Merge branch 'sycl' into work_group_static

Author: Victor Lomuller

.github/workflows/release-binaries.yml

@@ -436,7 +436,7 @@ jobs:
 
     - name: Attest Build Provenance
       id: provenance
-      uses: actions/attest-build-provenance@1c608d11d69870c2092266b3f9a6f3abbf17002c # v1.4.3
+      uses: actions/attest-build-provenance@ef244123eb79f2f7a7e75d99086184180e6d0018 # v1.4.4
       with:
         subject-path: ${{ needs.prepare.outputs.release-binary-filename }}
 

.github/workflows/release-sources.yml

@@ -92,7 +92,7 @@ jobs:
       - name: Attest Build Provenance
         if: github.event_name != 'pull_request'
         id: provenance
-        uses: actions/attest-build-provenance@1c608d11d69870c2092266b3f9a6f3abbf17002c # v1.4.3
+        uses: actions/attest-build-provenance@ef244123eb79f2f7a7e75d99086184180e6d0018 # v1.4.4
         with:
           subject-path: "*.xz"
       - if: github.event_name != 'pull_request'

.github/workflows/scorecard.yml

@@ -57,6 +57,6 @@ jobs:
 
       # Upload the results to GitHub's code scanning dashboard.
       - name: "Upload to code-scanning"
-        uses: github/codeql-action/upload-sarif@662472033e021d55d94146f66f6058822b0b39fd # v3.27.0
+        uses: github/codeql-action/upload-sarif@f09c1c0a94de965c15400f5634aa42fac8fb8f88 # v3.27.5
         with:
           sarif_file: results.sarif

.github/workflows/sycl-nightly.yml

@@ -198,7 +198,7 @@ jobs:
           echo "TAG=$(date +'%Y-%m-%d')-${GITHUB_SHA::7}" >> "$GITHUB_OUTPUT"
         fi
     - name: Upload binaries
-      uses: softprops/action-gh-release@e7a8f85e1c67a31e6ed99a94b41bd0b71bbee6b8
+      uses: softprops/action-gh-release@01570a1f39cb168c169c802c3bceb9e93fb10974
       with:
         files: |
           sycl_linux.tar.gz

clang/lib/Driver/ToolChains/SYCL.cpp

@@ -576,6 +576,29 @@ SYCL::getDeviceLibraries(const Compilation &C, const llvm::Triple &TargetTriple,
     }
   };
 
+  addLibraries(SYCLDeviceWrapperLibs);
+  if (IsSpirvAOT)
+    addLibraries(SYCLDeviceFallbackLibs);
+
+  bool NativeBfloatLibs;
+  bool NeedBfloatLibs = selectBfloatLibs(TargetTriple, C, NativeBfloatLibs);
+  if (NeedBfloatLibs) {
+    // Add native or fallback bfloat16 library.
+    if (NativeBfloatLibs)
+      addLibraries(SYCLDeviceBfloat16NativeLib);
+    else
+      addLibraries(SYCLDeviceBfloat16FallbackLib);
+  }
+
+  // Link in ITT annotations library unless fsycl-no-instrument-device-code
+  // is specified. This ensures that we are ABI-compatible with the
+  // instrumented device code, which was the default not so long ago.
+  if (Args.hasFlag(options::OPT_fsycl_instrument_device_code,
+                   options::OPT_fno_sycl_instrument_device_code, true))
+    addLibraries(SYCLDeviceAnnotationLibs);
+
+#if !defined(_WIN32)
+
   auto addSingleLibrary = [&](const DeviceLibOptInfo &Lib) {
     if (!DeviceLibLinkInfo[Lib.DeviceLibOption])
       return;
@@ -636,28 +659,6 @@ SYCL::getDeviceLibraries(const Compilation &C, const llvm::Triple &TargetTriple,
     return JIT;
   };
 
