resolve merge conflict

Author: kbobrovs

buildbot/configure.py

@@ -75,6 +75,11 @@ def do_configure(args):
         sycl_build_pi_hip_platform = args.hip_platform
         sycl_enabled_plugins.append("hip")
 
+    # all llvm compiler targets don't require 3rd party dependencies, so can be
+    # built/tested even if specific runtimes are not available
+    if args.enable_all_llvm_targets:
+        llvm_targets_to_build += ';NVPTX;AMDGPU'
+
     if args.werror or args.ci_defaults:
         sycl_werror = 'ON'
         xpti_enable_werror = 'ON'
@@ -210,6 +215,7 @@ def main():
     parser.add_argument("--hip-platform", type=str, choices=['AMD', 'NVIDIA'], default='AMD', help="choose hardware platform for HIP backend")
     parser.add_argument("--arm", action='store_true', help="build ARM support rather than x86")
     parser.add_argument("--enable-esimd-emulator", action='store_true', help="build with ESIMD emulation support")
+    parser.add_argument("--enable-all-llvm-targets", action='store_true', help="build compiler with all supported targets, it doesn't change runtime build")
     parser.add_argument("--no-assertions", action='store_true', help="build without assertions")
     parser.add_argument("--docs", action='store_true', help="build Doxygen documentation")
     parser.add_argument("--werror", action='store_true', help="Treat warnings as errors")

buildbot/dependency.conf

@@ -4,8 +4,8 @@ ocl_cpu_rt_ver=2022.13.3.0.16
 # https://github.com/intel/llvm/releases/download/2022-WW13/win-oclcpuexp-2022.13.3.0.16_rel.zip
 ocl_cpu_rt_ver_win=2022.13.3.0.16
 # Same GPU driver supports Level Zero and OpenCL
-# https://github.com/intel/compute-runtime/releases/tag/22.19.23136
-ocl_gpu_rt_ver=22.19.23136
+# https://github.com/intel/compute-runtime/releases/tag/22.20.23198
+ocl_gpu_rt_ver=22.20.23198
 # Same GPU driver supports Level Zero and OpenCL
 # https://downloadmirror.intel.com/723911/igfx_win_101.1404.zip
 ocl_gpu_rt_ver_win=101.1404
@@ -31,7 +31,7 @@ ocloc_ver_win=101.1404
 [DRIVER VERSIONS]
 cpu_driver_lin=2022.13.3.0.16
 cpu_driver_win=2022.13.3.0.16
-gpu_driver_lin=22.19.23136
+gpu_driver_lin=22.20.23198
 gpu_driver_win=101.1404
 fpga_driver_lin=2022.13.3.0.16
 fpga_driver_win=2022.13.3.0.16

clang/include/clang/Basic/LangOptions.h

@@ -470,6 +470,10 @@ class LangOptions : public LangOptionsBase {
   /// The seed used by the randomize structure layout feature.
   std::string RandstructSeed;
 
+  /// The name of the file to which the backend should save YAML optimization
+  /// records.
+  std::string OptRecordFile;
+
   LangOptions();
 
