Merge remote-tracking branch 'intel/sycl' into steffen/coverity_111125

Author: steffenlarsen

clang/include/clang/Driver/Options.td

@@ -1341,12 +1341,12 @@ def fgpu_sanitize : Flag<["-"], "fgpu-sanitize">, Group<f_Group>,
 def fno_gpu_sanitize : Flag<["-"], "fno-gpu-sanitize">, Group<f_Group>;
 
 def offload_compress : Flag<["--"], "offload-compress">,
-  HelpText<"Compress offload device binaries (HIP only)">;
+  HelpText<"Compress offload device binaries (HIP and SYCL only)">;
 def no_offload_compress : Flag<["--"], "no-offload-compress">;
 
 def offload_compression_level_EQ : Joined<["--"], "offload-compression-level=">,
   Flags<[HelpHidden]>,
-  HelpText<"Compression level for offload device binaries (HIP only)">;
+  HelpText<"Compression level for offload device binaries (HIP and SYCL only)">;
 
 def offload_jobs_EQ : Joined<["--"], "offload-jobs=">,
   HelpText<"Specify the number of threads to use for device offloading tasks "

clang/lib/Driver/ToolChains/Clang.cpp

@@ -10260,6 +10260,18 @@ void OffloadPackager::ConstructJob(Compilation &C, const JobAction &JA,
         "kind=" + Kind.str(),
     };
 
+    // When compiling like -fsycl-targets=spir64_gen -Xsycl-target-backend
+    // "-device pvc,bdw", the offloading arch will be "pvc,bdw", which
+    // contains a comma. Because the comma is used to separate fields
+    // within the --image option, we cannot pass arch=pvc,bdw directly.
+    // Instead, we pass it like arch=pvc,arch=bdw, then
+    // llvm-offload-binary joins them back to arch=pvc,bdw.
+    SmallVector<StringRef> Archs;
+    Arch.split(Archs, ',');
+    if (Archs.size() > 1) {
+      Parts[2] = "arch=" + llvm::join(Archs, ",arch=");
+    }
+
     if (TC->getDriver().isUsingOffloadLTO())
       for (StringRef Feature : FeatureArgs)
         Parts.emplace_back("feature=" + Feature.str());
@@ -10283,7 +10295,13 @@ void OffloadPackager::ConstructJob(Compilation &C, const JobAction &JA,
           AL += " ";
           AL += A;
         }
-        Parts.emplace_back(C.getArgs().MakeArgString(Twine(Opt) + AL));
+        // As mentioned earlier, we cannot pass a value with commas directly,
+        // but llvm-offload-binary joins multiple occurrences of the same
+        // option separated by commas, so we split the value on
+        // all commas and pass them as separate arguments.
+        for (StringRef Split : llvm::split(AL, ',')) {
+          Parts.emplace_back(C.getArgs().MakeArgString(Twine(Opt) + Split));
+        }
       };
       const ArgList &Args =
           C.getArgsForToolChain(nullptr, StringRef(), Action::OFK_SYCL);
@@ -10292,10 +10310,10 @@ void OffloadPackager::ConstructJob(Compilation &C, const JobAction &JA,
           static_cast<const toolchains::SYCLToolChain &>(*TC);
       SYCLTC.AddImpliedTargetArgs(TC->getTriple(), Args, BuildArgs, JA, *HostTC,
                                   Arch);
-      SYCLTC.TranslateBackendTargetArgs(TC->getTriple(), Args, BuildArgs, Arch);
+      SYCLTC.TranslateBackendTargetArgs(TC->getTriple(), Args, BuildArgs);
       createArgString("compile-opts=");
       BuildArgs.clear();
-      SYCLTC.TranslateLinkerTargetArgs(TC->getTriple(), Args, BuildArgs, Arch);
+      SYCLTC.TranslateLinkerTargetArgs(TC->getTriple(), Args, BuildArgs);
       createArgString("link-opts=");
     }
 

