diff --git a/.github/workflows/benchmark.yml b/.github/workflows/benchmark.yml
index c19ed8495..1c8fd04ed 100644
--- a/.github/workflows/benchmark.yml
+++ b/.github/workflows/benchmark.yml
@@ -1,149 +1,149 @@
-name: Benchmark
-
-on:
-  issue_comment:
-    types: created
-
-jobs:
-  benchmark_base:
-    # On pull requests and if the comment starts with `/run-benchmark`
-    if: github.event.issue.pull_request != null && startsWith(github.event.comment.body, '/run-benchmark')
-    runs-on: self-hosted
-
-    steps:
-      - name: Checkout code
-        uses: actions/checkout@v2
-        with:
-          ref: main
-
-      - name: Get base branch commit ID
-        id: get_base_commit
-        run: echo "BASE_COMMIT_ID=$(git rev-parse HEAD)" > base_commit_id.txt
-
-      - name: Upload base commit ID
-        uses: actions/upload-artifact@v3
-        with:
-          name: base-commit-id
-          path: base_commit_id.txt
-
-      - name: Set up Python
-        uses: actions/setup-python@v2
-        with:
-          python-version: '3.9'
-
-      - name: Create virtual environment
-        run: python -m venv bitblas_benchmark
-
-      - name: Activate virtual environment and install dependencies
-        run: |
-          source bitblas_benchmark/bin/activate
-          python -m pip install --upgrade pip
-          if [ -f requirements-dev.txt ]; then python -m pip install -r requirements-dev.txt; fi
-
-      - name: Install project in wheel mode
-        run: |
-          source bitblas_benchmark/bin/activate
-          python -m pip install .
+# name: Benchmark
+
+# on:
+#   issue_comment:
+#     types: created
+
+# jobs:
+#   benchmark_base:
+#     # On pull requests and if the comment starts with `/run-benchmark`
+#     if: github.event.issue.pull_request != null && startsWith(github.event.comment.body, '/run-benchmark')
+#     runs-on: self-hosted
+
+#     steps:
+#       - name: Checkout code
+#         uses: actions/checkout@v2
+#         with:
+#           ref: main
+
+#       - name: Get base branch commit ID
+#         id: get_base_commit
+#         run: echo "BASE_COMMIT_ID=$(git rev-parse HEAD)" > base_commit_id.txt
+
+#       - name: Upload base commit ID
+#         uses: actions/upload-artifact@v3
+#         with:
+#           name: base-commit-id
+#           path: base_commit_id.txt
+
+#       - name: Set up Python
+#         uses: actions/setup-python@v2
+#         with:
+#           python-version: '3.9'
+
+#       - name: Create virtual environment
+#         run: python -m venv bitblas_benchmark
+
+#       - name: Activate virtual environment and install dependencies
+#         run: |
+#           source bitblas_benchmark/bin/activate
+#           python -m pip install --upgrade pip
+#           if [ -f requirements-dev.txt ]; then python -m pip install -r requirements-dev.txt; fi
+
+#       - name: Install project in wheel mode
+#         run: |
+#           source bitblas_benchmark/bin/activate
+#           python -m pip install .
       
-      - name: Matmul Benchmark
-        run: |
-          source bitblas_benchmark/bin/activate
-          cd benchmark/operators
-          python ./benchmark_ops_matmul.py
-
-  benchmark_head:
-    # On pull requests and if the comment starts with `/run-benchmark`
-    if: github.event.issue.pull_request != null && startsWith(github.event.comment.body, '/run-benchmark')
-    runs-on: self-hosted
-    needs: [benchmark_base]
-
-    steps:
-      - name: Checkout PR branch code
-        uses: actions/checkout@v2
-        with:
-          fetch-depth: 0
-          ref: ${{ github.event.pull_request.head.ref }}
-
-      - name: Get PR branch commit ID
-        id: get_pr_commit
-        run: echo "PR_COMMIT_ID=$(git rev-parse HEAD)" > pr_commit_id.txt
-
-      - name: Upload PR commit ID
-        uses: actions/upload-artifact@v3
-        with:
-          name: pr-commit-id
-          path: pr_commit_id.txt
-
-      - name: Set up Python
-        uses: actions/setup-python@v2
-        with:
-          python-version: '3.9'
-
-      - name: Create virtual environment
-        run: python -m venv bitblas_benchmark
-
-      - name: Activate virtual environment and install dependencies
-        run: |
-          source bitblas_benchmark/bin/activate
-          python -m pip install --upgrade pip
-          if [ -f requirements-dev.txt ]; then python -m pip install -r requirements-dev.txt; fi
-
-      - name: Install project in wheel mode
-        run: |
-          source bitblas_benchmark/bin/activate
-          python -m pip install .