   /// Set language defaults for the given input language and

clang/include/clang/Driver/Options.td

@@ -2774,10 +2774,10 @@ def fsycl_device_code_split_esimd : Flag<["-"], "fsycl-device-code-split-esimd">
   Flags<[CoreOption]>, HelpText<"split ESIMD device code from SYCL into a separate device binary image (default). Has effect only for SPIR-based targets.">;
 def fno_sycl_device_code_split_esimd : Flag<["-"], "fno-sycl-device-code-split-esimd">,
   Flags<[CoreOption]>, HelpText<"do not split ESIMD and SYCL device code into separate device binary images. Has effect only for SPIR-based targets.">;
-def fsycl_instrument_device_code : Flag<["-"], "fsycl-instrument-device-code">,
-  Group<sycl_Group>, Flags<[CC1Option, CoreOption]>,
-  HelpText<"Add ITT instrumentation intrinsics calls">,
-  MarshallingInfoFlag<CodeGenOpts<"SPIRITTAnnotations">>;
+defm sycl_instrument_device_code : BoolFOption<"sycl-instrument-device-code",
+  CodeGenOpts<"SPIRITTAnnotations">, DefaultFalse,
+  PosFlag<SetTrue, [], "Add">, NegFlag<SetFalse, [], "Do not add">,
+  BothFlags<[CC1Option, CoreOption], " ITT instrumentation intrinsics calls">>;
 defm sycl_id_queries_fit_in_int: BoolFOption<"sycl-id-queries-fit-in-int",
   LangOpts<"SYCLValueFitInMaxInt">, DefaultTrue,
   PosFlag<SetTrue, [], "Assume">, NegFlag<SetFalse, [], "Do not assume">,
@@ -4858,6 +4858,12 @@ def fno_sycl_device_lib_EQ : CommaJoined<["-"], "fno-sycl-device-lib=">, Group<s
   Values<"libc, libm-fp32, libm-fp64, all">, HelpText<"Control exclusion of "
   "device libraries from device binary linkage. Valid arguments "
   "are libc, libm-fp32, libm-fp64, all">;
+def fsycl_device_lib_online_link : Flag<["-"], "fsycl-device-lib-online-link">,
+  Group<sycl_Group>, Flags<[NoArgumentUnused, CoreOption]>, HelpText<"Enables "
+  "sycl device library online link (experimental)">;
+def fno_sycl_device_lib_online_link : Flag<["-"], "fno-sycl-device-lib-online-link">,
+  Group<sycl_Group>, Flags<[NoArgumentUnused, CoreOption]>, HelpText<"Disables "
+  "sycl device library online link (experimental)">;
 def fsycl_fp32_prec_sqrt : Flag<["-"], "fsycl-fp32-prec-sqrt">, Group<sycl_Group>, Flags<[CC1Option]>,
   HelpText<"SYCL only. Specify that single precision floating-point sqrt is correctly rounded.">,
   MarshallingInfoFlag<CodeGenOpts<"SYCLFp32PrecSqrt">>;
@@ -5901,7 +5907,7 @@ def arcmt_action_EQ : Joined<["-"], "arcmt-action=">, Flags<[CC1Option, NoDriver
 
 def opt_record_file : Separate<["-"], "opt-record-file">,
   HelpText<"File name to use for YAML optimization record output">,
-  MarshallingInfoString<CodeGenOpts<"OptRecordFile">>;
+  MarshallingInfoString<LangOpts<"OptRecordFile">>;
 def opt_record_passes : Separate<["-"], "opt-record-passes">,
   HelpText<"Only record remark information for passes whose names match the given regular expression">;
 def opt_record_format : Separate<["-"], "opt-record-format">,

clang/lib/CodeGen/BackendUtil.cpp

@@ -934,7 +934,7 @@ void EmitAssemblyHelper::RunOptimizationPipeline(
   // Add SPIRITTAnnotations pass to the pass manager if
   // -fsycl-instrument-device-code option was passed. This option can be used
   // only with spir triple.
-  if (CodeGenOpts.SPIRITTAnnotations) {
+  if (LangOpts.SYCLIsDevice && CodeGenOpts.SPIRITTAnnotations) {
     assert(llvm::Triple(TheModule->getTargetTriple()).isSPIR() &&
            "ITT annotations can only be added to a module with spir target");
     MPM.addPass(SPIRITTAnnotationsPass());

clang/lib/CodeGen/CodeGenFunction.cpp

@@ -1602,26 +1602,27 @@ void CodeGenFunction::GenerateCode(GlobalDecl GD, llvm::Function *Fn,
 