clang/test/Driver/clang-linker-wrapper.cpp

@@ -63,13 +63,20 @@
 // CHK-CMDS-AOT-GEN: spirv-to-ir-wrapper{{.*}} -o [[FIRSTLLVMLINKIN:.*]].bc --llvm-spirv-opts --spirv-preserve-auxdata --spirv-target-env=SPV-IR --spirv-builtin-format=global
 // CHK-CMDS-AOT-GEN-NEXT: llvm-link{{.*}} --suppress-warnings [[FIRSTLLVMLINKIN]].bc -o [[FIRSTLLVMLINKOUT:.*]].bc
 // CHK-CMDS-AOT-GEN-NEXT: llvm-link{{.*}} -only-needed --suppress-warnings [[FIRSTLLVMLINKOUT]].bc {{.*}}.bc -o [[SECONDLLVMLINKOUT:.*]].bc
-// CHK-CMDS-AOT-GEN-NEXT: sycl-post-link{{.*}} SYCL_POST_LINK_OPTIONS -o [[SYCLPOSTLINKOUT:.*]].table [[SECONDLLVMLINKOUT]].bc
+// Check that target specified by -fsycl-targets is passed to sycl-post-link for filtering.
+// CHK-CMDS-AOT-GEN-NEXT: sycl-post-link{{.*}} SYCL_POST_LINK_OPTIONS -o intel_gpu_pvc,[[SYCLPOSTLINKOUT:.*]].table [[SECONDLLVMLINKOUT]].bc
 // CHK-CMDS-AOT-GEN-NEXT: llvm-spirv{{.*}} LLVM_SPIRV_OPTIONS -o {{.*}}
 // CHK-CMDS-AOT-GEN-NEXT: ocloc{{.*}} -output_no_suffix -spirv_input -device pvc{{.*}} -output {{.*}} -file {{.*}}
 // CHK-CMDS-AOT-GEN-NEXT: offload-wrapper: output: [[WRAPPEROUT:.*]].bc, input: {{.*}}, compile-opts: , link-opts:
 // CHK-CMDS-AOT-GEN-NEXT: clang{{.*}} -c -o [[LLCOUT:.*]].o [[WRAPPEROUT]].bc
 // CHK-CMDS-AOT-GEN-NEXT: "{{.*}}/ld" -- HOST_LINKER_FLAGS -dynamic-linker HOST_DYN_LIB -o a.out [[LLCOUT]].o HOST_LIB_PATH HOST_STAT_LIB {{.*}}.o
 
+// Check that when --gpu-tool-arg is specified in clang-linker-wrapper 
+// (happen when AOT device is specified via -Xsycl-target-backend '-device pvc' in clang), 
+// the target is not passed to sycl-post-link for filtering.
+// RUN: clang-linker-wrapper -sycl-embed-ir -sycl-device-libraries=%t1.devicelib.o -sycl-post-link-options="SYCL_POST_LINK_OPTIONS" -llvm-spirv-options="LLVM_SPIRV_OPTIONS" "--host-triple=x86_64-unknown-linux-gnu" "--gpu-tool-arg=-device pvc" "--linker-path=/usr/bin/ld" "--" HOST_LINKER_FLAGS "-dynamic-linker" HOST_DYN_LIB "-o" "a.out" HOST_LIB_PATH HOST_STAT_LIB %t1.o --dry-run 2>&1 | FileCheck -check-prefix=CHK-NO-CMDS-AOT-GEN %s
+// CHK-NO-CMDS-AOT-GEN-NOT: sycl-post-link{{.*}} -o intel_gpu_pv,{{.*}}
+
 /// Check for list of commands for standalone clang-linker-wrapper run for sycl (AOT for Intel CPU)
 // -------
 // Generate .o file as linker wrapper input.

clang/test/Driver/sycl-ftarget-compile-fast.cpp

@@ -8,7 +8,7 @@
 // RUN:   | FileCheck -check-prefix=TARGET_COMPILE_FAST_GEN %s
 
 // TARGET_COMPILE_FAST_GEN: llvm-offload-binary
-// TARGET_COMPILE_FAST_GEN: compile-opts={{.*}}-options -igc_opts 'PartitionUnit=1,SubroutineThreshold=50000'
+// TARGET_COMPILE_FAST_GEN: compile-opts={{.*}}-options -igc_opts 'PartitionUnit=1,compile-opts=SubroutineThreshold=50000'
 
 // RUN: %clang -### -target x86_64-unknown-linux-gnu -fsycl --offload-new-driver \
 // RUN:    -ftarget-compile-fast %s 2>&1 \

clang/test/Driver/sycl-offload-new-driver.c

@@ -188,7 +188,7 @@
 // RUN: %clangxx -fsycl -### -fsycl-targets=spir64_gen --offload-new-driver \
 // RUN:   -Xsycl-target-backend=spir64_gen "-device pvc,bdw" %s 2>&1 \
 // RUN:   | FileCheck -check-prefix COMMA_FILE %s
-// COMMA_FILE: llvm-offload-binary{{.*}} "--image=file={{.*}}pvc@bdw{{.*}},triple=spir64_gen-unknown-unknown,arch=pvc,bdw,kind=sycl,compile-opts=-device_options pvc -ze-intel-enable-auto-large-GRF-mode"
+// COMMA_FILE: llvm-offload-binary{{.*}} "--image=file={{.*}}pvc@bdw{{.*}},triple=spir64_gen-unknown-unknown,arch=pvc,arch=bdw,kind=sycl,compile-opts=-device_options pvc -ze-intel-enable-auto-large-GRF-mode -device pvc,compile-opts=bdw"
 
