Merge from 'main' to 'sycl-web' (35 commits)

  CONFLICT (content): Merge conflict in clang/lib/CodeGen/BackendUtil.cpp

.github/workflows/libcxx-build-and-test.yaml

@@ -23,8 +23,8 @@ on:
       - 'cmake/**'
       - '.github/workflows/libcxx-build-and-test.yaml'
   schedule:
-    # Run nightly at 8 AM UTC (or roughly 3 AM eastern)
-    - cron: '0 3 * * *'
+    # Run nightly at 08:00 UTC (aka 00:00 Pacific, aka 03:00 Eastern)
+    - cron: '0 8 * * *'
 
 permissions:
   contents: read # Default everything to read-only

clang/cmake/caches/Release.cmake

@@ -4,27 +4,53 @@
 
 # General Options
 set(LLVM_RELEASE_ENABLE_LTO THIN CACHE STRING "")
+set(LLVM_RELEASE_ENABLE_PGO ON CACHE BOOL "")
 
 set(CMAKE_BUILD_TYPE RELEASE CACHE STRING "")
 
 # Stage 1 Bootstrap Setup
 set(CLANG_ENABLE_BOOTSTRAP ON CACHE BOOL "")
-set(CLANG_BOOTSTRAP_TARGETS
-  clang
-  check-all
-  check-llvm
-  check-clang
-  test-suite
-  stage3
-  stage3-clang
-  stage3-check-all
-  stage3-check-llvm
-  stage3-check-clang
-  stage3-install
-  stage3-test-suite CACHE STRING "")
+if (LLVM_RELEASE_ENABLE_PGO)
+  set(CLANG_BOOTSTRAP_TARGETS
+    generate-profdata
+    stage2
+    stage2-clang
+    stage2-distribution
+    stage2-install
+    stage2-install-distribution
+    stage2-install-distribution-toolchain
+    stage2-check-all
+    stage2-check-llvm
+    stage2-check-clang
+    stage2-test-suite CACHE STRING "")
+else()
+  set(CLANG_BOOTSTRAP_TARGETS
+    clang
+    check-all
+    check-llvm
+    check-clang
+    test-suite
+    stage3
+    stage3-clang
+    stage3-check-all
+    stage3-check-llvm
+    stage3-check-clang
+    stage3-install
+    stage3-test-suite CACHE STRING "")
+endif()
 
 # Stage 1 Options
-set(LLVM_ENABLE_PROJECTS "clang" CACHE STRING "")
+set(STAGE1_PROJECTS "clang")
+set(STAGE1_RUNTIMES "")
+
+if (LLVM_RELEASE_ENABLE_PGO)
+  list(APPEND STAGE1_PROJECTS "lld")
+  list(APPEND STAGE1_RUNTIMES "compiler-rt")
+endif()
+
+set(LLVM_ENABLE_RUNTIMES ${STAGE1_RUNTIMES} CACHE STRING "")
+set(LLVM_ENABLE_PROJECTS ${STAGE1_PROJECTS} CACHE STRING "")
+
 set(LLVM_TARGETS_TO_BUILD Native CACHE STRING "")
 
 # Stage 2 Bootstrap Setup
@@ -37,11 +63,26 @@ set(BOOTSTRAP_CLANG_BOOTSTRAP_TARGETS
 
 # Stage 2 Options
 set(STAGE2_PROJECTS "clang")
-if (LLVM_RELEASE_ENABLE_LTO)
+set(STAGE2_RUNTIMES "")
+
+if (LLVM_RELEASE_ENABLE_LTO OR LLVM_RELEASE_ENABLE_PGO)
  list(APPEND STAGE2_PROJECTS "lld")
 endif()
+
+if (LLVM_RELEASE_ENABLE_PGO)
+  set(BOOTSTRAP_LLVM_BUILD_INSTRUMENTED IR CACHE STRING "")
+  list(APPEND STAGE2_RUNTIMES "compiler-rt")
+  set(BOOTSTRAP_LLVM_ENABLE_LTO ${LLVM_RELEASE_ENABLE_LTO})
+  if (LLVM_RELEASE_ENABLE_LTO)
+    set(BOOTSTRAP_LLVM_ENABLE_LLD ON CACHE BOOL "")
+  endif()
+endif()
+
 set(BOOTSTRAP_LLVM_ENABLE_PROJECTS ${STAGE2_PROJECTS} CACHE STRING "")
-set(BOOTSTRAP_LLVM_TARGETS_TO_BUILD Native CACHE STRING "")
+set(BOOTSTRAP_LLVM_ENABLE_RUNTIMES ${STAGE2_RUNTIMES} CACHE STRING "")
+if (NOT LLVM_RELEASE_ENABLE_PGO)
+  set(BOOTSTRAP_LLVM_TARGETS_TO_BUILD Native CACHE STRING "")
+endif()
 