   // Emit the standard function prologue.
   StartFunction(GD, ResTy, Fn, FnInfo, Args, Loc, BodyRange.getBegin());
-
-  SyclOptReportHandler &SyclOptReport = CGM.getDiags().getSYCLOptReport();
-  if (Fn && SyclOptReport.HasOptReportInfo(FD)) {
-    llvm::OptimizationRemarkEmitter ORE(Fn);
-    for (auto ORI : llvm::enumerate(SyclOptReport.GetInfo(FD))) {
-      llvm::DiagnosticLocation DL =
-          SourceLocToDebugLoc(ORI.value().KernelArgLoc);
-      StringRef NameInDesc = ORI.value().KernelArgDescName;
-      StringRef ArgType = ORI.value().KernelArgType;
-      StringRef ArgDesc = ORI.value().KernelArgDesc;
-      unsigned ArgSize = ORI.value().KernelArgSize;
-      StringRef ArgDecomposedField = ORI.value().KernelArgDecomposedField;
-
-      llvm::OptimizationRemark Remark("sycl", "Region", DL,
-                                      &Fn->getEntryBlock());
-      Remark << "Arg " << llvm::ore::NV("Argument", ORI.index()) << ":"
-             << ArgDesc << NameInDesc << "  (" << ArgDecomposedField
-             << "Type:" << ArgType << ", "
-             << "Size: " << llvm::ore::NV("Argument", ArgSize) << ")";
-      ORE.emit(Remark);
+  if (!getLangOpts().OptRecordFile.empty()) {
+    SyclOptReportHandler &SyclOptReport = CGM.getDiags().getSYCLOptReport();
+    if (Fn && SyclOptReport.HasOptReportInfo(FD)) {
+      llvm::OptimizationRemarkEmitter ORE(Fn);
+      for (auto ORI : llvm::enumerate(SyclOptReport.GetInfo(FD))) {
+        llvm::DiagnosticLocation DL =
+            SourceLocToDebugLoc(ORI.value().KernelArgLoc);
+        StringRef NameInDesc = ORI.value().KernelArgDescName;
+        StringRef ArgType = ORI.value().KernelArgType;
+        StringRef ArgDesc = ORI.value().KernelArgDesc;
+        unsigned ArgSize = ORI.value().KernelArgSize;
+        StringRef ArgDecomposedField = ORI.value().KernelArgDecomposedField;
+
+        llvm::OptimizationRemark Remark("sycl", "Region", DL,
+                                        &Fn->getEntryBlock());
+        Remark << "Arg " << llvm::ore::NV("Argument", ORI.index()) << ":"
+               << ArgDesc << NameInDesc << "  (" << ArgDecomposedField
+               << "Type:" << ArgType << ", "
+               << "Size: " << llvm::ore::NV("Argument", ArgSize) << ")";
+        ORE.emit(Remark);
+      }
     }
   }
 

clang/lib/Driver/Driver.cpp

@@ -4754,12 +4754,17 @@ class OffloadingActionBuilder final {
       SYCLTC->SYCLInstallation.getSYCLDeviceLibPath(LibLocCandidates);
       StringRef LibSuffix = isMSVCEnv ? ".obj" : ".o";
       using SYCLDeviceLibsList = SmallVector<DeviceLibOptInfo, 5>;
+
       const SYCLDeviceLibsList sycl_device_wrapper_libs = {
-          {"libsycl-crt", "libc"},
-          {"libsycl-complex", "libm-fp32"},
-          {"libsycl-complex-fp64", "libm-fp64"},
-          {"libsycl-cmath", "libm-fp32"},
-          {"libsycl-cmath-fp64", "libm-fp64"}};
+        {"libsycl-crt", "libc"},
+        {"libsycl-complex", "libm-fp32"},
+        {"libsycl-complex-fp64", "libm-fp64"},
+        {"libsycl-cmath", "libm-fp32"},
+        {"libsycl-cmath-fp64", "libm-fp64"},
+#if defined(_WIN32)
+        {"libsycl-msvc-math", "libm-fp32"},
+#endif
+      };
       // For AOT compilation, we need to link sycl_device_fallback_libs as
       // default too.
       const SYCLDeviceLibsList sycl_device_fallback_libs = {
@@ -4806,7 +4811,8 @@ class OffloadingActionBuilder final {
       addInputs(sycl_device_wrapper_libs);
       if (isSpirvAOT)
         addInputs(sycl_device_fallback_libs);
-      if (Args.hasArg(options::OPT_fsycl_instrument_device_code))
+      if (Args.hasFlag(options::OPT_fsycl_instrument_device_code,
+                       options::OPT_fno_sycl_instrument_device_code, true))
         addInputs(sycl_device_annotation_libs);
       return NumOfDeviceLibLinked != 0;
     }
@@ -4958,8 +4964,12 @@ class OffloadingActionBuilder final {
         // device libraries are only needed when current toolchain is using
         // AOT compilation.
         if (isSPIR) {
+          bool UseOnlineLink =
+              Args.hasFlag(options::OPT_fsycl_device_lib_online_link,
+                           options::OPT_fno_sycl_device_lib_online_link, false);
+          bool UseOfflineLink = isSpirvAOT || !UseOnlineLink;
           SYCLDeviceLibLinked = addSYCLDeviceLibs(
-              TC, FullLinkObjects, isSpirvAOT,
+              TC, FullLinkObjects, UseOfflineLink,
               C.getDefaultToolChain().getTriple().isWindowsMSVCEnvironment());
         }
 

clang/lib/Driver/ToolChains/Clang.cpp

@@ -4898,12 +4898,10 @@ void Clang::ConstructJob(Compilation &C, const JobAction &JA,
 
