feature" "+v8.1a" "-target-feature" "+crc" "-target-feature" "+fp-armv8" "-target-feature" "+lse" "-target-feature" "+rdm" "-target-feature" "+neon"
-// CHECK-V81A-FEATURE-3: "-cc1"{{.*}} "-triple" "aarch64{{.*}}" "-target-feature" "+v8.1a" "-target-feature" "+crc" "-target-feature" "+fp-armv8" "-target-feature" "+lse" "-target-feature" "-rdm" "-target-feature" "-neon"
+// CHECK-V81A-FEATURE-3: "-cc1"{{.*}} "-triple" "aarch64{{.*}}" "-target-feature" "+v8.1a" "-target-feature" "+crc" "-target-feature" "-fp-armv8" "-target-feature" "+lse" "-target-feature" "-rdm" "-target-feature" "-neon"
// ================== Check Memory Tagging Extensions (MTE).
// RUN: %clang -target arm64-none-linux-gnu -march=armv8.5-a+memtag -x c -E -dM %s -o - 2>&1 | FileCheck -check-prefix=CHECK-MEMTAG %s
diff --git a/clang/test/Sema/builtin-cpu-supports.c b/clang/test/Sema/builtin-cpu-supports.c
index 21a405c5a6bd7..51ee9661807f8 100644
--- a/clang/test/Sema/builtin-cpu-supports.c
+++ b/clang/test/Sema/builtin-cpu-supports.c
@@ -27,7 +27,7 @@ int main(void) {
(void)__builtin_cpu_supports("x86-64-v4");
(void)__builtin_cpu_supports("x86-64-v5"); // expected-warning {{invalid cpu feature string for builtin}}
#else
- if (__builtin_cpu_supports("neon"))
+ if (__builtin_cpu_supports("neon")) // expected-warning {{invalid cpu feature string for builtin}}
a("vsx");
if (__builtin_cpu_is("cortex-x3")) // expected-error {{builtin is not supported on this target}}
diff --git a/llvm/include/llvm/TargetParser/AArch64TargetParser.h b/llvm/include/llvm/TargetParser/AArch64TargetParser.h
index 0a3a3dd695c45..903c6f255f622 100644
--- a/llvm/include/llvm/TargetParser/AArch64TargetParser.h
+++ b/llvm/include/llvm/TargetParser/AArch64TargetParser.h
@@ -571,10 +571,20 @@ struct ExtensionSet {
// Add or remove a feature based on a modifier string. The string must be of
// the form "<name>" to enable a feature or "no<name>" to disable it. This
// will also enable or disable any features as required by the dependencies
- // between them.
- bool parseModifier(StringRef Modifier);
+ // between them. The function is used for command line option parsing.
+ bool parseCmdLineOptModifier(StringRef Modifier);
- void reconstructFromParsedFeatures(std::vector<std::string> &Features);
+ // Add or remove a feature based on a modifier string. The string must be of
+ // the form "<name>" to enable a feature or "no(-)<name>" to disable it. This
+ // will also enable or disable any features as required by the dependencies
+ // between them. The function is used for target(_version/_clones) attribute
+ // parsing.
+ bool parseAttributeModifier(StringRef Modifier);
+
+ // Constructs a new ExtensionSet by toggling the corresponding bits for every
+ // feature in the \p Features list without expanding their dependencies. Used
+ // for reconstructing an ExtensionSet from the output of toLLVMFeatures().
