From bf99fb8b72011cc3297599c918861d1a325ce7d1 Mon Sep 17 00:00:00 2001
From: Maksim Levental <maksim.levental@gmail.com>
Date: Thu, 29 Aug 2024 15:00:08 -0700
Subject: [PATCH 1/5] Simplify print_ir_aie2xclbin.sh (#724)

We don't actually need vitis and xrt should be sourced outside of this
script like in the ci.

Co-authored-by: Vivian <vivian.zhang@amd.com>
---
 .../print_ir_aie2xclbin.sh                    | 33 +------------------
 1 file changed, 1 insertion(+), 32 deletions(-)

diff --git a/build_tools/ci/print_ir_aie2xclbin/print_ir_aie2xclbin.sh b/build_tools/ci/print_ir_aie2xclbin/print_ir_aie2xclbin.sh
index edc70df08..76b06bb3a 100755
--- a/build_tools/ci/print_ir_aie2xclbin/print_ir_aie2xclbin.sh
+++ b/build_tools/ci/print_ir_aie2xclbin/print_ir_aie2xclbin.sh
@@ -9,25 +9,17 @@
 set -euo pipefail
 
 # Check for the number of provided arguments
-if [ "$#" -ne 3 ] && [ "$#" -ne 5 ]; then
+if [ "$#" -ne 3 ]; then
     echo -e "Illegal number of parameters: $#." \
             "\n For 2 parameters:" \
             "\n     1) <iree-compile-dir>" \
             "\n     2) <output-dir>" \
             "\n     3) <peano-install-dir>" \
-            "\n For 5 parameters:" \
-            "\n     1) <iree-compile-dir>" \
-            "\n     2) <output-dir>" \
-            "\n     3) <peano-install-dir>" \
-            "\n     4) <xrt-dir>" \
-            "\n     5) <vitis-install-dir>" \
             "\n Example (dependent on environment variables):" \
             "\n     ./print_ir_aie2xclbin.sh " \
             "\$IREE_BUILD_DIR/tools " \
             "results_dir_tmp "\
             "\$PEANO_INSTALL_DIR "\
-            "/opt/xilinx/xrt "\
-            "\$VITIS_INSTALL_PATH"
     exit 1
 fi
 
@@ -39,17 +31,6 @@ mkdir -p ${OUTPUT}
 if [ "$#" -eq 3 ]; then
   echo "Assuming that this is the 'CI case' as 3 parameters were provided."
   PEANO="$3"
-  XRT=/opt/xilinx/xrt
-  VITIS=/opt/Xilinx/Vitis/2024.2
-fi
-
-echo "chess-clang: $(find $VITIS -name chess-clang)"
-echo "xchesscc: $(find $VITIS -name xchesscc)"
-
-# The local set-paths-manually case:
-if [ "$#" -eq 5 ]; then
-  XRT="$4"
-  VITIS="$5"
 fi
 
 IREE_INSTALL_DIR="$1"
@@ -83,15 +64,6 @@ else
   exit 1
 fi
 
-if [ -d "${XRT}" ]; then
-  XRT=`realpath "${XRT}"`
-  source $XRT/setup.sh
-fi
-
-if [ -d "${VITIS}" ]; then
-  VITIS=${VITIS}
-fi
-
 # There might be a FileCheck program in the IREE_INSTALL_DIR. Check.
 # Do not fail if it is not there, we can also check if it already on PATH.
 if [ -x "${IREE_INSTALL_DIR}/bin/FileCheck" ]; then
@@ -113,7 +85,6 @@ ${SOURCE_MLIR_FILE} \
 --iree-hal-target-backends=amd-aie \
 --iree-amd-aie-peano-install-dir=${PEANO} \
 --iree-amd-aie-install-dir=${IREE_INSTALL_DIR} \
---iree-amd-aie-vitis-install-dir=${VITIS} \
 --iree-hal-dump-executable-files-to=${OUTPUT} \
 --aie2xclbin-print-ir-after-all \
 --aie2xclbin-print-ir-before-all \
@@ -178,7 +149,6 @@ ${SOURCE_MLIR_FILE} \
 --iree-hal-target-backends=amd-aie \
 --iree-amd-aie-peano-install-dir=${PEANO} \
 --iree-amd-aie-install-dir=${IREE_INSTALL_DIR} \
---iree-amd-aie-vitis-install-dir=${VITIS} \
 --iree-hal-dump-executable-intermediates-to=${OUTPUT} \
 --iree-hal-dump-executable-files-to=${OUTPUT} \
 --mlir-disable-threading \
@@ -201,7 +171,6 @@ ${SOURCE_MLIR_FILE} \
 --iree-hal-target-backends=amd-aie \
 --iree-amd-aie-peano-install-dir=${PEANO} \
 --iree-amd-aie-install-dir=${IREE_INSTALL_DIR} \
---iree-amd-aie-vitis-install-dir=${VITIS} \
 --iree-hal-dump-executable-intermediates-to=${OUTPUT} \
 --iree-hal-dump-executable-files-to=${OUTPUT} \
 --mlir-disable-threading \

From 1f32c6c1bcd4c58b92d1adc93f2b5348ec901c52 Mon Sep 17 00:00:00 2001
From: James Newling <james.newling@gmail.com>
Date: Thu, 29 Aug 2024 15:58:43 -0700
Subject: [PATCH 2/5] [CreateLogicalObjectFifoLink] Fail if accesses not
 contiguous  (#696)

I was looking into this pass because convolution lowering hits

```
error: 'amdaie.circular_dma_cpy_nd' op access pattern of strided operation overlaps with next one, which is not supported for now
```

But looking at the code now, it looks like it'll currently only work for
the the access patterns are perfectly contiguous? If that's the case, I
can look into fixing it (for both cases -- overlapping and
non-contiguous).

I left a comment in the code describing my thinking
---
 .../AMDAIECreateLogicalObjectFifoLink.cpp     | 39 ++++++++++++-------
 .../test/create_logical_objectfifo_link.mlir  |  6 +--
 2 files changed, 28 insertions(+), 17 deletions(-)

diff --git a/compiler/plugins/target/AMD-AIE/iree-amd-aie/Transforms/AMDAIECreateLogicalObjectFifoLink.cpp b/compiler/plugins/target/AMD-AIE/iree-amd-aie/Transforms/AMDAIECreateLogicalObjectFifoLink.cpp
index cfd347313..3ffcd6e6f 100644
--- a/compiler/plugins/target/AMD-AIE/iree-amd-aie/Transforms/AMDAIECreateLogicalObjectFifoLink.cpp
+++ b/compiler/plugins/target/AMD-AIE/iree-amd-aie/Transforms/AMDAIECreateLogicalObjectFifoLink.cpp
@@ -32,9 +32,9 @@ namespace mlir::iree_compiler::AMDAIE {
 /// these two access patterns overlap by both accessing elements in range [32,
 /// 63].
 template <CopyOpOperateOn OperateOn>
-LogicalResult checkForNoOverlappingAccessPatterns(
-    const SmallVector<std::pair<DoublyStridedCopyOpInterface, int64_t>>
-        &stridedOps) {
+LogicalResult checkForContiguousAccessPatterns(
+    ArrayRef<std::pair<DoublyStridedCopyOpInterface, int64_t>> stridedOps) {
+
   for (auto &&[i, stridedOpAndOffset] : llvm::enumerate(stridedOps)) {
     DoublyStridedCopyOpInterface stridedOp = stridedOpAndOffset.first;
     std::optional<int64_t> extent;
@@ -45,14 +45,26 @@ LogicalResult checkForNoOverlappingAccessPatterns(
     }
     if (!extent) {
       return stridedOp.emitOpError()
-             << "non-constant access extent is not supported";
+             << "has a non-constant access extent, which is not supported";
     }
     int64_t offset = stridedOpAndOffset.second;
-    if (i < (stridedOps.size() - 1) &&
-        (offset + extent.value()) > stridedOps[i + 1].second) {
-      return stridedOp.emitOpError()
-             << "access pattern of strided operation overlaps with next one, "
-                "which is not supported for now";
+    if (i < (stridedOps.size() - 1)) {
+      if (offset + extent.value() != stridedOps[i + 1].second) {
+        // TODO(newling) my understanding from the code is that the link
+        // operation effectively replaces the cumulative offset of each
+        // circular_dma_cpy_nd with the differential offset with
+        // the previous circular_dma_cpy_nd in the 'link' list.
+        //
+        // This however is hardcoded to a zero offset (later in the pass where
+        // discardAllNonZeroOffsets is called, offsets are set to zero). This
+        // effectively is constraining the link operation to only work with
+        // contiguous access patterns.
+        //
+        // Is this a bug?
+        return stridedOp.emitOpError()
+               << "has access pattern of which isn't contiguous with next one "
+                  "-- not currently supported.";
+      }
     }
   }
   return success();
@@ -81,9 +93,8 @@ LogicalResult createLogicalObjectFifoLink(
   for (Operation *userOp : logicalObjectFifo->getUsers()) {
     if (auto stridedOp = dyn_cast<DoublyStridedCopyOpInterface>(userOp)) {
       if (lastUserOp && stridedOp->getBlock() != lastUserOp->getBlock()) {
-        logicalObjectFifo->emitError(
-            "does have copy-like users not residing in the same block");
-        return failure();
+        return logicalObjectFifo->emitOpError(
+            "has copy-like users not residing in the same block");
       }
       auto sourceLogicalObjectFifo =
           dyn_cast<AMDAIE::LogicalObjectFifoFromMemrefOp>(
@@ -130,11 +141,11 @@ LogicalResult createLogicalObjectFifoLink(
   // Check that access patterns are not overlapping between consumers
   // respectively producers.
   if (failed(
-          checkForNoOverlappingAccessPatterns<CopyOpOperateOn::Target>(ins))) {
+          checkForContiguousAccessPatterns<CopyOpOperateOn::Target>(ins))) {
     return failure();
   }
   if (failed(
-          checkForNoOverlappingAccessPatterns<CopyOpOperateOn::Source>(outs))) {
+          checkForContiguousAccessPatterns<CopyOpOperateOn::Source>(outs))) {
     return failure();
   }
 
