[RISCV] Support __builtin_cpu_is

bolt/unittests/Core/MCPlusBuilder.cpp

@@ -90,15 +90,14 @@ INSTANTIATE_TEST_SUITE_P(AArch64, MCPlusBuilderTester,
                          ::testing::Values(Triple::aarch64));
 
 TEST_P(MCPlusBuilderTester, AliasX0) {
-  uint64_t AliasesX0[] = {AArch64::W0,    AArch64::W0_HI,
-                          AArch64::X0,    AArch64::W0_W1,
+  uint64_t AliasesX0[] = {AArch64::W0, AArch64::X0, AArch64::W0_W1,
                           AArch64::X0_X1, AArch64::X0_X1_X2_X3_X4_X5_X6_X7};
   size_t AliasesX0Count = sizeof(AliasesX0) / sizeof(*AliasesX0);
   testRegAliases(Triple::aarch64, AArch64::X0, AliasesX0, AliasesX0Count);
 }
 
 TEST_P(MCPlusBuilderTester, AliasSmallerX0) {
-  uint64_t AliasesX0[] = {AArch64::W0, AArch64::W0_HI, AArch64::X0};
+  uint64_t AliasesX0[] = {AArch64::W0, AArch64::X0};
   size_t AliasesX0Count = sizeof(AliasesX0) / sizeof(*AliasesX0);
   testRegAliases(Triple::aarch64, AArch64::X0, AliasesX0, AliasesX0Count, true);
 }

clang/docs/ReleaseNotes.rst

@@ -752,7 +752,6 @@ X86 Support
 - Support ISA of ``AMX-MOVRS``.
 - Support ISA of ``AMX-AVX512``.
 - Support ISA of ``AMX-TF32``.
-- Support ISA of ``MOVRS``.
 
 Arm and AArch64 Support
 ^^^^^^^^^^^^^^^^^^^^^^^

clang/include/clang/Basic/Attr.td

@@ -477,9 +477,6 @@ def TargetELF : TargetSpec {
 def TargetELFOrMachO : TargetSpec {
   let ObjectFormats = ["ELF", "MachO"];
 }
-def TargetIFuncSupport : TargetSpec {
-  let CustomCode = [{ Target.supportsIFunc() }];
-}
 def TargetWindowsArm64EC : TargetSpec {
   let CustomCode = [{ Target.getTriple().isWindowsArm64EC() }];
 }
@@ -1858,7 +1855,7 @@ def IBOutletCollection : InheritableAttr {
   let Documentation = [Undocumented];
 }
 
-def IFunc : Attr, TargetSpecificAttr<TargetIFuncSupport> {
+def IFunc : Attr, TargetSpecificAttr<TargetELFOrMachO> {
   let Spellings = [GCC<"ifunc">];
   let Args = [StringArgument<"Resolver">];
   let Subjects = SubjectList<[Function]>;

clang/include/clang/Basic/AttrDocs.td

@@ -6051,19 +6051,12 @@ declared entity. The entity must not have weak linkage; for example, in C++,
 it cannot be applied to a declaration if a definition at that location would be
 considered inline.
 
-Not all targets support this attribute:
-
-- ELF target support depends on both the linker and runtime linker, and is
-  available in at least lld 4.0 and later, binutils 2.20.1 and later, glibc
-  v2.11.1 and later, and FreeBSD 9.1 and later.
-- Mach-O targets support it, but with slightly different semantics: the resolver
-  is run at first call, instead of at load time by the runtime linker.
-- Windows target supports it on AArch64, but with different semantics: the
-  ``ifunc`` is replaced with a global function pointer, and the call is replaced
-  with an indirect call. The function pointer is initialized by a constructor
-  that calls the resolver.
-- Baremetal target supports it on AVR.
-- Other targets currently do not support this attribute.
+Not all targets support this attribute. ELF target support depends on both the
+linker and runtime linker, and is available in at least lld 4.0 and later,
+binutils 2.20.1 and later, glibc v2.11.1 and later, and FreeBSD 9.1 and later.
+Mach-O targets support it, but with slightly different semantics: the resolver
+is run at first call, instead of at load time by the runtime linker. Targets
+other than ELF and Mach-O currently do not support this attribute.
   }];
 }
 

clang/include/clang/Basic/BuiltinsX86.def

@@ -660,9 +660,6 @@ TARGET_BUILTIN(__builtin_ia32_vpdpbuud256, "V8iV8iV8iV8i", "ncV:256:", "avxvnnii
 TARGET_BUILTIN(__builtin_ia32_vpdpbuuds128, "V4iV4iV4iV4i", "ncV:128:", "avxvnniint8|avx10.2-256")
 TARGET_BUILTIN(__builtin_ia32_vpdpbuuds256, "V8iV8iV8iV8i", "ncV:256:", "avxvnniint8|avx10.2-256")
 