     // Forward -fsycl-instrument-device-code option to cc1. This option will
     // only be used for SPIR-V-based targets.
-    if (Arg *A =
-            Args.getLastArgNoClaim(options::OPT_fsycl_instrument_device_code))
-      if (Triple.isSPIR()) {
-        A->claim();
+    if (Triple.isSPIR())
+      if (Args.hasFlag(options::OPT_fsycl_instrument_device_code,
+                       options::OPT_fno_sycl_instrument_device_code, true))
         CmdArgs.push_back("-fsycl-instrument-device-code");
-      }
 
     if (!SYCLStdArg) {
       // The user had not pass SYCL version, thus we'll employ no-sycl-strict

clang/lib/Driver/ToolChains/SYCL.cpp

@@ -130,21 +130,16 @@ void SYCL::constructLLVMForeachCommand(Compilation &C, const JobAction &JA,
 // The list should match pre-built SYCL device library files located in
 // compiler package. Once we add or remove any SYCL device library files,
 // the list should be updated accordingly.
-static llvm::SmallVector<StringRef, 16> SYCLDeviceLibList{
-    "crt",
-    "cmath",
-    "cmath-fp64",
-    "complex",
-    "complex-fp64",
-    "itt-compiler-wrappers",
-    "itt-stubs",
-    "itt-user-wrappers",
-    "fallback-cassert",
-    "fallback-cstring",
-    "fallback-cmath",
-    "fallback-cmath-fp64",
-    "fallback-complex",
-    "fallback-complex-fp64"};
+
+static llvm::SmallVector<StringRef, 16> SYCLDeviceLibList {
+  "crt", "cmath", "cmath-fp64", "complex", "complex-fp64",
+#if defined(_WIN32)
+      "msvc-math",
+#endif
+      "itt-compiler-wrappers", "itt-stubs", "itt-user-wrappers",
+      "fallback-cassert", "fallback-cstring", "fallback-cmath",
+      "fallback-cmath-fp64", "fallback-complex", "fallback-complex-fp64"
+};
 
 const char *SYCL::Linker::constructLLVMLinkCommand(
     Compilation &C, const JobAction &JA, const InputInfo &Output,

clang/lib/Frontend/CompilerInvocation.cpp

@@ -1946,6 +1946,8 @@ bool CompilerInvocation::ParseCodeGenArgs(CodeGenOptions &Opts, ArgList &Args,
 
   bool NeedLocTracking = false;
 
+  Opts.OptRecordFile = LangOpts->OptRecordFile;
+
   if (!Opts.OptRecordFile.empty())
     NeedLocTracking = true;
 
@@ -3318,6 +3320,8 @@ void CompilerInvocation::GenerateLangArgs(const LangOptions &Opts,
       GenerateArg(Args, OPT_pic_is_pie, SA);
     for (StringRef Sanitizer : serializeSanitizerKinds(Opts.Sanitize))
       GenerateArg(Args, OPT_fsanitize_EQ, Sanitizer, SA);
+    if (!Opts.OptRecordFile.empty())
+      GenerateArg(Args, OPT_opt_record_file, Opts.OptRecordFile, SA);
 
     return;
   }
@@ -3617,6 +3621,12 @@ bool CompilerInvocation::ParseLangArgs(LangOptions &Opts, ArgList &Args,
     parseSanitizerKinds("-fsanitize=", Args.getAllArgValues(OPT_fsanitize_EQ),
                         Diags, Opts.Sanitize);
 
+    // OptRecordFile is used to generate the optimization record file should
+    // be set regardless of the input type.
+    if (Args.hasArg(OPT_opt_record_file))
+      Opts.OptRecordFile =
+          std::string(Args.getLastArgValue(OPT_opt_record_file));
+
     return Diags.getNumErrors() == NumErrorsBefore;
   }
 

