Merge branch 'main' into fix-peano-shift-ambiguity

.github/workflows/buildAndTest.yml

@@ -52,6 +52,7 @@ jobs:
       - name: Install Python packages
         run: |
           pip install -r python/requirements.txt
+          HOST_MLIR_PYTHON_PACKAGE_PREFIX=aie pip install -r python/requirements_extras.txt
 
       - name: Install packages
         run: |

.github/workflows/buildAndTestAieTools.yml

@@ -86,6 +86,7 @@ jobs:
             python -m venv aie-venv
             source aie-venv/bin/activate
             pip install -r python/requirements.txt
+            HOST_MLIR_PYTHON_PACKAGE_PREFIX=aie pip install -r python/requirements_extras.txt
 
             VERSION=$(utils/clone-llvm.sh --get-wheel-version)
             pip -q download mlir==$VERSION \

.github/workflows/buildAndTestAieToolsHsaBuildOnly.yml

@@ -111,6 +111,7 @@ jobs:
             python -m venv aie-venv
             source aie-venv/bin/activate
             pip install -r python/requirements.txt
+            HOST_MLIR_PYTHON_PACKAGE_PREFIX=aie pip install -r python/requirements_extras.txt
 
             VERSION=$(utils/clone-llvm.sh --get-wheel-version)
             pip -q download mlir==$VERSION \

.github/workflows/buildAndTestMulti.yml

@@ -94,6 +94,7 @@ jobs:
       - name: Install Python packages
         run: |
           pip install -r python/requirements.txt
+          HOST_MLIR_PYTHON_PACKAGE_PREFIX=aie pip install -r python/requirements_extras.txt
 
       - name: Setup Cpp
         uses: aminya/setup-cpp@v1

.github/workflows/buildAndTestPythons.yml

@@ -53,6 +53,7 @@ jobs:
       - name: Install Python packages
         run: |
           pip install -r python/requirements.txt
+          HOST_MLIR_PYTHON_PACKAGE_PREFIX=aie pip install -r python/requirements_extras.txt
 
       - name: Install packages
         run: |

.github/workflows/buildAndTestRyzenAI.yml

@@ -38,7 +38,7 @@ concurrency:
 env:
   DEBIAN_FRONTEND: noninteractive
   XILINXD_LICENSE_FILE: /opt/xilinx/Xilinx.lic
-  VITIS: /opt/Xilinx/Vitis/2023.2
+  VITIS: /opt/ryzen_ai-1.3.0/vitis_aie_essentials
   CMAKE_ARGS: |
             -DCMAKE_CXX_COMPILER_LAUNCHER=ccache \
             -DCMAKE_C_COMPILER_LAUNCHER=ccache \
@@ -79,6 +79,7 @@ jobs:
           python -m venv aie-venv
           source aie-venv/bin/activate
           pip install -r python/requirements.txt
+          HOST_MLIR_PYTHON_PACKAGE_PREFIX=aie pip install -r python/requirements_extras.txt
           pip install -r python/requirements_ml.txt
           pip install jupyter
           sed -i.bak 's/OUTPUT_TIMEOUT = 10/OUTPUT_TIMEOUT = 100/g' \
@@ -96,7 +97,7 @@ jobs:
           mkdir build
           pushd build
 
-          export PATH=/opt/Xilinx/Vitis/2023.2/bin:/opt/Xilinx/Vitis/2023.2/aietools/bin:$PATH
+          export PATH=$VITIS/bin:$VITIS/aietools/bin:$PATH
           cmake .. -G Ninja \
             -DPython3_EXECUTABLE=$(which python) \
             -DLLVM_EXTERNAL_LIT=$(which lit) \
@@ -110,7 +111,7 @@ jobs:
 
           popd
 
-  build-examples:
+  build-quick-setup:
     name: Run Examples on Ryzen AI
     runs-on: amd7940hs
     steps:
@@ -133,35 +134,15 @@ jobs:
 
           pip cache purge
           source /opt/xilinx/xrt/setup.sh
-          python -m venv aie-venv
-          source aie-venv/bin/activate
-          pip install -r python/requirements.txt
-          pip install -r python/requirements_ml.txt
-          pip install jupyter
-          sed -i.bak 's/OUTPUT_TIMEOUT = 10/OUTPUT_TIMEOUT = 100/g' \
-            $(python -c 'import site; print(site.getsitepackages()[0])')/jupyter_client/runapp.py
-
-          VERSION=$(utils/clone-llvm.sh --get-wheel-version)
-          pip -q download mlir==$VERSION \
-            -f https://github.com/Xilinx/mlir-aie/releases/expanded_assets/mlir-distro
-          unzip -q mlir-*.whl
-          # I have no clue why but the system clock on GHA containers is like 12 hours ahead.
-          # That means wheels have file with time stamps in the future which makes ninja loop
-          # forever when configuring. Set the time to some arbitrary stamp in the past just to be safe.
-          find mlir -exec touch -a -m -t 201108231405.14 {} \;
+          export PATH=$VITIS/bin:$VITIS/aietools/bin:$PATH
+          source utils/quick_setup.sh
+          # quick_setup changes directory to programming_examples, so we need to return to mlir-aie
+          cd ..
+          ./utils/build-mlir-aie-from-wheels.sh ./my_install/mlir build install ./my_install/llvm-aie
 
