Merge branch 'develop' into common-dims

ROCm · Sep 7, 2023 · 2d2ea3f · 2d2ea3f
2 parents 8171c0b + a0894c2
commit 2d2ea3f
Show file tree

Hide file tree

Showing 244 changed files with 16,767 additions and 11,999 deletions.
diff --git a/.github/workflows/ci.yaml b/.github/workflows/ci.yaml
@@ -55,6 +55,7 @@ jobs:
           -DMIGRAPHX_ENABLE_GPU=On \
           -DMIGRAPHX_ENABLE_CPU=On \
           -DMIGRAPHX_ENABLE_FPGA=On \
+          -DMIGRAPHX_ENABLE_MLIR=On \
           -DBUILD_DEV=On \
           -DROCM_ENABLE_GH_ANNOTATIONS=On \
           -DCLANG_TIDY_DEPEND_ON_TARGET=Off \
@@ -169,6 +170,71 @@ jobs:
         git config --global --add safe.directory /data
         python3 tools/format.py origin/${{ github.event_name == 'pull_request' && github.base_ref || 'develop' }}
 
+  sles:
+    runs-on: ROCM-Ubuntu
+    steps:
+    - uses: actions/checkout@v3
+      with:
+        fetch-depth: 0
+
+    # In this step, this action saves a list of existing images,
+    # the cache is created without them in the post run.
+    # It also restores the cache if it exists.
+    - name: Docker layer cache
+      uses: jpribyl/[email protected]
+      with:
+        key: docker-layer-caching-migraphx-sles-${{hashFiles('hip-clang.docker', '**/*requirements.txt', '**/install_prereqs.sh', 'rbuild.ini')}}
+        restore-keys:
+          docker-layer-caching-migraphx-sles-
+      # Ignore the failure of a step and avoid terminating the job.
+      continue-on-error: true
+
+    - name: Build the Docker image
+      run: docker build . --file tools/docker/sles.docker --tag migraphx-sles
+
+    - name: Restore cache files for ccache
+      uses: actions/cache/restore@v3 
+      id: ccache_restore
+      with:
+        path: ${{ github.workspace }}/ccache
+        key: ccache-sles-${{ github.ref }}
+        restore-keys: ccache-sles-
+
+    - name: Build migraphx
+      shell: bash -c "docker run -i -v=$GITHUB_WORKSPACE:/data -w /data migraphx-sles bash < {0}"
+      run: |
+        set -e
+        export CCACHE_COMPRESSLEVEL=10
+        export CCACHE_DIR=/data/ccache
+        export CCACHE_NOHASHDIR=true
+        export CCACHE_BASEDIR=/data
+        export CCACHE_MAXSIZE=1
+        mkdir build
+        cd build
+        CXX=/opt/rocm/llvm/bin/clang++ CC=/opt/rocm/llvm/bin/clang cmake \
+          -DBUILD_DEV=On \
+          -DCMAKE_CXX_COMPILER_LAUNCHER=/usr/local/bin/ccache \
+          -DCMAKE_C_COMPILER_LAUNCHER=/usr/local/bin/ccache \
+          ..
+        make -j$(nproc) tests driver
+
+    - name: Clear ccache cache before saving
+      if: ${{ steps.ccache_restore.outputs.cache-hit }}
+      shell: bash
+      env:
+        GITHUB_TOKEN: ${{ secrets.GITHUB_TOKEN }}
+      run: |
+        set +x
+        gh extension install actions/gh-actions-cache --pin v1.0.1
+        gh actions-cache delete ${{ steps.ccache_restore.outputs.cache-matched-key }} --confirm
+
+    - name: Save cache files for ccache
+      uses: actions/cache/save@v3 
+      if: always()
+      with:
+        path: ${{ github.workspace }}/ccache
+        key: ccache-sles-${{ github.ref }}
+
   pyflakes:
     runs-on: ubuntu-20.04
 
@@ -274,11 +340,10 @@ jobs:
         # This path is specific to Ubuntu
         path: ${{ github.workspace }}/cget
         # Look to see if there is a cache hit for the corresponding requirements file
-        key: ${{ matrix.os }}-cget-4-${{ hashFiles('requirements.txt', 'dev-requirements.txt') }}
+        key: ${{ matrix.os }}-cget-4-${{ hashFiles('requirements.txt', 'dev-requirements.txt', 'rbuild.ini') }}
         restore-keys: ${{ matrix.os }}-cget-4-
 
