Merge branch 'main' into dont_force_memspace1_toaie

.github/workflows/ci.yml → .github/workflows/ci-linux.yml

@@ -1,4 +1,4 @@
-name: CI
+name: CI Linux
 
 on:
   workflow_call:
@@ -9,72 +9,44 @@ on:
     branches:
       - main
 
-  schedule:
-      # At minute 0 past every 2nd hour. (see https://crontab.guru)
-      # this job is to keep the ccache cache warm
-      # specifically for the free windows runner
-    - cron: '0 */2 * * *'
-
 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-${{ github.event.number || github.sha }}
+  group: ci-build-test-cpp-linux-${{ github.event.number || github.sha }}
   cancel-in-progress: true
 
-defaults:
-  run:
-    # force bash for windows
-    shell: bash
-
 jobs:
   build_and_ctest:
-    name: Build and Test (${{ matrix.os }}, ASSERTIONS)
-    runs-on: ${{ matrix.runs-on }}
+    name: Build and Test (linux, ASSERTIONS)
+    runs-on: nod-ai-shared-cpubuilder-manylinux-x86_64
     strategy:
-      fail-fast: false
-      matrix:
-        include:
-          - os: linux
-            runs-on: nod-ai-shared-cpubuilder-manylinux-x86_64
-          - os: windows
-            runs-on: windows-2019
+      fail-fast: true
     env:
       CACHE_DIR: ${{ github.workspace }}/.container-cache
     steps:
+      - name: Set unified TZ
+        uses: szenius/set-timezone@v2.0
+        with:
+          # this is an arbitrary choice
+          timezoneLinux: "Asia/Singapore"
+          timezoneMacos: "Asia/Singapore"
+          timezoneWindows: "Singapore Standard Time"
+
       - name: Configure local git mirrors
-        if: ${{ matrix.runs-on == 'nod-ai-shared-cpubuilder-manylinux-x86_64' }}
         run: |
           /gitmirror/scripts/trigger_update_mirrors.sh
           /gitmirror/scripts/git_config.sh
-
       - name: "Checking out repository"
         uses: actions/checkout@8f4b7f84864484a7bf31766abe9204da3cbe65b3 # v3.5.0
         with:
           submodules: recursive
 
       - name: Install static libs
-        if: ${{ matrix.runs-on == 'nod-ai-shared-cpubuilder-manylinux-x86_64' }}
         run: |
           dnf install -y almalinux-release-devel
           yum remove -y openssl-devel zlib-devel || true
           yum install -y protobuf-devel protobuf-compiler
-
-      - name: Setup Cpp
-        uses: aminya/setup-cpp@v1
-        if: ${{ matrix.os == 'windows' }}
-        with:
-          compiler: msvc
-          vcvarsall: true
-          cmake: true
-          ninja: true
-          ccache: true
-
-      - uses: actions/setup-python@v4
-        if: ${{ matrix.os == 'windows' }}
-        with:
-          python-version: '3.11'
-
       - name: Sync source deps
         run: |
           python ./sync_deps.py
@@ -87,8 +59,8 @@ jobs:
         uses: actions/cache/restore@v3
         with:
           path: ${{ env.CACHE_DIR }}
-          key: "${{ matrix.os }}-build-test-cpp-asserts-v1-${{ github.sha }}"
-          restore-keys: "${{ matrix.os }}-build-test-cpp-"
+          key: linux-build-test-cpp-asserts-manylinux-v2-${{ github.sha }}
+          restore-keys: linux-build-test-cpp-
 
       - name: Build packages
         run: |
@@ -100,31 +72,33 @@ jobs:
         run: |
           rm -f iree-install/bin/clang*
           rm -f iree-install/bin/llvm-link*
-          tar cf iree-${{ matrix.os }}-dist.tar -C iree-install . -C ../iree-build tools/testing/e2e/iree-e2e-matmul-test
+          tar cf iree-dist-linux.tar -C iree-install . -C ../iree-build tools/testing/e2e/iree-e2e-matmul-test
 
       - name: Upload artifacts
         uses: actions/upload-artifact@v4
         if: ${{ !cancelled() }}
         with:
-          name: "${{ matrix.os }}_x86_64_release_packages"
-          path: "iree-${{ matrix.os }}-dist.tar"
+          name: linux_x86_64_release_packages
+          path: iree-dist-linux.tar
           if-no-files-found: warn
 
       - name: Save cache
         uses: actions/cache/save@v3
         if: ${{ !cancelled() }}
         with:
           path: ${{ env.CACHE_DIR }}
-          key: "${{ matrix.os }}-build-test-cpp-asserts-v1-${{ github.sha }}"
+          key: linux-build-test-cpp-asserts-manylinux-v2-${{ github.sha }}
 
   test_linux:
     name: E2E Test linux
-    runs-on: amd7940hs
     needs: build_and_ctest
     strategy:
-      fail-fast: true
+      fail-fast: false
+      matrix:
+        runs-on: [linux-phoenix-20240606, linux-phoenix-20240819]
+    runs-on: ${{ matrix.runs-on }}
     env:
-      XILINXD_LICENSE_FILE: /home/svcnod/Xilinx.lic
+      XILINXD_LICENSE_FILE: /opt/xilinx/Xilinx.lic
     steps:
       - name: "Checking out repository" # for test scripts
         uses: actions/checkout@8f4b7f84864484a7bf31766abe9204da3cbe65b3 # v3.5.0
@@ -139,7 +113,7 @@ jobs:
       - name: Extract artifact
         run: |
           mkdir iree-install
-          tar -xf iree-linux-dist.tar -C iree-install
+          tar -xf iree-dist-linux.tar -C iree-install
           bash build_tools/download_peano.sh
 