 /// Verify the arch value for the packager is populated with different
 /// scenarios for spir64_gen
@@ -212,6 +212,52 @@
 // RUN:   | FileCheck -check-prefix ARCH_CHECK %s
 // ARCH_CHECK: llvm-offload-binary{{.*}} "--image=file={{.*}}triple=spir64_gen-unknown-unknown,arch=bdw,kind=sycl{{.*}}"
 
+// Verify when a comma-separated list of architectures is provided in -device, they are
+// passed to llvm-offload-binary correctly.
+// RUN: %clangxx -fsycl -### -fsycl-targets=spir64_gen --offload-new-driver \
+// RUN:   -Xsycl-target-backend "-device pvc,bdw" %s 2>&1 \
+// RUN:   | FileCheck -check-prefix MULTI_ARCH %s
+// RUN: %clangxx -fsycl -### -fsycl-targets=spir64_gen --offload-new-driver \
+// RUN:   -Xsycl-target-backend=spir64_gen "-device pvc,bdw" %s 2>&1 \
+// RUN:   | FileCheck -check-prefix MULTI_ARCH %s
+// RUN: %clangxx -fsycl -### -fsycl-targets=spir64_gen --offload-new-driver \
+// RUN:   -Xs "-device pvc,bdw" %s 2>&1 \
+// RUN:   | FileCheck -check-prefix MULTI_ARCH %s
+// MULTI_ARCH: llvm-offload-binary{{.*}} "--image=file={{.*}}triple=spir64_gen-unknown-unknown,arch=pvc,arch=bdw,kind=sycl
+// MULTI_ARCH-SAME: compile-opts=-device_options pvc -ze-intel-enable-auto-large-GRF-mode -device pvc,compile-opts=bdw"
+
+// Verify that when an object produced by llvm-offload-binary with multiple Intel GPU architectures
+// clang-linker-wrapper will call ocloc with -device listing all architectures.
+// RUN: %clangxx -fsycl -fsycl-targets=spir64_gen --offload-new-driver \
+// RUN:  -Xsycl-target-backend=spir64_gen "-device pvc,bdw" -c %s -o %t_multiarch_test.o
+// RUN: clang-linker-wrapper --dry-run --linker-path=/usr/bin/ld \
+// RUN:  --host-triple=x86_64-unknown-linux-gnu %t_multiarch_test.o 2>&1 \
+// RUN:  | FileCheck -check-prefix=OCLOC_MULTI_ARCH %s
+// OCLOC_MULTI_ARCH: ocloc{{.*}}-device pvc,bdw
+
+// Verify for multiple targets with -Xsycl-target-backend= with commas in the values
+// are passed correctly to llvm-offload-binary.
+// RUN: %clangxx -fsycl -### --offload-new-driver \
+// RUN:  -fsycl-targets=nvptx64-nvidia-cuda,amdgcn-amd-amdhsa,spir64_gen \
+// RUN:  -Xsycl-target-backend=amdgcn-amd-amdhsa --offload-arch=gfx908,gfx1010 \
+// RUN:  -Xsycl-target-backend=nvptx64-nvidia-cuda --offload-arch=sm_86,sm_87,sm_89 \
+// RUN:  -Xsycl-target-backend=spir64_gen "-device pvc,bdw" \
+// RUN:  -Xsycl-target-linker=spir64_gen "-DFOO,BAR" \
+// RUN:  -nogpulib %s 2>&1 | FileCheck -check-prefix=MULTI_ARCH2 %s
+// MULTI_ARCH2: llvm-offload-binary{{.*}} "--image=file={{.*}}triple=amdgcn-amd-amdhsa,arch=gfx1010,kind=sycl,compile-opts=--offload-arch=gfx908,compile-opts=gfx1010"
+// MULTI_ARCH2-SAME: "--image=file={{.*}}triple=amdgcn-amd-amdhsa,arch=gfx908,kind=sycl,compile-opts=--offload-arch=gfx908,compile-opts=gfx1010"
+// MULTI_ARCH2-SAME: "--image=file={{.*}}triple=nvptx64-nvidia-cuda,arch=sm_86,kind=sycl,compile-opts=--offload-arch=sm_86,compile-opts=sm_87,compile-opts=sm_89"
+// MULTI_ARCH2-SAME: "--image=file={{.*}}triple=nvptx64-nvidia-cuda,arch=sm_87,kind=sycl,compile-opts=--offload-arch=sm_86,compile-opts=sm_87,compile-opts=sm_89"
+// MULTI_ARCH2-SAME: "--image=file={{.*}}triple=nvptx64-nvidia-cuda,arch=sm_89,kind=sycl,compile-opts=--offload-arch=sm_86,compile-opts=sm_87,compile-opts=sm_89"
+// MULTI_ARCH2-SAME: "--image=file={{.*}}triple=spir64_gen-unknown-unknown,arch=pvc,arch=bdw,kind=sycl,compile-opts=-device_options pvc -ze-intel-enable-auto-large-GRF-mode -device pvc,compile-opts=bdw,link-opts=-DFOO,link-opts=BAR"
+
+// Verify that the driver correctly handles link-opt and compile-opt values with commas
+// RUN: %clangxx -fsycl -### -fsycl-targets=spir64_gen --offload-new-driver \
+// RUN:   -Xsycl-target-backend "-device bdw -FOO a,b" \
+// RUN:   -Xsycl-target-linker "-BAR x,y" %s 2>&1 \
+// RUN:   | FileCheck -check-prefix COMMA_OPTS %s
+// COMMA_OPTS: llvm-offload-binary{{.*}} "--image=file={{.*}}triple=spir64_gen-unknown-unknown,arch=bdw,kind=sycl,compile-opts=-device bdw -FOO a,compile-opts=b,link-opts=-BAR x,link-opts=y"
+
 /// Verify that --cuda-path is passed to clang-linker-wrapper for SYCL offload
 // RUN: %clangxx -fsycl -### -fsycl-targets=nvptx64-nvidia-cuda -fno-sycl-libspirv \
 // RUN:          --cuda-gpu-arch=sm_20 --cuda-path=%S/Inputs/CUDA_80/usr/local/cuda %s \