     - name: Install dependencies
-      if: steps.deps_cache.outputs.cache-hit != 'true'
       run: rbuild prepare -d cget -s gh
 
     - name: Restore cache files for ccache

diff --git a/.github/workflows/sync-onnxrt-main.yaml b/.github/workflows/sync-onnxrt-main.yaml
@@ -47,6 +47,7 @@ jobs:
             onnxruntime
             dependancies
             automated 
+            skip bot checks
           assignees: TedThemistokleous
           reviewers: TedThemistokleous causten
           draft: false

diff --git a/CHANGELOG.md b/CHANGELOG.md
@@ -2,8 +2,49 @@
 
 Full documentation for MIGraphX is available at [MIGraphX Documentation](https://rocmdocs.amd.com/projects/AMDMIGraphX/en/latest/).
 
-## MIGraphX 2.5 for ROCm 5.5.0
+## MIGraphX 2.7 for ROCm 5.7.0
+### Added
+- Enabled hipRTC to not require dev packages for migraphx runtime and allow the ROCm install to be in a different directory than it was during build time
+- Add support for multi-target execution
+- Added Dynamic Batch support with C++/Python APIs
+- Add migraphx.create_argument to python API
+- Added dockerfile example for Ubuntu 22.04
+- Add TensorFlow supported ops in driver similar to exist onnx operator list
+- Add a MIGRAPHX_TRACE_MATCHES_FOR env variable to filter the matcher trace
+- Improved debugging by printing max,min,mean and stddev values for TRACE_EVAL = 2
+- use fast_math flag instead of ENV flag for GELU
+- Print message from driver if offload copy is set for compiled program
+### Optimizations
+- Optimized for ONNX Runtime 1.14.0
+- Improved compile times by only building for the GPU on the system
+- Improve performance of pointwise/reduction kernels when using NHWC layouts
+- Load specific version of the migraphx_py library
+- Annotate functions with the block size so the compiler can do a better job of optimizing 
+- Enable reshape on nonstandard shapes
+- Use half HIP APIs to compute max and min
+- Added support for broadcasted scalars to unsqueeze operator
+- Improved multiplies with dot operator
+- Handle broadcasts across dot and concat
+- Add verify namespace for better symbol resolution
+### Fixed
+- Resolved accuracy issues with FP16 resnet50
+- Update cpp generator to handle inf from  float
+- Fix assertion error during verify and make DCE work with tuples
+- Fix convert operation for NaNs
+- Fix shape typo in API test
+- Fix compile warnings for shadowing variable names
+- Add missing specialization for the `nullptr` for the hash function
+### Changed
+- Bumped version of half library to 5.6.0
+- Bumped CI to support rocm 5.6
+- Make building tests optional
+- replace np.bool with bool as per numpy request
+### Removed
+- Removed int8x4 rocBlas calls due to deprecation
+- removed std::reduce usage since not all OS' support it
+
 
+## MIGraphX 2.5 for ROCm 5.5.0
 ### Added
 - Y-Model feature to store tuning information with the optimized model
 - Added Python 3.10 bindings 
@@ -12,15 +53,11 @@ Full documentation for MIGraphX is available at [MIGraphX Documentation](https:/
 - Build support for ROCm MLIR
 - Added migraphx-driver flag to print optimizations in python (--python)
 - Added JIT implementation of the Gather and Pad operator which results in better handling of larger tensor sizes.
-
-
 ### Optimizations
 - Improved performance of Transformer based models
 - Improved performance of the Pad, Concat, Gather, and Pointwise operators
 - Improved onnx/pb file loading speed
 - Added general optimize pass which runs several passes such as simplify_reshapes/algebra and DCE in loop.
-
-
 ### Fixed
 - Improved parsing Tensorflow Protobuf files 
 - Resolved various accuracy issues with some onnx models
@@ -29,6 +66,5 @@ Full documentation for MIGraphX is available at [MIGraphX Documentation](https:/
 - Use --offload-arch instead of --cuda-gpu-arch for the HIP compiler
 - Changes inside JIT to use float accumulator for large reduce ops of half type to avoid overflow.
 - Changes inside JIT to temporarily use cosine to compute sine function.
-
 ### Changed
 - Changed version/location of 3rd party build dependencies to pick up fixes
diff --git a/CMakeLists.txt b/CMakeLists.txt
@@ -51,6 +51,7 @@ project(migraphx LANGUAGES C CXX)
 include(CTest)
 