 # Stage 3 Options
 set(BOOTSTRAP_BOOTSTRAP_LLVM_ENABLE_RUNTIMES "compiler-rt;libcxx;libcxxabi;libunwind" CACHE STRING "")

clang/lib/CodeGen/BackendUtil.cpp

@@ -1067,8 +1067,9 @@ void EmitAssemblyHelper::RunOptimizationPipeline(
       MPM.addPass(PB.buildO0DefaultPipeline(OptimizationLevel::O0,
                                       PrepareForLTO || PrepareForThinLTO));
     } else if (CodeGenOpts.FatLTO) {
-      assert(CodeGenOpts.UnifiedLTO && "FatLTO requires UnifiedLTO");
-      MPM.addPass(PB.buildFatLTODefaultPipeline(Level));
+      MPM.addPass(PB.buildFatLTODefaultPipeline(
+          Level, PrepareForThinLTO,
+          PrepareForThinLTO || shouldEmitRegularLTOSummary()));
     } else if (PrepareForThinLTO) {
       MPM.addPass(PB.buildThinLTOPreLinkDefaultPipeline(Level));
     } else if (PrepareForLTO) {
@@ -1172,9 +1173,8 @@ void EmitAssemblyHelper::RunOptimizationPipeline(
     if (!TheModule->getModuleFlag("EnableSplitLTOUnit"))
       TheModule->addModuleFlag(Module::Error, "EnableSplitLTOUnit",
                                uint32_t(CodeGenOpts.EnableSplitLTOUnit));
-    // FatLTO always means UnifiedLTO
-    if (!TheModule->getModuleFlag("UnifiedLTO"))
-      TheModule->addModuleFlag(Module::Error, "UnifiedLTO", uint32_t(1));
+    if (CodeGenOpts.UnifiedLTO && !TheModule->getModuleFlag("UnifiedLTO"))
+      TheModule->addModuleFlag(llvm::Module::Error, "UnifiedLTO", uint32_t(1));
   }
 
   // Print a textual, '-passes=' compatible, representation of pipeline if

clang/lib/Driver/ToolChains/Clang.cpp

@@ -5546,9 +5546,7 @@ void Clang::ConstructJob(Compilation &C, const JobAction &JA,
   bool UnifiedLTO = false;
   if (IsUsingLTO) {
     UnifiedLTO = Args.hasFlag(options::OPT_funified_lto,
-                              options::OPT_fno_unified_lto, Triple.isPS()) ||
-                 Args.hasFlag(options::OPT_ffat_lto_objects,
-                              options::OPT_fno_fat_lto_objects, false);
+                              options::OPT_fno_unified_lto, Triple.isPS());
     if (UnifiedLTO)
       CmdArgs.push_back("-funified-lto");
   }

clang/lib/Driver/ToolChains/Darwin.cpp

@@ -985,12 +985,12 @@ bool Darwin::hasBlocksRuntime() const {
 
 void Darwin::AddCudaIncludeArgs(const ArgList &DriverArgs,
                                 ArgStringList &CC1Args) const {
-  CudaInstallation.AddCudaIncludeArgs(DriverArgs, CC1Args);
+  CudaInstallation->AddCudaIncludeArgs(DriverArgs, CC1Args);
 }
 
 void Darwin::AddHIPIncludeArgs(const ArgList &DriverArgs,
                                ArgStringList &CC1Args) const {
-  RocmInstallation.AddHIPIncludeArgs(DriverArgs, CC1Args);
+  RocmInstallation->AddHIPIncludeArgs(DriverArgs, CC1Args);
 }
 
 // This is just a MachO name translation routine and there's no
@@ -3436,6 +3436,6 @@ SanitizerMask Darwin::getSupportedSanitizers() const {
 }
 
 void Darwin::printVerboseInfo(raw_ostream &OS) const {
-  CudaInstallation.print(OS);
-  RocmInstallation.print(OS);
+  CudaInstallation->print(OS);
+  RocmInstallation->print(OS);
 }

clang/lib/Driver/ToolChains/Darwin.h

@@ -10,6 +10,7 @@
 #define LLVM_CLANG_LIB_DRIVER_TOOLCHAINS_DARWIN_H
 
 #include "Cuda.h"
+#include "LazyDetector.h"
 #include "ROCm.h"
 #include "clang/Basic/DarwinSDKInfo.h"
 #include "clang/Basic/LangOptions.h"
@@ -321,8 +322,8 @@ class LLVM_LIBRARY_VISIBILITY Darwin : public MachO {
   /// The target variant triple that was specified (if any).
   mutable std::optional<llvm::Triple> TargetVariantTriple;
 
