Merge branch 'sycl' into oneapins

Signed-off-by: James Brodman <[email protected]>

Author: jbrodman

.github/CODEOWNERS

@@ -82,7 +82,7 @@ sycl/plugins/cuda/ @intel/llvm-reviewers-cuda
 
 sycl/doc/ @pvchupin @kbobrovs
 
-sycl/doc/extensions/ @mkinsner @jbrodman
+sycl/doc/extensions/ @intel/dpcpp-specification-reviewers
 
 xpti/ @tovinkere @andykaylor
 xptifw/ @tovinkere  @andykaylor

.github/workflows/clang-format.yml

@@ -9,16 +9,16 @@ jobs:
   build:
     runs-on: ubuntu-latest
     steps:
+    - name: Get clang-format first
+      run: sudo apt-get install -yqq clang-format-9
+
     - uses: actions/checkout@v2
       with:
         fetch-depth: 2
 
-    - name: Get clang-format first
-      run: sudo apt-get install -yqq clang-format-10
-
     - name: Run clang-format for the patch
       run: |
-        git diff -U0 --no-color ${GITHUB_SHA}^1 ${GITHUB_SHA} -- | ./clang/tools/clang-format/clang-format-diff.py -p1 -binary clang-format-10 > ./clang-format.patch
+        git diff -U0 --no-color ${GITHUB_SHA}^1 ${GITHUB_SHA} -- | ./clang/tools/clang-format/clang-format-diff.py -p1 -binary clang-format-9 > ./clang-format.patch
 
     # Add patch with formatting fixes to CI job artifacts
     - uses: actions/upload-artifact@v1

buildbot/dependency.conf

@@ -1,31 +1,33 @@
 [VERSIONS]
-# https://github.com/intel/llvm/releases/download/2020-WW20/oclcpuexp-2020.10.6.0.4_rel.tar.gz
-ocl_cpu_rt_ver=2020.10.6.0.04
-# https://github.com/intel/llvm/releases/download/2020-WW20/win-oclcpuexp-2020.10.6.0.4_rel.zip
-ocl_cpu_rt_ver_win=2020.10.6.0.04
+# https://github.com/intel/llvm/releases/download/2020-WW31/oclcpuexp-2020.10.7.0.15_rel.tar.gz
+ocl_cpu_rt_ver=2020.10.7.0.15
+# https://github.com/intel/llvm/releases/download/2020-WW31/win-oclcpuexp-2020.10.7.0.15_rel.zip
+ocl_cpu_rt_ver_win=2020.10.7.0.15
 # Same GPU driver supports Level Zero and OpenCL:
-# https://github.com/intel/compute-runtime/releases/tag/20.21.16886
-ocl_gpu_rt_ver=20.21.16886
+# https://github.com/intel/compute-runtime/releases/tag/20.25.17111
+ocl_gpu_rt_ver=20.29.17408
 # Same GPU driver supports Level Zero and OpenCL:
-# https://downloadmirror.intel.com/29616/a08/igfx_win10_100.8280.zip
-ocl_gpu_rt_ver_win=27.20.100.8280
+# https://downloadmirror.intel.com/29674/a08/igfx_win10_100.8336.zip
+ocl_gpu_rt_ver_win=27.20.100.8336
 intel_sycl_ver=build
 # https://github.com/oneapi-src/oneTBB/releases/download/v2020.2/tbb-2020.2-lin.tgz
-tbb_ver=20200121_111047
+tbb_ver=2021.1.8.515
 # https://github.com/oneapi-src/oneTBB/releases/download/v2020.2/tbb-2020.2-win.zip
-tbb_ver_win=20200124_000000
-# TODO: provide URL for Linux OpenCL emulator of FPGA device
-fpga_ver=20200216_000000
-# TODO: provide URL for Windows OpenCL emulator of FPGA device
-fpga_ver_win=20200216_000000
+tbb_ver_win=2021.1.8.514
+# https://github.com/intel/llvm/releases/download/2020-WW31/fpgaemu-2020.10.7.0.15_rel.tar.gz
+ocl_fpga_emu_ver=2020.10.7.0.15
+# https://github.com/intel/llvm/releases/download/2020-WW31/win-fpgaemu-2020.10.7.0.15_rel.zip
+ocl_fpga_emu_ver_win=2020.10.7.0.15
+fpga_ver=20200701_000002
+fpga_ver_win=20200701_000002
 