+ void reconstructFromParsedFeatures(const std::vector<std::string> &Features);
// Convert the set of enabled extension to an LLVM feature list, appending
// them to Features.
diff --git a/llvm/lib/Target/AArch64/AArch64Features.td b/llvm/lib/Target/AArch64/AArch64Features.td
index 8269cf19a210f..ffb899a301459 100644
--- a/llvm/lib/Target/AArch64/AArch64Features.td
+++ b/llvm/lib/Target/AArch64/AArch64Features.td
@@ -106,7 +106,7 @@ def FeatureFPARMv8 : Extension<"fp-armv8", "FPARMv8",
"Enable ARMv8 (FEAT_FP)", [],
"FEAT_FP", "+fp-armv8,+neon", 90>;
-let ArchExtKindSpelling = "AEK_SIMD", MArchName = "simd", MArchAlias = "neon" in
+let ArchExtKindSpelling = "AEK_SIMD", MArchName = "simd" in
def FeatureNEON : Extension<"neon", "NEON",
"Enable Advanced SIMD instructions (FEAT_AdvSIMD)", [FeatureFPARMv8],
"FEAT_SIMD", "+fp-armv8,+neon", 100>;
diff --git a/llvm/lib/TargetParser/AArch64TargetParser.cpp b/llvm/lib/TargetParser/AArch64TargetParser.cpp
index b448682cb6c01..3cbacf07243bc 100644
--- a/llvm/lib/TargetParser/AArch64TargetParser.cpp
+++ b/llvm/lib/TargetParser/AArch64TargetParser.cpp
@@ -220,6 +220,9 @@ void AArch64::ExtensionSet::disable(ArchExtKind E) {
Touched.set(E);
Enabled.reset(E);
+ if (E == AEK_SIMD)
+ disable(AEK_FP);
+
// Recursively disable all features that depends on this one.
for (auto Dep : ExtensionDependencies)
if (E == Dep.Earlier)
@@ -245,20 +248,39 @@ void AArch64::ExtensionSet::addArchDefaults(const ArchInfo &Arch) {
enable(E.ID);
}
-bool AArch64::ExtensionSet::parseModifier(StringRef Modifier) {
- LLVM_DEBUG(llvm::dbgs() << "parseModifier(" << Modifier << ")\n");
+bool AArch64::ExtensionSet::parseCmdLineOptModifier(StringRef Modifier) {
+ LLVM_DEBUG(llvm::dbgs() << "parseCmdLineOptModifier(" << Modifier << ")\n");
+
+ bool IsNegated = Modifier.starts_with("no");
+ StringRef ArchExt = IsNegated ? Modifier.drop_front(2) : Modifier;
+
+ if (auto AE = parseArchExtension(ArchExt)) {
+ if (AE->Feature.empty() || AE->NegFeature.empty())
+ return false;
+ if (IsNegated)
+ disable(AE->ID);
+ else
+ enable(AE->ID);
+ return true;
+ }
+ return false;
+}
+
+bool AArch64::ExtensionSet::parseAttributeModifier(StringRef Modifier) {
+ LLVM_DEBUG(llvm::dbgs() << "parseAttributeModifier(" << Modifier << ")\n");
size_t NChars = 0;
if (Modifier.starts_with("no-"))
NChars = 3;
else if (Modifier.starts_with("no"))
NChars = 2;
+ bool IsNegated = NChars != 0;
StringRef ArchExt = Modifier.drop_front(NChars);
if (auto AE = parseArchExtension(ArchExt)) {
if (AE->Feature.empty() || AE->NegFeature.empty())
return false;
- if (NChars)
+ if (IsNegated)
disable(AE->ID);
else
enable(AE->ID);
@@ -268,7 +290,7 @@ bool AArch64::ExtensionSet::parseModifier(StringRef Modifier) {
}
void AArch64::ExtensionSet::reconstructFromParsedFeatures(
- std::vector<std::string> &Features) {
+ const std::vector<std::string> &Features) {
assert(Touched.none() && "Bitset already initialized");
for (auto &F : Features) {
bool IsNegated = F[0] == '-';
