tilelang.intrinsics.mma_layout module

tilelang.intrinsics.mma_layout.get_swizzle_layout(row_idx, col_idx, row_size, dtype: Union[DataType, str], swizzle_bytes=None)

tilelang.intrinsics.mma_layout.ldmatrix_16x32_to_shared_16x32_layout_a(thread_id, local_id)

tilelang.intrinsics.mma_layout.ldmatrix_16x32_to_shared_16x32_layout_b(thread_id, local_id)

tilelang.intrinsics.mma_layout.ldmatrix_32x16_to_shared_16x32_layout_a(thread_id, local_id)

tilelang.intrinsics.mma_layout.ldmatrix_32x16_to_shared_16x32_layout_b(thread_id, local_id)

tilelang.intrinsics.mma_layout.ldmatrix_32x8_to_shared_16x16_layout(thread_id, local_id)

tilelang.intrinsics.mma_layout.ldmatrix_trans_32x8_to_shared_16x16_layout(thread_id, local_id)

tilelang.intrinsics.mma_layout.make_mma_swizzle_layout(shared_buf, is_smooth: bool = False)

tilelang.intrinsics.mma_layout.mma_32x8_to_shared_16x16_layout(thread_id, local_id)

tilelang.intrinsics.mma_layout.mma_store_32x8_to_shared_16x16_layout(thread_id, local_id)

tilelang.intrinsics.mma_layout.shared_16x16_to_mma_32x8_layout(i, j)

tilelang.intrinsics.mma_layout.shared_16x16_to_mma_32x8_layout_rs(i, j)

tilelang.intrinsics.mma_layout.shared_16x16_to_mma_32x8_layout_sr(i, j)

tilelang.intrinsics.mma_layout.shared_16x16_to_mma_32x8_layout_trans(i, j)

tilelang.intrinsics.mma_layout.shared_16x16_to_mma_32x8_smoothlayout(i, j)

tilelang.intrinsics.mma_layout.shared_16x32_to_mma_32x16_layout(i, j)

tilelang.intrinsics.mma_layout.shared_16x32_to_mma_32x16_smoothlayout(i, j)

tilelang.intrinsics.mma_layout.shared_32x16_to_mma_32x16_layout(i, j)

tilelang.intrinsics.mma_layout.shared_32x16_to_mma_32x16_smoothlayout(i, j)