Merge branch 'sycl' into sean/usm-normalized-fix

Author: Seanst98

.github/CODEOWNERS

@@ -139,9 +139,10 @@ sycl/include/sycl/ext/oneapi/experimental/graph.hpp @intel/sycl-graphs-reviewers
 sycl/source/detail/graph_impl.cpp @intel/sycl-graphs-reviewers
 sycl/source/detail/graph_impl.hpp @intel/sycl-graphs-reviewers
 sycl/unittests/Extensions/CommandGraph/ @intel/sycl-graphs-reviewers
-sycl/doc/design/CommandGraph.md @intel/sycl-graphs-reviewers
 sycl/test-e2e/Graph @intel/sycl-graphs-reviewers
+sycl/doc/design/CommandGraph.md @intel/sycl-graphs-reviewers
 sycl/doc/extensions/**/sycl_ext_oneapi_graph.asciidoc @intel/sycl-graphs-reviewers
+sycl/doc/syclgraph/ @intel/sycl-graphs-reviewers
 
 # syclcompat library
 sycl/**/syclcompat/ @intel/syclcompat-lib-reviewers

.github/workflows/sycl-docs.yml

@@ -1,4 +1,4 @@
-name: Generate Doxygen documentation
+name: Generate documentation
 
 on:
   schedule:
@@ -10,13 +10,21 @@ on:
     - '.github/workflows/sycl-docs.yml'
     - 'clang/docs/**'
     - 'sycl/doc/**'
+  push:
+    branches:
+    - sycl
+    paths:
+    - '.github/workflows/sycl-docs.yml'
+    - 'clang/docs/**'
+    - 'sycl/doc/**'
 
-permissions: read-all
+permissions:
+  contents: read
+  pages: write
+  id-token: write
 
 jobs:
   build:
-    permissions:
-      contents: write  # for Git to git push
     runs-on: ubuntu-latest
     if: github.repository == 'intel/llvm'
     steps:
@@ -25,38 +33,28 @@ jobs:
         path: repo
     - name: Install deps
       run: |
-        sudo apt-get install -y doxygen graphviz ssh ninja-build libhwloc-dev
+        sudo apt-get install -y graphviz ssh ninja-build libhwloc-dev
         sudo pip3 install -r repo/llvm/docs/requirements.txt
     - name: Build Docs
       run: |
         mkdir -p $GITHUB_WORKSPACE/build
         cd $GITHUB_WORKSPACE/build
         python $GITHUB_WORKSPACE/repo/buildbot/configure.py -w $GITHUB_WORKSPACE \
         -s $GITHUB_WORKSPACE/repo -o $GITHUB_WORKSPACE/build -t Release --docs
-        cmake --build . --target doxygen-sycl
-        cmake --build . --target doxygen-clang
         cmake --build . --target docs-sycl-html
         cmake --build . --target docs-clang-html
-    - name: Deploy
-      if: ${{ github.event_name == 'schedule' }}
-      env:
-        SSH_KEY: ${{secrets.ACTIONS_DEPLOY_KEY}}
-      run: |
-        mkdir -p ~/.ssh
-        echo "$SSH_KEY" > ~/.ssh/id_rsa
-        chmod 600 ~/.ssh/id_rsa
-        eval "$(ssh-agent -s)"
-        ssh-add -k ~/.ssh/id_rsa
-        git clone [email protected]:intel/llvm-docs.git docs
-        cd $GITHUB_WORKSPACE/docs
-        git rm -rf .
+        # Copy the generated docs to a separate directory for uploading.
+        mkdir $GITHUB_WORKSPACE/install_docs
+        cd $GITHUB_WORKSPACE/install_docs
+        mkdir clang
+        mv $GITHUB_WORKSPACE/build/tools/sycl/doc/html/* .
+        mv $GITHUB_WORKSPACE/build/tools/clang/docs/html/* clang/
         touch .nojekyll
-        yes | \cp -rf $GITHUB_WORKSPACE/build/tools/sycl/doc/html/* .
-        mv $GITHUB_WORKSPACE/build/tools/sycl/doc/doxygen/html doxygen/
-        mv $GITHUB_WORKSPACE/build/tools/clang/docs/html clang/
-        mv $GITHUB_WORKSPACE/build/tools/clang/docs/doxygen/html clang_doxygen/
-        git config --global user.name "iclsrc"
-        git config --global user.email "[email protected]"
-        git add .
-        git diff-index --quiet HEAD || git commit --amend -m "Update docs" -s
-        git push -f
+    # Upload the generated docs as an artifact and deploy to GitHub Pages.
+    - name: Upload artifact
+      uses: actions/upload-pages-artifact@v3
+      with:
+        path: ./install_docs
+    - name: Deploy to GitHub Pages
+      if: ${{ github.event_name == 'push' }}
+      uses: actions/deploy-pages@v4

README.md

@@ -42,7 +42,7 @@ source project. One of the examples is
 Features parity between this project and downstream projects is not guaranteed.
 