diff --git a/compiler/plugins/target/AMD-AIE/iree-amd-aie/Transforms/test/create_logical_objectfifo_link.mlir b/compiler/plugins/target/AMD-AIE/iree-amd-aie/Transforms/test/create_logical_objectfifo_link.mlir
index 2f915931c..67577eff1 100644
--- a/compiler/plugins/target/AMD-AIE/iree-amd-aie/Transforms/test/create_logical_objectfifo_link.mlir
+++ b/compiler/plugins/target/AMD-AIE/iree-amd-aie/Transforms/test/create_logical_objectfifo_link.mlir
@@ -79,7 +79,7 @@ func.func @link_multiple_inputs_with_overlapping_access(%arg0: memref<32x1024xi3
   %2 = amdaie.logicalobjectfifo.from_memref %arg2, {} : memref<8x8x4x8xi32, 2> -> !amdaie.logicalobjectfifo<memref<8x8x4x8xi32, 2>>
   %3 = amdaie.circular_dma_cpy_nd(%1[0] [1024] [1], %0[] [] []) : (!amdaie.logicalobjectfifo<memref<3x32x64xi32, 1>>, !amdaie.logicalobjectfifo<memref<32x1024xi32>>)
   %4 = amdaie.circular_dma_cpy_nd(%1[1, 0] [1, 1024] [2048, 1], %0[] [] []) : (!amdaie.logicalobjectfifo<memref<3x32x64xi32, 1>>, !amdaie.logicalobjectfifo<memref<32x1024xi32>>)
-  // expected-error @+1 {{access pattern of strided operation overlaps with next one}}
+  // expected-error @+1 {{op has access pattern of which isn't contiguous with next one}}
   %5 = amdaie.circular_dma_cpy_nd(%1[1, 0] [1, 1025] [1024, 1], %0[] [] []) : (!amdaie.logicalobjectfifo<memref<3x32x64xi32, 1>>, !amdaie.logicalobjectfifo<memref<32x1024xi32>>)
   %6 = amdaie.circular_dma_cpy_nd(%2[] [] [], %1[] [] []) : (!amdaie.logicalobjectfifo<memref<8x8x4x8xi32, 2>>, !amdaie.logicalobjectfifo<memref<3x32x64xi32, 1>>)
   return
@@ -151,7 +151,7 @@ func.func @link_multiple_outputs_with_overlapping_access(%arg0: memref<32x1024xi
   %3 = amdaie.circular_dma_cpy_nd(%1[] [] [], %0[] [] []) : (!amdaie.logicalobjectfifo<memref<3x32x64xi32, 1>>, !amdaie.logicalobjectfifo<memref<32x1024xi32>>)
   %4 = amdaie.circular_dma_cpy_nd(%2[] [] [], %1[1, 0] [1, 1024] [2048, 1]) : (!amdaie.logicalobjectfifo<memref<8x8x4x8xi32, 2>>, !amdaie.logicalobjectfifo<memref<3x32x64xi32, 1>>)
   %5 = amdaie.circular_dma_cpy_nd(%2[] [] [], %1[1, 0] [1, 1024] [1024, 1]) : (!amdaie.logicalobjectfifo<memref<8x8x4x8xi32, 2>>, !amdaie.logicalobjectfifo<memref<3x32x64xi32, 1>>)
-  // expected-error @+1 {{access pattern of strided operation overlaps with next one}}
+  // expected-error @+1 {{op has access pattern of which isn't contiguous with next one}}
   %6 = amdaie.circular_dma_cpy_nd(%2[] [] [], %1[0, 0] [32, 32] [64, 1]) : (!amdaie.logicalobjectfifo<memref<8x8x4x8xi32, 2>>, !amdaie.logicalobjectfifo<memref<3x32x64xi32, 1>>)
   return
 }
@@ -243,7 +243,7 @@ func.func @ensure_no_removal_of_offsets(%arg0: memref<32x1024xi32>, %arg1: memre
 
 func.func @link_different_blocks(%arg0: memref<32x1024xi32>, %arg1: memref<32x64xi32, 1>, %arg2: memref<8x8x4x8xi32, 2>) {
   %0 = amdaie.logicalobjectfifo.from_memref %arg0, {} :memref<32x1024xi32> -> !amdaie.logicalobjectfifo<memref<32x1024xi32>>
-  // expected-error @+2 {{does have copy-like users not residing in the same block}}
+  // expected-error @+2 {{has copy-like users not residing in the same block}}
   // expected-error @+1 {{couldn't create a link operation}}
   %1 = amdaie.logicalobjectfifo.from_memref %arg1, {} : memref<32x64xi32, 1> -> !amdaie.logicalobjectfifo<memref<32x64xi32, 1>>
   %2 = amdaie.logicalobjectfifo.from_memref %arg2, {} : memref<8x8x4x8xi32, 2> -> !amdaie.logicalobjectfifo<memref<8x8x4x8xi32, 2>>

From 3a8f9938c837989902db1c21b4d3f4b7178a8356 Mon Sep 17 00:00:00 2001
From: Maksim Levental <maksim.levental@gmail.com>
Date: Thu, 29 Aug 2024 17:02:45 -0700
Subject: [PATCH 3/5] Fix GHA caches (#719)

---
 .github/workflows/ci-linux.yml   | 40 ++++++++++++++++++++++++++++----
 .github/workflows/ci-macos.yml   | 40 +++++++++++++++++++++++---------
 .github/workflows/ci-windows.yml | 38 +++++++++++++++++++++---------
 build_tools/ci/build_test_cpp.sh |  3 ---
 4 files changed, 91 insertions(+), 30 deletions(-)

diff --git a/.github/workflows/ci-linux.yml b/.github/workflows/ci-linux.yml
index 100ea0eca..d0529f584 100644
--- a/.github/workflows/ci-linux.yml
+++ b/.github/workflows/ci-linux.yml
@@ -9,6 +9,11 @@ on:
     branches:
       - main
 
+  schedule:
+    # At minute 0 past every 6th hour. (see https://crontab.guru)
+    # this job is to keep the ccache cache warm
+    - cron: '0 */6 * * *'
+
 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
@@ -24,6 +29,8 @@ jobs:
       fail-fast: true
     env:
       CACHE_DIR: ${{ github.workspace }}/.container-cache
+      # either the PR number or `branch-N` where N always increments
+      CACHE_KEY: linux-build-test-cpp-asserts-manylinux-v2-${{ github.event.number || format('{0}-{1}', github.ref_name, github.run_number) }}
     steps:
       - name: Set unified TZ
         uses: szenius/set-timezone@v2.0
@@ -42,11 +49,11 @@ jobs:
         with:
           submodules: recursive
 
-      - name: Install static libs
+      - name: Install deps
         run: |
-          dnf install -y almalinux-release-devel
           yum remove -y openssl-devel zlib-devel || true
           yum install -y protobuf-devel protobuf-compiler
+
       - name: Sync source deps
         run: |
           python ./sync_deps.py
@@ -59,9 +66,18 @@ jobs:
         uses: actions/cache/restore@v3
         with:
           path: ${{ env.CACHE_DIR }}
-          key: linux-build-test-cpp-asserts-manylinux-v2-${{ github.sha }}
+          key:  ${{ env.CACHE_KEY }}
           restore-keys: linux-build-test-cpp-
 
+      - name: Configure ccache debug logs
+        run: |
+          # https://interrupt.memfault.com/blog/ccache-debugging
+          echo CCACHE_DEBUG=1 >> $GITHUB_ENV
+          echo CCACHE_DEBUGLEVEL=2 >> $GITHUB_ENV
+          echo CCACHE_DEBUGDIR="$HOME/ccache-debug" >> $GITHUB_ENV
+          echo CCACHE_SLOPPINESS=include_file_ctime,include_file_mtime,time_macros >> $GITHUB_ENV
+          echo CCACHE_COMPILERCHECK=content >> $GITHUB_ENV
+
       - name: Build packages
         run: |
           export cache_dir="${{ env.CACHE_DIR }}"
@@ -72,6 +88,20 @@ jobs:
         run: |
           tar cf iree-dist-linux.tar -C iree-install .
 
+      - name: Tar ccache logs
+        if: ${{ !cancelled() }}
+        run: |
+          ccache --show-stats -v
+          tar cf linux-ccache-logs.tar $CCACHE_DEBUGDIR
+
+      - name: Upload ccache logs
+        uses: actions/upload-artifact@v4
+        if: ${{ !cancelled() }}
+        with:
+          name: ccache_logs
+          path: linux-ccache-logs.tar
+          if-no-files-found: warn
+
       - name: Upload artifacts
         uses: actions/upload-artifact@v4
         if: ${{ !cancelled() }}
@@ -82,10 +112,10 @@ jobs:
 
       - name: Save cache
         uses: actions/cache/save@v3
-        if: ${{ !cancelled() }}
+        if: ${{ !cancelled() && github.event_name != 'pull_request' }}
         with:
           path: ${{ env.CACHE_DIR }}
-          key: linux-build-test-cpp-asserts-manylinux-v2-${{ github.sha }}
+          key: ${{ env.CACHE_KEY }}
 
   test_linux:
     name: E2E Test linux
diff --git a/.github/workflows/ci-macos.yml b/.github/workflows/ci-macos.yml
index eda8ee471..069831e34 100644
--- a/.github/workflows/ci-macos.yml
+++ b/.github/workflows/ci-macos.yml
@@ -9,10 +9,10 @@ on:
     branches:
       - main
 
+  schedule:
+    - cron: '0 */6 * * *'
+
 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-macos-${{ github.event.number || github.sha }}
   cancel-in-progress: true
 
@@ -26,6 +26,7 @@ jobs:
         runs-on: [macos-12, macos-14]
     env:
       CACHE_DIR: ${{ github.workspace }}/.container-cache
+      CACHE_KEY: ${{ matrix.runs-on }}-build-test-cpp-asserts-v1-${{ github.event.number || format('{0}-{1}', github.ref_name, github.run_number) }}
     steps:
       - name: Set unified TZ
         uses: szenius/set-timezone@v2.0
@@ -60,11 +61,16 @@ jobs:
         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: ${{ matrix.runs-on }}-build-test-cpp-asserts-v1-${{ github.sha }}-${{ github.event.repository.updated_at }}
+          key: ${{ env.CACHE_KEY }}
           restore-keys: ${{ matrix.runs-on }}-build-test-cpp-
 
+      - name: Configure ccache debug logs
+        run: |
+          echo CCACHE_DEBUG=1 >> $GITHUB_ENV
+          echo CCACHE_DEBUGLEVEL=2 >> $GITHUB_ENV
+          echo CCACHE_DEBUGDIR="$HOME/ccache-debug" >> $GITHUB_ENV
+          echo CCACHE_SLOPPINESS=include_file_ctime,include_file_mtime,time_macros >> $GITHUB_ENV
+
       - name: Build packages
         run: |
           export cache_dir="${{ env.CACHE_DIR }}"
@@ -73,9 +79,21 @@ jobs:
       - name: Create artifacts
         if: ${{ !cancelled() }}
         run: |
-          rm -f iree-install/bin/clang*
-          rm -f iree-install/bin/llvm-link*
-          tar cf iree-dist-${{ matrix.runs-on }}.tar -C iree-install . -C ../iree-build tools/testing/e2e/iree-e2e-matmul-test
+          tar cf iree-dist-${{ matrix.runs-on }}.tar -C iree-install .
+
+      - name: Tar ccache logs
+        if: ${{ !cancelled() }}
+        run: |
+          ccache --show-stats -v
+          tar cf macos-ccache-logs.tar $CCACHE_DEBUGDIR
+
+      - name: Upload ccache logs
+        uses: actions/upload-artifact@v4
+        if: ${{ !cancelled() }}
+        with:
+          name: ccache_logs
+          path: macos-ccache-logs.tar
+          if-no-files-found: warn
 
       - name: Upload artifacts
         uses: actions/upload-artifact@v4
@@ -87,7 +105,7 @@ jobs:
 
       - name: Save cache
         uses: actions/cache/save@v3
-        if: ${{ !cancelled() }}
+        if: ${{ !cancelled() && github.event_name != 'pull_request' }}
         with:
           path: ${{ env.CACHE_DIR }}
-          key: ${{ matrix.runs-on }}-build-test-cpp-asserts-v1-${{ github.sha }}-${{ github.event.repository.updated_at }}
+          key: ${{ env.CACHE_KEY }}
diff --git a/.github/workflows/ci-windows.yml b/.github/workflows/ci-windows.yml
index 4804d14eb..d7cb8a929 100644
--- a/.github/workflows/ci-windows.yml
+++ b/.github/workflows/ci-windows.yml
@@ -10,14 +10,9 @@ on:
       - main
 
   schedule:
-      # At minute 0 past every 12th hour. (see https://crontab.guru)
-      # this job is to keep the ccache cache warm
-    - cron: '0 */12 * * *'
+    - cron: '0 */6 * * *'
 
 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
 
@@ -34,6 +29,7 @@ jobs:
       fail-fast: true
     env:
       CACHE_DIR: ${{ github.workspace }}/.container-cache
+      CACHE_KEY: windows-build-test-cpp-asserts-v1-${{ github.event.number || format('{0}-{1}', github.ref_name, github.run_number) }}
     steps:
       - name: Set unified TZ
         uses: szenius/set-timezone@v2.0
@@ -73,11 +69,17 @@ jobs:
         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 }}
+          key: ${{ env.CACHE_KEY }}
           restore-keys: windows-build-test-cpp-
 
+      - name: Configure ccache debug logs
+        run: |
+          echo CCACHE_DEBUG=1 >> $GITHUB_ENV
+          echo CCACHE_DEBUGLEVEL=2 >> $GITHUB_ENV
+          echo CCACHE_DEBUGDIR="/c/ccache-debug" >> $GITHUB_ENV
+          echo CCACHE_SLOPPINESS=include_file_ctime,include_file_mtime,time_macros >> $GITHUB_ENV
+          echo CCACHE_COMPILERCHECK=content >> $GITHUB_ENV
+
       - name: Build packages
         run: |
           export cache_dir="${{ env.CACHE_DIR }}"
@@ -88,6 +90,20 @@ jobs:
         run: |
           tar cf iree-dist-windows.tar -C iree-install .
 