-// MOVRS
-TARGET_BUILTIN(__builtin_ia32_prefetchrs, "vvC*", "nc", "movrs")
-
 TARGET_BUILTIN(__builtin_ia32_gather3div2df, "V2dV2dvC*V2OiUcIi", "nV:128:", "avx512vl")
 TARGET_BUILTIN(__builtin_ia32_gather3div2di, "V2OiV2OivC*V2OiUcIi", "nV:128:", "avx512vl")
 TARGET_BUILTIN(__builtin_ia32_gather3div4df, "V4dV4dvC*V4OiUcIi", "nV:256:", "avx512vl")

clang/include/clang/Basic/BuiltinsX86_64.def

@@ -228,12 +228,6 @@ TARGET_BUILTIN(__builtin_ia32_aand64, "vv*SOi", "n", "raoint")
 TARGET_BUILTIN(__builtin_ia32_aor64, "vv*SOi", "n", "raoint")
 TARGET_BUILTIN(__builtin_ia32_axor64, "vv*SOi", "n", "raoint")
 
-// MOVRS
-TARGET_BUILTIN(__builtin_ia32_movrsqi, "ScvC*", "n", "movrs")
-TARGET_BUILTIN(__builtin_ia32_movrshi, "SsvC*", "n", "movrs")
-TARGET_BUILTIN(__builtin_ia32_movrssi, "SivC*", "n", "movrs")
-TARGET_BUILTIN(__builtin_ia32_movrsdi, "SLLivC*", "n", "movrs")
-
 // MOVRS and AVX10.2
 TARGET_BUILTIN(__builtin_ia32_vmovrsb128, "V16cV16cC*", "nV:128:", "movrs,avx10.2-256")
 TARGET_BUILTIN(__builtin_ia32_vmovrsb256, "V32cV32cC*", "nV:256:", "movrs,avx10.2-256")

clang/include/clang/Basic/TargetInfo.h

@@ -1497,10 +1497,6 @@ class TargetInfo : public TransferrableTargetInfo,
   bool supportsIFunc() const {
     if (getTriple().isOSBinFormatMachO())
       return true;
-    if (getTriple().isOSWindows() && getTriple().isAArch64())
-      return true;
-    if (getTriple().getArch() == llvm::Triple::ArchType::avr)
-      return true;
     return getTriple().isOSBinFormatELF() &&
            ((getTriple().isOSLinux() && !getTriple().isMusl()) ||
             getTriple().isOSFreeBSD());

clang/include/clang/Basic/arm_sve.td

@@ -1964,7 +1964,7 @@ let SVETargetGuard = "sve2,lut,bf16", SMETargetGuard = "sme2,lut,bf16" in {
 ////////////////////////////////////////////////////////////////////////////////
 // SVE2 - Optional
 
-let SVETargetGuard = "sve2,sve-aes", SMETargetGuard = InvalidMode in {
+let SVETargetGuard = "sve2-aes", SMETargetGuard = InvalidMode in {
 def SVAESD   : SInst<"svaesd[_{d}]",   "ddd", "Uc", MergeNone, "aarch64_sve_aesd", [IsOverloadNone]>;
 def SVAESIMC : SInst<"svaesimc[_{d}]", "dd",  "Uc", MergeNone, "aarch64_sve_aesimc", [IsOverloadNone]>;
 def SVAESE   : SInst<"svaese[_{d}]",   "ddd", "Uc", MergeNone, "aarch64_sve_aese", [IsOverloadNone]>;

clang/lib/Basic/Targets/AArch64.cpp

@@ -473,7 +473,7 @@ void AArch64TargetInfo::getTargetDefines(const LangOptions &Opts,
   if (HasSVE2p1)
     Builder.defineMacro("__ARM_FEATURE_SVE2p1", "1");
 
