Merge remote-tracking branch 'origin/sycl' into aspect_bf16

Author: aidan.belton

.github/CODEOWNERS

@@ -18,6 +18,9 @@ opencl/ @intel/llvm-reviewers-runtime
 # OpenCL "offline" compiler
 opencl/opencl-aot/ @intel/dpcpp-tools-reviewers
 
+# libclc library
+libclc/ @intel/llvm-reviewers-cuda
+
 # Device library
 libdevice/ @intel/llvm-reviewers-runtime
 

.github/workflows/sycl_linux_build_and_test.yml

@@ -52,6 +52,10 @@ on:
         type: string
         required: false
         default: "ghcr.io/intel/llvm/ubuntu2004_build:latest"
+      cuda_image:
+        type: string
+        required: false
+        default: "ghcr.io/intel/llvm/ubuntu2004_build:latest"
       lts_ref:
         type: string
         required: false

.github/workflows/sycl_nightly.yml

@@ -60,13 +60,3 @@ jobs:
         tags: |
           ghcr.io/${{ github.repository }}/sycl_ubuntu2004_nightly:no-drivers-${{ github.sha }}
           ghcr.io/${{ github.repository }}/sycl_ubuntu2004_nightly:no-drivers
-  ubuntu2004_build_test_new_pm:
-    # Default Linux building and LIT testing using new Pass Manager by default
-    if: github.repository == 'intel/llvm'
-    uses: ./.github/workflows/sycl_linux_build_and_test.yml
-    with:
-      build_cache_root: "/__w/"
-      build_cache_suffix: new_pm
-      build_artifact_suffix: new_pm
-      build_configure_extra_args: '--hip --cuda --cmake-opt=-DLLVM_ENABLE_NEW_PASS_MANAGER=ON'
-      lts_config: "hip_amdgpu;ocl_x64"

.github/workflows/sycl_precommit.yml

@@ -41,4 +41,4 @@ jobs:
       build_cache_size: "8G"
       build_artifact_suffix: "default"
       build_cache_suffix: "default"
-      lts_config: "hip_amdgpu;ocl_x64;ocl_gen9;l0_gen9;esimd_emu"
+      lts_config: "hip_amdgpu;ocl_x64;ocl_gen9;l0_gen9;esimd_emu;cuda"

.github/workflows/sycl_update_gpu_driver.yml

@@ -0,0 +1,58 @@
+name: Update GPU driver
+
+on:
+  schedule:
+    - cron:  '0 3 * * 2'
+  workflow_dispatch:
+
+jobs:
+  update_driver_linux:
+    runs-on: ubuntu-latest
+    if: github.repository == 'intel/llvm'
+    steps:
+    - uses: actions/checkout@v2
+    - name: Update dependencies file
+      run: |
+        version="$(python3 devops/scripts/update_drivers.py linux)"
+        echo 'NEW_DRIVER_VERSION='$version >> $GITHUB_ENV
+    - name: Create Pull Request
+      env:
+        BRANCH: ci/update_gpu_driver-linux-${{ env.NEW_DRIVER_VERSION }}
+        LLVMBOT_TOKEN: ${{ secrets.LLVM_MAIN_SYNC_BBSYCL_TOKEN }}
+        GITHUB_TOKEN: ${{ secrets.GITHUB_TOKEN }}
+      run: |
+        cd $GITHUB_WORKSPACE
+        # Set fake identity to fulfil git requirements
+        git config --global user.name "GitHub Actions"
+        git config --global user.email "[email protected]"
+        git checkout -B $BRANCH
+        git add -u
+        git commit -m "[GHA] Uplift GPU RT version for Linux CI" || exit 0   # exit if commit is empty
+        git push https://[email protected]/${{ github.repository }} ${BRANCH}
+        gh pr create --head $BRANCH --title "[GHA] Uplift GPU RT version for Linux CI" --body "Uplift GPU RT version for Linux to $NEW_DRIVER_VERSION"
+
+  update_driver_linux_staging:
+    runs-on: ubuntu-latest
+    if: github.repository == 'intel/llvm'
+    steps:
+    - uses: actions/checkout@v2
+    - name: Update dependencies file
+      run: |
+        version="$(python3 devops/scripts/update_drivers.py linux_staging)"
+        echo 'NEW_DRIVER_VERSION='$version >> $GITHUB_ENV
+    - name: Update sycl Branch
+      env:
+        BRANCH: ci/update_gpu_driver-linux_staging-${{ env.NEW_DRIVER_VERSION }}
+        LLVMBOT_TOKEN: ${{ secrets.LLVM_MAIN_SYNC_BBSYCL_TOKEN }}
+        GITHUB_TOKEN: ${{ secrets.GITHUB_TOKEN }}
+      run: |
+        cd $GITHUB_WORKSPACE
+        # Set fake identity to fulfil git requirements
+        git config --global user.name "GitHub Actions"
+        git config --global user.email "[email protected]"
+        git checkout -B $BRANCH
+        git add -u
+        git commit -m "[GHA] Uplift GPU RT version for Nightly Builds" || exit 0   # exit if commit is empty
+        git push https://[email protected]/${{ github.repository }} ${BRANCH}
+        gh pr create --head $BRANCH --title "[GHA] Uplift GPU RT version for Nightly Builds" --body "Uplift GPU RT version for Linux to $NEW_DRIVER_VERSION"
+

