Merge remote-tracking branch 'upstream/sycl' into urlog

Author: mateuszpn

.github/workflows/sycl-windows-build.yml

@@ -86,7 +86,6 @@ jobs:
     - name:  Detect hung tests
       if: always()
       shell: powershell
-      continue-on-error: true
       run: |
         Invoke-WebRequest -Uri "https://raw.githubusercontent.com/intel/llvm/refs/heads/sycl/devops/scripts/windows_detect_hung_tests.ps1" -OutFile "windows_detect_hung_tests.ps1"
         powershell.exe -File windows_detect_hung_tests.ps1

.github/workflows/sycl-windows-run-tests.yml

@@ -101,7 +101,6 @@ jobs:
     - name:  Detect hung tests
       if: always()
       shell: powershell
-      continue-on-error: true
       run: |
         Invoke-WebRequest -Uri "https://raw.githubusercontent.com/intel/llvm/refs/heads/sycl/devops/scripts/windows_detect_hung_tests.ps1" -OutFile "windows_detect_hung_tests.ps1"
         powershell.exe -File windows_detect_hung_tests.ps1
@@ -181,7 +180,7 @@ jobs:
       if: inputs.tests_selector == 'e2e'
       shell: bash {0}
       env:
-        LIT_OPTS: -v --no-progress-bar --show-unsupported --show-pass --show-xfail --max-time 3600 --time-tests --param print_features=True --param test-mode=${{ inputs.e2e_testing_mode }} ${{ inputs.extra_lit_opts }}
+        LIT_OPTS: -v --no-progress-bar --show-unsupported --show-pass --show-xfail --max-time ${{ inputs.e2e_testing_mode == 'run-only' && 1200 || 3600 }} --time-tests --param print_features=True --param test-mode=${{ inputs.e2e_testing_mode }} ${{ inputs.extra_lit_opts }}
       run: |
         # Run E2E tests.
         if [[ ${{inputs.compiler}} == 'icx' ]]; then

buildbot/configure.py

@@ -78,10 +78,6 @@ def do_configure(args, passthrough_args):
             sycl_enabled_backends.append("level_zero")
             sycl_enabled_backends.append("level_zero_v2")
 
-    # lld is needed on Windows or for the HIP adapter on AMD
-    if platform.system() == "Windows" or (args.hip and args.hip_platform == "AMD"):
-        llvm_enable_projects += ";lld"
-
     libclc_enabled = args.cuda or args.hip or args.native_cpu
     if libclc_enabled:
         llvm_enable_projects += ";libclc"
@@ -171,6 +167,10 @@ def do_configure(args, passthrough_args):
     if args.disable_preview_lib:
         sycl_preview_lib = "OFF"
 
+    # lld is needed on Windows or when building AMDGPU
+    if platform.system() == "Windows" or "AMDGPU" in llvm_targets_to_build:
+        llvm_enable_projects += ";lld"
+
     install_dir = os.path.join(abs_obj_dir, "install")
 
     cmake_cmd = [

clang/lib/CodeGen/CGBuiltin.cpp

@@ -3154,9 +3154,10 @@ RValue CodeGenFunction::EmitBuiltinExpr(const GlobalDecl GD, unsigned BuiltinID,
       GenerateIntrinsics =
           ConstWithoutErrnoOrExceptions && ErrnoOverridenToFalseWithOpt;
   }
-  if (GenerateIntrinsics &&
-      !(getLangOpts().SYCLIsDevice && (getTarget().getTriple().isNVPTX() ||
-                                       getTarget().getTriple().isAMDGCN()))) {
+  bool IsSYCLDeviceWithoutIntrinsics =
+      getLangOpts().SYCLIsDevice &&
+      (getTarget().getTriple().isNVPTX() || getTarget().getTriple().isAMDGCN());
+  if (GenerateIntrinsics && !IsSYCLDeviceWithoutIntrinsics) {
     switch (BuiltinIDIfNoAsmLabel) {
     case Builtin::BIacos:
     case Builtin::BIacosf:
@@ -4256,7 +4257,7 @@ RValue CodeGenFunction::EmitBuiltinExpr(const GlobalDecl GD, unsigned BuiltinID,
   case Builtin::BI__builtin_modf:
   case Builtin::BI__builtin_modff:
   case Builtin::BI__builtin_modfl:
-    if (Builder.getIsFPConstrained())
+    if (Builder.getIsFPConstrained() || IsSYCLDeviceWithoutIntrinsics)
       break; // TODO: Emit constrained modf intrinsic once one exists.
     return RValue::get(emitModfBuiltin(*this, E, Intrinsic::modf));
   case Builtin::BI__builtin_isgreater:

