Merge remote-tracking branch 'origin/sycl' into HEAD

Author: aelovikov-intel

.github/workflows/sycl-check-ready-to-merge-prs.yml

@@ -0,0 +1,54 @@
+# This workflow checks for ready-to-merge PRs - if a PR is open, not a draft,
+# passed all checks, and has been approved, it will ping @intel/llvm-gatekeepers
+# if this group has not already been mentioned or if the last mention was more
+# than $days days ago.
+
+name: Check ready-to-merge PRs
+
+on:
+  schedule:
+    - cron: '0 * * * *' # every hour
+  workflow_dispatch:
+
+permissions: read-all
+
+jobs:
+  notify-ready-prs:
+    permissions:
+      pull-requests: write
+    runs-on: ubuntu-latest
+    steps:
+      - name: Check
+        env:
+          GH_TOKEN: ${{ secrets.GITHUB_TOKEN }}
+        run: |
+          # Number of days before repeating the gatekeepers ping
+          days=3
+          days_in_seconds=$((days*24*60*60))
+
+          # Function to ping gatekeepers and print debug info
+          ping_gatekeepers() {
+            pr_number=$1
+            gh pr comment "$pr_number" --repo intel/llvm --body "@intel/llvm-gatekeepers please consider merging"
+            echo "Pinged @intel/llvm-gatekeepers for https://github.com/intel/llvm/pull/$pr_number"
+          }
+
+          # Get the list of suitable PRs
+          prs=$(gh pr list --search "is:open review:approved draft:no status:success" --repo intel/llvm --json number --jq '.[].number')
+          now=$(date -u +%s)
+          for pr in $prs; do
+            # Get the timestamp of the latest comment mentioning @intel/llvm-gatekeepers
+            latest_ts=$(gh pr view $pr --repo intel/llvm --json comments \
+              --jq '[.comments[] | select(.body | test("@intel/llvm-gatekeepers")) | .createdAt] | last')
+            # If there is no previous mention, ping the gatekeepers
+            if [[ -z "$latest_ts" ]]; then
+              ping_gatekeepers "$pr"
+            # If the latest mention is older than $days, ping the gatekeepers again
+            else
+              comment_time=$(date -u -d "$latest_ts" +%s)
+              age=$((now - comment_time))
+              if (( age >= days_in_seconds )); then
+                ping_gatekeepers "$pr"
+              fi
+            fi
+          done

buildbot/configure.py

@@ -82,6 +82,11 @@ def do_configure(args, passthrough_args):
     if libclc_enabled:
         llvm_enable_projects += ";libclc"
 
+    # DeviceRTL uses -fuse-ld=lld, so enable lld.
+    if args.offload:
+        llvm_enable_projects += ";lld"
+        sycl_enabled_backends.append("offload")
+
     if args.cuda:
         llvm_targets_to_build += ";NVPTX"
         libclc_targets_to_build = libclc_nvidia_target_names
@@ -210,6 +215,12 @@ def do_configure(args, passthrough_args):
         "-DSYCL_ENABLE_MAJOR_RELEASE_PREVIEW_LIB={}".format(sycl_preview_lib),
         "-DBUG_REPORT_URL=https://github.com/intel/llvm/issues",
     ]
+    if args.offload:
+        cmake_cmd.extend(
+            [
+                "-DUR_BUILD_ADAPTER_OFFLOAD=ON",
+            ]
+        )
 
     if libclc_enabled:
         cmake_cmd.extend(
@@ -340,6 +351,11 @@ def main():
         default="AMD",
         help="choose hardware platform for HIP backend",
     )
+    parser.add_argument(
+        "--offload",
+        action="store_true",
+        help="Enable UR liboffload adapter (experimental)",
+    )
     parser.add_argument(
         "--level_zero_adapter_version",
         type=str,

clang/lib/Driver/ToolChains/SYCL.cpp

@@ -69,8 +69,7 @@ const char *SYCLInstallationDetector::findLibspirvPath(
     SmallString<128> LibraryPath(Path);
     llvm::sys::path::append(LibraryPath, a, b, c, Basename);
 