+#       - name: Matmul Benchmark
+#         run: |
+#           source bitblas_benchmark/bin/activate
+#           cd benchmark/operators
+#           python ./benchmark_ops_matmul.py
+
+#   benchmark_head:
+#     # On pull requests and if the comment starts with `/run-benchmark`
+#     if: github.event.issue.pull_request != null && startsWith(github.event.comment.body, '/run-benchmark')
+#     runs-on: self-hosted
+#     needs: [benchmark_base]
+
+#     steps:
+#       - name: Checkout PR branch code
+#         uses: actions/checkout@v2
+#         with:
+#           fetch-depth: 0
+#           ref: ${{ github.event.pull_request.head.ref }}
+
+#       - name: Get PR branch commit ID
+#         id: get_pr_commit
+#         run: echo "PR_COMMIT_ID=$(git rev-parse HEAD)" > pr_commit_id.txt
+
+#       - name: Upload PR commit ID
+#         uses: actions/upload-artifact@v3
+#         with:
+#           name: pr-commit-id
+#           path: pr_commit_id.txt
+
+#       - name: Set up Python
+#         uses: actions/setup-python@v2
+#         with:
+#           python-version: '3.9'
+
+#       - name: Create virtual environment
+#         run: python -m venv bitblas_benchmark
+
+#       - name: Activate virtual environment and install dependencies
+#         run: |
+#           source bitblas_benchmark/bin/activate
+#           python -m pip install --upgrade pip
+#           if [ -f requirements-dev.txt ]; then python -m pip install -r requirements-dev.txt; fi
+
+#       - name: Install project in wheel mode
+#         run: |
+#           source bitblas_benchmark/bin/activate
+#           python -m pip install .
       
-      - name: Matmul Benchmark
-        run: |
-          source bitblas_benchmark/bin/activate
-          cd benchmark/operators
-          python ./benchmark_ops_matmul.py
-
-  benchmark_compare:
-    if: github.event.issue.pull_request != null && contains(github.event.comment.body, '/run-benchmark')
-    needs: [benchmark_base, benchmark_head]
-    runs-on: self-hosted
-
-    steps:
-      - name: Download commit IDs
-        uses: actions/download-artifact@v4.1.7
-        with:
-          name: base-commit-id
-          path: .
-
-      - name: Download PR commit ID
-        uses: actions/download-artifact@v4.1.7
-        with:
-          name: pr-commit-id
-          path: .
-
-      - name: Set up Python
-        uses: actions/setup-python@v2
-        with:
-          python-version: '3.9'
-
-      - name: Create virtual environment
-        run: python -m venv bitblas_benchmark
-
-      - name: Activate virtual environment and install dependencies
-        run: |
-          source bitblas_benchmark/bin/activate
-          python -m pip install --upgrade pip
-          if [ -f requirements-dev.txt ]; then python -m pip install -r requirements-dev.txt; fi
-
-      - name: Compare benchmark results
-        run: |
-          source bitblas_benchmark/bin/activate
-          cd benchmark/operators
-          python ./compare_benchmark.py --base $(cat base_commit_id.txt) --head $(cat pr_commit_id.txt) 2>&1 | tee compare_results.txt
-
-      - name: Authenticate GitHub CLI
-        env:
-          GITHUB_TOKEN: ${{ secrets.GITHUB_TOKEN }}
-        run: |
-          echo "${{ secrets.GITHUB_TOKEN }}" | gh auth login --with-token
-
-      - name: Post benchmark results
-        run: |
-          cat compare_results.txt
-          gh pr comment ${{ github.event.issue.number }} --body "$(cat compare_results.txt)"
+#       - name: Matmul Benchmark
+#         run: |
+#           source bitblas_benchmark/bin/activate
+#           cd benchmark/operators
+#           python ./benchmark_ops_matmul.py
+
+#   benchmark_compare:
+#     if: github.event.issue.pull_request != null && contains(github.event.comment.body, '/run-benchmark')
+#     needs: [benchmark_base, benchmark_head]
+#     runs-on: self-hosted
+
+#     steps:
+#       - name: Download commit IDs
+#         uses: actions/download-artifact@v4.1.7
+#         with:
+#           name: base-commit-id
+#           path: .