+      - name: Tar ccache logs
+        if: ${{ !cancelled() }}
+        run: |
+          ccache --show-stats -v
+          tar cf windows-ccache-logs.tar $CCACHE_DEBUGDIR
+
+      - name: Upload ccache logs
+        uses: actions/upload-artifact@v4
+        if: ${{ !cancelled() }}
+        with:
+          name: ccache_logs
+          path: windows-ccache-logs.tar
+          if-no-files-found: warn
+
       - name: Upload artifacts
         uses: actions/upload-artifact@v4
         if: ${{ !cancelled() }}
@@ -98,10 +114,10 @@ jobs:
 
       - name: Save cache
         uses: actions/cache/save@v3
-        if: ${{ !cancelled() }}
+        if: ${{ !cancelled() && github.event_name != 'pull_request' }}
         with:
           path: ${{ env.CACHE_DIR }}
-          key: windows-build-test-cpp-asserts-v1-${{ github.sha }}-${{ github.event.repository.updated_at }}
+          key: ${{ env.CACHE_KEY }}
 
   test_windows:
     name: E2E Test windows
diff --git a/build_tools/ci/build_test_cpp.sh b/build_tools/ci/build_test_cpp.sh
index 65c61332a..296dfc353 100644
--- a/build_tools/ci/build_test_cpp.sh
+++ b/build_tools/ci/build_test_cpp.sh
@@ -115,9 +115,6 @@ elif [[ "$OSTYPE" == "msys"* ]]; then
   ctest --test-dir "$build_dir" -R amd-aie --output-on-failure -j --repeat until-pass:5
 fi
 
-# Show ccache stats.
-ccache --show-stats
-
 rm -f "$install_dir"/bin/clang*
 rm -f "$install_dir"/bin/llvm-link*
 cp "$build_dir"/tools/testing/e2e/iree-e2e-matmul-test "$install_dir"/bin

From 7a97fb38dd86a0bde736e8e95ad4e33ca33b21bd Mon Sep 17 00:00:00 2001
From: Maksim Levental <maksim.levental@gmail.com>
Date: Thu, 29 Aug 2024 20:50:26 -0700
Subject: [PATCH 4/5] [Hot fix] Linux CI (#727)

---
 .github/workflows/ci-linux.yml   |  24 +---
 .github/workflows/ci-macos.yml   |  22 +---
 .github/workflows/ci-windows.yml |  23 +---
 build_tools/ci/build_test_cpp.sh |   3 +
 sync_deps.py                     | 189 +++++++++++++++----------------
 5 files changed, 100 insertions(+), 161 deletions(-)

diff --git a/.github/workflows/ci-linux.yml b/.github/workflows/ci-linux.yml
index d0529f584..968cf4630 100644
--- a/.github/workflows/ci-linux.yml
+++ b/.github/workflows/ci-linux.yml
@@ -69,18 +69,10 @@ jobs:
           key:  ${{ env.CACHE_KEY }}
           restore-keys: linux-build-test-cpp-
 
-      - name: Configure ccache debug logs
-        run: |
-          # https://interrupt.memfault.com/blog/ccache-debugging
-          echo CCACHE_DEBUG=1 >> $GITHUB_ENV
-          echo CCACHE_DEBUGLEVEL=2 >> $GITHUB_ENV
-          echo CCACHE_DEBUGDIR="$HOME/ccache-debug" >> $GITHUB_ENV
-          echo CCACHE_SLOPPINESS=include_file_ctime,include_file_mtime,time_macros >> $GITHUB_ENV
-          echo CCACHE_COMPILERCHECK=content >> $GITHUB_ENV
-
       - name: Build packages
         run: |
           export cache_dir="${{ env.CACHE_DIR }}"
+          export CCACHE_COMPILERCHECK="string:$(clang --version)"
           bash build_tools/ci/build_test_cpp.sh
 
       - name: Create artifacts
@@ -88,20 +80,6 @@ jobs:
         run: |
           tar cf iree-dist-linux.tar -C iree-install .
 
-      - name: Tar ccache logs
-        if: ${{ !cancelled() }}
-        run: |
-          ccache --show-stats -v
-          tar cf linux-ccache-logs.tar $CCACHE_DEBUGDIR
-
-      - name: Upload ccache logs
-        uses: actions/upload-artifact@v4
-        if: ${{ !cancelled() }}
-        with:
-          name: ccache_logs
-          path: linux-ccache-logs.tar
-          if-no-files-found: warn
-
       - name: Upload artifacts
         uses: actions/upload-artifact@v4
         if: ${{ !cancelled() }}
diff --git a/.github/workflows/ci-macos.yml b/.github/workflows/ci-macos.yml
index 069831e34..cf05c9376 100644
--- a/.github/workflows/ci-macos.yml
+++ b/.github/workflows/ci-macos.yml
@@ -64,16 +64,10 @@ jobs:
           key: ${{ env.CACHE_KEY }}
           restore-keys: ${{ matrix.runs-on }}-build-test-cpp-
 
-      - name: Configure ccache debug logs
-        run: |
-          echo CCACHE_DEBUG=1 >> $GITHUB_ENV
-          echo CCACHE_DEBUGLEVEL=2 >> $GITHUB_ENV
-          echo CCACHE_DEBUGDIR="$HOME/ccache-debug" >> $GITHUB_ENV
-          echo CCACHE_SLOPPINESS=include_file_ctime,include_file_mtime,time_macros >> $GITHUB_ENV
-
       - name: Build packages
         run: |
           export cache_dir="${{ env.CACHE_DIR }}"
+          export CCACHE_COMPILERCHECK="string:$(clang --version)"
           bash build_tools/ci/build_test_cpp.sh
 
       - name: Create artifacts
@@ -81,20 +75,6 @@ jobs:
         run: |
           tar cf iree-dist-${{ matrix.runs-on }}.tar -C iree-install .
 
-      - name: Tar ccache logs
-        if: ${{ !cancelled() }}
-        run: |
-          ccache --show-stats -v
-          tar cf macos-ccache-logs.tar $CCACHE_DEBUGDIR
-
-      - name: Upload ccache logs
-        uses: actions/upload-artifact@v4
-        if: ${{ !cancelled() }}
-        with:
-          name: ccache_logs
-          path: macos-ccache-logs.tar
-          if-no-files-found: warn
-
       - name: Upload artifacts
         uses: actions/upload-artifact@v4
         if: ${{ !cancelled() }}
diff --git a/.github/workflows/ci-windows.yml b/.github/workflows/ci-windows.yml
index d7cb8a929..de71a4f9a 100644
--- a/.github/workflows/ci-windows.yml
+++ b/.github/workflows/ci-windows.yml
@@ -72,17 +72,10 @@ jobs:
           key: ${{ env.CACHE_KEY }}
           restore-keys: windows-build-test-cpp-
 
-      - name: Configure ccache debug logs
-        run: |
-          echo CCACHE_DEBUG=1 >> $GITHUB_ENV
-          echo CCACHE_DEBUGLEVEL=2 >> $GITHUB_ENV
-          echo CCACHE_DEBUGDIR="/c/ccache-debug" >> $GITHUB_ENV
-          echo CCACHE_SLOPPINESS=include_file_ctime,include_file_mtime,time_macros >> $GITHUB_ENV
-          echo CCACHE_COMPILERCHECK=content >> $GITHUB_ENV
-
       - name: Build packages
         run: |
           export cache_dir="${{ env.CACHE_DIR }}"
+          export CCACHE_COMPILERCHECK="string:$(clang-cl.exe --version)"
           bash build_tools/ci/build_test_cpp.sh
 
       - name: Create artifacts
@@ -90,20 +83,6 @@ jobs:
         run: |
           tar cf iree-dist-windows.tar -C iree-install .
 
-      - name: Tar ccache logs
-        if: ${{ !cancelled() }}
-        run: |
-          ccache --show-stats -v
-          tar cf windows-ccache-logs.tar $CCACHE_DEBUGDIR
-
-      - name: Upload ccache logs
-        uses: actions/upload-artifact@v4
-        if: ${{ !cancelled() }}
-        with:
-          name: ccache_logs
-          path: windows-ccache-logs.tar
-          if-no-files-found: warn
-
       - name: Upload artifacts
         uses: actions/upload-artifact@v4
         if: ${{ !cancelled() }}
diff --git a/build_tools/ci/build_test_cpp.sh b/build_tools/ci/build_test_cpp.sh
index 296dfc353..38c98ca9d 100644
--- a/build_tools/ci/build_test_cpp.sh
+++ b/build_tools/ci/build_test_cpp.sh
@@ -36,10 +36,12 @@ elif [[ "$OSTYPE" == "msys"* ]]; then
   export CC=clang-cl.exe
   export CXX=clang-cl.exe
 fi
+
 export CCACHE_DIR="${cache_dir}/ccache"
 export CCACHE_MAXSIZE="700M"
 export CMAKE_C_COMPILER_LAUNCHER=ccache
 export CMAKE_CXX_COMPILER_LAUNCHER=ccache
+export CCACHE_SLOPPINESS=include_file_ctime,include_file_mtime,time_macros
 
 # Clear ccache stats.
 ccache -z
@@ -64,6 +66,7 @@ CMAKE_ARGS="\
   -DCMAKE_BUILD_TYPE=Release \
   -DCMAKE_INSTALL_PREFIX=$install_dir \
   -DCMAKE_INSTALL_LIBDIR=lib \
+  -DIREE_ERROR_ON_MISSING_SUBMODULES=OFF \
   -DIREE_ENABLE_ASSERTIONS=ON \
   -DIREE_BUILD_SAMPLES=OFF \
   -DIREE_BUILD_PYTHON_BINDINGS=ON \
diff --git a/sync_deps.py b/sync_deps.py
index 0d0e470c7..149b11533 100644
--- a/sync_deps.py
+++ b/sync_deps.py
@@ -2,122 +2,121 @@
 ### AUTO-GENERATED: DO NOT EDIT
 ### Casual developers and CI bots invoke this to do the most
 ### efficient checkout of dependencies.
-### Cross-repo project development should use the 
+### Cross-repo project development should use the
 ### 'shark-workspace' dev tool for more full featured setup.
 ### Update with: shark-workspace pin
 
 PINNED_VERSIONS = {
-  "iree": "60b65f30c932eaf967922785253a85a1aa14cebb",
+    "iree": "60b65f30c932eaf967922785253a85a1aa14cebb",
 }
 