-    if (Args.hasArgNoClaim(options::OPT__HASH_HASH_HASH) ||
-        D.getVFS().exists(LibraryPath))
+    if (D.getVFS().exists(LibraryPath))
       return Args.MakeArgString(LibraryPath);
 
     return nullptr;

clang/test/Driver/sycl-libspirv-toolchain.cpp

@@ -34,17 +34,16 @@
 // RUN: | FileCheck %s -DINSTALL_DIR=%{install_dir} -DRESOURCE_DIR=%{resource_dir} --check-prefixes=CHECK-DIR
 // CHECK-DIR: "-cc1"{{.*}} "-fsycl-is-device"{{.*}} "-mlink-builtin-bitcode" "[[INSTALL_DIR]]{{.*[\\/]}}remangled-{{.*}}.libspirv-nvptx64-nvidia-cuda.bc"
 //
-// The `-###` option disables file existence checks
+// If libspirv path doesn't exist, error is reported.
 // DEFINE: %{nonexistent_dir} = %/S/Inputs/SYCL/does_not_exist/lib/clang/resource_dir
-// RUN: %clang -### -ccc-install-dir %{nonexistent_dir} -fsycl -fsycl-targets=nvptx64-nvidia-cuda -nocudalib %s 2>&1 \
+// RUN: not %clang -### -ccc-install-dir %{nonexistent_dir} -fsycl -fsycl-targets=nvptx64-nvidia-cuda -nocudalib %s 2>&1 \
 // RUN: | FileCheck %s         -DDIR=%{nonexistent_dir} --check-prefixes=CHECK-HHH-NONEXISTENT
-// CHECK-HHH-NONEXISTENT: "-cc1"{{.*}} "-fsycl-is-device"{{.*}} "-mlink-builtin-bitcode" "[[DIR]]{{.*[\\/]}}remangled-{{.*}}.libspirv-nvptx64-nvidia-cuda.bc"
+// CHECK-HHH-NONEXISTENT: error: cannot find 'remangled-{{.*}}.libspirv-nvptx64-nvidia-cuda.bc'; provide path to libspirv library via '-fsycl-libspirv-path', or pass '-fno-sycl-libspirv' to build without linking with libspirv
 //
-// RUN: %clang -### -ccc-install-dir %{nonexistent_dir} -fsycl -fsycl-targets=amdgcn-amd-amdhsa -Xsycl-target-backend --offload-arch=gfx908 -nogpulib %s 2>&1 \
+// RUN: not %clang -### -ccc-install-dir %{nonexistent_dir} -fsycl -fsycl-targets=amdgcn-amd-amdhsa -Xsycl-target-backend --offload-arch=gfx908 -nogpulib %s 2>&1 \
 // RUN: | FileCheck %s         -DDIR=%{nonexistent_dir} --check-prefixes=CHECK-AMDGCN-HHH-NONEXISTENT
-// CHECK-AMDGCN-HHH-NONEXISTENT: "-cc1"{{.*}} "-fsycl-is-device"{{.*}} "-mlink-builtin-bitcode" "[[DIR]]{{.*[\\/]}}remangled-{{.*}}.libspirv-amdgcn-amd-amdhsa.bc"
+// CHECK-AMDGCN-HHH-NONEXISTENT: clang: error: cannot find 'remangled-{{.*}}.libspirv-amdgcn-amd-amdhsa.bc'; provide path to libspirv library via '-fsycl-libspirv-path', or pass '-fno-sycl-libspirv' to build without linking with libspirv
 //
-// `-fdriver-only` has no such special handling, so it will not find the file
 // RUN: not %clang -fdriver-only -ccc-install-dir %{nonexistent_dir} -fsycl -fsycl-targets=nvptx64-nvidia-cuda -nocudalib %s 2>&1 \
 // RUN: | FileCheck %s                      -DDIR=%{nonexistent_dir} --check-prefixes=CHECK-DO-NONEXISTENT
 // CHECK-DO-NONEXISTENT: error: cannot find 'remangled-{{.*}}.libspirv-nvptx64-nvidia-cuda.bc'; provide path to libspirv library via '-fsycl-libspirv-path', or pass '-fno-sycl-libspirv' to build without linking with libspirv