+
+#       - name: Download PR commit ID
+#         uses: actions/download-artifact@v4.1.7
+#         with:
+#           name: pr-commit-id
+#           path: .
+
+#       - name: Set up Python
+#         uses: actions/setup-python@v2
+#         with:
+#           python-version: '3.9'
+
+#       - name: Create virtual environment
+#         run: python -m venv bitblas_benchmark
+
+#       - name: Activate virtual environment and install dependencies
+#         run: |
+#           source bitblas_benchmark/bin/activate
+#           python -m pip install --upgrade pip
+#           if [ -f requirements-dev.txt ]; then python -m pip install -r requirements-dev.txt; fi
+
+#       - name: Compare benchmark results
+#         run: |
+#           source bitblas_benchmark/bin/activate
+#           cd benchmark/operators
+#           python ./compare_benchmark.py --base $(cat base_commit_id.txt) --head $(cat pr_commit_id.txt) 2>&1 | tee compare_results.txt
+
+#       - name: Authenticate GitHub CLI
+#         env:
+#           GITHUB_TOKEN: ${{ secrets.GITHUB_TOKEN }}
+#         run: |
+#           echo "${{ secrets.GITHUB_TOKEN }}" | gh auth login --with-token
+
+#       - name: Post benchmark results
+#         run: |
+#           cat compare_results.txt
+#           gh pr comment ${{ github.event.issue.number }} --body "$(cat compare_results.txt)"
diff --git a/3rdparty/tvm b/3rdparty/tvm
index c6be66d56..180359a9a 160000
--- a/3rdparty/tvm
+++ b/3rdparty/tvm
@@ -1 +1 @@
-Subproject commit c6be66d563695bfbaf4f3d46d312e82b6ad9be1d
+Subproject commit 180359a9ab7a04e6b3ad472a12295d9d3056fa95
diff --git a/bitblas/tl/mfma_macro_generator.py b/bitblas/tl/mfma_macro_generator.py
index b6ccc7b2a..2fe8c0061 100644
--- a/bitblas/tl/mfma_macro_generator.py
+++ b/bitblas/tl/mfma_macro_generator.py
@@ -146,7 +146,7 @@ def ldmatrix_a(self, A_local_buf, A_shared_buf, ki, thread_bindings, rk=0):
         WARP_SIZE = self.WARP_SIZE
         block_row_warps = self.block_row_warps
         block_col_warps = self.block_col_warps
-        warp_col_tiles = self.warp_col_tiles
+        warp_row_tiles = self.warp_row_tiles
         warp_cols = self.warp_cols
         chunk = self.chunk
         micro_size_x = self.micro_size_x
@@ -171,13 +171,13 @@ def _warp_ldmatrix_a(
                     for local_id in T.vectorized(local_size_a):
                         row, col = T.meta_var(reverse_index_map(tx, local_id))
                         l, r = (rk * chunk + ki * micro_size_k,
-                                tz * warp_col_tiles + i * micro_size_x)
+                                tz * warp_row_tiles + i * micro_size_x)
                         A_local_buf[i * local_size_a + local_id] = A_shared_buf[l + row, r + col]
             else:
                 for i in T.serial(warp_cols):
                     for local_id in T.vectorized(local_size_a):
                         row, col = T.meta_var(reverse_index_map(tx, local_id))
-                        l, r = (tz * warp_col_tiles + i * micro_size_x,
+                        l, r = (tz * warp_row_tiles + i * micro_size_x,
                                 rk * chunk + ki * micro_size_k)
                         A_local_buf[i * local_size_a + local_id] = A_shared_buf[l + row, r + col]