Merge

.github/workflows/ci.yaml

@@ -39,10 +39,8 @@ jobs:
 
     - name: Prepare timestamp
       id: cache_timestamp
-      shell: cmake -P {0}
-      run: |
-        string(TIMESTAMP current_date "%Y-%m-%d-%H;%M;%S" UTC)
-        message("::set-output name=timestamp::${current_date}")
+      shell: bash 
+      run: echo timestamp="$(date +'%Y-%m-%dT%H:%M:%S')" >> $GITHUB_OUTPUT
 
     - name: Cache files for tidy
       uses: pat-s/always-upload-cache@v3.0.11
@@ -65,6 +63,7 @@ jobs:
           -DMIGRAPHX_ENABLE_GPU=On \
           -DMIGRAPHX_ENABLE_CPU=On \
           -DMIGRAPHX_ENABLE_FPGA=On \
+          -DBUILD_DEV=On \
           -DROCM_ENABLE_GH_ANNOTATIONS=On \
           -DCLANG_TIDY_DEPEND_ON_TARGET=Off \
           -DCLANG_TIDY_CACHE=/data/tidy-cache \
@@ -88,10 +87,8 @@ jobs:
 
     - name: Prepare timestamp
       id: cache_timestamp
-      shell: cmake -P {0}
-      run: |
-        string(TIMESTAMP current_date "%Y-%m-%d-%H;%M;%S" UTC)
-        message("::set-output name=timestamp::${current_date}")
+      shell: bash 
+      run: echo timestamp="$(date +'%Y-%m-%dT%H:%M:%S')" >> $GITHUB_OUTPUT
 
     - name: Cache files for cppcheck
       uses: pat-s/always-upload-cache@v2.1.3
@@ -112,6 +109,7 @@ jobs:
         cd build
         CXX=/opt/rocm/llvm/bin/clang++ CC=/opt/rocm/llvm/bin/clang cmake \
           -DCPPCHECK_BUILD_DIR=/data/cppcheck-cache \
+          -DBUILD_DEV=On \
           -DROCM_ENABLE_GH_ANNOTATIONS=On \
           ..
         make -j2 cppcheck
@@ -233,12 +231,11 @@ jobs:
         python -m pip install --upgrade pip
         pip install https://github.com/RadeonOpenCompute/rbuild/archive/master.tar.gz
         rbuild prepare -d cget -s gh
+        sudo apt-get install -y lld
     - name: Prepare timestamp
       id: cache_timestamp
-      shell: cmake -P {0}
-      run: |
-        string(TIMESTAMP current_date "%Y-%m-%d-%H;%M;%S" UTC)
-        message("::set-output name=timestamp::${current_date}")
+      shell: bash 
+      run: echo timestamp="$(date +'%Y-%m-%dT%H:%M:%S')" >> $GITHUB_OUTPUT
 
     - name: Cache files for ccache
       # Ignore the failure of a step and avoid terminating the job.
@@ -262,10 +259,11 @@ jobs:
         rbuild build -d cget -s gh -T check \
           -DCMAKE_BUILD_TYPE=${{matrix.configuration}} \
           -DMIGRAPHX_ENABLE_PYTHON=${{matrix.configuration == 'release' && 'On' || 'Off'}} \
+          -DBUILD_DEV=On \
           -DCMAKE_CXX_FLAGS_DEBUG="-g1 -Os -fdebug-prefix-map=$PWD=. -fdebug-types-section -fno-omit-frame-pointer -fsanitize=undefined -fno-sanitize-recover=undefined" \
           -DCMAKE_CXX_FLAGS_CODECOV="-g1 -Og -fdebug-prefix-map=$PWD=. -fdebug-types-section -fprofile-arcs -ftest-coverage -fno-omit-frame-pointer" \
-          -DCMAKE_EXE_LINKER_FLAGS='-fuse-ld=gold' \
-          -DCMAKE_SHARED_LINKER_FLAGS='-fuse-ld=gold'
+          -DCMAKE_EXE_LINKER_FLAGS='-fuse-ld=lld' \
+          -DCMAKE_SHARED_LINKER_FLAGS='-fuse-ld=lld'
         ${{ github.workspace }}/cget/bin/ccache -s
 