clang/test/Driver/sycl-native-cpu.cpp

@@ -1,5 +1,5 @@
-// RUN: %clangxx -fsycl-device-only -fsycl-targets=native_cpu %s -### 2>&1 | FileCheck %s
-// RUN: %clangxx -fsycl-device-only -fsycl-targets=native_cpu --target=aarch64-unknown-linux-gnu %s -### 2>&1 | FileCheck %s --check-prefix=CHECK-AARCH64
+// RUN: %clangxx -fsycl-device-only -fsycl-targets=native_cpu -fno-sycl-libspirv %s -### 2>&1 | FileCheck %s
+// RUN: %clangxx -fsycl-device-only -fsycl-targets=native_cpu --target=aarch64-unknown-linux-gnu -fno-sycl-libspirv %s -### 2>&1 | FileCheck %s --check-prefix=CHECK-AARCH64
 
 
 // checks that the host triple is native_cpu, the device triple is set, and that the sycl-native-cpu LLVM option is set
@@ -8,29 +8,29 @@
 // checks that the target triples are set correctly when the target is set explicitly
 // CHECK-AARCH64: clang{{.*}}"-triple" "native_cpu"{{.*}}"-aux-triple" "aarch64-unknown-linux-gnu" {{.*}}"-D" "__SYCL_NATIVE_CPU__"
 
-// RUN:   %clang -### --target=x86_64-unknown-linux-gnu -fsycl -fsycl-targets=native_cpu -g %s 2>&1 | FileCheck -check-prefix=CHECK-LINUX %s
+// RUN:   %clang -### --target=x86_64-unknown-linux-gnu -fsycl -fsycl-targets=native_cpu -g -fno-sycl-libspirv %s 2>&1 | FileCheck -check-prefix=CHECK-LINUX %s
 // CHECK-LINUX: {{.*}}"-fsycl-is-device"{{.*}}"-dwarf-version=[[DVERSION:.*]]" "-debugger-tuning=gdb"
 // CHECK-LINUX-DAG: {{.*}}"-fsycl-is-host"{{.*}}"-dwarf-version=[[DVERSION]]" "-debugger-tuning=gdb"
 // CHECK-LINUX-NOT: codeview
 
-// RUN:   %clang -### --target=x86_64-windows-msvc -fsycl -fsycl-targets=native_cpu -g %s 2>&1 | FileCheck -check-prefix=CHECK-WIN %s
+// RUN:   %clang -### --target=x86_64-windows-msvc -fsycl -fsycl-targets=native_cpu -g -fno-sycl-libspirv %s 2>&1 | FileCheck -check-prefix=CHECK-WIN %s
 // CHECK-WIN: {{.*}}"-fsycl-is-device"{{.*}}"-gcodeview"
 // CHECK-WIN-DAG: {{.*}}"-fsycl-is-host"{{.*}}"-gcodeview"
 // CHECK-WIN-NOT: dwarf
 
