[llvm][TargetParser] Return StringMap from getHostCPUFeatures (#97824)

Previously this took a reference to a map and returned a bool to say
whether it succeeded. We can return a StringMap instead, as all callers
but 1 simply iterated the map if the bool was true, and passed in empty
maps as the starting point.

lldb's lit-cpuid did specifically check whether the call failed, but due
to the way the x86 routines work this works out the same as checking if
the returned map is empty.

Author: DavidSpickett

clang/lib/Driver/ToolChains/Arch/ARM.cpp

@@ -591,11 +591,9 @@ llvm::ARM::FPUKind arm::getARMTargetFeatures(const Driver &D,
 
   // Add CPU features for generic CPUs
   if (CPUName == "native") {
-    llvm::StringMap<bool> HostFeatures;
-    if (llvm::sys::getHostCPUFeatures(HostFeatures))
-      for (auto &F : HostFeatures)
-        Features.push_back(
-            Args.MakeArgString((F.second ? "+" : "-") + F.first()));
+    for (auto &F : llvm::sys::getHostCPUFeatures())
+      Features.push_back(
+          Args.MakeArgString((F.second ? "+" : "-") + F.first()));
   } else if (!CPUName.empty()) {
     // This sets the default features for the specified CPU. We certainly don't
     // want to override the features that have been explicitly specified on the

clang/lib/Driver/ToolChains/Arch/X86.cpp

@@ -131,11 +131,9 @@ void x86::getX86TargetFeatures(const Driver &D, const llvm::Triple &Triple,
   // If -march=native, autodetect the feature list.
   if (const Arg *A = Args.getLastArg(clang::driver::options::OPT_march_EQ)) {
     if (StringRef(A->getValue()) == "native") {
-      llvm::StringMap<bool> HostFeatures;
-      if (llvm::sys::getHostCPUFeatures(HostFeatures))
-        for (auto &F : HostFeatures)
-          Features.push_back(
-              Args.MakeArgString((F.second ? "+" : "-") + F.first()));
+      for (auto &F : llvm::sys::getHostCPUFeatures())
+        Features.push_back(
+            Args.MakeArgString((F.second ? "+" : "-") + F.first()));
     }
   }
 

lldb/utils/lit-cpuid/lit-cpuid.cpp

@@ -20,16 +20,15 @@ using namespace llvm;
 int main(int argc, char **argv) {
 #if defined(__i386__) || defined(_M_IX86) || \
     defined(__x86_64__) || defined(_M_X64)
-  StringMap<bool> features;
-
-  if (!sys::getHostCPUFeatures(features))
+  const StringMap<bool> features = sys::getHostCPUFeatures();
+  if (features.empty())
     return 1;
 
-  if (features["sse"])
+  if (features.lookup("sse"))
     outs() << "sse\n";
-  if (features["avx"])
+  if (features.lookup("avx"))
     outs() << "avx\n";
-  if (features["avx512f"])
+  if (features.lookup("avx512f"))
     outs() << "avx512f\n";
 #endif
 

llvm/include/llvm/TargetParser/Host.h

@@ -47,13 +47,12 @@ namespace sys {
   /// The particular format of the names are target dependent, and suitable for
   /// passing as -mattr to the target which matches the host.
   ///
-  /// \param Features - A string mapping feature names to either
-  /// true (if enabled) or false (if disabled). This routine makes no guarantees
-  /// about exactly which features may appear in this map, except that they are
-  /// all valid LLVM feature names.
-  ///
-  /// \return - True on success.
-  bool getHostCPUFeatures(StringMap<bool, MallocAllocator> &Features);
+  /// \return - A string map mapping feature names to either true (if enabled)
+  /// or false (if disabled). This routine makes no guarantees about exactly
+  /// which features may appear in this map, except that they are all valid LLVM
+  /// feature names. The map can be empty, for example if feature detection
+  /// fails.
+  const StringMap<bool, MallocAllocator> getHostCPUFeatures();
 
   /// This is a function compatible with cl::AddExtraVersionPrinter, which adds
   /// info about the current target triple and detected CPU.

llvm/lib/CodeGen/CommandFlags.cpp

@@ -624,12 +624,9 @@ std::string codegen::getFeaturesStr() {
   // This is necessary for x86 where the CPU might not support all the
   // features the autodetected CPU name lists in the target. For example,
   // not all Sandybridge processors support AVX.
-  if (getMCPU() == "native") {
-    StringMap<bool> HostFeatures;
-    if (sys::getHostCPUFeatures(HostFeatures))
-      for (const auto &[Feature, IsEnabled] : HostFeatures)
-        Features.AddFeature(Feature, IsEnabled);
-  }
+  if (getMCPU() == "native")
+    for (const auto &[Feature, IsEnabled] : sys::getHostCPUFeatures())
+      Features.AddFeature(Feature, IsEnabled);
 
   for (auto const &MAttr : getMAttrs())
     Features.AddFeature(MAttr);
@@ -644,12 +641,9 @@ std::vector<std::string> codegen::getFeatureList() {
   // This is necessary for x86 where the CPU might not support all the
   // features the autodetected CPU name lists in the target. For example,
   // not all Sandybridge processors support AVX.
-  if (getMCPU() == "native") {
-    StringMap<bool> HostFeatures;
-    if (sys::getHostCPUFeatures(HostFeatures))
-      for (const auto &[Feature, IsEnabled] : HostFeatures)
-        Features.AddFeature(Feature, IsEnabled);
-  }
+  if (getMCPU() == "native")
+    for (const auto &[Feature, IsEnabled] : sys::getHostCPUFeatures())
+      Features.AddFeature(Feature, IsEnabled);
 
   for (auto const &MAttr : getMAttrs())
     Features.AddFeature(MAttr);

llvm/lib/ExecutionEngine/Orc/JITTargetMachineBuilder.cpp

@@ -28,9 +28,7 @@ Expected<JITTargetMachineBuilder> JITTargetMachineBuilder::detectHost() {
   // Retrieve host CPU name and sub-target features and add them to builder.
   // Relocation model, code model and codegen opt level are kept to default
   // values.
-  llvm::StringMap<bool> FeatureMap;
-  llvm::sys::getHostCPUFeatures(FeatureMap);
-  for (auto &Feature : FeatureMap)
+  for (const auto &Feature : llvm::sys::getHostCPUFeatures())
     TMBuilder.getFeatures().AddFeature(Feature.first(), Feature.second);
 