clang/tools/clang-linker-wrapper/ClangLinkerWrapper.cpp

@@ -727,10 +727,24 @@ runSYCLPostLinkTool(ArrayRef<StringRef> InputFiles, const ArgList &Args) {
       createOutputFile(sys::path::filename(ExecutableName), "table");
   if (!TempFileOrErr)
     return TempFileOrErr.takeError();
+  std::string OutputPathWithArch = TempFileOrErr->str();
+
+  // Enable the driver to invoke sycl-post-link with the device architecture
+  // when Intel GPU targets are passed in -fsycl-targets.
+  // OPT_gpu_tool_arg_EQ is checked to ensure the device architecture is not
+  // passed through -Xsycl-target-backend=spir64_gen "-device <arch>" format
+  const llvm::Triple Triple(Args.getLastArgValue(OPT_triple_EQ));
+  StringRef Arch = Args.getLastArgValue(OPT_arch_EQ);
+  StringRef IsGPUTool = Args.getLastArgValue(OPT_gpu_tool_arg_EQ);
+
+  if (Triple.getSubArch() == llvm::Triple::SPIRSubArch_gen && !Arch.empty() &&
+      IsGPUTool.empty() && Arch != "*")
+    OutputPathWithArch = "intel_gpu_" + Arch.str() + "," + OutputPathWithArch;
+  else if (Triple.getSubArch() == llvm::Triple::SPIRSubArch_x86_64)
+    OutputPathWithArch = "spir64_x86_64," + OutputPathWithArch;
 