 ORIGINS = {
-  "iree": "https://github.com/iree-org/iree.git",
+    "iree": "https://github.com/iree-org/iree.git",
 }
 
-SUBMODULES = {
-  "iree": 1
-}
+SUBMODULES = {"iree"}
 
+EXCLUDED_SUBMODULES = {"third_party/torch-mlir"}
 
 
 ### Update support:
 
 import argparse
-from pathlib import Path
 import re
 import shlex
 import subprocess
+from pathlib import Path
 
 
 def main():
-  parser = argparse.ArgumentParser(description="Source deps sync")
-  parser.add_argument(
-      "--exclude-submodule",
-      nargs="*",
-      help="Exclude submodules by regex (relative to '{project}:{path})")
-  parser.add_argument("--exclude-dep",
-                      nargs="*",
-                      help="Excludes dependencies by regex")
-  parser.add_argument("--depth",
-                      type=int,
-                      default=0,
-                      help="Fetch revisions with --depth")
-  parser.add_argument("--submodules-depth",
-                      type=int,
-                      default=0,
-                      help="Update submodules with --depth")
-  args = parser.parse_args()
-
-  workspace_dir = Path(__file__).resolve().parent.parent
-  for repo_name, revision in PINNED_VERSIONS.items():
-    # Exclude this dep?
-    exclude_repo = False
-    for exclude_pattern in (args.exclude_dep or ()):
-      if re.search(exclude_pattern, repo_name):
-        exclude_repo = True
-    if exclude_repo:
-      print(f"Excluding {repo_name} based on --exclude-dep")
-      continue
-
-    print(f"Syncing {repo_name}")
-    repo_dir = workspace_dir / repo_name
-    if not repo_dir.exists():
-      # Shallow clone
-      print(f"  Cloning {repo_name}...")
-      repo_dir.mkdir()
-      run(["init"], repo_dir)
-      run(["remote", "add", "origin", ORIGINS[repo_name]], repo_dir)
-    # Checkout detached head.
-    fetch_args = ["fetch"]
-    if args.depth > 0:
-      fetch_args.extend(["--depth=1"])
-    fetch_args.extend(["origin", revision])
-    run(fetch_args, repo_dir)
-    run(["-c", "advice.detachedHead=false", "checkout", revision], repo_dir)
-    if SUBMODULES.get(repo_name):
-      print(f"  Initializing submodules for {repo_name}")
-      cp = run(["submodule", "status"],
-               repo_dir,
-               silent=True,
-               capture_output=True)
-      submodules = []
-      for submodule_status_line in cp.stdout.decode().splitlines():
-        submodule_status_parts = submodule_status_line.split()
-        submodule_path = submodule_status_parts[1]
-        exclude_submodule = False
-        for exclude_pattern in (args.exclude_submodule or ()):
-          if re.search(exclude_pattern, f"{repo_name}:{submodule_path}"):
-            exclude_submodule = True
-        if exclude_submodule:
-          print(f"  Excluding {submodule_path} based on --exclude-submodule")
-          continue
-        submodules.append(submodule_path)
-
-      update_args = ["submodule", "update", "--init"]
-      if args.submodules_depth > 0:
-        update_args.extend(["--depth", "1"])
-      update_args.extend(["--"])
-      update_args.extend(submodules)
-      run(update_args, repo_dir)
-
-
-def run(args,
-        cwd,
-        *,
-        capture_output: bool = False,
-        check: bool = True,
-        silent: bool = False):
-  args = ["git"] + args
-  args_text = ' '.join([shlex.quote(arg) for arg in args])
-  if not silent:
-    print(f"  [{cwd}]$ {args_text}")
-  cp = subprocess.run(args, cwd=str(cwd), capture_output=capture_output)
-  if check and cp.returncode != 0:
-    addl_info = f":\n({cp.stderr.decode()})" if capture_output else ""
-    raise RuntimeError(f"Git command failed: {args_text} (from {cwd})"
-                       f"{addl_info}")
-  return cp
+    parser = argparse.ArgumentParser(description="Source deps sync")
+    parser.add_argument(
+        "--exclude-submodule",
+        nargs="*",
+        help="Exclude submodules by regex (relative to '{project}:{path})",
+        default=(),
+    )
+    parser.add_argument(
+        "--exclude-dep", nargs="*", help="Excludes dependencies by regex"
+    )
+    parser.add_argument(
+        "--depth", type=int, default=0, help="Fetch revisions with --depth"
+    )
+    parser.add_argument(
+        "--submodules-depth", type=int, default=0, help="Update submodules with --depth"
+    )
+    args = parser.parse_args()
+
+    workspace_dir = Path(__file__).resolve().parent.parent
+    for repo_name, revision in PINNED_VERSIONS.items():
+        # Exclude this dep?
+        exclude_repo = False
+        for exclude_pattern in args.exclude_dep or ():
+            if re.search(exclude_pattern, repo_name):
+                exclude_repo = True
+        if exclude_repo:
+            print(f"Excluding {repo_name} based on --exclude-dep")
+            continue
+
+        print(f"Syncing {repo_name}")
+        repo_dir = workspace_dir / repo_name
+        if not repo_dir.exists():
+            # Shallow clone
+            print(f"  Cloning {repo_name}...")
+            repo_dir.mkdir()
+            run(["init"], repo_dir)
+            run(["remote", "add", "origin", ORIGINS[repo_name]], repo_dir)
+        # Checkout detached head.
+        fetch_args = ["fetch"]
+        if args.depth > 0:
+            fetch_args.extend(["--depth=1"])
+        fetch_args.extend(["origin", revision])
+        run(fetch_args, repo_dir)
+        run(["-c", "advice.detachedHead=false", "checkout", revision], repo_dir)
+        if repo_name in SUBMODULES:
+            print(f"  Initializing submodules for {repo_name}")
+            cp = run(
+                ["submodule", "status"], repo_dir, silent=True, capture_output=True
+            )
+            submodules = []
+            for submodule_status_line in cp.stdout.decode().splitlines():
+                submodule_status_parts = submodule_status_line.split()
+                submodule_path = submodule_status_parts[1]
+                exclude_submodule = False
+                for exclude_pattern in args.exclude_submodule + tuple(
+                    EXCLUDED_SUBMODULES
+                ):
+                    if re.search(exclude_pattern, f"{repo_name}:{submodule_path}"):
+                        exclude_submodule = True
+                if exclude_submodule:
+                    print(f"  Excluding {submodule_path} based on --exclude-submodule")
+                    continue
+                submodules.append(submodule_path)
+
+            update_args = ["submodule", "update", "--init"]
+            if args.submodules_depth > 0:
+                update_args.extend(["--depth", "1"])
+            update_args.extend(["--"])
+            update_args.extend(submodules)
+            print(update_args)
+            run(update_args, repo_dir)
+
+
+def run(
+    args, cwd, *, capture_output: bool = False, check: bool = True, silent: bool = False
+):
+    args = ["git"] + args
+    args_text = " ".join([shlex.quote(arg) for arg in args])
+    if not silent:
+        print(f"  [{cwd}]$ {args_text}")
+    cp = subprocess.run(args, cwd=str(cwd), capture_output=capture_output)
+    if check and cp.returncode != 0:
+        addl_info = f":\n({cp.stderr.decode()})" if capture_output else ""
+        raise RuntimeError(
+            f"Git command failed: {args_text} (from {cwd})" f"{addl_info}"
+        )
+    return cp
 
 
 if __name__ == "__main__":
-  main()
+    main()

From 9c4c167baf89a279888fba8db75907845946077c Mon Sep 17 00:00:00 2001
From: Vivian <vivian.zhang@amd.com>
Date: Thu, 29 Aug 2024 22:30:25 -0700
Subject: [PATCH 5/5] Set ObjectFifo pipeline as default and some cleanup
 (#720)

As discussed previously, we are now setting pack-peel and objectfifo
pipeline as default.

Some cleanup including:
- Remove cl flag `numCores` which was added for historical task, but is
no longer used.
- Rename sample tests to be more clear about the pipelines being used.
- Just keep a minimal number of ci tests for MLIR-AIR path.
---
 build_tools/ci/build_test_cpp.sh              |   2 +-
 build_tools/ci/cpu_comparison/run_test.py     |   6 +-
 .../print_ir_aie2xclbin.sh                    |   3 +-
 build_tools/ci/run_matmul_test.sh             | 222 ++----------------
 .../Transforms/AMDAIEAddLoweringStrategy.cpp  |   5 +-
 .../Transforms/KernelDispatch.cpp             |  75 +++---
 .../iree-amd-aie/Transforms/KernelDispatch.h  |   9 +-
 .../iree-amd-aie/Transforms/Passes.cpp        |   9 +-
 .../AMD-AIE/iree-amd-aie/Transforms/Passes.td |   8 +-
 .../Transforms/test/lowering_strategy.mlir    |   4 +-
 tests/samples/CMakeLists.txt                  |  12 +-
 ...line_e2e.mlir => conv2d_nhwc_air_e2e.mlir} |   2 +-
 ...matmul_elementwise_pack_peel_air_e2e.mlir} |   2 +-
 ...mul.mlir => matmul_pack_peel_air_e2e.mlir} |   2 +-
 ....mlir => matmul_pack_peel_objectfifo.mlir} |   0
 ...r => matmul_pack_peel_objectfifo_e2e.mlir} |   0
 ..._e2e.mlir => matmul_pad_pack_air_e2e.mlir} |   2 +-
 17 files changed, 84 insertions(+), 279 deletions(-)
 rename tests/samples/{conv_pipeline_e2e.mlir => conv2d_nhwc_air_e2e.mlir} (95%)
 rename tests/samples/{pack_peel_pipeline_matmul_elementwise.mlir => matmul_elementwise_pack_peel_air_e2e.mlir} (95%)
 rename tests/samples/{pack_peel_pipeline_matmul.mlir => matmul_pack_peel_air_e2e.mlir} (91%)
 rename tests/samples/{matmul_peeled_objectfifo.mlir => matmul_pack_peel_objectfifo.mlir} (100%)
 rename tests/samples/{matmul_peeled_objectfifo_e2e.mlir => matmul_pack_peel_objectfifo_e2e.mlir} (100%)
 rename tests/samples/{pad_pack_pipeline_e2e.mlir => matmul_pad_pack_air_e2e.mlir} (97%)

diff --git a/build_tools/ci/build_test_cpp.sh b/build_tools/ci/build_test_cpp.sh
index 38c98ca9d..0382601aa 100644
--- a/build_tools/ci/build_test_cpp.sh
+++ b/build_tools/ci/build_test_cpp.sh
@@ -112,7 +112,7 @@ echo "-----"
 if [[ "$OSTYPE" == "linux-gnu"* ]]; then
   ctest --test-dir "$build_dir" -R amd-aie --output-on-failure -j
 elif [[ "$OSTYPE" == "darwin"* ]]; then
-  ctest --test-dir "$build_dir" -R amd-aie -E "pack_peel_pipeline_matmul|conv_fill_spec_pad" --output-on-failure -j --repeat until-pass:5
+  ctest --test-dir "$build_dir" -R amd-aie -E "matmul_pack_peel_air_e2e|matmul_elementwise_pack_peel_air_e2e|conv_fill_spec_pad" --output-on-failure -j --repeat until-pass:5
 elif [[ "$OSTYPE" == "msys"* ]]; then
   # 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
diff --git a/build_tools/ci/cpu_comparison/run_test.py b/build_tools/ci/cpu_comparison/run_test.py
index 3ff4a05a5..09223a36e 100755
--- a/build_tools/ci/cpu_comparison/run_test.py
+++ b/build_tools/ci/cpu_comparison/run_test.py
@@ -627,7 +627,7 @@ def run(self, config):
         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)
+        aie_vs_llvm_cpu(config, test_name, tile_pipeline="pack-peel", lower_to_aie_pipeline="air", rtol=0, atol=0)
 