-  CudaInstallationDetector CudaInstallation;
-  RocmInstallationDetector RocmInstallation;
+  LazyDetector<CudaInstallationDetector> CudaInstallation;
+  LazyDetector<RocmInstallationDetector> RocmInstallation;
 
 private:
   void AddDeploymentTarget(llvm::opt::DerivedArgList &Args) const;

clang/lib/Driver/ToolChains/MSVC.cpp

@@ -582,24 +582,24 @@ bool MSVCToolChain::isPICDefaultForced() const {
 
 void MSVCToolChain::AddCudaIncludeArgs(const ArgList &DriverArgs,
                                        ArgStringList &CC1Args) const {
-  CudaInstallation.AddCudaIncludeArgs(DriverArgs, CC1Args);
+  CudaInstallation->AddCudaIncludeArgs(DriverArgs, CC1Args);
 }
 
 void MSVCToolChain::AddHIPIncludeArgs(const ArgList &DriverArgs,
                                       ArgStringList &CC1Args) const {
-  RocmInstallation.AddHIPIncludeArgs(DriverArgs, CC1Args);
+  RocmInstallation->AddHIPIncludeArgs(DriverArgs, CC1Args);
 }
 
 void MSVCToolChain::AddHIPRuntimeLibArgs(const ArgList &Args,
                                          ArgStringList &CmdArgs) const {
   CmdArgs.append({Args.MakeArgString(StringRef("-libpath:") +
-                                     RocmInstallation.getLibPath()),
+                                     RocmInstallation->getLibPath()),
                   "amdhip64.lib"});
 }
 
 void MSVCToolChain::printVerboseInfo(raw_ostream &OS) const {
-  CudaInstallation.print(OS);
-  RocmInstallation.print(OS);
+  CudaInstallation->print(OS);
+  RocmInstallation->print(OS);
 }
 
 std::string

clang/lib/Driver/ToolChains/MSVC.h

@@ -11,6 +11,7 @@
 
 #include "AMDGPU.h"
 #include "Cuda.h"
+#include "LazyDetector.h"
 #include "clang/Driver/Compilation.h"
 #include "clang/Driver/Tool.h"
 #include "clang/Driver/ToolChain.h"
@@ -142,8 +143,8 @@ class LLVM_LIBRARY_VISIBILITY MSVCToolChain : public ToolChain {
   std::optional<llvm::StringRef> WinSdkDir, WinSdkVersion, WinSysRoot;
   std::string VCToolChainPath;
   llvm::ToolsetLayout VSLayout = llvm::ToolsetLayout::OlderVS;
-  CudaInstallationDetector CudaInstallation;
-  RocmInstallationDetector RocmInstallation;
+  LazyDetector<CudaInstallationDetector> CudaInstallation;
+  LazyDetector<RocmInstallationDetector> RocmInstallation;
 };
 
 } // end namespace toolchains

clang/lib/Driver/ToolChains/MinGW.cpp

@@ -614,17 +614,17 @@ SanitizerMask toolchains::MinGW::getSupportedSanitizers() const {
 
 void toolchains::MinGW::AddCudaIncludeArgs(const ArgList &DriverArgs,
                                            ArgStringList &CC1Args) const {
-  CudaInstallation.AddCudaIncludeArgs(DriverArgs, CC1Args);
+  CudaInstallation->AddCudaIncludeArgs(DriverArgs, CC1Args);
 }
 
 void toolchains::MinGW::AddHIPIncludeArgs(const ArgList &DriverArgs,
                                           ArgStringList &CC1Args) const {
-  RocmInstallation.AddHIPIncludeArgs(DriverArgs, CC1Args);
+  RocmInstallation->AddHIPIncludeArgs(DriverArgs, CC1Args);
 }
 
 void toolchains::MinGW::printVerboseInfo(raw_ostream &OS) const {
-  CudaInstallation.print(OS);
-  RocmInstallation.print(OS);
+  CudaInstallation->print(OS);
+  RocmInstallation->print(OS);
 }
 