 Project documentation is available at:
-[DPC++ Documentation](https://intel.github.io/llvm-docs/).
+[DPC++ Documentation](https://intel.github.io/llvm/).
 
 ### How to use DPC++
 

buildbot/configure.py

@@ -116,7 +116,6 @@ def do_configure(args):
         llvm_enable_assertions = "OFF"
 
     if args.docs:
-        llvm_enable_doxygen = "ON"
         llvm_enable_sphinx = "ON"
 
     if args.shared_libs:

clang/include/clang/Basic/DiagnosticSemaKinds.td

@@ -12433,6 +12433,12 @@ def err_sycl_kernel_incorrectly_named : Error<
   "'-fsycl-unnamed-lambda' to enable unnamed kernel lambdas"
   "}0">;
 
+// SYCL free function kernels extension.
+def note_free_function_kernel_param_type_not_fwd_declarable : Note<
+  "%0 is not forward declarable">;
+def note_free_function_kernel_param_type_not_supported : Note<
+  "%0 is not yet supported as a free function kernel parameter">;
+
 def err_sycl_kernel_not_function_object
     : Error<"kernel parameter must be a lambda or function object">;
 def err_sycl_restrict : Error<

clang/include/clang/Driver/Options.td

@@ -7049,9 +7049,16 @@ def fsycl_remove_unused_external_funcs : Flag<["-"], "fsycl-remove-unused-extern
 def fno_sycl_remove_unused_external_funcs : Flag<["-"], "fno-sycl-remove-unused-external-funcs">,
   HelpText<"Prevent removal of unused `SYCL_EXTERNAL` functions">;
 def fsycl_allow_device_dependencies : Flag<["-"], "fsycl-allow-device-dependencies">,
-  HelpText<"Allow dependencies between device code images">;
+  HelpText<"Deprecated: please use -fsycl-allow-device-image-dependencies instead.">,
+  Flags<[Deprecated]>;
 def fno_sycl_allow_device_dependencies : Flag<["-"], "fno-sycl-allow-device-dependencies">,
-  HelpText<"Do not allow dependencies between device code images (default)">;
+  HelpText<"Deprecated: please use -fno-sycl-allow-device-image-dependencies instead.">,
+  Flags<[Deprecated]>;
+
+defm sycl_allow_device_image_dependencies: BoolOptionWithoutMarshalling<"f", "sycl-allow-device-image-dependencies",
+  PosFlag<SetTrue, [], [ClangOption], "Allow dependencies between device code images">,
+  NegFlag<SetFalse, [], [ClangOption], "Do not allow dependencies between device code images (default)">>;
+
 def fsycl_dump_device_code_EQ : Joined<["-"], "fsycl-dump-device-code=">,
   Flags<[NoXarchOption]>,
   HelpText<"Dump device code into the user provided directory.">;

clang/lib/CodeGen/CGBuiltin.cpp

@@ -2738,7 +2738,8 @@ RValue CodeGenFunction::EmitBuiltinExpr(const GlobalDecl GD, unsigned BuiltinID,
           ConstWithoutErrnoOrExceptions && ErrnoOverridenToFalseWithOpt;
   }
   if (GenerateIntrinsics &&
-      !(getLangOpts().SYCLIsDevice && getTarget().getTriple().isNVPTX())) {
+      !(getLangOpts().SYCLIsDevice && (getTarget().getTriple().isNVPTX() ||
+                                       getTarget().getTriple().isAMDGCN()))) {
     switch (BuiltinIDIfNoAsmLabel) {
     case Builtin::BIacos:
     case Builtin::BIacosf:

clang/lib/Driver/Driver.cpp

@@ -5555,7 +5555,7 @@ class OffloadingActionBuilder final {
         // AOT compilation.
         bool SYCLDeviceLibLinked = false;
         Action *NativeCPULib = nullptr;
-        if (IsSPIR || IsNVPTX || IsSYCLNativeCPU) {
+        if (IsSPIR || IsNVPTX || IsAMDGCN || IsSYCLNativeCPU) {
           bool UseJitLink =
               IsSPIR &&
               Args.hasFlag(options::OPT_fsycl_device_lib_jit_link,

clang/lib/Driver/ToolChains/Clang.cpp

@@ -10801,10 +10801,17 @@ static void addArgs(ArgStringList &DstArgs, const llvm::opt::ArgList &Alloc,
   }
 }
 
-static bool allowDeviceDependencies(const llvm::opt::ArgList &TCArgs) {
+static bool allowDeviceImageDependencies(const llvm::opt::ArgList &TCArgs) {
+  // deprecated
   if (TCArgs.hasFlag(options::OPT_fsycl_allow_device_dependencies,
                      options::OPT_fno_sycl_allow_device_dependencies, false))
     return true;
+
+  // preferred
+  if (TCArgs.hasFlag(options::OPT_fsycl_allow_device_image_dependencies,
+                     options::OPT_fno_sycl_allow_device_image_dependencies, false))
+    return true;
+
   return false;
 }
 