         if config.xdna_datetime and config.xdna_datetime < 20240801:
             for name in [
@@ -657,10 +657,10 @@ def run(self, config):
         )
         if config.vitis_dir:
             aie_vs_llvm_cpu(
-                config, test_name, tile_pipeline="pack-peel", use_ukernel=True
+                config, test_name, tile_pipeline="pack-peel", lower_to_aie_pipeline="air", use_ukernel=True
             )
         aie_vs_llvm_cpu(
-            config, test_name, tile_pipeline="pack-peel", use_ukernel=False
+            config, test_name, tile_pipeline="pack-peel", lower_to_aie_pipeline="air", use_ukernel=False
         )
 
 
diff --git a/build_tools/ci/print_ir_aie2xclbin/print_ir_aie2xclbin.sh b/build_tools/ci/print_ir_aie2xclbin/print_ir_aie2xclbin.sh
index 76b06bb3a..39ea12f1d 100755
--- a/build_tools/ci/print_ir_aie2xclbin/print_ir_aie2xclbin.sh
+++ b/build_tools/ci/print_ir_aie2xclbin/print_ir_aie2xclbin.sh
@@ -93,7 +93,6 @@ ${SOURCE_MLIR_FILE} \
 --mlir-print-ir-after-all \
 --mlir-print-ir-module-scope \
 --mlir-disable-threading \
---iree-amdaie-tile-pipeline=pad-pack \
 -o ${OUTPUT}/test_artefact.vmfb \
 --iree-amd-aie-show-invoked-commands"
 
@@ -147,6 +146,7 @@ IREE_COMPILE_COMMAND="${IREE_COMPILE_EXE} \
 ${SOURCE_MLIR_FILE} \
 --compile-mode=hal-executable \
 --iree-hal-target-backends=amd-aie \
+--iree-amdaie-lower-to-aie-pipeline=air \
 --iree-amd-aie-peano-install-dir=${PEANO} \
 --iree-amd-aie-install-dir=${IREE_INSTALL_DIR} \
 --iree-hal-dump-executable-intermediates-to=${OUTPUT} \
@@ -169,6 +169,7 @@ IREE_COMPILE_COMMAND="${IREE_COMPILE_EXE} \
 ${SOURCE_MLIR_FILE} \
 --compile-mode=hal-executable \
 --iree-hal-target-backends=amd-aie \
+--iree-amdaie-lower-to-aie-pipeline=air \
 --iree-amd-aie-peano-install-dir=${PEANO} \
 --iree-amd-aie-install-dir=${IREE_INSTALL_DIR} \
 --iree-hal-dump-executable-intermediates-to=${OUTPUT} \
diff --git a/build_tools/ci/run_matmul_test.sh b/build_tools/ci/run_matmul_test.sh
index ad7bead47..32fcac59f 100755
--- a/build_tools/ci/run_matmul_test.sh
+++ b/build_tools/ci/run_matmul_test.sh
@@ -536,74 +536,29 @@ run_matmul_test \
     --use_ukernel "0" \
     --num_repeat_runs "2"
 
+###################################################################
+# MLIR-AIR Matmul tests
+###################################################################
+
 if [ -d "$VITIS" ]; then
   run_matmul_test \
       --name_prefix "ukern" \
+      --lower_to_aie_pipeline "air" \
+      --tile_pipeline "pad-pack" \
       --lhs_rhs_type "bf16" \
       --acc_type "f32" \
       --m "256"  --k "256" --n "256" \
       --use_ukernel "1"
 fi
 
-# Disabled until the following issue is resolved:
-# https://github.com/Xilinx/llvm-aie/issues/102
-#
-# run_matmul_test \
-#   --name_prefix "transpose_int32" \
-#   --lhs_rhs_type "i32" \
-#   --acc_type "i32" \
-#   --m "8" --n "16" --k "32" \
-#   --do_transpose_rhs "1"
-
-
-run_matmul_test \
-  --name_prefix "transpose_i8_i32" \
-  --lhs_rhs_type "i8" \
-  --acc_type "i32" \
-  --m "16" --n "32" --k "64" \
-  --do_transpose_rhs "1"
-
-run_matmul_test \
-  --name_prefix "transpose_bf16" \
-  --lhs_rhs_type "bf16" \
-  --acc_type "f32" \
-  --m "256" --n "256" --k "256" \
-  --do_transpose_rhs "1"
-
-# The below matmul case passes with
-# tile_sizes = [[1, 1], [0, 0, 250], [1, 1], [0, 0, 2]], packedSizes = [1, 1, 5]
-# but fails with tile_sizes = [[1, 1], [0, 0, 200], [1, 1], [0, 0, 1]], packedSizes = [1, 1, 8],
-# with the error LLVM ERROR: unable to legalize instruction: %152:_(<2 x s32>) = G_FMUL %148:_, %150:_ (in function: core_0_2)
-# The later is what a more vectorization friendly packing looks like so this test is expected failing the test here.
-# TODO: check if the test will pass with a more recent llvm-aie and if it doesnt, report it upstream.
-# Disabled until the following issue is resolved:
-# https://github.com/Xilinx/llvm-aie/issues/102
-# run_matmul_test \
-#    --name_prefix "failure_0" \
-#    --lhs_rhs_type "i32" \
-#    --acc_type "i32" \
-#    --m "1"  --n "1" --k "1000" \
-#    --expect_compile_failure "1"
-
-# The below matmul case passes with
-# tile_sizes = [52, 52], [0, 0, 63], [26, 26], [0, 0, 3], packedSizes = [2, 2, 7]
-# but fails with tile_sizes = [[52, 52], [0, 0, 63], [4, 4], [0, 0, 3]], packedSizes = [4, 4, 7],
-# in AIRHerdPlacementPass with the error No valid placement found
-# The later is what a more vectorization friendly packing looks like so we are expected failing the test here.
-# We should fix this failure.
-# run_matmul_test \
-#    --name_prefix "failure_0" \
-#    --lhs_rhs_type "i32" \
-#    --acc_type "i32" \
-#    --m "52"  --n "52" --k "63" \
-#    --expect_compile_failure "1"
-
 # Example of a run with a group of 2+ matmuls. Currently this test is passed
 # the flag '--num_repeat_runs 0" as there is currently an issue with the runtime if
 # multiple matmuls are run in the same test. TODO(newling/nmeshram): Document
 # this issue.
 run_matmul_test \
     --name_prefix "multiple_matmuls" \
+    --lower_to_aie_pipeline "air" \
+    --tile_pipeline "pad-pack" \
     --lhs_rhs_type "i32" \
     --acc_type "i32" \
     --m "512,8,16" \
@@ -612,105 +567,28 @@ run_matmul_test \
     --num_repeat_runs "0"
 
 run_matmul_test \
-    --name_prefix "small" \
-    --lhs_rhs_type "i32" \
-    --acc_type "i32" \
-    --m "16"  --n "16" --k "8"
-
-run_matmul_test \
-    --name_prefix "small" \
-    --lhs_rhs_type "i32" \
-    --acc_type "i32" \
-    --m "8"  --n "32" --k "16"
-
-# Disabled until the following issue is resolved:
-# https://github.com/Xilinx/llvm-aie/issues/102
-# run_matmul_test \
-#     --name_prefix "small" \
-#     --lhs_rhs_type "i32" \
-#     --acc_type "i32" \
-#     --m "9"  --n "7" --k "16"
-
-run_matmul_test \
-    --name_prefix "large" \
-    --lhs_rhs_type "i32" \
-    --acc_type "i32" \
-    --m "64"  --n "64" --k "128"
-
-run_matmul_test \
-    --name_prefix "large" \
-    --lhs_rhs_type "i32" \
-    --acc_type "i32" \
-    --m "512"  --n "512" --k "512"
-
-run_matmul_test \
-    --name_prefix "int8" \
-    --lhs_rhs_type "i8" \
-    --acc_type "i32" \
-    --m "64"  --n "64" --k "64"
-
-run_matmul_test \
-    --name_prefix "bf16_2304" \
-    --lhs_rhs_type "bf16" \
-    --acc_type "f32" \
-    --m "128"  --n "128" --k "2304"
+  --name_prefix "transpose_i8_i32" \
+  --lower_to_aie_pipeline "air" \
+  --tile_pipeline "pad-pack" \
+  --lhs_rhs_type "i8" \
+  --acc_type "i32" \
+  --m "16" --n "32" --k "64" \
+  --do_transpose_rhs "1"
 
 run_matmul_test \
-    --name_prefix "packPeel" \
+    --name_prefix "packPeel_i32" \
     --tile_pipeline "pack-peel" \
     --lhs_rhs_type "i32" \
     --acc_type "i32" \
     --m "64"  --n "64" --k "128"
 
-# We're seeing intermittent numerical errors in these 3 tests,
-# needs investigation. TODO(newling/yzhang93): Add more info.
-# Appears to be only pack-peel pipeline with bf16->f32.
-# Using 'num_repeat_runs=0' flag to avoid running the numerical test.
-#################################################################
-
-
-# TODO: compilation error with the below test.
-#
-# error: 'aie.dma_bd' op Cannot give more than 3 dimensions for step sizes and wraps in this  tile (got 4 dimensions).
-#
-# The config generated with the current strategy is:
-#
-# packing_config = #amdaie.packing_config<packing_config =
-#   [{packedSizes = [64, 64, 64],
-#     transposePackIndices = [1],
-#     unpackEmpty = [false],
-#     innerPerm = [[1, 0]],
-#     outerPerm = [[0, 1]]},
-#     {
-#       packedSizes = [0, 0, 0, 4, 4, 8],
-#       transposePackIndices = [0, 1, 2],
-#       unpackEmpty = [false, false, true],
-#       innerPerm = [[0, 1], [1, 0], [0, 1]],
-#       outerPerm = [[0, 1, 3, 2], [0, 1, 3, 2], [0, 1, 3, 2]]}]>
-#     }
-run_matmul_test \
-    --name_prefix "packPeel" \
-    --tile_pipeline "pack-peel" \
-    --lhs_rhs_type "bf16" \
-    --acc_type "f32" \
-    --m "64"  --n "64" --k "128" \
-    --num_repeat_runs "0"
-
 run_matmul_test \
-    --name_prefix "packPeelLarge" \
+    --name_prefix "packPeel_bf16" \
     --tile_pipeline "pack-peel" \
     --lhs_rhs_type "bf16" \
     --acc_type "f32" \
     --m "512"  --n "512" --k "512"
 
-run_matmul_test \
-    --name_prefix "packPeel2304" \
-    --tile_pipeline "pack-peel" \
-    --lhs_rhs_type "bf16" \
-    --acc_type "f32" \
-    --m "128"  --n "128" --k "2304"
-
-
 run_matmul_test \
   --name_prefix "packPeel_t_bf16" \
   --tile_pipeline "pack-peel" \
@@ -719,56 +597,6 @@ run_matmul_test \
   --m "128" --n "256" --k "512" \
   --do_transpose_rhs "1"
 