 [DRIVER VERSIONS]
-cpu_driver_lin=2020.10.6.0.04
-cpu_driver_win=2020.10.6.0.04
-gpu_driver_lin=20.21.16886
-gpu_driver_win=27.20.100.8280
-fpga_driver_lin=2020.9.2.0
-fpga_driver_win=2020.9.2.0
+cpu_driver_lin=2020.10.7.0.15
+cpu_driver_win=2020.10.7.0.15
+gpu_driver_lin=20.29.17408
+gpu_driver_win=27.20.100.8336
+fpga_driver_lin=2020.10.7.0.15
+fpga_driver_win=2020.10.7.0.15
 # NVidia CUDA driver
-# http://developer.download.nvidia.com/compute/cuda/repos/ubuntu1804/x86_64/cuda-drivers_450.36.06-1_amd64.deb
-nvidia_gpu_driver_lin=450.36.06
+# TODO provide URL for CUDA driver
+nvidia_gpu_driver_lin=435.21

clang/lib/Driver/Driver.cpp

@@ -5740,6 +5740,7 @@ InputInfo Driver::BuildJobsForActionNoCache(
 
   // Only use pipes when there is exactly one input.
   InputInfoList InputInfos;
+  bool JobForPreprocessToStdout = false;
   for (const Action *Input : Inputs) {
     // Treat dsymutil and verify sub-jobs as being at the top-level too, they
     // shouldn't get temporary output names.
@@ -5750,6 +5751,11 @@ InputInfo Driver::BuildJobsForActionNoCache(
         C, Input, TC, BoundArch, SubJobAtTopLevel, MultipleArchs, LinkingOutput,
         CachedResults, A->getOffloadingDeviceKind()));
   }
+  // Check if we are in sub-work for preprocessing for host side. If so we will
+  // add another job to print information to terminal later.
+  if (!AtTopLevel && A->getKind() == Action::PreprocessJobClass &&
+      C.getJobs().size() == 1)
+    JobForPreprocessToStdout = true;
 
   // Always use the first input as the base input.
   const char *BaseInput = InputInfos[0].getBaseInput();
@@ -5784,6 +5790,7 @@ InputInfo Driver::BuildJobsForActionNoCache(
 
   // Determine the place to write output to, if any.
   InputInfo Result;
+  InputInfo ResultForPreprocessToStdout;
   InputInfoList UnbundlingResults;
   if (auto *UA = dyn_cast<OffloadUnbundlingJobAction>(JA)) {
     // If we have an unbundling job, we need to create results for all the
@@ -5945,6 +5952,8 @@ InputInfo Driver::BuildJobsForActionNoCache(
                                              AtTopLevel, MultipleArchs,
                                              OffloadingPrefix),
                        BaseInput);
+    if (JobForPreprocessToStdout)
+      ResultForPreprocessToStdout = InputInfo(A, "-", BaseInput);
   }
 
   if (CCCPrintBindings && !CCGenDiagnostics) {
@@ -5967,12 +5976,19 @@ InputInfo Driver::BuildJobsForActionNoCache(
       llvm::errs() << "] \n";
     }
   } else {
-    if (UnbundlingResults.empty())
+    if (UnbundlingResults.empty()) {
       T->ConstructJob(
           C, *JA, Result, InputInfos,
           C.getArgsForToolChain(TC, BoundArch, JA->getOffloadingDeviceKind()),
           LinkingOutput);
-    else
+      // Add another job to print information to terminal for host side.
+      if (JobForPreprocessToStdout) {
+        T->ConstructJob(
+            C, *JA, ResultForPreprocessToStdout, InputInfos,
+            C.getArgsForToolChain(TC, BoundArch, JA->getOffloadingDeviceKind()),
+            LinkingOutput);
+      }
+    } else
       T->ConstructJobMultipleOutputs(
           C, *JA, UnbundlingResults, InputInfos,
           C.getArgsForToolChain(TC, BoundArch, JA->getOffloadingDeviceKind()),

clang/test/Driver/print-internal-defines-for-sycl.c

@@ -0,0 +1,5 @@
+// Test if clang is able to print internal defines in SYCL mode
+//
+// RUN: %clangxx -fsycl -dM -E -x c++ %s 2>&1 \
+// RUN: | FileCheck --check-prefix CHECK-PRINT-INTERNAL-DEFINES %s
+// CHECK-PRINT-INTERNAL-DEFINES: #define

sycl/include/CL/sycl.hpp

@@ -37,6 +37,11 @@
 #include <CL/sycl/platform.hpp>
 #include <CL/sycl/pointers.hpp>
 #include <CL/sycl/program.hpp>
+#include <CL/sycl/properties/accessor_properties.hpp>
+#include <CL/sycl/properties/buffer_properties.hpp>
+#include <CL/sycl/properties/context_properties.hpp>
+#include <CL/sycl/properties/image_properties.hpp>
+#include <CL/sycl/properties/queue_properties.hpp>
 #include <CL/sycl/queue.hpp>
 #include <CL/sycl/range.hpp>
 #include <CL/sycl/sampler.hpp>

sycl/include/CL/sycl/ONEAPI/group_algorithm.hpp

@@ -138,6 +138,12 @@ template <typename Ptr, typename T>
 using EnableIfIsPointer =
     cl::sycl::detail::enable_if_t<cl::sycl::detail::is_pointer<Ptr>::value, T>;
 