   SmallVector<StringRef, 8> CmdArgs;
   CmdArgs.push_back(*SYCLPostLinkPath);
-  const llvm::Triple Triple(Args.getLastArgValue(OPT_triple_EQ));
   Arg *SYCLDeviceLibLoc = Args.getLastArg(OPT_sycl_device_library_location_EQ);
   if (SYCLDeviceLibLoc && !Triple.isSPIRAOT()) {
     std::string SYCLDeviceLibLocParam = SYCLDeviceLibLoc->getValue();
@@ -748,7 +762,7 @@ runSYCLPostLinkTool(ArrayRef<StringRef> InputFiles, const ArgList &Args) {
   SYCLPostLinkOptions.split(CmdArgs, " ", /* MaxSplit = */ -1,
                             /* KeepEmpty = */ false);
   CmdArgs.push_back("-o");
-  CmdArgs.push_back(*TempFileOrErr);
+  CmdArgs.push_back(Args.MakeArgString(OutputPathWithArch));
   for (auto &File : InputFiles)
     CmdArgs.push_back(File);
   if (Error Err = executeCommands(*SYCLPostLinkPath, CmdArgs))
@@ -943,24 +957,29 @@ static void addSYCLBackendOptions(const ArgList &Args,
   if (IsCPU) {
     BackendOptions.split(CmdArgs, " ", /*MaxSplit=*/-1, /*KeepEmpty=*/false);
   } else {
-    // ocloc -options args need to be comma separated, e.g. `-options
-    // "-g,-cl-opt-disable"`. Otherwise, only the first arg is processed by
-    // ocloc as an arg for -options, and the rest are processed as standalone
-    // flags, possibly leading to errors.
+    // ocloc -options takes arguments in the form of '-options "-g
+    // -cl-opt-disable"' where each argument is separated with spaces.
     // split function here returns a pair with everything before the separator
     // ("-options") in the first member of the pair, and everything after the
     // separator in the second part of the pair. The separator is not included
     // in any of them.
     auto [BeforeOptions, AfterOptions] = BackendOptions.split("-options ");
     // Only add if not empty, an empty arg can lead to ocloc errors.
-    if (!BeforeOptions.empty())
-      CmdArgs.push_back(BeforeOptions);
+    if (!BeforeOptions.empty()) {
+      SmallVector<StringRef, 8> BeforeArgs;
+      BeforeOptions.split(BeforeArgs, " ", /*MaxSplit=*/-1,
+                          /*KeepEmpty=*/false);
+      for (const auto &string : BeforeArgs) {
+        CmdArgs.push_back(string);
+      }
+    }
     if (!AfterOptions.empty()) {
-      // Separator not included by the split function, so explicitly added here.
       CmdArgs.push_back("-options");
-      std::string Replace = AfterOptions.str();
-      std::replace(Replace.begin(), Replace.end(), ' ', ',');
-      CmdArgs.push_back(Args.MakeArgString(Replace));
+      // Split the options string by spaces and rejoin to normalize whitespace
+      SmallVector<StringRef, 8> AfterArgs;
+      AfterOptions.split(AfterArgs, " ", /*MaxSplit=*/-1, /*KeepEmpty=*/false);
+      std::string JoinedOptions = llvm::join(AfterArgs, " ");
+      CmdArgs.push_back(Args.MakeArgString(JoinedOptions));
     }
   }
 
@@ -1441,9 +1460,8 @@ static Expected<StringRef> linkDevice(ArrayRef<StringRef> InputFiles,
   if (ExtractedDeviceLibFiles.empty()) {
     // TODO: Add NVPTX when ready
     if (Triple.isSPIROrSPIRV())
-      return createStringError(
-          inconvertibleErrorCode(),
-          " SYCL device library file list cannot be empty.");
+      WithColor::warning(errs(), LinkerExecutable)
+          << "SYCL device library file list is empty\n";
     return *LinkedFile;
   }
 

devops/actions/run-tests/benchmark/action.yml

@@ -261,13 +261,14 @@ runs:
         ${{ inputs.dry_run == 'true' && '--dry-run' || '' }} \
 
       echo "::endgroup::"
-      
-      # Run integration tests
-      # NOTE: Each integration test prints its own group name as part of test script
-      export LLVM_BENCHMARKS_UNIT_TESTING=1
-      export COMPUTE_BENCHMARKS_BUILD_PATH=$WORKDIR/compute-benchmarks-build
-      python3 ./devops/scripts/benchmarks/tests/test_integration.py
 
+      # Run benchmarks' integration tests
+      # NOTE: Each integration test prints its own group name as part of test script
+      if [ '${{ github.event_name == 'pull_request' }}' = 'true' ]; then
+        export LLVM_BENCHMARKS_UNIT_TESTING=1
+        export COMPUTE_BENCHMARKS_BUILD_PATH=$WORKDIR/compute-benchmarks-build
+        python3 ./devops/scripts/benchmarks/tests/test_integration.py
+      fi
   - name: Cache changes and upload github summary
     if: always()
     shell: bash

devops/scripts/benchmarks/benches/compute.py

@@ -61,8 +61,8 @@ def git_url(self) -> str:
         return "https://github.com/intel/compute-benchmarks.git"
 
     def git_hash(self) -> str:
-        # Nov 7, 2025
-        return "d985da634fc1a9416ca0bd067cfb9886b02d0211"
+        # Nov 17, 2025
+        return "932ae79f7cca7e156285fc10a59610927c769e89"
 
     def setup(self) -> None:
         if options.sycl is None:
@@ -860,6 +860,7 @@ def _bin_args(self, run_trace: TracingType = TracingType.NONE) -> list[str]:
             "--multiplier=1",
             "--vectorSize=1",
             "--lws=256",
+            "--prefetch=0",
         ]