Merge remote-tracking branch 'origin/sycl' into sycl-max-wg-size-kernel-props

Author: frasercrmck

.github/workflows/sycl-docs.yml

@@ -1,4 +1,4 @@
-name: Generate Doxygen documentation
+name: Generate documentation
 
 on:
   schedule:
@@ -25,16 +25,14 @@ 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
@@ -52,9 +50,7 @@ jobs:
         git rm -rf .
         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 .

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/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/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/test/Driver/sycl-deprecated.cpp

@@ -3,4 +3,6 @@
 // RUN: %clangxx -fno-sycl-explicit-simd %s -### 2>&1 | FileCheck %s -DOPTION=-fno-sycl-explicit-simd
 // RUN: %clangxx -fsycl -fsycl-use-bitcode %s -### 2>&1 | FileCheck %s -DOPTION=-fsycl-use-bitcode
 // RUN: %clangxx -fsycl -fno-sycl-use-bitcode %s -### 2>&1 | FileCheck %s -DOPTION=-fno-sycl-use-bitcode
+// RUN: %clangxx -fsycl -fsycl-allow-device-dependencies %s -### 2>&1 | FileCheck %s -DOPTION=-fsycl-allow-device-dependencies
+// RUN: %clangxx -fsycl -fno-sycl-allow-device-dependencies %s -### 2>&1 | FileCheck %s -DOPTION=-fno-sycl-allow-device-dependencies
 // CHECK: option '[[OPTION]]' is deprecated and will be removed in a future release

clang/test/Driver/sycl-offload-old-model.cpp

@@ -174,16 +174,16 @@
 // RUN: %clang -### -target x86_64-unknown-linux-gnu -fsycl --no-offload-new-driver -fsycl-targets=spir64_gen %s 2>&1 | FileCheck -check-prefix=CHECK_SYCL_POST_LINK_OPT_PASS %s
 // CHECK_SYCL_POST_LINK_OPT_PASS: sycl-post-link{{.*}}emit-only-kernels-as-entry-points
 // RUN: %clang -### -target x86_64-unknown-linux-gnu -fsycl --no-offload-new-driver -fsycl-targets=spir64_gen -fno-sycl-remove-unused-external-funcs %s 2>&1 | FileCheck -check-prefix=CHECK_SYCL_POST_LINK_OPT_NO_PASS %s
-// RUN: %clang -### -target x86_64-unknown-linux-gnu -fsycl --no-offload-new-driver -fsycl-targets=spir64_gen -fsycl-allow-device-dependencies %s 2>&1 | FileCheck -check-prefix=CHECK_SYCL_POST_LINK_OPT_NO_PASS %s
+// RUN: %clang -### -target x86_64-unknown-linux-gnu -fsycl --no-offload-new-driver -fsycl-targets=spir64_gen -fsycl-allow-device-image-dependencies %s 2>&1 | FileCheck -check-prefix=CHECK_SYCL_POST_LINK_OPT_NO_PASS %s
 // CHECK_SYCL_POST_LINK_OPT_NO_PASS-NOT: sycl-post-link{{.*}}emit-only-kernels-as-entry-points
 