clang/lib/Driver/ToolChains/SYCL.cpp

@@ -603,7 +603,15 @@ SYCL::getDeviceLibraries(const Compilation &C, const llvm::Triple &TargetTriple,
       // For DG2, we just use libsycl-msan as placeholder.
       {"libsycl-msan", "internal"},
       {"libsycl-msan-pvc", "internal"}};
-  const SYCLDeviceLibsList SYCLDeviceTsanLibs = {{"libsycl-tsan", "internal"}};
+  const SYCLDeviceLibsList SYCLDeviceTsanLibs = {
+      {"libsycl-tsan", "internal"},
+      {"libsycl-tsan-cpu", "internal"},
+      // Currently, we only provide aot tsan libdevice for PVC and CPU.
+      // For DG2, we just use libsycl-tsan as placeholder.
+      // TODO: replace "libsycl-tsan" with "libsycl-tsan-dg2" when DG2
+      // AOT support is added.
+      {"libsycl-tsan", "internal"},
+      {"libsycl-tsan-pvc", "internal"}};
 #endif
 
   const SYCLDeviceLibsList SYCLNativeCpuDeviceLibs = {
@@ -759,7 +767,7 @@ SYCL::getDeviceLibraries(const Compilation &C, const llvm::Triple &TargetTriple,
   else if (SanitizeVal == "memory")
     addSingleLibrary(SYCLDeviceMsanLibs[sanitizer_lib_idx]);
   else if (SanitizeVal == "thread")
-    addLibraries(SYCLDeviceTsanLibs);
+    addSingleLibrary(SYCLDeviceTsanLibs[sanitizer_lib_idx]);
 #endif
 
   if (isSYCLNativeCPU(TargetTriple))
@@ -883,6 +891,8 @@ static llvm::SmallVector<StringRef, 16> SYCLDeviceLibList{
     "msan-pvc",
     "msan-cpu",
     "tsan",
+    "tsan-pvc",
+    "tsan-cpu",
 #endif
     "imf",
     "imf-fp64",

clang/lib/Frontend/InitPreprocessor.cpp

@@ -1535,18 +1535,25 @@ static void InitializePredefinedMacros(const TargetInfo &TI,
     if (LangOpts.GPURelocatableDeviceCode)
       Builder.defineMacro("SYCL_EXTERNAL", "__attribute__((sycl_device))");
 
-    const llvm::Triple &DeviceTriple = TI.getTriple();
-    const llvm::Triple::SubArchType DeviceSubArch = DeviceTriple.getSubArch();
-    if (DeviceTriple.isNVPTX() || DeviceTriple.isAMDGPU() ||
-        (DeviceTriple.isSPIR() &&
-         DeviceSubArch != llvm::Triple::SPIRSubArch_fpga) ||
+    // This gets called twice, once with TI set to the host TargetInfo, once
+    // with TI set to the device TargetInfo.
+    const llvm::Triple &Triple = TI.getTriple();
+    const llvm::Triple::SubArchType SubArch = Triple.getSubArch();
+    if (Triple.isNVPTX() || Triple.isAMDGPU() ||
+        (Triple.isSPIR() && SubArch != llvm::Triple::SPIRSubArch_fpga) ||
         LangOpts.SYCLIsNativeCPU)
       Builder.defineMacro("SYCL_USE_NATIVE_FP_ATOMICS");
     // Enable generation of USM address spaces for FPGA.
-    if (DeviceSubArch == llvm::Triple::SPIRSubArch_fpga) {
+    if (SubArch == llvm::Triple::SPIRSubArch_fpga) {
       Builder.defineMacro("__ENABLE_USM_ADDR_SPACE__");
       Builder.defineMacro("SYCL_DISABLE_FALLBACK_ASSERT");
     }
+
+    if (Triple.isWindowsMSVCEnvironment()) {
+      // MSVC inline definitions of stdio functions should not be used for SYCL
+      // device code.
+      Builder.defineMacro("_NO_CRT_STDIO_INLINE");
+    }
   } else if (LangOpts.SYCLIsHost && LangOpts.SYCLESIMDBuildHostCode) {
     Builder.defineMacro("__ESIMD_BUILD_HOST_CODE");
   }