-###################################################################
-
-run_matmul_test \
-    --name_prefix "mm2" \
-    --lhs_rhs_type "bf16" \
-    --acc_type "f32" \
-    --m "308"  --k "9728" --n "2432"
-
-run_matmul_test \
-    --name_prefix "mm3" \
-    --lhs_rhs_type "bf16" \
-    --acc_type "f32" \
-    --m "308"  --k "2432" --n "2432"
-
-run_matmul_test \
-     --name_prefix "mm4" \
-     --lhs_rhs_type "bf16" \
-     --acc_type "f32" \
-     --m "308"  --k "2432" --n "7296"
-
-run_matmul_test \
-     --name_prefix "mm5" \
-     --lhs_rhs_type "bf16" \
-     --acc_type "f32" \
-     --m "8192" --k "2432" --n "9728"
-
-run_matmul_test \
-    --name_prefix "mm6" \
-    --lhs_rhs_type "bf16" \
-    --acc_type "f32" \
-    --m "308"  --k "2432" --n "9728"
-
-run_matmul_test \
-    --name_prefix "mm7" \
-    --lhs_rhs_type "bf16" \
-    --acc_type "f32" \
-    --m "8192" --k "2432" --n "2432"
-
-run_matmul_test \
-     --name_prefix "mm8" \
-     --lhs_rhs_type "bf16" \
-     --acc_type "f32" \
-     --m "8192" --k "9728" --n "2432"
-
-run_matmul_test \
-    --name_prefix "mm9" \
-    --lhs_rhs_type "bf16" \
-    --acc_type "f32" \
-    --m "8192" --k "2432" --n "7296"
-
 ###################################################################
 # ObjectFifo Matmul tests
 ###################################################################
@@ -805,7 +633,7 @@ i32_shapes_medium=(
 )
 
 run_matmul_test_on_shapes ${i32_shapes_small[@]} \
-    --name_prefix "small" \
+    --name_prefix "small_i32" \
     --lower_to_aie_pipeline "objectFifo" \
     --tile_pipeline "pack-peel" \
     --lhs_rhs_type "i32" \
@@ -820,7 +648,7 @@ if [ "$OSTYPE" != "msys" ]; then
 fi
 
 run_matmul_test_on_shapes ${i32_shapes_medium[@]} \
-    --name_prefix "medium" \
+    --name_prefix "medium_i32" \
     --lower_to_aie_pipeline "objectFifo" \
     --tile_pipeline "pack-peel" \
     --lhs_rhs_type "i32" \
@@ -852,7 +680,7 @@ bf16_ukernel_shapes_medium=(
 )
 
 run_matmul_test_on_shapes ${bf16_i8_shapes_small[@]} \
-    --name_prefix "small" \
+    --name_prefix "small_bf16" \
     --lower_to_aie_pipeline "objectFifo" \
     --tile_pipeline "pack-peel" \
     --lhs_rhs_type "bf16" \
@@ -860,7 +688,7 @@ run_matmul_test_on_shapes ${bf16_i8_shapes_small[@]} \
     --num_repeat_runs "2"
 
 run_matmul_test_on_shapes ${bf16_i8_shapes_medium[@]} \
-    --name_prefix "medium" \
+    --name_prefix "medium_bf16" \
     --lower_to_aie_pipeline "objectFifo" \
     --tile_pipeline "pack-peel" \
     --lhs_rhs_type "bf16" \
@@ -869,7 +697,7 @@ run_matmul_test_on_shapes ${bf16_i8_shapes_medium[@]} \
 
 # i8 Matmul tests.
 run_matmul_test_on_shapes ${bf16_i8_shapes_small[@]} \
-    --name_prefix "small" \
+    --name_prefix "small_i8" \
     --lower_to_aie_pipeline "objectFifo" \
     --tile_pipeline "pack-peel" \
     --lhs_rhs_type "i8" \
@@ -877,7 +705,7 @@ run_matmul_test_on_shapes ${bf16_i8_shapes_small[@]} \
     --num_repeat_runs "2"
 
 run_matmul_test_on_shapes ${bf16_i8_shapes_medium[@]} \
-    --name_prefix "medium" \
+    --name_prefix "medium_i8" \
     --lower_to_aie_pipeline "objectFifo" \
     --tile_pipeline "pack-peel" \
     --lhs_rhs_type "i8" \
@@ -886,7 +714,7 @@ run_matmul_test_on_shapes ${bf16_i8_shapes_medium[@]} \
 
 if [ -d "$VITIS" ]; then
   run_matmul_test_on_shapes ${bf16_ukernel_shapes_small[@]} \
-      --name_prefix "small" \
+      --name_prefix "small_ukern" \
       --lower_to_aie_pipeline "objectFifo" \
       --tile_pipeline "pack-peel" \
       --lhs_rhs_type "bf16" \
@@ -895,7 +723,7 @@ if [ -d "$VITIS" ]; then
       --use_ukernel "1"
 
   run_matmul_test_on_shapes ${bf16_ukernel_shapes_medium[@]} \
-      --name_prefix "medium" \
+      --name_prefix "medium_ukern" \
       --lower_to_aie_pipeline "objectFifo" \
       --tile_pipeline "pack-peel" \
       --lhs_rhs_type "bf16" \
diff --git a/compiler/plugins/target/AMD-AIE/iree-amd-aie/Transforms/AMDAIEAddLoweringStrategy.cpp b/compiler/plugins/target/AMD-AIE/iree-amd-aie/Transforms/AMDAIEAddLoweringStrategy.cpp
index 4ca8649f7..19f9fa9d4 100644
--- a/compiler/plugins/target/AMD-AIE/iree-amd-aie/Transforms/AMDAIEAddLoweringStrategy.cpp
+++ b/compiler/plugins/target/AMD-AIE/iree-amd-aie/Transforms/AMDAIEAddLoweringStrategy.cpp
@@ -64,13 +64,10 @@ void AMDAIELoweringStrategyPass::runOnOperation() {
     }
   }
 
-  // To simplify development, the number of cores can be passed as a flag during
-  // compilation. In the future these parameters could be read from file.
-  struct AIEConfig cfg = {numCores};
   for (auto funcOp : moduleOp.getOps<FunctionOpInterface>()) {
     // Set the strategy with default heuristics.
     if (failed(initAIELaunchConfig(funcOp, usePassPipeline,
-                                   useLowerToAIEPipeline, cfg))) {
+                                   useLowerToAIEPipeline))) {
       funcOp.emitOpError("failed to have a lowering configuration set for it.");
       return signalPassFailure();
     }
diff --git a/compiler/plugins/target/AMD-AIE/iree-amd-aie/Transforms/KernelDispatch.cpp b/compiler/plugins/target/AMD-AIE/iree-amd-aie/Transforms/KernelDispatch.cpp
index f8f4773ef..1d564be37 100644
--- a/compiler/plugins/target/AMD-AIE/iree-amd-aie/Transforms/KernelDispatch.cpp
+++ b/compiler/plugins/target/AMD-AIE/iree-amd-aie/Transforms/KernelDispatch.cpp
@@ -312,8 +312,7 @@ static SmallVector<int64_t> setInnerPermB(bool isMatmulTransposeB) {
 
 static LogicalResult setRootConfigForPackPeelPipeline(
     mlir::FunctionOpInterface entryPointFn, linalg::LinalgOp linalgOp,
-    LowerToAIEPassPipeline useLowerToAIEPipeline, AIEConfig cfg,
-    bool isMatmulTransposeB) {
+    LowerToAIEPassPipeline useLowerToAIEPipeline, bool isMatmulTransposeB) {
   bool isObjectFifo =
       useLowerToAIEPipeline == LowerToAIEPassPipeline::ObjectFifo;
   auto maybePackPeelTiling =
@@ -389,7 +388,7 @@ static LogicalResult setRootConfigForPackPeelPipeline(
 
 static LogicalResult setRootConfigForPadPackPipeline(
     mlir::FunctionOpInterface entryPointFn, linalg::LinalgOp linalgOp,
-    AIEConfig cfg, bool isMatmulTransposeB) {
+    bool isMatmulTransposeB) {
   auto maybePadPackTiling = ParameterSetting::create(
       linalgOp, /*isPackPeel=*/false, /*isObjectFifo=*/false);
   if (failed(maybePadPackTiling)) return failure();
@@ -445,8 +444,7 @@ static LogicalResult setRootConfigForPadPackPipeline(
 //===----------------------------------------------------------------------===//
 
 static LogicalResult setRootConfigForConvDecomposePipeline(
-    mlir::FunctionOpInterface entryPointFn, linalg::LinalgOp linalgOp,
-    AIEConfig cfg) {
+    mlir::FunctionOpInterface entryPointFn, linalg::LinalgOp linalgOp) {
   FailureOr<std::array<uint32_t, 3>> maybeInstructionSize =
       getMatmulInstructionSize(linalgOp);
   int64_t OW = 4;
@@ -606,13 +604,13 @@ static bool isMatmulTransposeB(linalg::GenericOp genericOp) {
 /// transposition.
 static LogicalResult setTransposeLikeOpRootConfig(
     mlir::FunctionOpInterface entryPointFn, linalg::LinalgOp linalgOp,
-    TilePassPipeline passPipeline, LowerToAIEPassPipeline useLowerToAIEPipeline,
-    AIEConfig cfg) {
+    TilePassPipeline passPipeline,
+    LowerToAIEPassPipeline useLowerToAIEPipeline) {
   if (passPipeline == TilePassPipeline::PackPeelPipeline)
     return setRootConfigForPackPeelPipeline(entryPointFn, linalgOp,
-                                            useLowerToAIEPipeline, cfg, true);
+                                            useLowerToAIEPipeline, true);
   else if (passPipeline == TilePassPipeline::PadPackPipeline)
-    return setRootConfigForPadPackPipeline(entryPointFn, linalgOp, cfg, true);
+    return setRootConfigForPadPackPipeline(entryPointFn, linalgOp, true);
   return linalgOp.emitError(
       "Unhandled pass pipeline in setTransposeLikeOpRootConfig.");
 }
@@ -621,17 +619,16 @@ static LogicalResult setTransposeLikeOpRootConfig(
 // Root Configurations
 //===----------------------------------------------------------------------===//
 
-static LogicalResult setRootConfig(mlir::FunctionOpInterface entryPointFn,
-                                   linalg::GenericOp genericOp,
-                                   TilePassPipeline passPipeline,
-                                   LowerToAIEPassPipeline useLowerToAIEPipeline,
-                                   AIEConfig cfg) {
+static LogicalResult setRootConfig(
+    mlir::FunctionOpInterface entryPointFn, linalg::GenericOp genericOp,
+    TilePassPipeline passPipeline,
+    LowerToAIEPassPipeline useLowerToAIEPipeline) {
   assert(!getLoweringConfig<IREE::Codegen::LoweringConfigAttr>(genericOp) &&
          "expected lowering_config is not set");
 
   if (isMatmulTransposeB(genericOp) &&
       succeeded(setTransposeLikeOpRootConfig(
-          entryPointFn, genericOp, passPipeline, useLowerToAIEPipeline, cfg))) {
+          entryPointFn, genericOp, passPipeline, useLowerToAIEPipeline))) {
     return success();
   }
 
@@ -640,18 +637,16 @@ static LogicalResult setRootConfig(mlir::FunctionOpInterface entryPointFn,
 