 // checks that -sycl-opt is not enabled by default on NativeCPU so that the full llvm optimization is enabled
 // Also check that we pass the expected backend options.
-// RUN:   %clang -fsycl -fsycl-targets=native_cpu --target=aarch64-unknown-linux-gnu -march=armv9.4-a -### %s 2>&1 | FileCheck -check-prefix=CHECK-OPTS %s
+// RUN:   %clang -fsycl -fsycl-targets=native_cpu --target=aarch64-unknown-linux-gnu -march=armv9.4-a -fno-sycl-libspirv -### %s 2>&1 | FileCheck -check-prefix=CHECK-OPTS %s
 // CHECK-OPTS: clang{{.*}}"-triple" "native_cpu"{{.*}}"-aux-triple" "[[TRIPLE:[^"]*]]"
 // CHECK-OPTS: clang{{.*}}"-triple" "[[TRIPLE]]"{{.*}}"-fsycl-is-device"
 // CHECK-OPTS-NOT: -sycl-opt
 // CHECK-OPTS-SAME: "-Wno-override-module" "-mllvm" "-sycl-native-cpu-backend"
 // CHECK-OPTS-SAME: "-aux-target-feature" "+v9.4a"
 
-// RUN: %clangxx -fsycl -fsycl-targets=spir64 %s -### 2>&1 | FileCheck -check-prefix=CHECK-NONATIVECPU %s
+// RUN: %clangxx -fsycl -fsycl-targets=spir64 -fno-sycl-libspirv %s -### 2>&1 | FileCheck -check-prefix=CHECK-NONATIVECPU %s
 // CHECK-NONATIVECPU-NOT: "-D" "__SYCL_NATIVE_CPU__"
 
 // Checking that coverage testing options are accepted by native_cpu, and that device and host compilation invocations receive the same options
-// RUN: %clangxx -fsycl -fsycl-targets=native_cpu -Werror -fno-profile-instr-generate -fprofile-instr-generate -fno-coverage-mapping -fcoverage-mapping -### %s 2>&1 | FileCheck %s --check-prefix=CHECK_COV_INVO
+// RUN: %clangxx -fsycl -fsycl-targets=native_cpu -Werror -fno-profile-instr-generate -fprofile-instr-generate -fno-coverage-mapping -fcoverage-mapping -fno-sycl-libspirv -Wno-unsafe-libspirv-not-linked -### %s 2>&1 | FileCheck %s --check-prefix=CHECK_COV_INVO
 // CHECK_COV_INVO:{{.*}}clang{{.*}}"-fsycl-is-device"{{.*}} "-D" "__SYCL_NATIVE_CPU__"{{.*}}"-fprofile-instrument=clang"{{.*}}"-fcoverage-mapping" "-fcoverage-compilation-dir={{.*}}"
 // CHECK_COV_INVO:{{.*}}clang{{.*}}"-fsycl-is-host"{{.*}}"-fprofile-instrument=clang"{{.*}}"-fcoverage-mapping" "-fcoverage-compilation-dir={{.*}}"

devops/dependencies-igc-dev.json

@@ -1,10 +1,10 @@
 {
   "linux": {
     "igc_dev": {
-      "github_tag": "igc-dev-47c1f6e",
-      "version": "47c1f6e",
-      "updated_at": "2025-08-20T15:46:36Z",
-      "url": "https://api.github.com/repos/intel/intel-graphics-compiler/actions/artifacts/3810490997/zip",
+      "github_tag": "igc-dev-46629d9",
+      "version": "46629d9",
+      "updated_at": "2025-08-30T10:44:04Z",
+      "url": "https://api.github.com/repos/intel/intel-graphics-compiler/actions/artifacts/3889106305/zip",
       "root": "{DEPS_ROOT}/opencl/runtime/linux/oclgpu"
     }
   }

devops/dependencies.json

@@ -19,9 +19,9 @@
       "root": "{DEPS_ROOT}/opencl/runtime/linux/oclgpu"
     },
     "level_zero": {
-      "github_tag": "v1.22.4",
-      "version": "v1.22.4",
-      "url": "https://github.com/oneapi-src/level-zero/releases/tag/v1.22.4",
+      "github_tag": "v1.24.2",
+      "version": "v1.24.2",
+      "url": "https://github.com/oneapi-src/level-zero/releases/tag/v1.24.2",
       "root": "{DEPS_ROOT}/opencl/runtime/linux/oclgpu"
     },
     "tbb": {

devops/scripts/benchmarks/PERFORMANCE_TUNING.md

@@ -6,38 +6,16 @@ For framework-specific information, see [README.md](README.md) and [CONTRIB.md](
 ## Table of Contents
 