clang/test/Driver/Inputs/SYCL/lib/libsycl-tsan-cpu.bc

@@ -415,3 +415,26 @@
 // SYCL_DEVICE_TSAN_MACRO-SAME: "USE_SYCL_DEVICE_TSAN"
 // SYCL_DEVICE_TSAN_MACRO: llvm-link{{.*}} "-only-needed"
 // SYCL_DEVICE_TSAN_MACRO-SAME: "{{.*}}libsycl-tsan.bc"
+
+/// ###########################################################################
+/// test behavior of linking libsycl-tsan-pvc for PVC target AOT compilation when tsan flag is applied.
+// RUN: %clangxx -fsycl -fsycl-targets=intel_gpu_pvc --no-offload-new-driver %s --sysroot=%S/Inputs/SYCL \
+// RUN: -Xarch_device -fsanitize=thread -### 2>&1 | FileCheck %s -check-prefix=SYCL_DEVICE_LIB_TSAN_PVC
+// RUN: %clangxx -fsycl -fsycl-targets=spir64_gen -Xsycl-target-backend "-device pvc" --no-offload-new-driver %s \
+// RUN: --sysroot=%S/Inputs/SYCL -Xarch_device -fsanitize=thread -### 2>&1 | FileCheck %s -check-prefix=SYCL_DEVICE_LIB_TSAN_PVC
+// RUN: %clangxx -fsycl -fsycl-targets=spir64_gen -Xsycl-target-backend=spir64_gen "-device pvc" --no-offload-new-driver %s \
+// RUN: --sysroot=%S/Inputs/SYCL -Xarch_device -fsanitize=thread -### 2>&1 | FileCheck %s -check-prefix=SYCL_DEVICE_LIB_TSAN_PVC
+// RUN: %clangxx -fsycl -fsycl-targets=spir64_gen -Xsycl-target-backend "-device 12.60.7" --no-offload-new-driver %s \
+// RUN: --sysroot=%S/Inputs/SYCL -Xarch_device -fsanitize=thread -### 2>&1 | FileCheck %s -check-prefix=SYCL_DEVICE_LIB_TSAN_PVC
+// RUN: %clangxx -fsycl -fsycl-targets=spir64_gen -Xs "-device 12.60.7" --no-offload-new-driver %s --sysroot=%S/Inputs/SYCL \
+// RUN: -Xarch_device -fsanitize=thread -### 2>&1 | FileCheck %s -check-prefix=SYCL_DEVICE_LIB_TSAN_PVC
+// SYCL_DEVICE_LIB_TSAN_PVC: llvm-link{{.*}} "-only-needed" "{{.*}}libsycl-crt.bc"
+// SYCL_DEVICE_LIB_TSAN_PVC-SAME: "{{.*}}libsycl-tsan-pvc.bc"
+
+
+/// ###########################################################################
+/// test behavior of linking libsycl-tsan-cpu for CPU target AOT compilation when tsan flag is applied.
+// RUN: %clangxx -fsycl -fsycl-targets=spir64_x86_64 --no-offload-new-driver %s --sysroot=%S/Inputs/SYCL \
+// RUN: -Xarch_device -fsanitize=thread -### 2>&1 | FileCheck %s -check-prefix=SYCL_DEVICE_LIB_TSAN_CPU
+// SYCL_DEVICE_LIB_TSAN_CPU: llvm-link{{.*}} "-only-needed" "{{.*}}libsycl-crt.bc"
+// SYCL_DEVICE_LIB_TSAN_CPU-SAME: "{{.*}}libsycl-tsan-cpu.bc"

clang/test/Driver/Inputs/SYCL/lib/libsycl-tsan-pvc.bc

@@ -401,3 +401,15 @@
 // SYCL_DEVICE_TSAN_MACRO: "-cc1"
 // SYCL_DEVICE_TSAN_MACRO-SAME: "USE_SYCL_DEVICE_TSAN"
 // SYCL_DEVICE_TSAN_MACRO: libsycl-tsan.new.o
+
+/// test behavior of tsan libdevice linking when -fsanitize=thread is available for AOT targets
+// RUN: %clangxx -fsycl -fsycl-targets=intel_gpu_pvc --offload-new-driver %s --sysroot=%S/Inputs/SYCL \
+// RUN: -fsanitize=thread -### 2>&1 | FileCheck %s -check-prefix=SYCL_DEVICE_LIB_TSAN_PVC
+// SYCL_DEVICE_LIB_TSAN_PVC: clang-linker-wrapper{{.*}} "-sycl-device-libraries
+// SYCL_DEVICE_LIB_TSAN_PVC-SAME: {{.*}}libsycl-tsan-pvc.new.o
+
+/// test behavior of tsan libdevice linking when -fsanitize=thread is available for AOT targets
+// RUN: %clangxx -fsycl -fsycl-targets=spir64_x86_64 --offload-new-driver %s --sysroot=%S/Inputs/SYCL \
+// RUN: -fsanitize=thread -### 2>&1 | FileCheck %s -check-prefix=SYCL_DEVICE_LIB_TSAN_CPU
+// SYCL_DEVICE_LIB_TSAN_CPU: clang-linker-wrapper{{.*}} "-sycl-device-libraries
+// SYCL_DEVICE_LIB_TSAN_CPU-SAME: {{.*}}libsycl-tsan-cpu.new.o