diff --git a/llvm/unittests/TargetParser/TargetParserTest.cpp b/llvm/unittests/TargetParser/TargetParserTest.cpp
index 797d7dfbca200..babb958ffc799 100644
--- a/llvm/unittests/TargetParser/TargetParserTest.cpp
+++ b/llvm/unittests/TargetParser/TargetParserTest.cpp
@@ -2303,7 +2303,7 @@ TEST_P(AArch64ExtensionDependenciesBaseArchTestFixture,
llvm::AArch64::ExtensionSet Extensions;
Extensions.addArchDefaults(Params.Arch);
for (auto M : Params.Modifiers) {
- bool success = Extensions.parseModifier(M);
+ bool success = Extensions.parseCmdLineOptModifier(M);
EXPECT_TRUE(success);
}
std::vector<StringRef> Features;
@@ -2337,7 +2337,7 @@ TEST_P(AArch64ExtensionDependenciesBaseCPUTestFixture,
EXPECT_TRUE(CPU);
Extensions.addCPUDefaults(*CPU);
for (auto M : Params.Modifiers) {
- bool success = Extensions.parseModifier(M);
+ bool success = Extensions.parseCmdLineOptModifier(M);
EXPECT_TRUE(success);
}
std::vector<StringRef> Features;
>From b8c1ef71dae418fdaf7e9bd811d77a3eaf6f109d Mon Sep 17 00:00:00 2001
From: Alexandros Lamprineas <alexandros.lamprineas at arm.com>
Date: Wed, 5 Jun 2024 14:38:39 +0100
Subject: [PATCH 3/3] Changes from last revision:
* clang format
* Do not make AEK_FP depend on AEK_SIMD.
---
clang/lib/AST/ASTContext.cpp | 4 ++--
clang/test/CodeGen/aarch64-cpu-supports-target.c | 2 +-
clang/test/CodeGen/aarch64-targetattr.c | 6 +++---
.../test/Preprocessor/aarch64-target-features.c | 16 ++++++++++++++--
llvm/lib/TargetParser/AArch64TargetParser.cpp | 8 --------
5 files changed, 20 insertions(+), 16 deletions(-)
diff --git a/clang/lib/AST/ASTContext.cpp b/clang/lib/AST/ASTContext.cpp
index d452f5fec0991..f5a79e9a6e3be 100644
--- a/clang/lib/AST/ASTContext.cpp
+++ b/clang/lib/AST/ASTContext.cpp
@@ -13665,8 +13665,8 @@ QualType ASTContext::getCorrespondingSignedFixedPointType(QualType Ty) const {
}
// Given a list of FMV features, add each of their backend features to the list.
-static std::vector<std::string>
-getFMVBackendFeaturesFor(const llvm::SmallVectorImpl<StringRef> &FMVFeatStrings) {
+static std::vector<std::string> getFMVBackendFeaturesFor(
+ const llvm::SmallVectorImpl<StringRef> &FMVFeatStrings) {
std::vector<std::string> BackendFeats;
for (StringRef F : FMVFeatStrings) {
if (auto FMVExt = llvm::AArch64::parseArchExtension(F)) {
diff --git a/clang/test/CodeGen/aarch64-cpu-supports-target.c b/clang/test/CodeGen/aarch64-cpu-supports-target.c
index 4a8f7cf8595ca..28187bcf74533 100644
--- a/clang/test/CodeGen/aarch64-cpu-supports-target.c
+++ b/clang/test/CodeGen/aarch64-cpu-supports-target.c
@@ -49,4 +49,4 @@ int test_versions() {
}
// CHECK: attributes #0 = { noinline nounwind optnone "no-trapping-math"="true" "stack-protector-buffer-size"="8" }
// CHECK: attributes #1 = { noinline nounwind optnone "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-features"="+fp-armv8,+neon" }
-// CHECK: attributes #2 = { noinline nounwind optnone "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-features"="+fp-armv8,+fullfp16,+neon,+sve" }