     - name: Upload code coverage
@@ -329,12 +327,11 @@ jobs:
         python -m pip install --upgrade pip
         pip install https://github.com/RadeonOpenCompute/rbuild/archive/master.tar.gz
         rbuild prepare -d cget -s gh
+        sudo apt-get install -y lld
     - name: Prepare timestamp
       id: cache_timestamp
-      shell: cmake -P {0}
-      run: |
-        string(TIMESTAMP current_date "%Y-%m-%d-%H;%M;%S" UTC)
-        message("::set-output name=timestamp::${current_date}")
+      shell: bash
+      run: echo timestamp="$(date +'%Y-%m-%dT%H:%M:%S')" >> $GITHUB_OUTPUT
 
     - name: Cache files for ccache
       # Ignore the failure of a step and avoid terminating the job.
@@ -358,10 +355,11 @@ jobs:
         rbuild build -d cget -s gh -T check \
           -DCMAKE_BUILD_TYPE=${{matrix.configuration}} \
           -DMIGRAPHX_ENABLE_PYTHON=${{matrix.configuration == 'release' && 'On' || 'Off'}} \
+          -DBUILD_DEV=On \
           -DCMAKE_CXX_FLAGS_DEBUG="-g1 -Os -fdebug-prefix-map=$PWD=. -fdebug-types-section -fno-omit-frame-pointer -fsanitize=undefined -fno-sanitize-recover=undefined" \
           -DCMAKE_CXX_FLAGS_CODECOV="-g1 -Og -fdebug-prefix-map=$PWD=. -fdebug-types-section -fprofile-arcs -ftest-coverage -fno-omit-frame-pointer" \
-          -DCMAKE_EXE_LINKER_FLAGS='-fuse-ld=gold' \
-          -DCMAKE_SHARED_LINKER_FLAGS='-fuse-ld=gold' \
+          -DCMAKE_EXE_LINKER_FLAGS='-fuse-ld=lld' \
+          -DCMAKE_SHARED_LINKER_FLAGS='-fuse-ld=lld' \
           -DMIGRAPHX_ENABLE_FPGA=On
         ${{ github.workspace }}/cget/bin/ccache -s
 

.github/workflows/history.yaml

@@ -20,6 +20,10 @@ on:
         description: Repository where benchmark utils are stored
         required: true
         default: "ROCmSoftwarePlatform/migraphx-benchmark-utils"
+      organization:
+        description: Organization based on which location of files will be different
+        required: true
+        default: "AMD"
 
 jobs:
   release:
@@ -29,6 +33,7 @@ jobs:
       end_date: ${{ github.event.inputs.end_date || 'yyyy-mm-dd' }}
       history_repo: ${{ github.event.inputs.history_repo || 'ROCmSoftwarePlatform/migraphx-reports' }}
       benchmark_utils_repo: ${{ github.event.inputs.benchmark_utils_repo || 'ROCmSoftwarePlatform/migraphx-benchmark-utils' }}
+      organization: ${{ github.event.inputs.organization || 'AMD' }}
     secrets:
       gh_token: ${{ secrets.MIGRAPHX_BOT_TOKEN }}
       mail_user: ${{ secrets.MAIL_USERNAME }}

.github/workflows/performance.yaml

@@ -29,6 +29,10 @@ on:
         description: Last N results
         required: true
         default: '10'
+      model_timeout:
+        description: If model in performance test script passes this threshold, it will be skipped
+        required: true
+        default: '30m'
       flags:
         description: -m for Max value; -s for Std dev; -r for Threshold file
         required: true
@@ -46,6 +50,7 @@ jobs:
       performance_reports_repo: ${{ github.event.inputs.performance_reports_repo || 'ROCmSoftwarePlatform/migraphx-reports' }}
       benchmark_utils_repo: ${{ github.event.inputs.benchmark_utils_repo || 'ROCmSoftwarePlatform/migraphx-benchmark-utils' }}
       organization: ${{ github.event.inputs.organization || 'AMD' }}
+      model_timeout: ${{ github.event.inputs.model_timeout || '30m' }}
     secrets:
       gh_token: ${{ secrets.MIGRAPHX_BOT_TOKEN }}
       mail_user: ${{ secrets.MAIL_USERNAME }}