 find_package(ROCM REQUIRED)
+find_package(Threads REQUIRED)
 
 find_path(HALF_INCLUDE_DIR half.hpp PATH_SUFFIXES half)
 if (NOT HALF_INCLUDE_DIR)
@@ -227,14 +228,12 @@ rocm_enable_cppcheck(
         shadowVar
         shadowVariable
         unsafeClassDivZero
+        # Disable because of too many FPs
+        arithOperationsOnVoidPointer
         definePrefix:*test/include/test.hpp
         ctuOneDefinitionRuleViolation:*test/*
         useSmartPointer:*src/api/api.cpp
         useSmartPointer:*make_shared_array.hpp
-        constParameter:*src/targets/gpu/*.cpp
-        constParameter:*src/targets/gpu/*.hpp
-        # Suppress mlir_conv.cpp since this file will be deleted
-        *:*src/targets/gpu/mlir_conv.cpp
     FORCE
     INCONCLUSIVE
     RULE_FILE
@@ -251,10 +250,13 @@ rocm_enable_cppcheck(
         ${CMAKE_CURRENT_SOURCE_DIR}/src/targets/gpu/kernels/include
         ${CMAKE_CURRENT_SOURCE_DIR}/test/include
     DEFINE
+        MIGRAPHX_MLIR=1
         CPPCHECK=1
         __device__=
         __host__=
         __global__=
+    UNDEFINE
+        MIGRAPHX_USE_CLANG_TIDY
 )
 
 enable_testing()

diff --git a/Dockerfile b/Dockerfile
@@ -77,6 +77,9 @@ ADD dev-requirements.txt /dev-requirements.txt
 ADD requirements.txt /requirements.txt
 ADD rbuild.ini /rbuild.ini
 
+# Temporarily install a new cmake until switching to ubuntu 22.04
+RUN pip3 install cmake==3.22.1
+
 COPY ./tools/install_prereqs.sh /
 RUN /install_prereqs.sh /usr/local / && rm /install_prereqs.sh
 RUN test -f /usr/local/hash || exit 1
@@ -98,6 +101,9 @@ RUN cget -p $PREFIX install facebook/[email protected] -X subdir -DCMAKE_DIR=build/cma
 RUN cget -p $PREFIX install [email protected] -DENABLE_TESTING=OFF
 RUN cget -p /opt/cmake install kitware/[email protected]
 
+# Install MLIR
+ADD mlir-requirements.txt /mlir-requirements.txt
+RUN cget -p /usr/local install -f /mlir-requirements.txt
 
 COPY ./test/onnx/.onnxrt-commit /
 
@@ -113,9 +119,6 @@ RUN git clone --single-branch --branch ${ONNXRUNTIME_BRANCH} --recursive ${ONNXR
 
 ADD tools/build_and_test_onnxrt.sh /onnxruntime/build_and_test_onnxrt.sh
 
-# Use the /opt/cmake install because LLVM/MLIR need cmake >= 3.20
-RUN env PATH=/opt/cmake/bin:$PATH cget -p /usr/local install ROCmSoftwarePlatform/rocMLIR@1ad9d6df32acc6d29d58e8ed6710e36746d0a4d6 -DBUILD_FAT_LIBROCKCOMPILER=On
-
 ENV MIOPEN_FIND_DB_PATH=/tmp/miopen/find-db
 ENV MIOPEN_USER_DB_PATH=/tmp/miopen/user-db
 ENV LD_LIBRARY_PATH=$PREFIX/lib