+// CHECK: attributes #2 = { noinline nounwind optnone "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-features"="+fp-armv8,+fullfp16,+sve" }
diff --git a/clang/test/CodeGen/aarch64-targetattr.c b/clang/test/CodeGen/aarch64-targetattr.c
index c9272c235d4f6..644e6a692c3be 100644
--- a/clang/test/CodeGen/aarch64-targetattr.c
+++ b/clang/test/CodeGen/aarch64-targetattr.c
@@ -96,11 +96,11 @@ void minusarch() {}
// CHECK: attributes #6 = { {{.*}} "target-cpu"="generic" }
// CHECK: attributes #7 = { {{.*}} "tune-cpu"="generic" }
// CHECK: attributes #8 = { {{.*}} "target-cpu"="neoverse-n1" "target-features"="+aes,+crc,+dotprod,+fp-armv8,+fullfp16,+lse,+neon,+ras,+rcpc,+rdm,+sha2,+spe,+ssbs,+v8.1a,+v8.2a,+v8a" "tune-cpu"="cortex-a710" }
-// CHECK: attributes #9 = { {{.*}} "target-features"="+fp-armv8,+fullfp16,+neon,+sve" "tune-cpu"="cortex-a710" }
+// CHECK: attributes #9 = { {{.*}} "target-features"="+fp-armv8,+fullfp16,+sve" "tune-cpu"="cortex-a710" }
// CHECK: attributes #10 = { {{.*}} "target-cpu"="neoverse-v1" "target-features"="+aes,+bf16,+complxnum,+crc,+dotprod,+fp-armv8,+fp16fml,+fullfp16,+i8mm,+jsconv,+lse,+neon,+pauth,+rand,+ras,+rcpc,+rdm,+sha2,+sha3,+sm4,+spe,+ssbs,+sve,+sve2,+v8.1a,+v8.2a,+v8.3a,+v8.4a,+v8a" }
// CHECK: attributes #11 = { {{.*}} "target-cpu"="neoverse-v1" "target-features"="+aes,+bf16,+complxnum,+crc,+dotprod,+fp-armv8,+fp16fml,+fullfp16,+i8mm,+jsconv,+lse,+neon,+pauth,+rand,+ras,+rcpc,+rdm,+sha2,+sha3,+sm4,+spe,+ssbs,+v8.1a,+v8.2a,+v8.3a,+v8.4a,+v8a,-sve" }
-// CHECK: attributes #12 = { {{.*}} "target-features"="+fp-armv8,+fullfp16,+neon,+sve" }
-// CHECK: attributes #13 = { {{.*}} "target-features"="+fp-armv8,+fullfp16,+neon" }
+// CHECK: attributes #12 = { {{.*}} "target-features"="+fp-armv8,+fullfp16,+sve" }
+// CHECK: attributes #13 = { {{.*}} "target-features"="+fp-armv8,+fullfp16" }
// CHECK: attributes #14 = { {{.*}} "target-cpu"="neoverse-n1" "target-features"="+aes,+bf16,+complxnum,+crc,+dotprod,+fp-armv8,+fullfp16,+i8mm,+jsconv,+lse,+neon,+pauth,+ras,+rcpc,+rdm,+sha2,+spe,+ssbs,+sve,+sve2,+v8.1a,+v8.2a,+v8.3a,+v8.4a,+v8.5a,+v8.6a,+v8a" "tune-cpu"="cortex-a710" }
// CHECK: attributes #15 = { {{.*}} "branch-target-enforcement"="true" "guarded-control-stack"="true" {{.*}} "target-features"="+aes,+bf16,+complxnum,+crc,+dotprod,+fp-armv8,+fullfp16,+i8mm,+jsconv,+lse,+neon,+pauth,+ras,+rcpc,+rdm,+sha2,+spe,+ssbs,+sve,+sve2,+v8.1a,+v8.2a,+v8.3a,+v8.4a,+v8.5a,+v8.6a,+v8a" "tune-cpu"="cortex-a710" }
// CHECK-NOT: attributes #16 = {{.*}} "target-features"
diff --git a/clang/test/Preprocessor/aarch64-target-features.c b/clang/test/Preprocessor/aarch64-target-features.c
index 0d5dd35e7632f..82304a15a04a3 100644
--- a/clang/test/Preprocessor/aarch64-target-features.c
+++ b/clang/test/Preprocessor/aarch64-target-features.c
@@ -261,6 +261,18 @@
// CHECK-FULLFP16-VECTOR-SCALAR: #define __ARM_FP 0xE
// CHECK-FULLFP16-VECTOR-SCALAR: #define __ARM_FP16_FORMAT_IEEE 1
+// +fp16fml+nosimd doesn't make sense as the fp16fml instructions all require SIMD.