-  if (HasSVE2 && HasSVEAES)
+  if (HasSVE2 && HasSVE2AES)
     Builder.defineMacro("__ARM_FEATURE_SVE2_AES", "1");
 
   if (HasSVE2 && HasSVE2BitPerm)
@@ -769,7 +769,7 @@ bool AArch64TargetInfo::hasFeature(StringRef Feature) const {
       .Case("f32mm", FPU & SveMode && HasMatmulFP32)
       .Case("f64mm", FPU & SveMode && HasMatmulFP64)
       .Case("sve2", FPU & SveMode && HasSVE2)
-      .Case("sve-aes", HasSVEAES)
+      .Case("sve2-pmull128", FPU & SveMode && HasSVE2AES)
       .Case("sve2-bitperm", FPU & SveMode && HasSVE2BitPerm)
       .Case("sve2-sha3", FPU & SveMode && HasSVE2SHA3)
       .Case("sve2-sm4", FPU & SveMode && HasSVE2SM4)
@@ -861,10 +861,12 @@ bool AArch64TargetInfo::handleTargetFeatures(std::vector<std::string> &Features,
       HasSVE2 = true;
       HasSVE2p1 = true;
     }
-    if (Feature == "+sve-aes") {
+    if (Feature == "+sve2-aes") {
       FPU |= NeonMode;
+      FPU |= SveMode;
       HasFullFP16 = true;
-      HasSVEAES = true;
+      HasSVE2 = true;
+      HasSVE2AES = true;
     }
     if (Feature == "+sve2-sha3") {
       FPU |= NeonMode;

clang/lib/Basic/Targets/AArch64.h

@@ -78,7 +78,7 @@ class LLVM_LIBRARY_VISIBILITY AArch64TargetInfo : public TargetInfo {
   bool HasBFloat16 = false;
   bool HasSVE2 = false;
   bool HasSVE2p1 = false;
-  bool HasSVEAES = false;
+  bool HasSVE2AES = false;
   bool HasSVE2SHA3 = false;
   bool HasSVE2SM4 = false;
   bool HasSVEB16B16 = false;

clang/lib/Basic/Targets/RISCV.cpp

@@ -508,3 +508,10 @@ bool RISCVTargetInfo::validateGlobalRegisterVariable(
   }
   return false;
 }
+
+bool RISCVTargetInfo::validateCpuIs(StringRef CPUName) const {
+  assert(getTriple().isOSLinux() &&
+         "__builtin_cpu_is() is only supported for Linux.");
+
+  return llvm::RISCV::hasValidCPUModel(CPUName);
+}

clang/lib/Basic/Targets/RISCV.h

@@ -128,8 +128,10 @@ class RISCVTargetInfo : public TargetInfo {
   }
 
   bool supportsCpuSupports() const override { return getTriple().isOSLinux(); }
+  bool supportsCpuIs() const override { return getTriple().isOSLinux(); }
   bool supportsCpuInit() const override { return getTriple().isOSLinux(); }
   bool validateCpuSupports(StringRef Feature) const override;
+  bool validateCpuIs(StringRef CPUName) const override;
   bool isValidFeatureName(StringRef Name) const override;
 
   bool validateGlobalRegisterVariable(StringRef RegName, unsigned RegSize,

clang/lib/CodeGen/CGBuiltin.cpp

@@ -67,6 +67,7 @@
 #include "llvm/Support/ScopedPrinter.h"
 #include "llvm/TargetParser/AArch64TargetParser.h"
 #include "llvm/TargetParser/RISCVISAInfo.h"
+#include "llvm/TargetParser/RISCVTargetParser.h"
 #include "llvm/TargetParser/X86TargetParser.h"
 #include <optional>
 #include <sstream>
@@ -22505,6 +22506,47 @@ Value *CodeGenFunction::EmitHexagonBuiltinExpr(unsigned BuiltinID,
   return nullptr;
 }
 