diff --git a/Jenkinsfile b/Jenkinsfile
@@ -26,7 +26,7 @@ def rocmtestnode(Map conf) {
             rm -rf build
             mkdir build
             cd build
-            cmake -DCMAKE_C_COMPILER_LAUNCHER=ccache -DCMAKE_CXX_COMPILER_LAUNCHER=ccache -DBUILD_DEV=On ${flags} ..
+            cmake -DCMAKE_C_COMPILER_LAUNCHER=ccache -DCMAKE_CXX_COMPILER_LAUNCHER=ccache -DBUILD_DEV=On -DCMAKE_EXECUTE_PROCESS_COMMAND_ECHO=STDOUT ${flags} ..
             git diff
             git diff-index --quiet HEAD || (echo "Git repo is not clean after running cmake." && exit 1)
             make -j\$(nproc) generate VERBOSE=1
@@ -90,7 +90,7 @@ def rocmnodename(name) {
     } else if(name == "mi100+") {
         node_name = "${rocmtest_name} && (gfx908 || gfx90a) && !vm";
     } else if(name == "cdna") {
-        node_name = "${rocmtest_name} && (gfx908 || gfx90a || vega) && !vm";
+        node_name = "${rocmtest_name} && (gfx908 || gfx90a || vega20) && !vm";
     } else if(name == "nogpu") {
         node_name = "${rocmtest_name} && nogpu";
     }
@@ -114,6 +114,10 @@ rocmtest clang_debug: rocmnode('cdna') { cmake_build ->
         cmake_build(flags: "-DCMAKE_BUILD_TYPE=release")
         stash includes: 'build/*.deb', name: 'migraphx-package'
     }
+// }, hidden_symbols: rocmnode('cdna') { cmake_build ->
+//     stage('Hidden symbols') {
+//         cmake_build(flags: "-DMIGRAPHX_ENABLE_PYTHON=Off -DMIGRAPHX_ENABLE_GPU=On -DMIGRAPHX_ENABLE_CPU=On -DCMAKE_CXX_VISIBILITY_PRESET=hidden -DCMAKE_C_VISIBILITY_PRESET=hidden")
+//     }
 }, all_targets_debug : rocmnode('cdna') { cmake_build ->
     stage('All targets Release') {
         cmake_build(flags: "-DCMAKE_BUILD_TYPE=release -DMIGRAPHX_ENABLE_GPU=On -DMIGRAPHX_ENABLE_CPU=On -DMIGRAPHX_ENABLE_FPGA=On") 

diff --git a/cmake/Embed.cmake b/cmake/Embed.cmake
@@ -24,11 +24,7 @@
 find_program(EMBED_LD ld)
 find_program(EMBED_OBJCOPY objcopy)
 
-if(LINUX)
-    option(EMBED_USE_LD "Use ld to embed data files" ON)
-else()
-    option(EMBED_USE_LD "Use ld to embed data files" OFF)
-endif()
+option(EMBED_USE_LD "Use ld to embed data files" OFF)
 
 function(wrap_string)
     set(options)
@@ -60,8 +56,8 @@ endfunction()
 
 function(generate_embed_source EMBED_NAME)
     set(options)
-    set(oneValueArgs SRC HEADER)
-    set(multiValueArgs OBJECTS SYMBOLS)
+    set(oneValueArgs SRC HEADER RELATIVE)
+    set(multiValueArgs OBJECTS SYMBOLS FILES)
 
     cmake_parse_arguments(PARSE "${options}" "${oneValueArgs}" "${multiValueArgs}" ${ARGN})
 
@@ -78,6 +74,8 @@ function(generate_embed_source EMBED_NAME)
     foreach(idx RANGE ${LEN})
         list(GET PARSE_SYMBOLS ${idx} SYMBOL)
         list(GET PARSE_OBJECTS ${idx} OBJECT)
+        list(GET PARSE_FILES ${idx} FILE)
+
         set(START_SYMBOL "_binary_${SYMBOL}_start")
         set(END_SYMBOL "_binary_${SYMBOL}_end")
         if(EMBED_USE_LD)
@@ -92,9 +90,11 @@ function(generate_embed_source EMBED_NAME)
             ")
         endif()
 
-        # TODO: Should use NAME_WLE
-        get_filename_component(BASE_NAME "${OBJECT}" NAME)
-        string(REGEX REPLACE ".[A-Za-z0-9_]+$" "" BASE_NAME ${BASE_NAME})
+        if(PARSE_RELATIVE)
+            file(RELATIVE_PATH BASE_NAME ${PARSE_RELATIVE} "${FILE}")
+        else()
+            get_filename_component(BASE_NAME "${FILE}" NAME)
+        endif()
 
         string(APPEND INIT_KERNELS "
             { \"${BASE_NAME}\", { ${START_SYMBOL}, ${END_SYMBOL}} },
@@ -162,6 +162,11 @@ function(embed_file OUTPUT_FILE OUTPUT_SYMBOL FILE)
 endfunction()
 