+template <typename T>
+using EnableIfIsTriviallyCopyable = cl::sycl::detail::enable_if_t<
+    std::is_trivially_copyable<T>::value &&
+        !cl::sycl::detail::is_vector_arithmetic<T>::value,
+    T>;
+
 // EnableIf shorthands for algorithms that depend on type and an operator
 template <typename T, typename BinaryOperation>
 using EnableIfIsScalarArithmeticNativeOp = cl::sycl::detail::enable_if_t<
@@ -286,8 +292,8 @@ EnableIfIsPointer<Ptr, bool> none_of(Group g, Ptr first, Ptr last,
 }
 
 template <typename Group, typename T>
-EnableIfIsScalarArithmetic<T> broadcast(Group, T x,
-                                        typename Group::id_type local_id) {
+EnableIfIsTriviallyCopyable<T> broadcast(Group, T x,
+                                         typename Group::id_type local_id) {
   static_assert(sycl::detail::is_generic_group<Group>::value,
                 "Group algorithms only support the sycl::group and "
                 "ONEAPI::sub_group class.");
@@ -323,7 +329,7 @@ EnableIfIsVectorArithmetic<T> broadcast(Group g, T x,
 }
 
 template <typename Group, typename T>
-EnableIfIsScalarArithmetic<T>
+EnableIfIsTriviallyCopyable<T>
 broadcast(Group g, T x, typename Group::linear_id_type linear_local_id) {
   static_assert(sycl::detail::is_generic_group<Group>::value,
                 "Group algorithms only support the sycl::group and "
@@ -363,7 +369,7 @@ broadcast(Group g, T x, typename Group::linear_id_type linear_local_id) {
 }
 
 template <typename Group, typename T>
-EnableIfIsScalarArithmetic<T> broadcast(Group g, T x) {
+EnableIfIsTriviallyCopyable<T> broadcast(Group g, T x) {
   static_assert(sycl::detail::is_generic_group<Group>::value,
                 "Group algorithms only support the sycl::group and "
                 "ONEAPI::sub_group class.");

sycl/include/CL/sycl/accessor.hpp

@@ -22,6 +22,8 @@
 #include <CL/sycl/id.hpp>
 #include <CL/sycl/image.hpp>
 #include <CL/sycl/pointers.hpp>
+#include <CL/sycl/properties/accessor_properties.hpp>
+#include <CL/sycl/property_list.hpp>
 #include <CL/sycl/sampler.hpp>
 
 /// \file accessor.hpp

sycl/include/CL/sycl/detail/property_helper.hpp

@@ -0,0 +1,83 @@
+//==--------- property_helper.hpp --- SYCL property helper -----------------==//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+
+#pragma once
+
+#include <CL/sycl/detail/common.hpp>
+
+__SYCL_INLINE_NAMESPACE(cl) {
+namespace sycl {
+
+namespace detail {
+
+// All properties are split here to dataless properties and properties with
+// data. A dataless property is one which has no data stored in it. A property
+// with data is one which has data stored in it and usually provides and access
+// to it. For dataless property we just store a bool which indicates if a
+// property is set or not. For properties with data we store a pointer to the
+// base class because we do not know the size of such properties beforehand.
+
+// List of all dataless properties' IDs
+enum DataLessPropKind {
+  BufferUseHostPtr = 0,
+  ImageUseHostPtr,
+  QueueEnableProfiling,
+  InOrder,
+  NoInit,
+  BufferUsePinnedHostMemory,
+  UsePrimaryContext,
+  DataLessPropKindSize
+};
+
+// List of all properties with data IDs
+enum PropWithDataKind {
+  BufferUseMutex = 0,
+  BufferContextBound,
+  ImageUseMutex,
+  ImageContextBound,
+  PropWithDataKindSize
+};
+
+// Base class for dataless properties, needed to check that the type of an
+// object passed to the property_list is a property.
+class DataLessPropertyBase {};
+
+// Helper class for the dataless properties. Every such property is supposed
+// to inherit from it. The ID template parameter should be one from
+// DataLessPropKind.
+template <int ID> class DataLessProperty : DataLessPropertyBase {
+public:
+  static constexpr int getKind() { return ID; }
+};
+
+// Base class for properties with data, needed to check that the type of an
+// object passed to the property_list is a property and for checking if two
+// properties with data are of the same type.
+class PropertyWithDataBase {
+public:
+  PropertyWithDataBase(int ID) : MID(ID) {}
+  bool isSame(int ID) const { return ID == MID; }
+  virtual ~PropertyWithDataBase() = default;
+
+private:
+  int MID = -1;
+};
+
+// Helper class for the properties with data. Every such property is supposed
+// to inherit from it. The ID template parameter should be one from
+// PropWithDataKind.
+template <int ID> class PropertyWithData : public PropertyWithDataBase {
+public:
+  PropertyWithData() : PropertyWithDataBase(ID) {}
+  static int getKind() { return ID; }
+};
+
+} // namespace detail
+
+} // namespace sycl
+} // __SYCL_INLINE_NAMESPACE(cl)