.github/workflows/sync-onnxrt-main.yaml

 name: Onnxruntime main weekly sync
 on:
   schedule:
-    - cron: "05 17 * * 1"
+    - cron: '07 17 * * 5'
 
 jobs:
-  runs-on: ubuntu-latest
-  sync:
+  createPullRequest:
+    name: Update and create pull request
+    runs-on: ubuntu-latest
     steps:
+      - name: get_date
+        run: echo todays_date="$(date +'%Y-%m-%d')" >> $GITHUB_ENV
+
+      - name: extract_sha1 
+        run: echo onnxsha="$(git ls-remote https://github.com/microsoft/onnxruntime.git HEAD | awk '{print $1}')" >> $GITHUB_ENV
+
+      - name: echo_sha1
+        run: echo ${{ env.onnxsha }}
+
       - uses: actions/checkout@v3
-        with:
+        with: 
           ref: develop
-          path: ../
 
-  get_date:
-    steps:
-      - run: echo "::set-output name=date::$(date +'%Y-%m-%d')"
+      - name: update_file
+        run:  echo ${{ env.onnxsha }} > test/onnx/.onnxrt-commit
 
-  update_file:
-    needs: [sync get_date]
-    steps:
-      - run: git clone https://github.com/microsoft/onnxruntime.git && cd onnxruntime && git rev-parse HEAD >> ../test/onnx/.onnxrt-commit
-
-  Add_commit:
-    needs: update_file
-    steps:
-      - name: Add & Commit
-        uses: EndBug/add-and-commit@v9.1.1
+      - name: Make changes to pull request
+        uses: peter-evans/create-pull-request@v4
         with:
-          new_branch: onnxruntime-sync-${{ steps.date.outputs.date }}
-          add: ../test/onnx/.onnxrt-commit
-          message: Update Onnxruntime commit to latest release
-          default_author: github_actions
-          push: true
-  PR:
-    needs: Add_commit
-    steps:
-      - name: GitHub Action for creating Pull Requests
-        uses: devops-infra/action-pull-request@v0.5.3
-        with:
-          github_token: ${{ secrets.GITHUB_TOKEN }}
-          title: Sync Onnxruntime main
-          reviewer: pfultz2, causten
-          assignee: TedThemistokleous
-          label: automatic, onnxruntime
-          target_branch: develop
+          token: ${{ secrets.GITHUB_TOKEN }}
+          commit-message: Update onnxruntime main ${{ env.onnxsha }}
+          committer: Github <noreply@github.com>
+          author: ${{ github.actor }} <${{ github.actor }}@users.noreply.github.com>
+          signoff: false
+          branch: onnxruntime-sync-${{ env.todays_date }}
+          delete-branch: true
+          title: 'Onnxruntime Weekly Sync ${{ env.todays_date }}'
+          body: |
+            Update CI point for Onnxruntime builds
+            - Updated with changes from ${{ env.todays_date }}
+            - Auto-generated by [create-pull-request][1]
+            - Update .onnxrt-commit to ${{ env.onnxsha }}
+
+            [1]: https://github.com/peter-evans/create-pull-request
+          labels: |
+            onnxruntime
+            dependancies
+            automated 
+          assignees: TedThemistokleous
+          reviewers: TedThemistokleous causten
+          draft: false
+          base: develop
\ No newline at end of file

CMakeLists.txt

@@ -46,6 +46,8 @@ else()
     set(CMAKE_INSTALL_PREFIX "/opt/rocm" CACHE PATH "")
 endif()
 
+set(CMAKE_BUILD_RPATH "${CMAKE_BINARY_DIR}/lib")
+
 project(migraphx)
 find_package(ROCM REQUIRED)
 
@@ -65,8 +67,10 @@ set(CMAKE_EXTRA_INCLUDE_FILES)
 
 include(ROCMSetupVersion)
 
-rocm_setup_version(VERSION 2.6)
-set(MIGRAPHX_SO_VERSION ${PROJECT_VERSION_MAJOR}.${PROJECT_VERSION_MINOR})
+option(BUILD_DEV "Build for development purpose only" OFF)
+
+rocm_setup_version(VERSION 2.6.0)
+set(MIGRAPHX_SO_VERSION ${PROJECT_VERSION_MAJOR}.${PROJECT_VERSION_MINOR}.${PROJECT_VERSION_PATCH})
 
 option( BUILD_SHARED_LIBS "Build as a shared library" ON )
 