 /// Check selective passing of -allow-device-image-dependencies to sycl-post-link tool
-// RUN: %clang -### -target x86_64-unknown-linux-gnu -fsycl --no-offload-new-driver -fsycl-targets=spir64_fpga -fsycl-allow-device-dependencies %s 2>&1 | FileCheck -check-prefix=CHECK_SYCL_POST_LINK_ADID_PASS %s
-// RUN: %clang -### -target x86_64-unknown-linux-gnu -fsycl --no-offload-new-driver -fsycl-targets=spir64_gen -fsycl-allow-device-dependencies %s 2>&1 | FileCheck -check-prefix=CHECK_SYCL_POST_LINK_ADID_PASS %s
-// RUN: %clang -### -target x86_64-unknown-linux-gnu -fsycl --no-offload-new-driver -fsycl-targets=spir64_gen -fno-sycl-allow-device-dependencies %s 2>&1 | FileCheck -check-prefix=CHECK_SYCL_POST_LINK_ADID_NO_PASS %s
-// RUN: %clang_cl -### -fsycl --no-offload-new-driver -fsycl-targets=spir64_fpga -fsycl-allow-device-dependencies %s 2>&1 | FileCheck -check-prefix=CHECK_SYCL_POST_LINK_ADID_PASS %s
-// RUN: %clang_cl -### -fsycl --no-offload-new-driver -fsycl-targets=spir64_gen -fsycl-allow-device-dependencies %s 2>&1 | FileCheck -check-prefix=CHECK_SYCL_POST_LINK_ADID_PASS %s
-// RUN: %clang_cl -### -fsycl --no-offload-new-driver -fsycl-targets=spir64_gen -fno-sycl-allow-device-dependencies %s 2>&1 | FileCheck -check-prefix=CHECK_SYCL_POST_LINK_ADID_NO_PASS %s
+// RUN: %clang -### -target x86_64-unknown-linux-gnu -fsycl --no-offload-new-driver -fsycl-targets=spir64_fpga -fsycl-allow-device-image-dependencies %s 2>&1 | FileCheck -check-prefix=CHECK_SYCL_POST_LINK_ADID_PASS %s
+// RUN: %clang -### -target x86_64-unknown-linux-gnu -fsycl --no-offload-new-driver -fsycl-targets=spir64_gen -fsycl-allow-device-image-dependencies %s 2>&1 | FileCheck -check-prefix=CHECK_SYCL_POST_LINK_ADID_PASS %s
+// RUN: %clang -### -target x86_64-unknown-linux-gnu -fsycl --no-offload-new-driver -fsycl-targets=spir64_gen -fno-sycl-allow-device-image-dependencies %s 2>&1 | FileCheck -check-prefix=CHECK_SYCL_POST_LINK_ADID_NO_PASS %s
+// RUN: %clang_cl -### -fsycl --no-offload-new-driver -fsycl-targets=spir64_fpga -fsycl-allow-device-image-dependencies %s 2>&1 | FileCheck -check-prefix=CHECK_SYCL_POST_LINK_ADID_PASS %s
+// RUN: %clang_cl -### -fsycl --no-offload-new-driver -fsycl-targets=spir64_gen -fsycl-allow-device-image-dependencies %s 2>&1 | FileCheck -check-prefix=CHECK_SYCL_POST_LINK_ADID_PASS %s
+// RUN: %clang_cl -### -fsycl --no-offload-new-driver -fsycl-targets=spir64_gen -fno-sycl-allow-device-image-dependencies %s 2>&1 | FileCheck -check-prefix=CHECK_SYCL_POST_LINK_ADID_NO_PASS %s
 
 // CHECK_SYCL_POST_LINK_ADID_PASS: sycl-post-link{{.*}}allow-device-image-dependencies
 // CHECK_SYCL_POST_LINK_ADID_NO_PASS-NOT: sycl-post-link{{.*}}allow-device-image-dependencies

libdevice/cmath_wrapper_fp64.cpp

@@ -395,7 +395,7 @@ short _Exp(double *px, double y,
   if (*px < -HUGE_EXP || y == 0.0) // certain underflow
     *px = 0.0;
   else if (HUGE_EXP < *px) { // certain overflow
-    *px = _Inf._Double;
+    *px = _Inf._Double * (y < 0.F ? -1.F : 1.F);
     ret = _INFCODE;
   } else { // xexp won't overflow
     double g = *px * invln2;

sycl/cmake/modules/FetchUnifiedRuntime.cmake

@@ -116,15 +116,14 @@ if(SYCL_UR_USE_FETCH_CONTENT)
       CACHE PATH "Path to external '${name}' adapter source dir" FORCE)
   endfunction()
 
-  set(UNIFIED_RUNTIME_REPO "https://github.com/frasercrmck/unified-runtime.git")
-    
-  # commit cabf128094eff9ff7b79bdff559640a8a111f0c3
-  # Merge: a96fcbc5 15bca3b6
-  # Author: Omar Ahmed <[email protected]>
-  # Date:   Mon Aug 19 16:20:45 2024 +0100
-  #     Merge pull request #1984 from rafbiels/rafbiels/cuda-stream-race-cond
-  #     Fix race condition in CUDA stream creation
-  set(UNIFIED_RUNTIME_TAG 6eb5208bad81457d367dbb5b51ad266cda7cf7c1)
+  set(UNIFIED_RUNTIME_REPO "https://github.com/oneapi-src/unified-runtime.git")
+  # commit 7ecf64d60c31cd72bd88588498536d067bad59d6
+  # Merge: 17aa04d3 6eb5208b
+  # Author: aarongreig <[email protected]>
+  # Date:   Wed Sep 25 11:14:47 2024 +0100
+  #     Merge pull request #1996 from frasercrmck/ur-max-wg-size-props
+  #     Add two new properties to ur_kernel_group_info_t
+  set(UNIFIED_RUNTIME_TAG 7ecf64d60c31cd72bd88588498536d067bad59d6)
 