 - [Overview](#overview)
-- [System Configuration](#system-configuration)
 - [CPU Tuning](#cpu-tuning)
 - [GPU Configuration](#gpu-configuration)
+- [Driver Version](#driver-version)
 - [Perf Configuration](#perf-configuration)
 - [Environment Variables](#environment-variables)
 
 ## Overview
 
 Performance benchmarking requires a stable and optimized system environment to produce reliable and reproducible results. This guide covers essential system tuning steps for reducing run-to-run variance in benchmark results.
 
-## System Configuration
-
-### Kernel Parameters
-
-Add the following to `/etc/default/grub` in `GRUB_CMDLINE_LINUX`:
-```
-# Disable CPU frequency scaling
-# intel_pstate=disable
-
-# Isolate CPUs for benchmark workloads (example: reserve cores 2-7), preventing other processes
-# from using them.
-# isolcpus=2-7
-
-GRUB_CMDLINE_LINUX="intel_pstate=disable isolcpus=2-7 <other_options>"
-```
-
-Update GRUB and reboot:
-```bash
-sudo update-grub
-sudo reboot
-```
-
 ## CPU Tuning
 
 ### CPU Frequency Scaling
@@ -66,7 +44,7 @@ After=multi-user.target
 
 [Service]
 Type=oneshot
-ExecStart=/usr/bin/cpupower frequency-set --governor performance && sysctl --system
+ExecStart=/usr/bin/cpupower frequency-set --governor performance
 
 [Install]
 WantedBy=multi-user.target
@@ -99,7 +77,12 @@ cat /sys/class/drm/card1/device/vendor  # Should be 0x8086 for Intel
 cat /sys/class/drm/card1/device/device  # Device ID
 ```
 
-Verify the max frequency is set to the true max. For Arc B580, the maximum frequency is 2850 MHz. To see this value, run “cat /sys/class/drm/card1/device/tile0/gt0/freq0/max_freq”. If the above value is not equal to the max frequency, set it as such:
+Verify the max frequency is set to the true max. For Arc B580, the maximum frequency is 2850 MHz. To see this value, run:
+```bash
+cat /sys/class/drm/card1/device/tile0/gt0/freq0/max_freq
+```
+
+If the above value is not equal to the max frequency, set it as such:
 ```bash
 # Arc B580 (Battlemage)
 echo 2850 > /sys/class/drm/card1/device/tile0/gt0/freq0/max_freq
@@ -118,9 +101,9 @@ max_freq=$(cat /sys/class/drm/card1/gt_max_freq_mhz)
 echo $max_freq | sudo tee /sys/class/drm/card1/gt_min_freq_mhz
 ```
 
-The result can be verified using tools such as oneprof or unitrace to track frequency over time for some arbitrary benchmark (many iterations of a small problem size is recommended). The frequency should remain fixed assuming thermal throttling does not occur.
+The result can be verified using tools such as `oneprof` or `unitrace` to track frequency over time for some arbitrary benchmark (many iterations of a small problem size is recommended). The frequency should remain fixed assuming thermal throttling does not occur.
 
-## Driver version
+## Driver Version
 Make sure you are using the latest driver (Ubuntu)
 ```bash
 sudo apt update && sudo apt upgrade
@@ -145,6 +128,7 @@ Make the setting persistent across reboots by adding it to sysctl configuration:
 echo 'kernel.perf_event_paranoid = -1' | sudo tee -a /etc/sysctl.d/99-perf.conf
 
 # Apply immediately
+sudo sysctl kernel.perf_event_paranoid=-1
 sudo sysctl -p
 ```