Dockerfile

@@ -110,7 +110,7 @@ RUN git clone --single-branch --branch ${ONNXRUNTIME_BRANCH} --recursive ${ONNXR
 
 ADD tools/build_and_test_onnxrt.sh /onnxruntime/build_and_test_onnxrt.sh
 
-RUN cget -p /usr/local install ROCmSoftwarePlatform/rocMLIR@78b706fe9879587ab98b6614ae539265374a3fae -DBUILD_MIXR_TARGET=On -DLLVM_ENABLE_ZSTD=Off -DLLVM_ENABLE_THREADS=Off
+RUN cget -p /usr/local install ROCmSoftwarePlatform/rocMLIR@acb727b348086b58a7f261b32c0e4f0686a4c0ee -DBUILD_MIXR_TARGET=On -DLLVM_ENABLE_ZSTD=Off -DLLVM_ENABLE_THREADS=Off
 
 ENV MIOPEN_FIND_DB_PATH=/tmp/miopen/find-db
 ENV MIOPEN_USER_DB_PATH=/tmp/miopen/user-db

Jenkinsfile

@@ -28,7 +28,7 @@ def rocmtestnode(Map conf) {
             rm -rf build
             mkdir build
             cd build
-            cmake -DCMAKE_C_COMPILER_LAUNCHER=ccache -DCMAKE_CXX_COMPILER_LAUNCHER=ccache ${flags} ..
+            cmake -DCMAKE_C_COMPILER_LAUNCHER=ccache -DCMAKE_CXX_COMPILER_LAUNCHER=ccache -DBUILD_DEV=On ${flags} ..
             make -j\$(nproc) generate all doc package check VERBOSE=1
         """
         echo cmd
@@ -116,9 +116,11 @@ rocmtest clang_debug: rocmnode('vega') { cmake_build ->
     }
 }, mlir_debug: rocmnode('vega') { cmake_build ->
     stage('MLIR Debug') {
-        def sanitizers = "undefined"
-        def debug_flags = "-g -O2 -fsanitize=${sanitizers} -fno-sanitize-recover=${sanitizers}"
-        cmake_build(flags: "-DCMAKE_BUILD_TYPE=debug -DMIGRAPHX_ENABLE_PYTHON=Off -DMIGRAPHX_ENABLE_MLIR=On -DCMAKE_CXX_FLAGS_DEBUG='${debug_flags}' -DCMAKE_C_FLAGS_DEBUG='${debug_flags}'")
+        withEnv(['MIGRAPHX_ENABLE_MLIR=1']) {
+            def sanitizers = "undefined"
+            def debug_flags = "-g -O2 -fsanitize=${sanitizers} -fno-sanitize-recover=${sanitizers}"
+            cmake_build(flags: "-DCMAKE_BUILD_TYPE=debug -DMIGRAPHX_ENABLE_PYTHON=Off -DMIGRAPHX_ENABLE_MLIR=On -DCMAKE_CXX_FLAGS_DEBUG='${debug_flags}' -DCMAKE_C_FLAGS_DEBUG='${debug_flags}'")
+        }
     }
 }, clang_asan: rocmnode('nogpu') { cmake_build ->
     stage('Clang ASAN') {

doc/src/driver/compile.rst

@@ -32,6 +32,10 @@ Disable fast math optimization
 
 Perform an exhaustive search to find the fastest version of generated kernels for selected backend
 
+.. options:: --split-single-dyn-dim
+
+Enable the split single dynamic dimension pass
+
 .. option::  --fp16
 
 Quantize for fp16

doc/src/driver/read.rst

@@ -24,7 +24,7 @@ Load as MIGraphX JSON
 
 .. option::  --batch [unsigned int] (Default: 1)
 
-Set batch size for model
+For a static model, set batch size. For a dynamic batch model, sets the batch size at runtime.
 