-          mkdir build
+          # build is created by the build-mlir-aie-from-wheels.sh script
           pushd build
 
-          export PATH=$VITIS/bin:$VITIS/aietools/bin:$PATH
-          cmake .. -G Ninja \
-            -DPython3_EXECUTABLE=$(which python) \
-            -DLLVM_EXTERNAL_LIT=$(which lit) \
-            -DCMAKE_INSTALL_PREFIX=$PWD/../mlir_aie \
-            -DCMAKE_MODULE_PATH=$PWD/../cmake/modulesXilinx \
-            -DMLIR_DIR=$PWD/../mlir/lib/cmake/mlir \
-            $CMAKE_ARGS
-
           ninja install
           ninja check-reference-designs
           ninja check-programming-guide

.github/workflows/buildAndTestRyzenAISw.yml

@@ -0,0 +1,134 @@
+name: Build and Test with Ryzen AI Sofware
+
+on:
+  push:
+    branches:
+      - main
+      - ryzen-ai-sw-test
+  pull_request:
+  workflow_dispatch:
+    inputs:
+      AIE_COMMIT:
+        description: 'which aie commit to build'
+        type: string
+        required: false
+        default: ''
+      LIT_FILTER:
+        description: 'LIT_FILTER for tests (regex of paths in test/)'
+        type: string
+        required: false
+        default: ''
+
+defaults:
+  run:
+    shell: bash
+
+concurrency:
+  # A PR number if a pull request and otherwise the commit hash. This cancels
+  # queued and in-progress runs for the same PR (presubmit) or commit
+  # (postsubmit).
+  group: ci-build-test-ryzenai-tools-linux-${{ github.event.number || github.sha }}
+  cancel-in-progress: true
+
+env:
+  DEBIAN_FRONTEND: noninteractive
+
+jobs:
+  build-repo:
+    name: Build and Test with Ryzen AI Software
+
+    runs-on: ubuntu-latest
+
+    steps:
+
+      - name: Free disk space
+        uses: descriptinc/free-disk-space@main
+        with:
+          tool-cache: true
+          android: true
+          dotnet: true
+          haskell: true
+          large-packages: true
+          swap-storage: false
+
+      - name: Docker prune
+        shell: bash
+        run: |
+          docker system prune -a -f
+
+      - uses: actions/checkout@v4
+        with:
+          submodules: "true"
+      - uses: uraimo/run-on-arch-action@v2.5.0
+        name: Build and Test
+        id: runcmd
+        with:
+          distro: none
+          base_image: ghcr.io/xilinx/mlir-aie/ubuntu22-ryzenai-1.3.0ea
+          githubToken: ${{ github.token }}
+          dockerRunArgs: |
+            --mac-address ${{ secrets.XILINX_MAC }}
+          run: |
+
+            MLIR_DIR=$PWD
+            pushd /workspace
+
+            # this is the inverse of `base64 -w 1000000 Xilinx.lic`
+            # the -w ("wrap after 1000000 cols") is so that there are no spaces in the XILINX_LIC env var
+            echo -n "${{ secrets.XILINX_LIC }}" | base64 --decode > /workspace/Xilinx.lic
+            export XILINXD_LICENSE_FILE=/workspace/Xilinx.lic
+
+            ccrypt -d -K ${{ secrets.RYZEN_AI_SW_KEY }} /workspace/ryzen_ai-1.3.0ea1.tgz.cpt
+            tar xvf /workspace/ryzen_ai-1.3.0ea1.tgz
+            pushd /workspace/ryzen_ai-1.3.0
+            ./install_ryzen_ai_1_3.sh -a yes -p /workspace/venv -l
+            popd
+
+            export LD_LIBRARY_PATH=
+            export PYTHONPATH=
+            source /workspace/venv/bin/activate
+            source /opt/xilinx/xrt/setup.sh
+            popd
+            
+            if [ x"${{ inputs.AIE_COMMIT }}" != x"" ]; then
+              git reset --hard ${{ inputs.AIE_COMMIT }}
+            fi
+
+            pip install onnxruntime==1.17.0
+            pip install -r python/requirements.txt
+            pip install -r python/requirements_ml.txt
+            HOST_MLIR_PYTHON_PACKAGE_PREFIX=aie pip install -r python/requirements_extras.txt
+
+            VERSION=$(utils/clone-llvm.sh --get-wheel-version)
+
+            cd /workspace
+            pip -q download mlir==$VERSION \
+              -f https://github.com/Xilinx/mlir-aie/releases/expanded_assets/mlir-distro
+            unzip -q mlir-*.whl
+
+            mkdir -p /workspace/install
+            mkdir -p /workspace/build
+
+            pushd /workspace/build
+
+            cmake $MLIR_DIR -G Ninja \
+                -DCMAKE_CXX_COMPILER_LAUNCHER=ccache \
+                -DCMAKE_C_COMPILER_LAUNCHER=ccache \
+                -DCMAKE_EXE_LINKER_FLAGS_INIT="-fuse-ld=lld" \
+                -DCMAKE_MODULE_LINKER_FLAGS_INIT="-fuse-ld=lld" \
+                -DCMAKE_SHARED_LINKER_FLAGS_INIT="-fuse-ld=lld" \
+                -DXRT_ROOT=/opt/xilinx/xrt \
+                -DAIE_ENABLE_PYTHON_PASSES=OFF \
+                -DAIE_ENABLE_XRT_PYTHON_BINDINGS=ON \
+                -DAIE_INCLUDE_INTEGRATION_TESTS=OFF \
+                -DAIE_VITIS_COMPOONENTS=AIE2 \
+                -DPython3_EXECUTABLE=$(which python) \
+                -DLLVM_EXTERNAL_LIT=$(which lit) \
+                -DCMAKE_INSTALL_PREFIX=/workspace/install \
+                -DCMAKE_MODULE_PATH=$MLIR_DIR/cmake/modulesXilinx \
+                -DMLIR_DIR=/workspace/mlir/lib/cmake/mlir
+
+            ninja install
+            ninja check-aie
+
+            popd