 /// Sets the lowering configuration for dispatch region with root op that
 /// implements the contraction operation interface.
-static LogicalResult setRootConfig(mlir::FunctionOpInterface entryPointFn,
-                                   linalg::ContractionOpInterface contractionOp,
-                                   TilePassPipeline passPipeline,
-                                   LowerToAIEPassPipeline useLowerToAIEPipeline,
-                                   AIEConfig cfg) {
+static LogicalResult setRootConfig(
+    mlir::FunctionOpInterface entryPointFn,
+    linalg::ContractionOpInterface contractionOp, TilePassPipeline passPipeline,
+    LowerToAIEPassPipeline useLowerToAIEPipeline) {
   assert(!getLoweringConfig<IREE::Codegen::LoweringConfigAttr>(contractionOp) &&
          "expected lowering_config is not set");
   auto linalgOp = cast<linalg::LinalgOp>(contractionOp.getOperation());
   if (isa<linalg::MatmulTransposeBOp>(linalgOp)) {
-    if (succeeded(setTransposeLikeOpRootConfig(entryPointFn, linalgOp,
-                                               passPipeline,
-                                               useLowerToAIEPipeline, cfg))) {
+    if (succeeded(setTransposeLikeOpRootConfig(
+            entryPointFn, linalgOp, passPipeline, useLowerToAIEPipeline))) {
       return success();
     }
     return failure();
@@ -672,31 +667,30 @@ static LogicalResult setRootConfig(mlir::FunctionOpInterface entryPointFn,
   // approach which will have different tile sizes and pass pipelines
   if (passPipeline == TilePassPipeline::PackPeelPipeline)
     return setRootConfigForPackPeelPipeline(entryPointFn, linalgOp,
-                                            useLowerToAIEPipeline, cfg, false);
+                                            useLowerToAIEPipeline, false);
   if (passPipeline == TilePassPipeline::PadPackPipeline)
-    return setRootConfigForPadPackPipeline(entryPointFn, linalgOp, cfg, false);
+    return setRootConfigForPadPackPipeline(entryPointFn, linalgOp, false);
   return linalgOp.emitError("Unhandled pass pipeline in setRootConfig.");
 }
 
 static LogicalResult setConvRootConfig(mlir::FunctionOpInterface entryPointFn,
                                        linalg::ConvolutionOpInterface convOp,
-                                       TilePassPipeline passPipeline,
-                                       AIEConfig cfg) {
+                                       TilePassPipeline passPipeline) {
   assert(!getLoweringConfig<IREE::Codegen::LoweringConfigAttr>(convOp) &&
          "expected lowering_config is not set");
   auto linalgOp = cast<linalg::LinalgOp>(convOp.getOperation());
 
   // Current tiling strategy is based on llvm-cpu ConvTileAndDecomposeExpert.
   if (passPipeline == TilePassPipeline::ConvDecomposePipeline)
-    return setRootConfigForConvDecomposePipeline(entryPointFn, linalgOp, cfg);
+    return setRootConfigForConvDecomposePipeline(entryPointFn, linalgOp);
   return linalgOp.emitError("Unhandled pass pipeline in setConvRootConfig.");
 }
 
 /// Redirects to methods that set the configuration based on operation type.
 static LogicalResult setRootConfigImpl(
     mlir::FunctionOpInterface entryPointFn, Operation *op,
-    TilePassPipeline passPipeline, LowerToAIEPassPipeline useLowerToAIEPipeline,
-    AIEConfig cfg) {
+    TilePassPipeline passPipeline,
+    LowerToAIEPassPipeline useLowerToAIEPipeline) {
   auto setRootConfigFn = [&](Operation *op) -> LogicalResult {
     return TypeSwitch<Operation *, LogicalResult>(op)
         // TODO (nmeshram): This is very limited for now, plan is to
@@ -706,15 +700,15 @@ static LogicalResult setRootConfigImpl(
         .Case<linalg::Conv2DNhwcHwcfOp, linalg::Conv2DNchwFchwOp,
               linalg::Conv2DNhwcHwcfQOp, linalg::DepthwiseConv2DNhwcHwcOp>(
             [&](auto op) {
-              return setConvRootConfig(entryPointFn, op, passPipeline, cfg);
+              return setConvRootConfig(entryPointFn, op, passPipeline);
             })
         .Case<linalg::GenericOp>([&](auto op) {
           return setRootConfig(entryPointFn, op, passPipeline,
-                               useLowerToAIEPipeline, cfg);
+                               useLowerToAIEPipeline);
         })
         .Case<linalg::ContractionOpInterface>([&](auto op) {
           return setRootConfig(entryPointFn, op, passPipeline,
-                               useLowerToAIEPipeline, cfg);
+                               useLowerToAIEPipeline);
         })
         .Default([&](Operation *op) { return success(); });
   };
@@ -724,8 +718,8 @@ static LogicalResult setRootConfigImpl(
 /// Sets the translation information to use for a dispatch region.
 static LogicalResult setTranslationInfoAndRootConfig(
     mlir::FunctionOpInterface entryPointFn, ArrayRef<Operation *> computeOps,
-    TilePassPipeline passPipeline, LowerToAIEPassPipeline useLowerToAIEPipeline,
-    AIEConfig cfg) {
+    TilePassPipeline passPipeline,
+    LowerToAIEPassPipeline useLowerToAIEPipeline) {
   // Make sure that lowering_config is not preset on any compute ops.
   for (auto computeOp : computeOps) {
     if (getLoweringConfig<IREE::Codegen::LoweringConfigAttr>(computeOp))
@@ -741,7 +735,7 @@ static LogicalResult setTranslationInfoAndRootConfig(
     return entryPointFn.emitError("Case with no root ops not yet supported.");
 
   if (failed(setRootConfigImpl(entryPointFn, rootOperation, passPipeline,
-                               useLowerToAIEPipeline, cfg)))
+                               useLowerToAIEPipeline)))
     return failure();
   return success();
 }
@@ -750,10 +744,9 @@ static LogicalResult setTranslationInfoAndRootConfig(
 // Entry Point
 //===----------------------------------------------------------------------===//
 
-LogicalResult initAIELaunchConfig(FunctionOpInterface funcOp,
-                                  TilePassPipeline passPipeline,
-                                  LowerToAIEPassPipeline useLowerToAIEPipeline,
-                                  AIEConfig cfg) {
+LogicalResult initAIELaunchConfig(
+    FunctionOpInterface funcOp, TilePassPipeline passPipeline,
+    LowerToAIEPassPipeline useLowerToAIEPipeline) {
   if (getTranslationInfo(funcOp)) return success();
 
   // TODO (nmeshram): Need a default pipeline for control flow cases.
@@ -762,7 +755,7 @@ LogicalResult initAIELaunchConfig(FunctionOpInterface funcOp,
 
   SmallVector<Operation *> computeOps = getComputeOps(funcOp);
   if (failed(setTranslationInfoAndRootConfig(funcOp, computeOps, passPipeline,
-                                             useLowerToAIEPipeline, cfg)))
+                                             useLowerToAIEPipeline)))
     return failure();
 
   // The root configuration setting introduces `tensor.dim` operations.
diff --git a/compiler/plugins/target/AMD-AIE/iree-amd-aie/Transforms/KernelDispatch.h b/compiler/plugins/target/AMD-AIE/iree-amd-aie/Transforms/KernelDispatch.h
index 3afec4f7d..879f0882c 100644
--- a/compiler/plugins/target/AMD-AIE/iree-amd-aie/Transforms/KernelDispatch.h
+++ b/compiler/plugins/target/AMD-AIE/iree-amd-aie/Transforms/KernelDispatch.h
@@ -35,16 +35,9 @@ enum class PeelingType { First, Last, FirstLast };
 /// Enum for operands to be bufferized to allocation.
 enum class BufferizeOperand { InputOutput, Input, Output, DefOp };
 
-/// Struct specifying the number of cores to use. This will be replaced
-/// by a more versatile handling in the future.
-struct AIEConfig {
-  int32_t num_cores;
-};
-
 LogicalResult initAIELaunchConfig(FunctionOpInterface funcOp,
                                   TilePassPipeline usePassPipeline,
-                                  LowerToAIEPassPipeline useLowerToAIEPipeline,
-                                  AIEConfig cfg);
+                                  LowerToAIEPassPipeline useLowerToAIEPipeline);
 
 }  // namespace mlir::iree_compiler::AMDAIE
 
diff --git a/compiler/plugins/target/AMD-AIE/iree-amd-aie/Transforms/Passes.cpp b/compiler/plugins/target/AMD-AIE/iree-amd-aie/Transforms/Passes.cpp
index cb3d87425..d5243c86d 100644
--- a/compiler/plugins/target/AMD-AIE/iree-amd-aie/Transforms/Passes.cpp
+++ b/compiler/plugins/target/AMD-AIE/iree-amd-aie/Transforms/Passes.cpp
@@ -52,7 +52,7 @@ static llvm::cl::opt<LowerToAIEPassPipeline> clUseLowerToAIEPipeline(
                      clEnumValN(LowerToAIEPassPipeline::ObjectFifo,
                                 "objectFifo",
                                 "Use the IREE lowering to objectFifos")),
-    llvm::cl::init(LowerToAIEPassPipeline::AIR));
+    llvm::cl::init(LowerToAIEPassPipeline::ObjectFifo));
 
 /// Command line option for selecting the lowering pipeline to use tiling
 /// computations and packing data.
@@ -69,11 +69,7 @@ static llvm::cl::opt<TilePassPipeline> clUseTilePipeline(
         clEnumValN(TilePassPipeline::ConvDecomposePipeline, "conv-decompose",
                    "Use the conv-decompose based lowering strategy for "
                    "convolution interface ops")),
-    llvm::cl::init(TilePassPipeline::PadPackPipeline));
-
-static llvm::cl::opt<int32_t> clNumCores(
-    "iree-amdaie-num-cores",
-    llvm::cl::desc("Choose the number of cores to use"), llvm::cl::init(1));
+    llvm::cl::init(TilePassPipeline::PackPeelPipeline));
 