 .. option::  --nhwc
 
@@ -46,6 +46,14 @@ Trim instructions from the end (Default: 0)
 
 Dim of a parameter (format: "@name d1 d2 dn")
 
+.. options:: --dyn-input-dim [std::vector<std::string>]
+
+Set dynamic dimensions of a parameter using JSON formatting (format "@name" "dynamic_dimension_json")
+
+.. options:: --default-dyn-dim
+
+Set the default dynamic dimension (format {min:x, max:y, optimals:[o1,o2,...]})
+
 .. option::  --optimize, -O
 
 Optimize when reading

examples/migraphx/cpp_parse_load_save/CMakeLists.txt

@@ -27,7 +27,7 @@ project (PLS)
 set (CMAKE_CXX_STANDARD 14)
 set (EXAMPLE parse_load_save)
 
-list (APPEND CMAKE_PREFIX_PATH /opt/rocm/hip /opt/rocm)
+list (APPEND CMAKE_PREFIX_PATH /opt/rocm)
 find_package (migraphx)
 
 message("source file: " ${EXAMPLE}.cpp " ---> bin: " ${EXAMPLE})

examples/migraphx/custom_op_miopen_kernel/CMakeLists.txt

@@ -27,7 +27,7 @@ project (custom_miopen_kernel)
 set (CMAKE_CXX_STANDARD 14)
 set (EXAMPLE custom_op_miopen_kernel)
 
-list (APPEND CMAKE_PREFIX_PATH /opt/rocm/hip /opt/rocm)
+list (APPEND CMAKE_PREFIX_PATH /opt/rocm)
 find_package (migraphx REQUIRED)
 find_package (miopen REQUIRED)
 

examples/migraphx/custom_op_rocblas_kernel/CMakeLists.txt

@@ -28,7 +28,7 @@ set (CMAKE_CXX_STANDARD 14)
 set (EXAMPLE custom_op_rocblas_kernel)
 
 
-list (APPEND CMAKE_PREFIX_PATH /opt/rocm/hip /opt/rocm)
+list (APPEND CMAKE_PREFIX_PATH /opt/rocm)
 find_package (migraphx REQUIRED)
 find_package (rocblas REQUIRED)
 

examples/migraphx/migraphx_driver/README.md

@@ -29,7 +29,7 @@ See below for a comprehensive list of commands and option arguments, as well as
 | --tf                                     | Load file as a tensorflow graph                           |
 | --migraphx                               | Load file as a migraphx graph                             |
 | --migraphx-json                          | Load file as a migraphx JSON graph                        |
-| --batch                                  | Set batch size for the model                              |
+| --batch                                  | For a static model, set batch size. For a dynamic batch model, sets the batch size at runtime.|
 | --nhwc                                   | Treat tensorflow format as nhwc                           |
 | --nchw                                   | Treat tensorflow format as nchw                           |
 | --skip-unknown-operators                 | Skip unknown operators when parsing and continue to parse |
@@ -44,12 +44,16 @@ See below for a comprehensive list of commands and option arguments, as well as
 | --output \| -o                           | Output to file                                            |
 | --fill0                                  | Fill parameter with 0s                                    |
 | --fill1                                  | Fill parameter with 1s                                    |
+| --input-dim                              | Set static dimensions of a parameter                      |
+| --dyn-input-dim                          | Set dynamic dimensions of a parameter                     |