clang/lib/Sema/SemaDeclAttr.cpp

@@ -7623,14 +7623,17 @@ SYCLIntelFPGAMaxConcurrencyAttr *Sema::MergeSYCLIntelFPGAMaxConcurrencyAttr(
   // Check to see if there's a duplicate attribute with different values
   // already applied to the declaration.
   if (const auto *DeclAttr = D->getAttr<SYCLIntelFPGAMaxConcurrencyAttr>()) {
-    const auto *DeclExpr = dyn_cast<ConstantExpr>(DeclAttr->getNThreadsExpr());
-    const auto *MergeExpr = dyn_cast<ConstantExpr>(A.getNThreadsExpr());
-    if (DeclExpr && MergeExpr &&
-        DeclExpr->getResultAsAPSInt() != MergeExpr->getResultAsAPSInt()) {
-      Diag(DeclAttr->getLoc(), diag::warn_duplicate_attribute) << &A;
-      Diag(A.getLoc(), diag::note_previous_attribute);
+    if (const auto *DeclExpr =
+            dyn_cast<ConstantExpr>(DeclAttr->getNThreadsExpr())) {
+      if (const auto *MergeExpr = dyn_cast<ConstantExpr>(A.getNThreadsExpr())) {
+        if (DeclExpr->getResultAsAPSInt() != MergeExpr->getResultAsAPSInt()) {
+          Diag(DeclAttr->getLoc(), diag::warn_duplicate_attribute) << &A;
+          Diag(A.getLoc(), diag::note_previous_attribute);
+        }
+        // Do not add a duplicate attribute.
+        return nullptr;
+      }
     }
-    return nullptr;
   }
 
   return ::new (Context)
@@ -7654,14 +7657,21 @@ void Sema::AddSYCLIntelFPGAMaxConcurrencyAttr(Decl *D,
       return;
     }
 
+    // Check to see if there's a duplicate attribute with different values
+    // already applied to the declaration.
     if (const auto *DeclAttr = D->getAttr<SYCLIntelFPGAMaxConcurrencyAttr>()) {
-      const auto *DeclExpr =
-          dyn_cast<ConstantExpr>(DeclAttr->getNThreadsExpr());
-      if (DeclExpr && ArgVal != DeclExpr->getResultAsAPSInt()) {
-        Diag(CI.getLoc(), diag::warn_duplicate_attribute) << CI;
-        Diag(DeclAttr->getLoc(), diag::note_previous_attribute);
+      // If the other attribute argument is instantiation dependent, we won't
+      // have converted it to a constant expression yet and thus we test
+      // whether this is a null pointer.
+      if (const auto *DeclExpr =
+              dyn_cast<ConstantExpr>(DeclAttr->getNThreadsExpr())) {
+        if (ArgVal != DeclExpr->getResultAsAPSInt()) {
+          Diag(CI.getLoc(), diag::warn_duplicate_attribute) << CI;
+          Diag(DeclAttr->getLoc(), diag::note_previous_attribute);
+        }
+        // Drop the duplicate attribute.
+        return;
       }
-      return;
     }
   }
 

clang/lib/Sema/SemaSYCL.cpp

@@ -3669,17 +3669,29 @@ void Sema::ConstructOpenCLKernel(FunctionDecl *KernelCallerFunc,
   SyclOptReportCreator opt_report(*this, kernel_decl, KernelObj->getLocation());
 
   KernelObjVisitor Visitor{*this};
-  Visitor.VisitRecordBases(KernelObj, kernel_decl, kernel_body, int_header,
-                           int_footer, opt_report);
-  Visitor.VisitRecordFields(KernelObj, kernel_decl, kernel_body, int_header,
-                            int_footer, opt_report);
+
+  // Visit handlers to generate information for optimization record only if
+  // optimization record is saved.
+  if (!getLangOpts().OptRecordFile.empty()) {
+    Visitor.VisitRecordBases(KernelObj, kernel_decl, kernel_body, int_header,
+                             int_footer, opt_report);
+    Visitor.VisitRecordFields(KernelObj, kernel_decl, kernel_body, int_header,
+                              int_footer, opt_report);
+  } else {
+    Visitor.VisitRecordBases(KernelObj, kernel_decl, kernel_body, int_header,
+                             int_footer);
+    Visitor.VisitRecordFields(KernelObj, kernel_decl, kernel_body, int_header,
+                              int_footer);
+  }
 
   if (ParmVarDecl *KernelHandlerArg =
           getSyclKernelHandlerArg(KernelCallerFunc)) {
     kernel_decl.handleSyclKernelHandlerType();
     kernel_body.handleSyclKernelHandlerType(KernelHandlerArg);
     int_header.handleSyclKernelHandlerType(KernelHandlerArg->getType());
-    opt_report.handleSyclKernelHandlerType();
+
+    if (!getLangOpts().OptRecordFile.empty())
+      opt_report.handleSyclKernelHandlerType();
   }
 }