+Value *CodeGenFunction::EmitRISCVCpuIs(const CallExpr *E) {
+  const Expr *CPUExpr = E->getArg(0)->IgnoreParenCasts();
+  StringRef CPUStr = cast<clang::StringLiteral>(CPUExpr)->getString();
+  return EmitRISCVCpuIs(CPUStr);
+}
+
+Value *CodeGenFunction::EmitRISCVCpuIs(StringRef CPUStr) {
+  llvm::Type *Int32Ty = Builder.getInt32Ty();
+  llvm::Type *Int64Ty = Builder.getInt64Ty();
+  llvm::StructType *StructTy = llvm::StructType::get(Int32Ty, Int64Ty, Int64Ty);
+  llvm::Constant *RISCVCPUModel =
+      CGM.CreateRuntimeVariable(StructTy, "__riscv_cpu_model");
+  cast<llvm::GlobalValue>(RISCVCPUModel)->setDSOLocal(true);
+
+  auto loadRISCVCPUID = [&](unsigned Index) {
+    Value *Ptr = Builder.CreateStructGEP(StructTy, RISCVCPUModel, Index);
+    Value *CPUID = Builder.CreateAlignedLoad(StructTy->getTypeAtIndex(Index),
+                                             Ptr, llvm::MaybeAlign());
+    return CPUID;
+  };
+
+  const llvm::RISCV::CPUModel CPUModel = llvm::RISCV::getCPUModel(CPUStr);
+
+  // Compare mvendorid.
+  Value *VendorID = loadRISCVCPUID(0);
+  Value *Result =
+      Builder.CreateICmpEQ(VendorID, Builder.getInt32(CPUModel.MVendorID));
+
+  // Compare marchid.
+  Value *ArchID = loadRISCVCPUID(1);
+  Result = Builder.CreateAnd(
+      Result, Builder.CreateICmpEQ(ArchID, Builder.getInt64(CPUModel.MArchID)));
+
+  // Compare mimplid.
+  Value *ImplID = loadRISCVCPUID(2);
+  Result = Builder.CreateAnd(
+      Result, Builder.CreateICmpEQ(ImplID, Builder.getInt64(CPUModel.MImpID)));
+
+  return Result;
+}
+
 Value *CodeGenFunction::EmitRISCVBuiltinExpr(unsigned BuiltinID,
                                              const CallExpr *E,
                                              ReturnValueSlot ReturnValue) {
@@ -22513,6 +22555,8 @@ Value *CodeGenFunction::EmitRISCVBuiltinExpr(unsigned BuiltinID,
     return EmitRISCVCpuSupports(E);
   if (BuiltinID == Builtin::BI__builtin_cpu_init)
     return EmitRISCVCpuInit();
+  if (BuiltinID == Builtin::BI__builtin_cpu_is)
+    return EmitRISCVCpuIs(E);
 
   SmallVector<Value *, 4> Ops;
   llvm::Type *ResultType = ConvertType(E->getType());

clang/lib/CodeGen/CodeGenFunction.h

@@ -4730,6 +4730,8 @@ class CodeGenFunction : public CodeGenTypeCache {
   llvm::Value *EmitRISCVCpuSupports(const CallExpr *E);
   llvm::Value *EmitRISCVCpuSupports(ArrayRef<StringRef> FeaturesStrs);
   llvm::Value *EmitRISCVCpuInit();
+  llvm::Value *EmitRISCVCpuIs(const CallExpr *E);
+  llvm::Value *EmitRISCVCpuIs(StringRef CPUStr);
 
   void AddAMDGPUFenceAddressSpaceMMRA(llvm::Instruction *Inst,
                                       const CallExpr *E);

clang/lib/Driver/ToolChains/PS4CPU.cpp

@@ -183,7 +183,6 @@ void tools::PS4cpu::Linker::ConstructJob(Compilation &C, const JobAction &JA,
     CmdArgs.push_back(
         Args.MakeArgString(Twine("-lto-debug-options=") + LTOArgs));
 
-  // Sanitizer runtimes must be supplied before all other objects and libs.
   if (!Args.hasArg(options::OPT_nostdlib, options::OPT_nodefaultlibs))
     TC.addSanitizerArgs(Args, CmdArgs, "-l", "");
 
@@ -361,70 +360,27 @@ void tools::PS5cpu::Linker::ConstructJob(Compilation &C, const JobAction &JA,
   if (StringRef Jobs = getLTOParallelism(Args, D); !Jobs.empty())
     AddLTOFlag(Twine("jobs=") + Jobs);
 