+// However, as +fp16fml implies +fp16 there is a set of defines that we would expect.
+// RUN: %clang --target=aarch64 -march=armv8-a+fp16fml+nosimd -x c -E -dM %s -o - | FileCheck -match-full-lines --check-prefix=CHECK-FULLFP16-SCALAR %s
+// RUN: %clang --target=aarch64 -march=armv8-a+fp16+nosimd -x c -E -dM %s -o - | FileCheck -match-full-lines --check-prefix=CHECK-FULLFP16-SCALAR %s
+// RUN: %clang --target=aarch64 -march=armv8.4-a+fp16fml+nosimd -x c -E -dM %s -o - | FileCheck -match-full-lines --check-prefix=CHECK-FULLFP16-SCALAR %s
+// RUN: %clang --target=aarch64 -march=armv8.4-a+fp16+nosimd -x c -E -dM %s -o - | FileCheck -match-full-lines --check-prefix=CHECK-FULLFP16-SCALAR %s
+// CHECK-FULLFP16-SCALAR-NOT: #define __ARM_FEATURE_FP16_FML 1
+// CHECK-FULLFP16-SCALAR: #define __ARM_FEATURE_FP16_SCALAR_ARITHMETIC 1
+// CHECK-FULLFP16-SCALAR-NOT: #define __ARM_FEATURE_FP16_VECTOR_ARITHMETIC 1
+// CHECK-FULLFP16-SCALAR: #define __ARM_FP 0xE
+// CHECK-FULLFP16-SCALAR: #define __ARM_FP16_FORMAT_IEEE 1
+
// RUN: %clang --target=aarch64 -march=armv8.2-a -x c -E -dM %s -o - | FileCheck -match-full-lines --check-prefix=CHECK-FULLFP16-NOFML-VECTOR-SCALAR %s
// RUN: %clang --target=aarch64 -march=armv8.2-a+nofp16 -x c -E -dM %s -o - | FileCheck -match-full-lines --check-prefix=CHECK-FULLFP16-NOFML-VECTOR-SCALAR %s
// RUN: %clang --target=aarch64 -march=armv8.2-a+nofp16fml -x c -E -dM %s -o - | FileCheck -match-full-lines --check-prefix=CHECK-FULLFP16-NOFML-VECTOR-SCALAR %s
@@ -452,7 +464,7 @@
// RUN: %clang -target aarch64 -mcpu=cortex-a53+noSIMD -### -c %s 2>&1 | FileCheck -check-prefix=CHECK-MCPU-3 %s
// CHECK-MCPU-1: "-cc1"{{.*}} "-triple" "aarch64{{.*}}" "-target-feature" "-aes"{{.*}} "-target-feature" "-sha2"
// CHECK-MCPU-2: "-cc1"{{.*}} "-triple" "aarch64{{.*}}" "-target-feature" "+crc" "-target-feature" "+fp-armv8" "-target-feature" "+neon"
-// CHECK-MCPU-3: "-cc1"{{.*}} "-triple" "aarch64{{.*}}" "-target-feature" "+v8a" "-target-feature" "-aes" "-target-feature" "+crc" "-target-feature" "-fp-armv8" "-target-feature" "-sha2" "-target-feature" "-neon"
+// CHECK-MCPU-3: "-cc1"{{.*}} "-triple" "aarch64{{.*}}" "-target-feature" "+v8a" "-target-feature" "-aes" "-target-feature" "+crc" "-target-feature" "+fp-armv8" "-target-feature" "-sha2" "-target-feature" "-neon"
// RUN: %clang -target aarch64 -mcpu=cyclone+nocrc+nocrypto -march=armv8-a -### -c %s 2>&1 | FileCheck -check-prefix=CHECK-MCPU-MARCH %s