   TMBuilder.setCPU(std::string(llvm::sys::getHostCPUName()));

llvm/lib/Target/TargetMachineC.cpp

@@ -363,11 +363,8 @@ char *LLVMGetHostCPUName(void) {
 
 char *LLVMGetHostCPUFeatures(void) {
   SubtargetFeatures Features;
-  StringMap<bool> HostFeatures;
-
-  if (sys::getHostCPUFeatures(HostFeatures))
-    for (const auto &[Feature, IsEnabled] : HostFeatures)
-      Features.AddFeature(Feature, IsEnabled);
+  for (const auto &[Feature, IsEnabled] : sys::getHostCPUFeatures())
+    Features.AddFeature(Feature, IsEnabled);
 
   return strdup(Features.getString().c_str());
 }

llvm/lib/TargetParser/Host.cpp

@@ -1708,12 +1708,13 @@ VendorSignatures getVendorSignature(unsigned *MaxLeaf) {
 
 #if defined(__i386__) || defined(_M_IX86) || \
     defined(__x86_64__) || defined(_M_X64)
-bool sys::getHostCPUFeatures(StringMap<bool> &Features) {
+const StringMap<bool> sys::getHostCPUFeatures() {
   unsigned EAX = 0, EBX = 0, ECX = 0, EDX = 0;
   unsigned MaxLevel;
+  StringMap<bool> Features;
 
   if (getX86CpuIDAndInfo(0, &MaxLevel, &EBX, &ECX, &EDX) || MaxLevel < 1)
-    return false;
+    return Features;
 
   getX86CpuIDAndInfo(1, &EAX, &EBX, &ECX, &EDX);
 
@@ -1901,13 +1902,14 @@ bool sys::getHostCPUFeatures(StringMap<bool> &Features) {
   Features["avx10.1-512"] =
       Features["avx10.1-256"] && HasLeaf24 && ((EBX >> 18) & 1);
 
-  return true;
+  return Features;
 }
 #elif defined(__linux__) && (defined(__arm__) || defined(__aarch64__))
-bool sys::getHostCPUFeatures(StringMap<bool> &Features) {
+const StringMap<bool> sys::getHostCPUFeatures() {
+  StringMap<bool> Features;
   std::unique_ptr<llvm::MemoryBuffer> P = getProcCpuinfoContent();
   if (!P)
-    return false;
+    return Features;
 
   SmallVector<StringRef, 32> Lines;
   P->getBuffer().split(Lines, "\n");
@@ -1970,38 +1972,42 @@ bool sys::getHostCPUFeatures(StringMap<bool> &Features) {
     Features["crypto"] = true;
 #endif
 
-  return true;
+  return Features;
 }
 #elif defined(_WIN32) && (defined(__aarch64__) || defined(_M_ARM64))
-bool sys::getHostCPUFeatures(StringMap<bool> &Features) {
+const StringMap<bool> sys::getHostCPUFeatures() {
+  StringMap<bool> Features;
+
   if (IsProcessorFeaturePresent(PF_ARM_NEON_INSTRUCTIONS_AVAILABLE))
     Features["neon"] = true;
   if (IsProcessorFeaturePresent(PF_ARM_V8_CRC32_INSTRUCTIONS_AVAILABLE))
     Features["crc"] = true;
   if (IsProcessorFeaturePresent(PF_ARM_V8_CRYPTO_INSTRUCTIONS_AVAILABLE))
     Features["crypto"] = true;
 
-  return true;
+  return Features;
 }
 #elif defined(__linux__) && defined(__loongarch__)
 #include <sys/auxv.h>
-bool sys::getHostCPUFeatures(StringMap<bool> &Features) {
+const StringMap<bool> sys::getHostCPUFeatures() {
   unsigned long hwcap = getauxval(AT_HWCAP);
   bool HasFPU = hwcap & (1UL << 3); // HWCAP_LOONGARCH_FPU
   uint32_t cpucfg2 = 0x2;
   __asm__("cpucfg %[cpucfg2], %[cpucfg2]\n\t" : [cpucfg2] "+r"(cpucfg2));
 
+  StringMap<bool> Features;
+
   Features["f"] = HasFPU && (cpucfg2 & (1U << 1)); // CPUCFG.2.FP_SP
   Features["d"] = HasFPU && (cpucfg2 & (1U << 2)); // CPUCFG.2.FP_DP
 
   Features["lsx"] = hwcap & (1UL << 4);  // HWCAP_LOONGARCH_LSX
   Features["lasx"] = hwcap & (1UL << 5); // HWCAP_LOONGARCH_LASX
   Features["lvz"] = hwcap & (1UL << 9);  // HWCAP_LOONGARCH_LVZ
 
-  return true;
+  return Features;
 }
 #else
-bool sys::getHostCPUFeatures(StringMap<bool> &Features) { return false; }
+const StringMap<bool> sys::getHostCPUFeatures() { return {}; }
 #endif
 
 #if __APPLE__