+  if (!Args.hasArg(options::OPT_nostdlib, options::OPT_nodefaultlibs))
+    TC.addSanitizerArgs(Args, CmdArgs, "-l", "");
+
   TC.AddFilePathLibArgs(Args, CmdArgs);
   Args.addAllArgs(CmdArgs, {options::OPT_L, options::OPT_T_Group,
                             options::OPT_s, options::OPT_t});
 
   if (Args.hasArg(options::OPT_Z_Xlinker__no_demangle))
     CmdArgs.push_back("--no-demangle");
 
-  // Sanitizer runtimes must be supplied before all other objects and libs.
-  if (!Args.hasArg(options::OPT_nostdlib, options::OPT_nodefaultlibs))
-    TC.addSanitizerArgs(Args, CmdArgs, "-l", "");
-
-  const bool AddStartFiles =
-      !Relocatable &&
-      !Args.hasArg(options::OPT_nostartfiles, options::OPT_nostdlib);
-
-  auto AddCRTObject = [&](const char *Name) {
-    CmdArgs.push_back(Args.MakeArgString(TC.GetFilePath(Name)));
-  };
-
-  if (AddStartFiles) {
-    if (!Shared)
-      AddCRTObject("crt1.o");
-    AddCRTObject("crti.o");
-    AddCRTObject(Shared   ? "crtbeginS.o"
-                 : Static ? "crtbeginT.o"
-                          : "crtbegin.o");
-  }
-
   AddLinkerInputs(TC, Inputs, Args, CmdArgs, JA);
 
-  if (!Relocatable &&
-      !Args.hasArg(options::OPT_nodefaultlibs, options::OPT_nostdlib)) {
-
-    if (UseJMC) {
-      CmdArgs.push_back("--push-state");
-      CmdArgs.push_back("--whole-archive");
-      CmdArgs.push_back("-lSceJmc_nosubmission");
-      CmdArgs.push_back("--pop-state");
-    }
-
-    if (Args.hasArg(options::OPT_pthread))
-      CmdArgs.push_back("-lpthread");
-
-    if (Static) {
-      if (!Args.hasArg(options::OPT_nostdlibxx))
-        CmdArgs.push_back("-lstdc++");
-      if (!Args.hasArg(options::OPT_nolibc)) {
-        CmdArgs.push_back("-lm");
-        CmdArgs.push_back("-lc");
-      }
-
-      CmdArgs.push_back("-lcompiler_rt");
-      CmdArgs.push_back("-lkernel");
-    } else {
-      // The C and C++ libraries are combined.
-      if (!Args.hasArg(options::OPT_nolibc, options::OPT_nostdlibxx))
-        CmdArgs.push_back("-lc_stub_weak");
-
-      CmdArgs.push_back("-lkernel_stub_weak");
-    }
+  if (Args.hasArg(options::OPT_pthread)) {
+    CmdArgs.push_back("-lpthread");
   }
-  if (AddStartFiles) {
-    AddCRTObject(Shared ? "crtendS.o" : "crtend.o");
-    AddCRTObject("crtn.o");
+
+  if (UseJMC) {
+    CmdArgs.push_back("--push-state");
+    CmdArgs.push_back("--whole-archive");
+    CmdArgs.push_back("-lSceJmc_nosubmission");
+    CmdArgs.push_back("--pop-state");
   }
 
   if (Args.hasArg(options::OPT_fuse_ld_EQ)) {

clang/lib/Headers/CMakeLists.txt

@@ -233,7 +233,6 @@ set(x86_files
   movdirintrin.h
   movrs_avx10_2_512intrin.h
   movrs_avx10_2intrin.h
-  movrsintrin.h
   mwaitxintrin.h
   nmmintrin.h
   pconfigintrin.h

clang/lib/Headers/immintrin.h

@@ -605,10 +605,6 @@ _storebe_i64(void * __P, long long __D) {
 #include <movdirintrin.h>
 #endif
 
-#if !defined(__SCE__) || __has_feature(modules) || defined(__MOVRS__)
-#include <movrsintrin.h>
-#endif
-
 #if !defined(__SCE__) || __has_feature(modules) ||                             \
     (defined(__AVX10_2__) && defined(__MOVRS__))
 #include <movrs_avx10_2intrin.h>