+| --default-dyn-dim                        | Set default dynamic dimension                             |
 | --gpu                                    | Compile on the gpu                                        |
 | --cpu                                    | Compile on the cpu                                        |
 | --ref                                    | Compile on the reference implementation                   |
 | --enable-offload-copy                    | Enable implicit offload copying                           |
 | --disable-fast-math                      | Disable fast math optimization                            |
 | --exhaustive-tune                        | Enable exhaustive search to find fastest kernel           |
+| --split-single-dyn-dim                   | Enable split_single_dyn_dim compiler pass                 |
 | --fp16                                   | Quantize for fp16                                         |
 | --int8                                   | Quantize for int8                                         |
 | --tolerance                              | Tolerance for errors                                      |
@@ -88,7 +92,7 @@ batch_norm_inference
 broadcast
 capture
 ceil
-check_context::migraphx::version_1::gpu::context
+check_context::migraphx::gpu::context
 clip
 concat
 contiguous
@@ -304,7 +308,7 @@ $ /opt/rocm/bin/migraphx-driver run --onnx simple_graph.onnx
 ```
 Compiling ... 
 Reading: simple_graph.onnx
-@0 = check_context::migraphx::version_1::gpu::context -> float_type, {}, {}
+@0 = check_context::migraphx::gpu::context -> float_type, {}, {}
 @1 = hip::hip_allocate_memory[shape=float_type, {256}, {1},id=scratch] -> float_type, {256}, {1}
 @2 = hip::hip_copy_literal[id=@literal:1] -> float_type, {784, 128}, {128, 1}
 x:0 = @param:x:0 -> float_type, {1, 28, 28}, {784, 28, 1}
@@ -327,7 +331,7 @@ x:0 = @param:x:0 -> float_type, {1, 28, 28}, {784, 28, 1}
 @18 = @return(@17)
 
 Allocating params ... 
-@0 = check_context::migraphx::version_1::gpu::context -> float_type, {}, {}
+@0 = check_context::migraphx::gpu::context -> float_type, {}, {}
 @1 = hip::hip_allocate_memory[shape=float_type, {256}, {1},id=scratch] -> float_type, {256}, {1}
 @2 = hip::hip_copy_literal[id=@literal:1] -> float_type, {784, 128}, {128, 1}
 x:0 = @param:x:0 -> float_type, {1, 28, 28}, {784, 28, 1}
@@ -399,7 +403,7 @@ $ /opt/rocm/bin/migraphx-driver compile --gpu --fp16 simple_graph.pb
 ```
 Compiling ... 
 Reading: simple_graph.pb
-@0 = check_context::migraphx::version_1::gpu::context -> float_type, {}, {}
+@0 = check_context::migraphx::gpu::context -> float_type, {}, {}
 @1 = hip::hip_allocate_memory[shape=float_type, {456}, {1},id=scratch] -> float_type, {456}, {1}
 @2 = hip::hip_copy_literal[id=@literal:0] -> half_type, {784, 128}, {128, 1}
 @3 = load[offset=256,end=1824](@1) -> half_type, {1, 28, 28}, {784, 28, 1}
@@ -502,7 +506,7 @@ x = @param:x -> float_type, {1, 28, 28}, {784, 28, 1}
 @18 = ref::softmax[axis=1](@17) -> float_type, {1, 10}, {10, 1}
 @19 = ref::identity(@18) -> float_type, {1, 10}, {10, 1}
 