.github/workflows/sycl_windows_build_and_test.yml

@@ -38,7 +38,7 @@ jobs:
         IF NOT EXIST D:\github\_work\cache MKDIR D:\github\_work\cache
         IF NOT EXIST D:\github\_work\cache\sycl_${{inputs.build_cache_suffix}} MKDIR D:\github\_work\cache\${{inputs.build_cache_suffix}}
         python.exe src/buildbot/configure.py -o build ^
-          --ci-default ^
+          --ci-defaults ^
           --cmake-opt="-DCMAKE_C_COMPILER=cl" ^
           --cmake-opt="-DCMAKE_CXX_COMPILER=cl" ^
           --cmake-opt="-DCMAKE_INSTALL_PREFIX=%GITHUB_WORKSPACE%\install" ^

bolt/README.md

@@ -180,7 +180,7 @@ Once you have `perf.fdata` ready, you can use it for optimizations with
 BOLT. Assuming your environment is setup to include the right path, execute
 `llvm-bolt`:
 ```
-$ llvm-bolt <executable> -o <executable>.bolt -data=perf.fdata -reorder-blocks=cache+ -reorder-functions=hfsort -split-functions=2 -split-all-cold -split-eh -dyno-stats
+$ llvm-bolt <executable> -o <executable>.bolt -data=perf.fdata -reorder-blocks=ext-tsp -reorder-functions=hfsort -split-functions=2 -split-all-cold -split-eh -dyno-stats
 ```
 
 If you do need an updated debug info, then add `-update-debug-sections` option

bolt/docs/OptimizingClang.md

@@ -64,7 +64,7 @@ Notice that we are passing `clang-7` to `perf2bolt` which is the real binary tha
 the generated profile:
 ```bash
 $ llvm-bolt $CPATH/clang-7 -o $CPATH/clang-7.bolt -b clang-7.yaml \
-    -reorder-blocks=cache+ -reorder-functions=hfsort+ -split-functions=3 \
+    -reorder-blocks=ext-tsp -reorder-functions=hfsort+ -split-functions=3 \
     -split-all-cold -dyno-stats -icf=1 -use-gnu-stack
 ```
 The output will look similar to the one below:

bolt/docs/doxygen.cfg.in

@@ -38,7 +38,7 @@ PROJECT_NAME           = "BOLT"
 # could be handy for archiving the generated documentation or if some version
 # control system is used.
 
-PROJECT_NUMBER         =
+PROJECT_NUMBER         = @PACKAGE_VERSION@
 
 # Using the PROJECT_BRIEF tag one can provide an optional one line description
 # for a project that appears at the top of each page and should give viewer a

bolt/include/bolt/Core/BinaryContext.h

@@ -82,6 +82,13 @@ inline raw_ostream &operator<<(raw_ostream &OS, const SegmentInfo &SegInfo) {
   return OS;
 }
 