// RUN: %clang -target aarch64 -march=armv8-a -mcpu=cyclone+nocrc+nocrypto -### -c %s 2>&1 | FileCheck -check-prefix=CHECK-MCPU-MARCH %s
@@ -482,7 +494,7 @@
// RUN: %clang -target aarch64 -march=armv8.1a+noSIMD -### -c %s 2>&1 | FileCheck -check-prefix=CHECK-V81A-FEATURE-3 %s
// CHECK-V81A-FEATURE-1: "-cc1"{{.*}} "-triple" "aarch64{{.*}}" "-target-feature" "+v8.1a" "-target-feature" "+aes" "-target-feature" "+crc" "-target-feature" "+crypto" "-target-feature" "+fp-armv8" "-target-feature" "+lse" "-target-feature" "+rdm" "-target-feature" "+sha2" "-target-feature" "+neon"
// CHECK-V81A-FEATURE-2: "-cc1"{{.*}} "-triple" "aarch64{{.*}}" "-target-feature" "+v8.1a" "-target-feature" "+crc" "-target-feature" "+fp-armv8" "-target-feature" "+lse" "-target-feature" "+rdm" "-target-feature" "+neon"
-// CHECK-V81A-FEATURE-3: "-cc1"{{.*}} "-triple" "aarch64{{.*}}" "-target-feature" "+v8.1a" "-target-feature" "+crc" "-target-feature" "-fp-armv8" "-target-feature" "+lse" "-target-feature" "-rdm" "-target-feature" "-neon"
+// CHECK-V81A-FEATURE-3: "-cc1"{{.*}} "-triple" "aarch64{{.*}}" "-target-feature" "+v8.1a" "-target-feature" "+crc" "-target-feature" "+fp-armv8" "-target-feature" "+lse" "-target-feature" "-rdm" "-target-feature" "-neon"
// ================== Check Memory Tagging Extensions (MTE).
// RUN: %clang -target arm64-none-linux-gnu -march=armv8.5-a+memtag -x c -E -dM %s -o - 2>&1 | FileCheck -check-prefix=CHECK-MEMTAG %s
diff --git a/llvm/lib/TargetParser/AArch64TargetParser.cpp b/llvm/lib/TargetParser/AArch64TargetParser.cpp
index 3cbacf07243bc..941c4ab1e3855 100644
--- a/llvm/lib/TargetParser/AArch64TargetParser.cpp
+++ b/llvm/lib/TargetParser/AArch64TargetParser.cpp
@@ -174,11 +174,6 @@ void AArch64::ExtensionSet::enable(ArchExtKind E) {
Touched.set(E);
Enabled.set(E);
- // These depend on each other, meaning you can't have one without the other.
- // We don't want this to be in ExtensionDependencies to avoid infinite loops.
- if (E == AEK_FP)
- enable(AEK_SIMD);
-
// Recursively enable all features that this one depends on. This handles all
// of the simple cases, where the behaviour doesn't depend on the base
// architecture version.
@@ -220,9 +215,6 @@ void AArch64::ExtensionSet::disable(ArchExtKind E) {
Touched.set(E);
Enabled.reset(E);
- if (E == AEK_SIMD)
- disable(AEK_FP);
-
// Recursively disable all features that depends on this one.
for (auto Dep : ExtensionDependencies)
if (E == Dep.Earlier)
More information about the cfe-commits
mailing list