.github/workflows/buildRyzenWheels.yml

@@ -52,6 +52,7 @@ jobs:
           python -m venv aie-venv
           source aie-venv/bin/activate
           pip install -r python/requirements.txt
+          HOST_MLIR_PYTHON_PACKAGE_PREFIX=aie pip install -r python/requirements_extras.txt
 
           VERSION=$(utils/clone-llvm.sh --get-wheel-version)
           pip -q download mlir==$VERSION \
@@ -116,9 +117,7 @@ jobs:
       fail-fast: false
       matrix:
         python_version: [
-#          "3.8", "3.9",
-          "3.10",
-#          "3.11", "3.12"
+          "3.10", "3.12",
         ]
 
     steps:
@@ -158,6 +157,7 @@ jobs:
           python -m venv aie-venv
           source aie-venv/bin/activate
           pip install -r python/requirements.txt
+          HOST_MLIR_PYTHON_PACKAGE_PREFIX=aie pip install -r python/requirements_extras.txt
           source aie-venv/bin/activate
           
           export MLIR_INSTALL_ABS_PATH=$PWD/mlir

.github/workflows/lintAndFormat.yml

@@ -43,6 +43,7 @@ jobs:
           
           pip install cmake==3.27.9 numpy psutil pybind11 rich pkginfo lit PyYAML requests
           pip install -r python/requirements.txt
+          HOST_MLIR_PYTHON_PACKAGE_PREFIX=aie pip install -r python/requirements_extras.txt
 
       - name: Get MLIR
         id: mlir-wheels
@@ -204,6 +205,7 @@ jobs:
         run: |
           pip install cmake numpy psutil pybind11 rich lit
           pip install -r python/requirements.txt
+          HOST_MLIR_PYTHON_PACKAGE_PREFIX=aie pip install -r python/requirements_extras.txt
 
       - name: Install packages
         run: sudo apt-get install -y ninja-build clang lld llvm libelf-dev
@@ -236,6 +238,7 @@ jobs:
         if: steps.changed-files.outputs.changed-files != ''
         run: |
           pip install -r python/requirements.txt
+          HOST_MLIR_PYTHON_PACKAGE_PREFIX=aie pip install -r python/requirements_extras.txt
 
       - name: Build and generate coverage (Release)
         if: steps.changed-files.outputs.changed-files != ''

.github/workflows/mlirAIEDistro.yml

@@ -228,6 +228,7 @@ jobs:
         export PIP_NO_BUILD_ISOLATION=false
         
         pip install -r requirements.txt
+        HOST_MLIR_PYTHON_PACKAGE_PREFIX=aie pip install -r python/requirements_extras.txt
         pip install importlib-metadata
         CIBW_ARCHS=${{ matrix.ARCH }} MATRIX_OS=${{ matrix.OS }} ./scripts/download_mlir.sh
         
@@ -360,6 +361,7 @@ jobs:
         with:
           sparse-checkout: |
             python/requirements.txt
+            python/requirements_extras.txt
 