+// AArch64-specific symbol markers used to delimit code/data in .text.
+enum class MarkerSymType : char {
+  NONE = 0,
+  CODE,
+  DATA,
+};
+
 enum class MemoryContentsType : char {
   UNKNOWN = 0,             /// Unknown contents.
   POSSIBLE_JUMP_TABLE,     /// Possibly a non-PIC jump table.
@@ -549,6 +556,9 @@ class BinaryContext {
 
   std::unique_ptr<MCDisassembler> DisAsm;
 
+  /// Symbolic disassembler.
+  std::unique_ptr<MCDisassembler> SymbolicDisAsm;
+
   std::unique_ptr<MCAsmBackend> MAB;
 
   /// Indicates if relocations are available for usage.
@@ -662,6 +672,11 @@ class BinaryContext {
            TheTriple->getArch() == llvm::Triple::x86_64;
   }
 
+  // AArch64-specific functions to check if symbol is used to delimit
+  // code/data in .text. Code is marked by $x, data by $d.
+  MarkerSymType getMarkerType(const SymbolRef &Symbol) const;
+  bool isMarker(const SymbolRef &Symbol) const;
+
   /// Iterate over all BinaryData.
   iterator_range<binary_data_const_iterator> getBinaryData() const {
     return make_range(BinaryDataMap.begin(), BinaryDataMap.end());
@@ -1187,18 +1202,20 @@ class BinaryContext {
                         uint64_t Offset = 0,
                         const BinaryFunction *Function = nullptr,
                         bool PrintMCInst = false, bool PrintMemData = false,
-                        bool PrintRelocations = false) const;
+                        bool PrintRelocations = false,
+                        StringRef Endl = "\n") const;
 
   /// Print a range of instructions.
   template <typename Itr>
   uint64_t
   printInstructions(raw_ostream &OS, Itr Begin, Itr End, uint64_t Offset = 0,
                     const BinaryFunction *Function = nullptr,
                     bool PrintMCInst = false, bool PrintMemData = false,
-                    bool PrintRelocations = false) const {
+                    bool PrintRelocations = false,
+                    StringRef Endl = "\n") const {
     while (Begin != End) {
       printInstruction(OS, *Begin, Offset, Function, PrintMCInst, PrintMemData,
-                       PrintRelocations);
+                       PrintRelocations, Endl);
       Offset += computeCodeSize(Begin, Begin + 1);
       ++Begin;
     }

bolt/include/bolt/Core/BinaryDomTree.h

@@ -0,0 +1,57 @@
+//==- bolt/Core/BinaryDomTree.h - Dominator Tree at low-level IR -*- C++ -*-==//
+//
+// 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
+//
+//===----------------------------------------------------------------------===//
+//
+// This file defines the BinaryDomTree class, which represents a dominator tree
+// in the CFG of a binary function.
+//
+//===----------------------------------------------------------------------===//
+
+#ifndef BOLT_CORE_BINARY_DOMTREE_H
+#define BOLT_CORE_BINARY_DOMTREE_H
+
+#include "llvm/Support/GenericDomTreeConstruction.h"
+
+namespace llvm {
+namespace bolt {
+
+class BinaryBasicBlock;
+using BinaryDomTreeNode = DomTreeNodeBase<BinaryBasicBlock>;
+using BinaryDominatorTree = DomTreeBase<BinaryBasicBlock>;
+
+} // namespace bolt
+
+// BinaryDominatorTree GraphTraits specializations.
+template <>
+struct GraphTraits<bolt::BinaryDomTreeNode *>
+    : public DomTreeGraphTraitsBase<bolt::BinaryDomTreeNode,
+                                    bolt::BinaryDomTreeNode::iterator> {};
+
+template <>
+struct GraphTraits<const bolt::BinaryDomTreeNode *>
+    : public DomTreeGraphTraitsBase<const bolt::BinaryDomTreeNode,
+                                    bolt::BinaryDomTreeNode::const_iterator> {};
+
+template <>
+struct GraphTraits<bolt::BinaryDominatorTree *>
+    : public GraphTraits<bolt::BinaryDomTreeNode *> {
+  static NodeRef getEntryNode(bolt::BinaryDominatorTree *DT) {
+    return DT->getRootNode();
+  }
+
+  static nodes_iterator nodes_begin(bolt::BinaryDominatorTree *N) {
+    return df_begin(getEntryNode(N));
+  }
+
+  static nodes_iterator nodes_end(bolt::BinaryDominatorTree *N) {
+    return df_end(getEntryNode(N));
+  }
+};
+
+} // namespace llvm
+
+#endif

bolt/include/bolt/Core/BinaryFunction.h

@@ -667,9 +667,6 @@ class BinaryFunction {
                                            uint64_t Offset,
                                            uint64_t &TargetAddress);
 