 // Include directories for various hosts:

clang/lib/Driver/ToolChains/MinGW.h

@@ -11,6 +11,7 @@
 
 #include "Cuda.h"
 #include "Gnu.h"
+#include "LazyDetector.h"
 #include "ROCm.h"
 #include "clang/Driver/Tool.h"
 #include "clang/Driver/ToolChain.h"
@@ -102,8 +103,8 @@ class LLVM_LIBRARY_VISIBILITY MinGW : public ToolChain {
   Tool *buildAssembler() const override;
 
 private:
-  CudaInstallationDetector CudaInstallation;
-  RocmInstallationDetector RocmInstallation;
+  LazyDetector<CudaInstallationDetector> CudaInstallation;
+  LazyDetector<RocmInstallationDetector> RocmInstallation;
 
   std::string Base;
   std::string GccLibDir;

clang/lib/Frontend/CompilerInvocation.cpp

@@ -1868,20 +1868,6 @@ bool CompilerInvocation::ParseCodeGenArgs(CodeGenOptions &Opts, ArgList &Args,
     if (Args.hasArg(OPT_funified_lto))
       Opts.PrepareForThinLTO = true;
   }
-  if (Arg *A = Args.getLastArg(options::OPT_ffat_lto_objects,
-                               options::OPT_fno_fat_lto_objects)) {
-    if (A->getOption().matches(options::OPT_ffat_lto_objects)) {
-      if (Arg *Uni = Args.getLastArg(options::OPT_funified_lto,
-                                     options::OPT_fno_unified_lto)) {
-        if (Uni->getOption().matches(options::OPT_fno_unified_lto))
-          Diags.Report(diag::err_drv_incompatible_options)
-              << A->getAsString(Args) << "-fno-unified-lto";
-      } else
-        Diags.Report(diag::err_drv_argument_only_allowed_with)
-            << A->getAsString(Args) << "-funified-lto";
-    }
-  }
-
   if (Arg *A = Args.getLastArg(OPT_fthinlto_index_EQ)) {
     if (IK.getLanguage() != Language::LLVM_IR)
       Diags.Report(diag::err_drv_argument_only_allowed_with)

clang/test/Analysis/malloc.mm

@@ -116,17 +116,17 @@ void testUseAfterFree() {
 }
 
 void testNoCopy() {
-  char *p = (char *)calloc(sizeof(int), 1);
+  char *p = (char *)calloc(1, sizeof(int));
   CustomData *w = [CustomData somethingNoCopy:p]; // no-warning
 }
 
 void testFreeWhenDone() {
-  char *p = (char *)calloc(sizeof(int), 1);
+  char *p = (char *)calloc(1, sizeof(int));
   CustomData *w = [CustomData something:p freeWhenDone:1]; // no-warning
 }
 
 void testFreeWhenDonePositive() {
-  char *p = (char *)calloc(sizeof(int), 1);
+  char *p = (char *)calloc(1, sizeof(int));
   CustomData *w = [CustomData something:p freeWhenDone:0]; // expected-warning{{leak}}
 }
 

clang/test/Analysis/uninit-vals.m

@@ -158,7 +158,7 @@ Point makePoint(float x, float y) {
 }
 
 void PR14765_test(void) {
-  Circle *testObj = calloc(sizeof(Circle), 1);
+  Circle *testObj = calloc(1, sizeof(Circle));
 
   clang_analyzer_eval(testObj->size == 0); // expected-warning{{TRUE}}
                                            // expected-note@-1{{TRUE}}
@@ -207,7 +207,7 @@ IntPoint makeIntPoint(int x, int y) {
 }
 
 void PR14765_test_int(void) {
-  IntCircle *testObj = calloc(sizeof(IntCircle), 1);
+  IntCircle *testObj = calloc(1, sizeof(IntCircle));
 
   clang_analyzer_eval(testObj->size == 0); // expected-warning{{TRUE}}
                                            // expected-note@-1{{TRUE}}
@@ -311,7 +311,7 @@ void testLargeStructsNotCopiedPerField(void) {
 }
 
 void testSmallStructInLargerStruct(void) {
-  IntCircle2D *testObj = calloc(sizeof(IntCircle2D), 1);
+  IntCircle2D *testObj = calloc(1, sizeof(IntCircle2D));
 
   clang_analyzer_eval(testObj->size == 0); // expected-warning{{TRUE}}
                                            // expected-note@-1{{TRUE}}

clang/test/CodeGen/alloc-size.c

@@ -137,7 +137,7 @@ void test5(void) {
   // CHECK: store i32 36
   gi = OBJECT_SIZE_BUILTIN(&data->t[1], 3);
 
-  struct Data *const arr = my_calloc(sizeof(*data), 2);
+  struct Data *const arr = my_calloc(2, sizeof(*data));
   // CHECK: store i32 96
   gi = OBJECT_SIZE_BUILTIN(arr, 0);
   // CHECK: store i32 96
@@ -171,7 +171,7 @@ void test6(void) {
   // CHECK: store i32 11
   gi = OBJECT_SIZE_BUILTIN(data->end, 3);
 
-  struct Data *const arr = my_calloc(sizeof(*arr) + 5, 3);
+  struct Data *const arr = my_calloc(3, sizeof(*arr) + 5);
   // AFAICT, GCC treats malloc and calloc identically. So, we should do the
   // same.
   //