@@ -10835,7 +10842,7 @@ static void getNonTripleBasedSYCLPostLinkOpts(const ToolChain &TC,
                      options::OPT_fsycl_esimd_force_stateless_mem, false))
     addArgs(PostLinkArgs, TCArgs, {"-lower-esimd-force-stateless-mem=false"});
 
-  if (allowDeviceDependencies(TCArgs))
+  if (allowDeviceImageDependencies(TCArgs))
     addArgs(PostLinkArgs, TCArgs, {"-allow-device-image-dependencies"});
 }
 
@@ -10853,7 +10860,7 @@ static bool shouldEmitOnlyKernelsAsEntryPoints(const ToolChain &TC,
     return true;
   // When supporting dynamic linking, non-kernels in a device image can be
   // called.
-  if (allowDeviceDependencies(TCArgs))
+  if (allowDeviceImageDependencies(TCArgs))
     return false;
   if (Triple.isNVPTX() || Triple.isAMDGPU())
     return false;

clang/lib/Driver/ToolChains/SYCL.cpp

@@ -165,9 +165,9 @@ static bool selectBfloatLibs(const llvm::Triple &Triple, const Compilation &C,
 
   // spir64 target is actually JIT compilation, so we defer selection of
   // bfloat16 libraries to runtime. For AOT we need libraries, but skip
-  // for Nvidia.
-  NeedLibs =
-      Triple.getSubArch() != llvm::Triple::NoSubArch && !Triple.isNVPTX();
+  // for Nvidia and AMD.
+  NeedLibs = Triple.getSubArch() != llvm::Triple::NoSubArch &&
+             !Triple.isNVPTX() && !Triple.isAMDGCN();
   UseNative = false;
   if (NeedLibs && Triple.getSubArch() == llvm::Triple::SPIRSubArch_gen &&
       C.hasOffloadToolChain<Action::OFK_SYCL>()) {
@@ -212,9 +212,9 @@ SYCL::getDeviceLibraries(const Compilation &C, const llvm::Triple &TargetTriple,
   SmallVector<std::string, 8> LibraryList;
   const llvm::opt::ArgList &Args = C.getArgs();
 
-  // For NVPTX we only use one single bitcode library and ignore
+  // For NVPTX and AMDGCN we only use one single bitcode library and ignore
   // manually specified SYCL device libraries.
-  bool IgnoreSingleLibs = TargetTriple.isNVPTX();
+  bool IgnoreSingleLibs = TargetTriple.isNVPTX() || TargetTriple.isAMDGCN();
 
   struct DeviceLibOptInfo {
     StringRef DeviceLibName;
@@ -278,6 +278,9 @@ SYCL::getDeviceLibraries(const Compilation &C, const llvm::Triple &TargetTriple,
   if (TargetTriple.isNVPTX() && IgnoreSingleLibs)
     LibraryList.push_back(Args.MakeArgString("devicelib--cuda.bc"));
 
+  if (TargetTriple.isAMDGCN() && IgnoreSingleLibs)
+    LibraryList.push_back(Args.MakeArgString("devicelib--amd.bc"));
+
   if (IgnoreSingleLibs)
     return LibraryList;
 
@@ -1812,22 +1815,18 @@ void SYCLToolChain::AddSYCLIncludeArgs(const clang::driver::Driver &Driver,
                                        const ArgList &DriverArgs,
                                        ArgStringList &CC1Args) {
   // Add the SYCL header search locations in the specified order.
-  //   ../include/sycl
   //   ../include/sycl/stl_wrappers
   //   ../include
   SmallString<128> IncludePath(Driver.Dir);
   llvm::sys::path::append(IncludePath, "..");
   llvm::sys::path::append(IncludePath, "include");
-  SmallString<128> SYCLPath(IncludePath);
-  llvm::sys::path::append(SYCLPath, "sycl");
   // This is used to provide our wrappers around STL headers that provide
   // additional functions/template specializations when the user includes those
   // STL headers in their programs (e.g., <complex>).
-  SmallString<128> STLWrappersPath(SYCLPath);
+  SmallString<128> STLWrappersPath(IncludePath);
+  llvm::sys::path::append(STLWrappersPath, "sycl");
   llvm::sys::path::append(STLWrappersPath, "stl_wrappers");
   CC1Args.push_back("-internal-isystem");
-  CC1Args.push_back(DriverArgs.MakeArgString(SYCLPath));
-  CC1Args.push_back("-internal-isystem");
   CC1Args.push_back(DriverArgs.MakeArgString(STLWrappersPath));
   CC1Args.push_back("-internal-isystem");
   CC1Args.push_back(DriverArgs.MakeArgString(IncludePath));