-  addLibraries(SYCLDeviceWrapperLibs);
-  if (IsSpirvAOT)
-    addLibraries(SYCLDeviceFallbackLibs);
-
-  bool NativeBfloatLibs;
-  bool NeedBfloatLibs = selectBfloatLibs(TargetTriple, C, NativeBfloatLibs);
-  if (NeedBfloatLibs) {
-    // Add native or fallback bfloat16 library.
-    if (NativeBfloatLibs)
-      addLibraries(SYCLDeviceBfloat16NativeLib);
-    else
-      addLibraries(SYCLDeviceBfloat16FallbackLib);
-  }
-
-  // Link in ITT annotations library unless fsycl-no-instrument-device-code
-  // is specified. This ensures that we are ABI-compatible with the
-  // instrumented device code, which was the default not so long ago.
-  if (Args.hasFlag(options::OPT_fsycl_instrument_device_code,
-                   options::OPT_fno_sycl_instrument_device_code, true))
-    addLibraries(SYCLDeviceAnnotationLibs);
-
-#if !defined(_WIN32)
   std::string SanitizeVal;
   size_t sanitizer_lib_idx = getSingleBuildTarget();
   if (Arg *A = Args.getLastArg(options::OPT_fsanitize_EQ,

devops/cts_exclude_filter_L0_GPU

@@ -1,2 +1,2 @@
-# CMPLRLLVM-61839
-multi_ptr
+# CMPLRTST-26179
+device

devops/cts_exclude_filter_OCL_CPU

@@ -2,5 +2,3 @@
 math_builtin_api
 # https://github.com/intel/llvm/issues/13574
 hierarchical
-# CMPLRLLVM-61839
-multi_ptr

libclc/amdgcn-amdhsa/libspirv/atomic/atomic_helpers.h

@@ -19,6 +19,7 @@ extern int __oclc_amdgpu_reflect(__constant char *);
                                    OUT_ORDER)                                  \
   {                                                                            \
     switch (IN_SCOPE) {                                                        \
+    case Invocation:                                                           \
     case Subgroup:                                                             \
       OUT_SCOPE = __HIP_MEMORY_SCOPE_WAVEFRONT;                                \
       break;                                                                   \

libclc/ptx-nvidiacl/libspirv/atomic/atomic_cmpxchg.cl

@@ -16,6 +16,7 @@ _CLC_OVERLOAD _CLC_DECL void __spirv_MemoryBarrier(unsigned int, unsigned int);
 #define __CLC_NVVM_ATOMIC_CAS_IMPL_ORDER(TYPE, TYPE_NV, TYPE_MANGLED_NV, OP,   \
                                          ADDR_SPACE, ADDR_SPACE_NV, ORDER)     \
   switch (scope) {                                                             \
+  case Invocation:                                                             \
   case Subgroup:                                                               \
   case Workgroup: {                                                            \
     if (__clc_nvvm_reflect_arch() >= 600) {                                    \
@@ -44,6 +45,7 @@ _CLC_OVERLOAD _CLC_DECL void __spirv_MemoryBarrier(unsigned int, unsigned int);
 #define __CLC_NVVM_ATOMIC_CAS_IMPL_ACQUIRE_FENCE(                              \
     TYPE, TYPE_NV, TYPE_MANGLED_NV, OP, ADDR_SPACE, ADDR_SPACE_NV)             \
   switch (scope) {                                                             \
+  case Invocation:                                                             \
   case Subgroup:                                                               \
   case Workgroup: {                                                            \
     if (__clc_nvvm_reflect_arch() >= 600) {                                    \

libclc/ptx-nvidiacl/libspirv/atomic/atomic_helpers.h

@@ -18,6 +18,7 @@ _CLC_OVERLOAD _CLC_DECL void __spirv_MemoryBarrier(unsigned int, unsigned int);
 #define __CLC_NVVM_ATOMIC_IMPL_ORDER(TYPE, TYPE_NV, TYPE_MANGLED_NV, OP,       \
                                      ADDR_SPACE, ADDR_SPACE_NV, ORDER)         \
   switch (scope) {                                                             \
+  case Invocation:                                                             \
   case Subgroup:                                                               \
   case Workgroup: {                                                            \
     if (__clc_nvvm_reflect_arch() >= 600) {                                    \
@@ -46,6 +47,7 @@ _CLC_OVERLOAD _CLC_DECL void __spirv_MemoryBarrier(unsigned int, unsigned int);
 #define __CLC_NVVM_ATOMIC_IMPL_ACQUIRE_FENCE(TYPE, TYPE_NV, TYPE_MANGLED_NV,   \
                                              OP, ADDR_SPACE, ADDR_SPACE_NV)    \
   switch (scope) {                                                             \
+  case Invocation:                                                             \
   case Subgroup:                                                               \
   case Workgroup: {                                                            \
     if (__clc_nvvm_reflect_arch() >= 600) {                                    \