       - name: Create venv and install dependencies

.github/workflows/ci-windows.yml

@@ -0,0 +1,106 @@
+name: CI Windows
+
+on:
+  workflow_call:
+  workflow_dispatch:
+  pull_request:
+  merge_group:
+  push:
+    branches:
+      - main
+
+  schedule:
+      # At minute 0 past every 4nd hour. (see https://crontab.guru)
+      # this job is to keep the ccache cache warm
+    - cron: '0 */4 * * *'
+
+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-cpp-windows-${{ github.event.number || github.sha }}
+  cancel-in-progress: true
+
+defaults:
+  run:
+    # force bash for windows
+    shell: bash
+
+jobs:
+  build_and_ctest:
+    name: Build and Test (windows, ASSERTIONS)
+    runs-on: windows-2019
+    strategy:
+      fail-fast: true
+    env:
+      CACHE_DIR: ${{ github.workspace }}/.container-cache
+    steps:
+      - name: Set unified TZ
+        uses: szenius/set-timezone@v2.0
+        with:
+          # this is an arbitrary choice
+          timezoneLinux: "Asia/Singapore"
+          timezoneMacos: "Asia/Singapore"
+          timezoneWindows: "Singapore Standard Time"
+
+      - name: "Checking out repository"
+        uses: actions/checkout@8f4b7f84864484a7bf31766abe9204da3cbe65b3 # v3.5.0
+        with:
+          submodules: recursive
+
+      - name: Setup Cpp
+        uses: aminya/setup-cpp@v1
+        with:
+          compiler: msvc
+          vcvarsall: true
+          cmake: true
+          ninja: true
+          ccache: true
+
+      - uses: actions/setup-python@v4
+        with:
+          python-version: '3.11'
+
+      - name: Sync source deps
+        run: |
+          python ./sync_deps.py
+
+      - name: Python deps
+        run: |
+          pip install "numpy<2" pyyaml "pybind11[global]==2.10.3" nanobind
+
+      - name: Enable cache
+        uses: actions/cache/restore@v3
+        with:
+          path: ${{ env.CACHE_DIR }}
+          # without datetime stamps you'll get collisions for the cache warming runs
+          # ("Failed to save: Unable to reserve cache with key ..., another job may be creating this cache.")
+          key: windows-build-test-cpp-asserts-v1-${{ github.sha }}-${{ github.event.repository.updated_at }}
+          restore-keys: windows-build-test-cpp-
+
+      - name: Build packages
+        run: |
+          export cache_dir="${{ env.CACHE_DIR }}"
+          bash build_tools/ci/build_test_cpp.sh
+
+      - name: Create artifacts
+        if: ${{ !cancelled() }}
+        run: |
+          rm -f iree-install/bin/clang*
+          rm -f iree-install/bin/llvm-link*
+          tar cf iree-dist-windows.tar -C iree-install . -C ../iree-build tools/testing/e2e/iree-e2e-matmul-test.exe
+
+      - name: Upload artifacts
+        uses: actions/upload-artifact@v4
+        if: ${{ !cancelled() }}
+        with:
+          name: windows_x86_64_release_packages
+          path: iree-dist-windows.tar
+          if-no-files-found: warn
+
+      - name: Save cache
+        uses: actions/cache/save@v3
+        if: ${{ !cancelled() }}
+        with:
+          path: ${{ env.CACHE_DIR }}
+          key: windows-build-test-cpp-asserts-v1-${{ github.sha }}-${{ github.event.repository.updated_at }}

build_tools/ci/build_test_cpp.sh

@@ -26,6 +26,8 @@ echo "Using python: $python"
 
 # https://stackoverflow.com/a/8597411/9045206
 # note: on windows (git-bash) result is "msys"
+# well only if you have apparently the right version of git-bash installed
+# https://stackoverflow.com/a/72164385
 if [[ "$OSTYPE" == "linux-gnu"* ]]; then
   export CMAKE_TOOLCHAIN_FILE="$this_dir/linux_default_toolchain.cmake"
   export CC=clang
@@ -86,7 +88,12 @@ cmake --build "$build_dir" --target iree-install-dist
 
 echo "CTest"
 echo "-----"
-ctest --test-dir "$build_dir" -R amd-aie --output-on-failure -j
+if [[ "$OSTYPE" == "linux-gnu"* ]]; then
+  ctest --test-dir "$build_dir" -R amd-aie --output-on-failure -j
+else
+  # hack while windows is flaky to get past failing tests
+  ctest --test-dir "$build_dir" -R amd-aie --output-on-failure -j --repeat until-pass:5
+fi
 