       - uses: actions/download-artifact@v3
         with:
@@ -374,6 +376,7 @@ jobs:
         shell: bash
         run: |
           pip install -r python/requirements.txt
+          HOST_MLIR_PYTHON_PACKAGE_PREFIX=aie pip install -r python/requirements_extras.txt
           unzip -o -q dist/mlir_aie\*.whl
           
           if [ x"${{ matrix.ENABLE_RTTI }}" == x"ON" ]; then
@@ -383,6 +386,8 @@ jobs:
           fi
           
           python -c 'import aie.ir'
+          python -c 'import aie.extras'
+          python -c 'import aie.helpers'
 
   upload_wheels:
 

aie_kernels/CMakeLists.txt

@@ -13,7 +13,11 @@ function(add_aie_kernels arch)
     if(EXISTS "${PROJECT_SOURCE_DIR}/tools/chess-clang/xchesscc_wrapper")
         add_custom_target(${arch}_kernels ALL DEPENDS ${CMAKE_CURRENT_BINARY_DIR}/mm.o)
         add_custom_command(OUTPUT ${CMAKE_CURRENT_BINARY_DIR}/mm.o
-                      COMMAND ${PROJECT_SOURCE_DIR}/tools/chess-clang/xchesscc_wrapper ${arch} -I ${VITIS_AIETOOLS_DIR}/include -c ${CMAKE_CURRENT_SOURCE_DIR}/mm.cc -o ${CMAKE_CURRENT_BINARY_DIR}/mm.o
+                      COMMAND ${PROJECT_SOURCE_DIR}/tools/chess-clang/xchesscc_wrapper ${arch}
+                        -aietools ${VITIS_AIETOOLS_DIR}
+                        -I ${VITIS_AIETOOLS_DIR}/include
+                        -c ${CMAKE_CURRENT_SOURCE_DIR}/mm.cc
+                        -o ${CMAKE_CURRENT_BINARY_DIR}/mm.o
                       DEPENDS ${CMAKE_CURRENT_SOURCE_DIR}/mm.cc)
         add_dependencies(aie-kernels ${arch}_kernels)
 

aie_kernels/aie2/cascade_mm.cc

@@ -34,7 +34,7 @@ void matmul_scalar_cascade_put_only(T_in *a, T_in *b, T_out *c) {
         running_sum += a[row * colA + i] * b[i * colB + col];
       }
       v16int32 v16 = undef_v16int32();
-      v16 = upd_elem(v16, 0, running_sum);
+      v16 = upd_elem(v16, 0, (int)running_sum);
       put_mcd(v16);
     }
   }
@@ -51,7 +51,7 @@ void matmul_scalar_cascade_get_only(T_in *a, T_in *b, T_out *c) {
         running_sum += a[row * colA + i] * b[i * colB + col];
       }
       v16int32 v16 = get_scd_v16int32();
-      running_sum += ext_elem(v16, 0);
+      running_sum += ext_elem(v16, 0U);
       c[row * colB + col] += running_sum;
     }
   }
@@ -68,8 +68,8 @@ void matmul_scalar_cascade_put_get(T_in *a, T_in *b, T_out *c) {
         running_sum += a[row * colA + i] * b[i * colB + col];
       }
       v16int32 v16 = get_scd_v16int32();
-      running_sum += ext_elem(v16, 0);
-      v16 = upd_elem(v16, 0, running_sum);
+      running_sum += ext_elem(v16, 0U);
+      v16 = upd_elem(v16, 0, (int)running_sum);
       put_mcd(v16);
     }
   }

aie_kernels/aie2/reduce_add.cc

@@ -27,15 +27,15 @@ static void _reduce_add_vector(int32_t *restrict in, int32_t *restrict out,
       running_total = test;
     }
   after_vector = running_total;
-  v16int32 first = shift_bytes(after_vector, after_vector, 32);
+  v16int32 first = shift_bytes(after_vector, after_vector, 32U);
   v16int32 second = add(after_vector, first);
-  v16int32 second_shift = shift_bytes(second, second, 16);
+  v16int32 second_shift = shift_bytes(second, second, 16U);
   v16int32 third = add(second, second_shift);
-  v16int32 third_shift = shift_bytes(third, third, 8);
+  v16int32 third_shift = shift_bytes(third, third, 8U);
   v16int32 fourth = add(third, third_shift);
-  v16int32 fourth_shift = shift_bytes(fourth, fourth, 4);
+  v16int32 fourth_shift = shift_bytes(fourth, fourth, 4U);
   v16int32 fifth = add(fourth, fourth_shift);
-  int32_t last = extract_elem(fifth, 0);
+  int32_t last = extract_elem(fifth, 0U);
   *(int32_t *)out = last;
   return;
 }