   set(UMF_BUILD_EXAMPLES OFF CACHE INTERNAL "EXAMPLES")
   # Due to the use of dependentloadflag and no installer for UMF and hwloc we need

sycl/doc/extensions/supported/sycl_ext_intel_esimd/README.md

@@ -16,7 +16,6 @@ vectorization is expressed explicitly in the code by the programmer.
 Please see the additional resources on the Intel DPC++ project's github:
 
 1) [ESIMD Extension Specification](./sycl_ext_intel_esimd.md)
-1) [ESIMD API/doxygen reference](https://intel.github.io/llvm-docs/doxygen/group__sycl__esimd.html)
 1) [Examples](./examples/README.md)
 1) [ESIMD end-to-end LIT tests](https://github.com/intel/llvm/blob/sycl/sycl/test-e2e/ESIMD/)
 1) [Implementation and API Restrictions](./sycl_ext_intel_esimd.md#implementation-restrictions)

sycl/doc/extensions/supported/sycl_ext_intel_esimd/sycl_ext_intel_esimd.md

@@ -38,7 +38,6 @@
   - [Other restrictions](#other-restrictions)
 
 ## Other content:
-* [ESIMD API/doxygen reference](https://intel.github.io/llvm-docs/doxygen/group__sycl__esimd.html)
 * [Examples](./examples/README.md)
 * [ESIMD LIT tests - working code examples](https://github.com/intel/llvm/blob/sycl/sycl/test-e2e/ESIMD/)
 
@@ -162,7 +161,6 @@ The element type must either be a vectorizable type or the `sycl::half` type.
 The set of vectorizable types is the
 set of fundamental SYCL arithmetic types excluding `bool`. The length of the
 vector is the second template parameter.
-See the complete [API reference](https://intel.github.io/llvm-docs/doxygen/classcl_1_1____ESIMD__NS_1_1simd.html#details) for the `simd` class for more details.
 
 ESIMD compiler back-end does the best it can to map each `simd` class object to a
 contiguous block of registers in the general register file (GRF).
@@ -351,8 +349,6 @@ reduction operations are supported:
 - maximum
 - minimum
 
-See more details on the API documentation [page TODO](https://intel.github.io/llvm-docs/doxygen).
-
 ### Memory access APIs
 
 Explicit SIMD memory access interface is quite different from the standard SYCL
@@ -598,9 +594,6 @@ Many memory access APIs accept offsets as arguments, which are used to determine
 actual memory location for the access. Offsets are always expressed in bytes
 rather than element units.
 
-See more details in the API documentation
-[page TODO](https://intel.github.io/llvm-docs/doxygen).
-
 ### Math operations
 
 #### Extended math
@@ -649,10 +642,6 @@ ESIMD supports the following non-standard math functions implemented in hardware
 - Plane equation - `plane`. Solves a component-wise plane equation
   `w = p*u + q*v + r` where `u`, `v`, `w` are vectors and `p`, `q`, `r` are scalars.
 
-
-See more details in the API documentation
-[page TODO](https://intel.github.io/llvm-docs/doxygen).
-
 ### Dot Product Accumulate Systolic - `DPAS` API
 
 DPAS is the matrix multiply-add-and-accumulate operation performed on limited size matrices/tiles.
@@ -870,9 +859,6 @@ There are other useful miscellaneous APIs provided by ESIMD.
 - Find least significant set bit - `fbl`.
 - Find most significant set bit - `fbh`.
 
-See more details in the API documentation
-[page TODO](https://intel.github.io/llvm-docs/doxygen).
-
 <br>
 
 ### Private Global Variables.