-  DenseMap<const MCInst *, SmallVector<MCInst *, 4>>
-  computeLocalUDChain(const MCInst *CurInstr);
-
   BinaryFunction &operator=(const BinaryFunction &) = delete;
   BinaryFunction(const BinaryFunction &) = delete;
 
@@ -836,6 +833,15 @@ class BinaryFunction {
     return make_range(JumpTables.begin(), JumpTables.end());
   }
 
+  /// Return relocation associated with a given \p Offset in the function,
+  /// or nullptr if no such relocation exists.
+  const Relocation *getRelocationAt(uint64_t Offset) const {
+    assert(CurrentState == State::Empty &&
+           "Relocations unavailable in the current function state.");
+    auto RI = Relocations.find(Offset);
+    return (RI == Relocations.end()) ? nullptr : &RI->second;
+  }
+
   /// Returns the raw binary encoding of this function.
   ErrorOr<ArrayRef<uint8_t>> getData() const;
 
@@ -1951,11 +1957,6 @@ class BinaryFunction {
     return ColdLSDASymbol;
   }
 
-  /// True if the symbol is a mapping symbol used in AArch64 to delimit
-  /// data inside code section.
-  bool isDataMarker(const SymbolRef &Symbol, uint64_t SymbolSize) const;
-  bool isCodeMarker(const SymbolRef &Symbol, uint64_t SymbolSize) const;
-
   void setOutputDataAddress(uint64_t Address) { OutputDataOffset = Address; }
 
   uint64_t getOutputDataAddress() const { return OutputDataOffset; }

bolt/include/bolt/Core/BinaryLoop.h

@@ -15,18 +15,13 @@
 #ifndef BOLT_CORE_BINARY_LOOP_H
 #define BOLT_CORE_BINARY_LOOP_H
 
-#include "llvm/ADT/DepthFirstIterator.h"
 #include "llvm/Analysis/LoopInfoImpl.h"
-#include "llvm/Support/GenericDomTreeConstruction.h"
 
 namespace llvm {
 namespace bolt {
 
 class BinaryBasicBlock;
 
-using BinaryDomTreeNode = DomTreeNodeBase<BinaryBasicBlock>;
-using BinaryDominatorTree = DomTreeBase<BinaryBasicBlock>;
-
 class BinaryLoop : public LoopBase<BinaryBasicBlock, BinaryLoop> {
 public:
   BinaryLoop() : LoopBase<BinaryBasicBlock, BinaryLoop>() {}
@@ -62,35 +57,4 @@ class BinaryLoopInfo : public LoopInfoBase<BinaryBasicBlock, BinaryLoop> {
 } // namespace bolt
 } // namespace llvm
 
-namespace llvm {
-
-// BinaryDominatorTree GraphTraits specializations.
-template <>
-struct GraphTraits<bolt::BinaryDomTreeNode *>
-    : public DomTreeGraphTraitsBase<bolt::BinaryDomTreeNode,
-                                    bolt::BinaryDomTreeNode::iterator> {};
-
-template <>
-struct GraphTraits<const bolt::BinaryDomTreeNode *>
-    : public DomTreeGraphTraitsBase<const bolt::BinaryDomTreeNode,
-                                    bolt::BinaryDomTreeNode::const_iterator> {};
-
-template <>
-struct GraphTraits<bolt::BinaryDominatorTree *>
-    : public GraphTraits<bolt::BinaryDomTreeNode *> {
-  static NodeRef getEntryNode(bolt::BinaryDominatorTree *DT) {
-    return DT->getRootNode();
-  }
-
-  static nodes_iterator nodes_begin(bolt::BinaryDominatorTree *N) {
-    return df_begin(getEntryNode(N));
-  }
-
-  static nodes_iterator nodes_end(bolt::BinaryDominatorTree *N) {
-    return df_end(getEntryNode(N));
-  }
-};
-
-} // namespace llvm
-
 #endif