 # Show ccache stats.
 ccache --show-stats

build_tools/ci/cpu_comparison/run_test.py

@@ -306,6 +306,15 @@ def __init__(
         if xrt_hash:
             self.xrt_hash = xrt_hash[0]
 
+        xdna_datetime = re.findall(
+            # eg 2.18.0_20240606
+            r"amdxdna\s+:\s\d\.\d+\.\d+_(\d+)",
+            xrt_info,
+            flags=re.MULTILINE | re.IGNORECASE,
+        )
+        if xdna_datetime:
+            self.xdna_datetime = int(xdna_datetime[0])
+
         # Try and get the peano commit hash. This is a bit of a hack, if it fails
         # peano_commit_has is left as "undetermined".
         self.peano_commit_hash = "undetermined"
@@ -589,38 +598,45 @@ def run(self, config):
         test_files_dir = config.file_dir / "test_files"
         output_dir = config.output_dir
 
-        for name in [
-            "two_matmul_switching",
-            "matmul_f32_8_8_4",
-            "matmul_f32_8_4_8",
-        ]:
-            aie_vs_llvm_cpu(config, test_files_dir / f"{name}.mlir")
-
-        aie_vs_llvm_cpu(
-            config,
-            test_files_dir / "three_matmuls.mlir",
-            function_name="three_$mm$",
-        )
-
-        # Test(s) of the form matmul(A,B) where A:MxK, B:KxN
-        test_name = output_dir / "test_from_template.mlir"
-        template_name = matmul_template_dir / "matmul_MxK_KxN.mlir"
-        generate_matmul_test(test_name, template_name, 32, 32, 64, "bf16", "f32")
-        aie_vs_llvm_cpu(config, test_name)
-
-        # Test(s) of the form matmul(A,B) + C where A:MxK, B:KxN, C:N
-        test_name = output_dir / "test_from_template_bias_N.mlir"
-        template_name = matmul_template_dir / "matmul_bias_MxK_KxN_N.mlir"
-        generate_matmul_test(test_name, template_name, 1024, 1024, 512, "bf16", "f32")
-        aie_vs_llvm_cpu(config, test_name, tile_pipeline="pack-peel", use_ukernel=True)
-        aie_vs_llvm_cpu(config, test_name, tile_pipeline="pack-peel", use_ukernel=False)
-
         # Test(s) of the form matmul(A,B) + C where A:MxK, B:KxN, C:MxN
         test_name = output_dir / "test_from_template_full_bias.mlir"
         template_name = matmul_template_dir / "matmul_bias_MxK_KxN_MxN.mlir"
         generate_matmul_test(test_name, template_name, 128, 128, 256, "i32", "i32")
         aie_vs_llvm_cpu(config, test_name, tile_pipeline="pack-peel", rtol=0, atol=0)
 
+        if config.xdna_datetime and config.xdna_datetime < 20240819:
+            for name in [
+                "two_matmul_switching",
+                "matmul_f32_8_8_4",
+                "matmul_f32_8_4_8",
+            ]:
+                aie_vs_llvm_cpu(config, test_files_dir / f"{name}.mlir")
+
+            aie_vs_llvm_cpu(
+                config,
+                test_files_dir / "three_matmuls.mlir",
+                function_name="three_$mm$",
+            )
+
+            # Test(s) of the form matmul(A,B) where A:MxK, B:KxN
+            test_name = output_dir / "test_from_template.mlir"
+            template_name = matmul_template_dir / "matmul_MxK_KxN.mlir"
+            generate_matmul_test(test_name, template_name, 32, 32, 64, "bf16", "f32")
+            aie_vs_llvm_cpu(config, test_name)
+
+            # Test(s) of the form matmul(A,B) + C where A:MxK, B:KxN, C:N
+            test_name = output_dir / "test_from_template_bias_N.mlir"
+            template_name = matmul_template_dir / "matmul_bias_MxK_KxN_N.mlir"
+            generate_matmul_test(
+                test_name, template_name, 1024, 1024, 512, "bf16", "f32"
+            )
+            aie_vs_llvm_cpu(
+                config, test_name, tile_pipeline="pack-peel", use_ukernel=True
+            )
+            aie_vs_llvm_cpu(
+                config, test_name, tile_pipeline="pack-peel", use_ukernel=False
+            )
+
 
 class SmokeSet(TestSet):
     def __init__(self):

compiler/plugins/target/AMD-AIE/aie/AMDAIECreatePathFindFlows.cpp

@@ -6,15 +6,12 @@
 
 #include <algorithm>
 #include <cassert>
-#include <list>
 #include <set>
 
 #include "AIEDialect.h"
 #include "Passes.h"
 #include "iree-amd-aie/aie_runtime/iree_aie_router.h"
 #include "iree-amd-aie/aie_runtime/iree_aie_runtime.h"
-#include "llvm/ADT/DenseMapInfo.h"
-#include "llvm/Support/raw_os_ostream.h"
 #include "mlir/Pass/Pass.h"
 #include "mlir/Transforms/DialectConversion.h"