 static llvm::cl::opt<std::string> clPathToUkernels(
     "iree-amdaie-path-to-ukernels",
@@ -553,7 +549,6 @@ void buildAMDAIETransformPassPipeline(OpPassManager &variantPassManager,
     AMDAIELoweringStrategyOptions options;
     options.usePassPipeline = clUseTilePipeline;
     options.useLowerToAIEPipeline = clUseLowerToAIEPipeline;
-    options.numCores = clNumCores;
     modulePassManager.addPass(createAMDAIELoweringStrategyPass(options));
   }
   modulePassManager.addPass(createLowerExecutableUsingTransformDialectPass());
diff --git a/compiler/plugins/target/AMD-AIE/iree-amd-aie/Transforms/Passes.td b/compiler/plugins/target/AMD-AIE/iree-amd-aie/Transforms/Passes.td
index 9f6560870..c1ec4c15c 100644
--- a/compiler/plugins/target/AMD-AIE/iree-amd-aie/Transforms/Passes.td
+++ b/compiler/plugins/target/AMD-AIE/iree-amd-aie/Transforms/Passes.td
@@ -290,7 +290,7 @@ def AMDAIELowerExecutableTarget :
   let options = [
     Option<"usePassPipeline", "use-pass-pipeline",
       "mlir::iree_compiler::AMDAIE::TilePassPipeline",
-      /*default=*/"mlir::iree_compiler::AMDAIE::TilePassPipeline::PadPackPipeline",
+      /*default=*/"mlir::iree_compiler::AMDAIE::TilePassPipeline::PackPeelPipeline",
       "Pass pipeline to use while lowering to AIR dialect",
       [{::llvm::cl::values(
         clEnumValN(mlir::iree_compiler::AMDAIE::TilePassPipeline::PackPeelPipeline, "pack-peel",
@@ -310,7 +310,7 @@ def AMDAIELoweringStrategy :
   let options = [
     Option<"usePassPipeline", "use-pass-pipeline",
       "mlir::iree_compiler::AMDAIE::TilePassPipeline",
-      /*default=*/"mlir::iree_compiler::AMDAIE::TilePassPipeline::PadPackPipeline",
+      /*default=*/"mlir::iree_compiler::AMDAIE::TilePassPipeline::PackPeelPipeline",
       "Pass pipeline to use while lowering to AIR dialect",
       [{::llvm::cl::values(
         clEnumValN(mlir::iree_compiler::AMDAIE::TilePassPipeline::PackPeelPipeline, "pack-peel",
@@ -320,11 +320,9 @@ def AMDAIELoweringStrategy :
         clEnumValN(mlir::iree_compiler::AMDAIE::TilePassPipeline::ConvDecomposePipeline, "conv-decompose",
                    "Use the conv-decompose based lowering strategy for convolution interface ops.")
       )}]>,
-    Option<"numCores", "num-cores", "int32_t", /*default=*/"1",
-      "Choose the number of cores to use">,
     Option<"useLowerToAIEPipeline", "use-lower-to-aie-pipeline",
       "mlir::iree_compiler::AMDAIE::LowerToAIEPassPipeline",
-      /*default=*/"mlir::iree_compiler::AMDAIE::LowerToAIEPassPipeline::AIR",
+      /*default=*/"mlir::iree_compiler::AMDAIE::LowerToAIEPassPipeline::ObjectFifo",
       "Lowering pass pipeline to use",
       [{::llvm::cl::values(
         clEnumValN(mlir::iree_compiler::AMDAIE::LowerToAIEPassPipeline::ObjectFifo, "objectFifo",
diff --git a/compiler/plugins/target/AMD-AIE/iree-amd-aie/Transforms/test/lowering_strategy.mlir b/compiler/plugins/target/AMD-AIE/iree-amd-aie/Transforms/test/lowering_strategy.mlir
index 872d627a2..8c7546d78 100644
--- a/compiler/plugins/target/AMD-AIE/iree-amd-aie/Transforms/test/lowering_strategy.mlir
+++ b/compiler/plugins/target/AMD-AIE/iree-amd-aie/Transforms/test/lowering_strategy.mlir
@@ -1,5 +1,5 @@
-// RUN: iree-opt --split-input-file --pass-pipeline='builtin.module(iree-amdaie-lowering-strategy{use-pass-pipeline=pad-pack})' %s | FileCheck %s --check-prefix=CHECK-PAD-PACK
-// RUN: iree-opt --split-input-file --pass-pipeline='builtin.module(iree-amdaie-lowering-strategy{use-pass-pipeline=pack-peel})' %s | FileCheck %s --check-prefix=CHECK-PACK-PEEL
+// RUN: iree-opt --split-input-file --pass-pipeline='builtin.module(iree-amdaie-lowering-strategy{use-lower-to-aie-pipeline=air use-pass-pipeline=pad-pack})' %s | FileCheck %s --check-prefix=CHECK-PAD-PACK
+// RUN: iree-opt --split-input-file --pass-pipeline='builtin.module(iree-amdaie-lowering-strategy{use-lower-to-aie-pipeline=air use-pass-pipeline=pack-peel})' %s | FileCheck %s --check-prefix=CHECK-PACK-PEEL
 
 // CHECK-PAD-PACK{LITERAL}: #config = #iree_codegen.lowering_config<tile_sizes = [[64, 64], [0, 0, 256], [16, 16], [0, 0, 2]]>
 // CHECK-PAD-PACK{LITERAL}: #packingConfig = #amdaie.packing_config<packing_config = [{packedSizes = [4, 4, 8], transposePackIndices = [0, 1, 2], unpackEmpty = [false, false, true], innerPerm = [[0, 1], [1, 0], [0, 1]], outerPerm = [[1, 0], [1, 0], [1, 0]]}]>
diff --git a/tests/samples/CMakeLists.txt b/tests/samples/CMakeLists.txt
index 33d0e5646..618409664 100644
--- a/tests/samples/CMakeLists.txt
+++ b/tests/samples/CMakeLists.txt
@@ -8,12 +8,12 @@ iree_lit_test_suite(
   NAME
     lit
   SRCS
-    "conv_pipeline_e2e.mlir"
-    "matmul_peeled_objectfifo.mlir"
-    "matmul_peeled_objectfifo_e2e.mlir"
-    "pack_peel_pipeline_matmul.mlir"
-    "pack_peel_pipeline_matmul_elementwise.mlir"
-    "pad_pack_pipeline_e2e.mlir"
+    "conv2d_nhwc_air_e2e.mlir"
+    "matmul_elementwise_pack_peel_air_e2e.mlir"
+    "matmul_pack_peel_air_e2e.mlir"
+    "matmul_pack_peel_objectfifo.mlir"
+    "matmul_pack_peel_objectfifo_e2e.mlir"
+    "matmul_pad_pack_air_e2e.mlir"
     "xdna_oplib_plugin.mlir"
   TOOLS
     ${IREE_LLD_TARGET}
diff --git a/tests/samples/conv_pipeline_e2e.mlir b/tests/samples/conv2d_nhwc_air_e2e.mlir
similarity index 95%
rename from tests/samples/conv_pipeline_e2e.mlir
rename to tests/samples/conv2d_nhwc_air_e2e.mlir
index ffe8222f8..2b005150a 100644
--- a/tests/samples/conv_pipeline_e2e.mlir
+++ b/tests/samples/conv2d_nhwc_air_e2e.mlir
@@ -1,4 +1,4 @@
-// RUN: iree-compile --iree-hal-target-backends=amd-aie --compile-to=executable-sources %s | iree-opt --pass-pipeline="builtin.module(hal.executable(hal.executable.variant(iree-hal-translate-target-executable-variants{target=amd-aie})))" --iree-amdaie-tile-pipeline=conv-decompose --split-input-file | FileCheck %s
+// RUN: iree-compile --iree-hal-target-backends=amd-aie --compile-to=executable-sources %s | iree-opt --pass-pipeline="builtin.module(hal.executable(hal.executable.variant(iree-hal-translate-target-executable-variants{target=amd-aie})))" --iree-amdaie-tile-pipeline=conv-decompose --iree-amdaie-lower-to-aie-pipeline=air --split-input-file | FileCheck %s
 
 func.func @conv_2d_nhwc_hwcf(%arg0: tensor<2x14x14x32xi32>, %arg1: tensor<3x3x32x64xi32>) -> tensor<2x12x12x64xi32> {
   %cst = arith.constant 0 : i32
diff --git a/tests/samples/pack_peel_pipeline_matmul_elementwise.mlir b/tests/samples/matmul_elementwise_pack_peel_air_e2e.mlir
similarity index 95%
rename from tests/samples/pack_peel_pipeline_matmul_elementwise.mlir
rename to tests/samples/matmul_elementwise_pack_peel_air_e2e.mlir
index c99b3b269..2f666db91 100644
--- a/tests/samples/pack_peel_pipeline_matmul_elementwise.mlir
+++ b/tests/samples/matmul_elementwise_pack_peel_air_e2e.mlir
@@ -1,4 +1,4 @@
-// RUN: iree-compile --iree-hal-target-backends=amd-aie --compile-to=executable-sources %s | iree-opt --pass-pipeline="builtin.module(hal.executable(hal.executable.variant(iree-hal-translate-target-executable-variants{target=amd-aie})))" --iree-amdaie-tile-pipeline=pack-peel --iree-amdaie-matmul-elementwise-fusion --split-input-file | FileCheck %s
+// RUN: iree-compile --iree-hal-target-backends=amd-aie --compile-to=executable-sources %s | iree-opt --pass-pipeline="builtin.module(hal.executable(hal.executable.variant(iree-hal-translate-target-executable-variants{target=amd-aie})))" --iree-amdaie-lower-to-aie-pipeline=air --iree-amdaie-tile-pipeline=pack-peel --iree-amdaie-matmul-elementwise-fusion --split-input-file | FileCheck %s
 
 func.func @matmul_elementwise_i32(%lhs: tensor<1024x512xi32>, %rhs: tensor<512x1024xi32>, %ele: tensor<1024x1024xi32>) -> tensor<1024x1024xi32>
 {
diff --git a/tests/samples/pack_peel_pipeline_matmul.mlir b/tests/samples/matmul_pack_peel_air_e2e.mlir
similarity index 91%
rename from tests/samples/pack_peel_pipeline_matmul.mlir
rename to tests/samples/matmul_pack_peel_air_e2e.mlir
index a626a2132..e29ded73e 100644
--- a/tests/samples/pack_peel_pipeline_matmul.mlir
+++ b/tests/samples/matmul_pack_peel_air_e2e.mlir
@@ -1,4 +1,4 @@
-// RUN: iree-compile --iree-hal-target-backends=amd-aie --compile-to=executable-sources %s | iree-opt --pass-pipeline="builtin.module(hal.executable(hal.executable.variant(iree-hal-translate-target-executable-variants{target=amd-aie})))" --iree-amdaie-tile-pipeline=pack-peel --split-input-file | FileCheck %s
+// RUN: iree-compile --iree-hal-target-backends=amd-aie --compile-to=executable-sources %s | iree-opt --pass-pipeline="builtin.module(hal.executable(hal.executable.variant(iree-hal-translate-target-executable-variants{target=amd-aie})))" --iree-amdaie-lower-to-aie-pipeline=air --iree-amdaie-tile-pipeline=pack-peel --split-input-file | FileCheck %s
 
 func.func @matmul_i8_i32(%lhs: tensor<32x16xi8>, %rhs: tensor<16x32xi8>) -> tensor<32x32xi32>
 {
diff --git a/tests/samples/matmul_peeled_objectfifo.mlir b/tests/samples/matmul_pack_peel_objectfifo.mlir
similarity index 100%
rename from tests/samples/matmul_peeled_objectfifo.mlir
rename to tests/samples/matmul_pack_peel_objectfifo.mlir
diff --git a/tests/samples/matmul_peeled_objectfifo_e2e.mlir b/tests/samples/matmul_pack_peel_objectfifo_e2e.mlir
similarity index 100%
rename from tests/samples/matmul_peeled_objectfifo_e2e.mlir
rename to tests/samples/matmul_pack_peel_objectfifo_e2e.mlir
diff --git a/tests/samples/pad_pack_pipeline_e2e.mlir b/tests/samples/matmul_pad_pack_air_e2e.mlir
similarity index 97%
rename from tests/samples/pad_pack_pipeline_e2e.mlir
rename to tests/samples/matmul_pad_pack_air_e2e.mlir
index 14bdcb04c..90ef20392 100644
--- a/tests/samples/pad_pack_pipeline_e2e.mlir
+++ b/tests/samples/matmul_pad_pack_air_e2e.mlir
@@ -1,4 +1,4 @@
-// RUN: iree-compile --iree-hal-target-backends=amd-aie --compile-to=executable-sources %s | iree-opt --pass-pipeline="builtin.module(hal.executable(hal.executable.variant(iree-hal-translate-target-executable-variants{target=amd-aie})))" --iree-amdaie-tile-pipeline=pad-pack --split-input-file | FileCheck %s --check-prefix=CPP
+// RUN: iree-compile --iree-hal-target-backends=amd-aie --compile-to=executable-sources %s | iree-opt --pass-pipeline="builtin.module(hal.executable(hal.executable.variant(iree-hal-translate-target-executable-variants{target=amd-aie})))" --iree-amdaie-tile-pipeline=pad-pack --iree-amdaie-lower-to-aie-pipeline=air --split-input-file | FileCheck %s --check-prefix=CPP
 
 // This test demonstrates Pad-Pack pipeline based e2e lowering.