-@0 = check_context::migraphx::version_1::gpu::context -> float_type, {}, {}
+@0 = check_context::migraphx::gpu::context -> float_type, {}, {}
 @1 = hip::hip_allocate_memory[shape=float_type, {256}, {1},id=scratch] -> float_type, {256}, {1}
 @2 = hip::hip_copy_literal[id=@literal:3] -> float_type, {784, 128}, {128, 1}
 x = @param:x -> float_type, {1, 28, 28}, {784, 28, 1}
@@ -538,7 +542,7 @@ $ /opt/rocm/bin/migraphx-driver perf simple_graph.pb
 ```
 Compiling ... 
 Reading: simple_graph.pb
-@0 = check_context::migraphx::version_1::gpu::context -> float_type, {}, {}
+@0 = check_context::migraphx::gpu::context -> float_type, {}, {}
 @1 = hip::hip_allocate_memory[shape=float_type, {256}, {1},id=scratch] -> float_type, {256}, {1}
 @2 = hip::hip_copy_literal[id=@literal:3] -> float_type, {784, 128}, {128, 1}
 @3 = load[offset=0,end=512](@1) -> float_type, {1, 128}, {128, 1}
@@ -561,7 +565,7 @@ output = @param:output -> float_type, {1, 10}, {10, 1}
 
 Allocating params ... 
 Running performance report ... 
-@0 = check_context::migraphx::version_1::gpu::context -> float_type, {}, {}: 0.00057782ms, 1%
+@0 = check_context::migraphx::gpu::context -> float_type, {}, {}: 0.00057782ms, 1%
 @1 = hip::hip_allocate_memory[shape=float_type, {256}, {1},id=scratch] -> float_type, {256}, {1}: 0.000295ms, 1%
 @2 = hip::hip_copy_literal[id=@literal:3] -> float_type, {784, 128}, {128, 1}: 0.00027942ms, 1%
 @3 = load[offset=0,end=512](@1) -> float_type, {1, 128}, {128, 1}: 0.000232ms, 1%
@@ -591,7 +595,7 @@ hip::hip_copy_literal: 0.00186824ms, 1%
 load: 0.0016288ms, 1%
 @param: 0.0013428ms, 1%
 broadcast: 0.00118042ms, 1%
-check_context::migraphx::version_1::gpu::context: 0.00057782ms, 1%
+check_context::migraphx::gpu::context: 0.00057782ms, 1%
 reshape: 0.00033842ms, 1%
 hip::hip_allocate_memory: 0.000295ms, 1%
 

examples/vision/cpp_mnist/CMakeLists.txt

@@ -27,7 +27,7 @@ project (CAI)
 set (CMAKE_CXX_STANDARD 14)
 set (EXAMPLE mnist_inference)
 
-list (APPEND CMAKE_PREFIX_PATH /opt/rocm/hip /opt/rocm)
+list (APPEND CMAKE_PREFIX_PATH /opt/rocm)
 find_package (migraphx)
 
 message("source file: " ${EXAMPLE}.cpp " ---> bin: " ${EXAMPLE})

src/CMakeLists.txt

@@ -50,6 +50,7 @@ add_library(migraphx
     env.cpp
     file_buffer.cpp
     fuse_pointwise.cpp
+    fuse_reduce.cpp
     generate.cpp
     inline_module.cpp
     insert_pad.cpp
@@ -73,6 +74,7 @@ add_library(migraphx
     process.cpp
     program.cpp
     propagate_constant.cpp
+    promote_literals.cpp
     quantization.cpp
     quantize_fp16.cpp
     quantize_int8.cpp
@@ -91,6 +93,7 @@ add_library(migraphx
     shape.cpp
     simplify_algebra.cpp
     simplify_reshapes.cpp
+    split_single_dyn_dim.cpp
     tmp_dir.cpp
     value.cpp
     verify_args.cpp
@@ -251,7 +254,7 @@ find_package(PkgConfig)
 pkg_check_modules(SQLITE3 REQUIRED IMPORTED_TARGET sqlite3)
 target_link_libraries(migraphx PRIVATE PkgConfig::SQLITE3)
 
-find_package(msgpackc-cxx)
+find_package(msgpackc-cxx QUIET)
 if(NOT msgpackc-cxx_FOUND)
     find_package(msgpack REQUIRED)
 endif()
@@ -292,6 +295,10 @@ if(HAVE_HALF_EXPR)
     target_compile_definitions(migraphx PUBLIC -DHAS_HALF_V1)
 endif()
 
+if(BUILD_DEV)
+    target_compile_definitions(migraphx PUBLIC -DBUILD_DEV)
+endif()
+
 rocm_export_targets(
   TARGETS migraphx::migraphx_c
   NAMESPACE migraphx::

src/api/CMakeLists.txt

@@ -26,10 +26,13 @@ add_library(migraphx_c
     api.cpp
 )
 set_target_properties(migraphx_c PROPERTIES EXPORT_NAME c)
-rocm_set_soversion(migraphx_c 3.0)
+
+# migraphx_c is stable API interface library. SO version of this should be 
+# bumped when binary compatibility is broken. 
+rocm_set_soversion(migraphx_c 3.0)   
 
 rocm_clang_tidy_check(migraphx_c)
-target_link_libraries(migraphx_c PRIVATE migraphx migraphx_tf migraphx_onnx migraphx_all_targets)
+target_link_libraries(migraphx_c PRIVATE migraphx migraphx_tf migraphx_onnx)
 
 rocm_install_targets(
   TARGETS migraphx_c

src/api/api.cpp

@@ -32,7 +32,6 @@
 #include <migraphx/register_target.hpp>
 #include <migraphx/generate.hpp>
 #include <migraphx/quantization.hpp>
-#include <migraphx/ref/target.hpp>
 #include <migraphx/load_save.hpp>
 #include <migraphx/make_op.hpp>
 #include <migraphx/register_op.hpp>

src/common.cpp

@@ -89,8 +89,8 @@ std::vector<shape::dynamic_dimension> compute_broadcasted_dyn_dims(shape s0, sha
             }
             else if(a == 1 or b == 1)
             {
-                // setting opt to 0, may need to be changed