 function(add_embed_library EMBED_NAME)
+    set(options)
+    set(oneValueArgs RELATIVE)
+    set(multiValueArgs)
+    cmake_parse_arguments(PARSE "${options}" "${oneValueArgs}" "${multiValueArgs}" ${ARGN})
+
     file(MAKE_DIRECTORY ${CMAKE_CURRENT_BINARY_DIR}/embed)
     file(MAKE_DIRECTORY ${CMAKE_CURRENT_BINARY_DIR}/embed/${EMBED_NAME})
     set(EMBED_DIR ${CMAKE_CURRENT_BINARY_DIR}/embed/${EMBED_NAME})
@@ -171,15 +176,26 @@ function(add_embed_library EMBED_NAME)
     set(OUTPUT_FILES)
     set(SYMBOLS)
     message(STATUS "Embedding files")
-    foreach(FILE ${ARGN})
+    foreach(FILE ${PARSE_UNPARSED_ARGUMENTS})
         embed_file(OUTPUT_FILE OUTPUT_SYMBOL ${FILE})
         list(APPEND OUTPUT_FILES ${OUTPUT_FILE})
         list(APPEND SYMBOLS ${OUTPUT_SYMBOL})
     endforeach()
     message(STATUS "Generating embedding library ${EMBED_NAME}")
-    generate_embed_source(${EMBED_NAME} SRC ${SRC_FILE} HEADER ${HEADER_FILE} OBJECTS ${OUTPUT_FILES} SYMBOLS ${SYMBOLS})
-    add_library(${EMBED_NAME} STATIC ${OUTPUT_FILES} "${SRC_FILE}")
-    target_include_directories(${EMBED_NAME} PUBLIC "${EMBED_DIR}/include")
-    target_compile_options(${EMBED_NAME} PRIVATE -Wno-reserved-identifier -Wno-extern-initializer -Wno-missing-variable-declarations)
-    set_target_properties(${EMBED_NAME} PROPERTIES POSITION_INDEPENDENT_CODE On)
+    generate_embed_source(${EMBED_NAME} SRC ${SRC_FILE} HEADER ${HEADER_FILE} OBJECTS ${OUTPUT_FILES} SYMBOLS ${SYMBOLS} RELATIVE ${PARSE_RELATIVE} FILES ${PARSE_UNPARSED_ARGUMENTS})
+
+    set(INTERNAL_EMBED_LIB embed_lib_${EMBED_NAME})
+    add_library(${INTERNAL_EMBED_LIB} OBJECT "${SRC_FILE}")
+    target_include_directories(${INTERNAL_EMBED_LIB} PRIVATE "${EMBED_DIR}/include")
+    target_compile_options(${INTERNAL_EMBED_LIB} PRIVATE -Wno-reserved-identifier -Wno-extern-initializer -Wno-missing-variable-declarations)
+    set_target_properties(${INTERNAL_EMBED_LIB} PROPERTIES POSITION_INDEPENDENT_CODE On)
+
+    add_library(${EMBED_NAME} INTERFACE)
+    if(EMBED_USE_LD)
+        target_sources(${EMBED_NAME} INTERFACE ${OUTPUT_FILES})
+    else()
+        target_sources(${INTERNAL_EMBED_LIB} PRIVATE ${OUTPUT_FILES})
+    endif()
+    target_sources(${EMBED_NAME} INTERFACE $<TARGET_OBJECTS:${INTERNAL_EMBED_LIB}>)
+    target_include_directories(${EMBED_NAME} INTERFACE "${EMBED_DIR}/include")
 endfunction()
diff --git a/cmake/ExportHeader.cmake b/cmake/ExportHeader.cmake
@@ -29,7 +29,10 @@ endif()
 include(GenerateExportHeader)
 
 function(migraphx_generate_export_header TARGET)
-    cmake_parse_arguments(PARSE "" "DIRECTORY" "" ${ARGN})
+    set(options)
+    set(oneValueArgs DIRECTORY)
+    set(multiValueArgs)
+    cmake_parse_arguments(PARSE "${options}" "${oneValueArgs}" "${multiValueArgs}" ${ARGN})
     if(PARSE_DIRECTORY)
         set(__directory ${PARSE_DIRECTORY})
     else()