-                return shape::dynamic_dimension{std::max(a.min, b.min), std::max(a.max, b.max), 0};
+                // setting optimals to empty, may need to be changed
+                return shape::dynamic_dimension{std::max(a.min, b.min), std::max(a.max, b.max)};
             }
             else
             {
@@ -148,10 +148,8 @@ shape common_shape(const std::vector<shape>& shapes)
     return {compute_common_types(shapes), compute_common_lens(shapes)};
 }
 
-instruction_ref insert_common_op(module& m,
-                                 instruction_ref ins,
-                                 const operation& op,
-                                 std::vector<instruction_ref> inputs)
+std::vector<instruction_ref>
+insert_common_args(module& m, instruction_ref ins, std::vector<instruction_ref> inputs)
 {
     if(std::any_of(
            inputs.cbegin(), inputs.cend(), [](auto input) { return input->get_shape().dynamic(); }))
@@ -210,7 +208,20 @@ instruction_ref insert_common_op(module& m,
             return input;
         });
     }
-    return m.insert_instruction(ins, op, inputs);
+    return inputs;
+}
+
+std::vector<instruction_ref> add_common_args(module& m, std::vector<instruction_ref> inputs)
+{
+    return insert_common_args(m, m.end(), std::move(inputs));
+}
+
+instruction_ref insert_common_op(module& m,
+                                 instruction_ref ins,
+                                 const operation& op,
+                                 std::vector<instruction_ref> inputs)
+{
+    return m.insert_instruction(ins, op, insert_common_args(m, ins, std::move(inputs)));
 }
 
 instruction_ref add_common_op(module& m, const operation& op, std::vector<instruction_ref> inputs)

src/cpp_generator.cpp

@@ -106,6 +106,13 @@ cpp_generator::function& cpp_generator::function::set_generic_types(const module
     return *this;
 }
 
+cpp_generator::function& cpp_generator::function::add_generic_param(const std::string& pname)
+{
+    params.push_back({pname, "T" + pname});
+    tparams.push_back("class T" + pname);
+    return *this;
+}
+
 struct cpp_generator_impl
 {
     std::stringstream fs{};
@@ -182,7 +189,8 @@ std::string cpp_generator::generate_point_op(const operation& op,
 
 std::string cpp_generator::str() const { return impl->fs.str(); }
 
-cpp_generator::function cpp_generator::generate_module(const module& m)
+cpp_generator::function cpp_generator::generate_module(const module& m,
+                                                       const generate_module_callback& g)
 {
     function f;
     auto name = transform_string(m.name(), [](char c) {
@@ -195,13 +203,7 @@ cpp_generator::function cpp_generator::generate_module(const module& m)
             if(ins->name() == "@literal")
                 return shape::cpp_type(ins->get_shape().type()) + "(" +
                        ins->get_literal().to_string() + ")";
-            std::vector<std::string> args;
-            std::transform(ins->inputs().begin(),
-                           ins->inputs().end(),
-                           std::back_inserter(args),
-                           [&](auto i) { return names.at(i); });
-
-            auto s = this->generate_point_op(ins->get_operator(), args);
+            auto s = g(ins, names);
             if(impl->fresult)
                 return impl->fresult(ins->get_shape()) + '(' + s + ')';
             else
@@ -210,6 +212,24 @@ cpp_generator::function cpp_generator::generate_module(const module& m)
     return f;
 }
 
+std::vector<std::string>
+cpp_generator::to_args(const std::vector<instruction_ref>& inputs,
+                       const std::unordered_map<instruction_ref, std::string>& names)
+{
+    std::vector<std::string> args;
+    std::transform(inputs.begin(), inputs.end(), std::back_inserter(args), [&](auto i) {
+        return names.at(i);
+    });
+    return args;
+}
+
+cpp_generator::function cpp_generator::generate_module(const module& m)
+{
+    return this->generate_module(m, [&](auto ins, const auto& names) {
+        return this->generate_point_op(ins->get_operator(), to_args(ins->inputs(), names));
+    });
+}
+
 std::string cpp_generator::create_function(const cpp_generator::function& f)
 {
     impl->function_count++;

src/driver/CMakeLists.txt

@@ -42,6 +42,7 @@ add_custom_command(
 )
 set_directory_properties(PROPERTIES ADDITIONAL_MAKE_CLEAN_FILES ${CMAKE_RUNTIME_OUTPUT_DIRECTORY}/driver)
 rocm_clang_tidy_check(driver)
+
 target_link_libraries(driver migraphx_all_targets migraphx_onnx migraphx_tf)
 
 rocm_install_targets(