tilelang.intrinsics package

Submodules

• tilelang.intrinsics.mfma_layout module
  • make_mfma_swizzle_layout()
  • shared_16x16_to_ldmatrix_64x4_layout()
  • shared_16x16_to_local_64x4_layout_A()
  • shared_16x16_to_local_64x4_layout_B()
  • shared_16x16_to_local_64x4_layout_C()
  • shared_16x32_to_local_64x8_layout_A()
  • shared_16x32_to_local_64x8_layout_B()
  • shared_16x4_to_local_64x1_layout_A()
  • shared_4x16_to_local_64x1_layout_B()
  • thread_id_shared_access_64x1_to_16x4_layout_A()
  • thread_id_shared_access_64x1_to_4x16_layout_B()
  • thread_id_shared_access_64x4_to_16x16_layout_A()
  • thread_id_shared_access_64x4_to_16x16_layout_B()
  • thread_id_shared_access_64x4_to_16x16_layout_C_m_n()
  • thread_id_shared_access_64x4_to_16x16_layout_C_n_m()
  • thread_id_shared_access_64x8_to_16x32_layout_A()
  • thread_id_shared_access_64x8_to_16x32_layout_B()
• tilelang.intrinsics.mfma_macro_generator module
  • MatrixCoreIntrinEmitter
    • MatrixCoreIntrinEmitter.M_DIM
    • MatrixCoreIntrinEmitter.N_DIM
    • MatrixCoreIntrinEmitter.WARP_SIZE
    • MatrixCoreIntrinEmitter.dtype_abbrv
    • MatrixCoreIntrinEmitter.extract_thread_binding()
    • MatrixCoreIntrinEmitter.get_ldmatrix_index_map()
    • MatrixCoreIntrinEmitter.is_m_first
    • MatrixCoreIntrinEmitter.k_pack
    • MatrixCoreIntrinEmitter.ldmatrix_a()
    • MatrixCoreIntrinEmitter.ldmatrix_b()
    • MatrixCoreIntrinEmitter.mfma()
    • MatrixCoreIntrinEmitter.stmatrix()
• tilelang.intrinsics.mma_layout module
  • get_swizzle_layout()
  • ldmatrix_16x32_to_shared_16x32_layout_a()
  • ldmatrix_16x32_to_shared_16x32_layout_b()
  • ldmatrix_32x16_to_shared_16x32_layout_a()
  • ldmatrix_32x16_to_shared_16x32_layout_b()
  • ldmatrix_32x8_to_shared_16x16_layout()
  • ldmatrix_trans_32x8_to_shared_16x16_layout()
  • make_mma_swizzle_layout()
  • mma_32x8_to_shared_16x16_layout()
  • mma_store_32x8_to_shared_16x16_layout()
  • shared_16x16_to_mma_32x8_layout()
  • shared_16x16_to_mma_32x8_layout_rs()
  • shared_16x16_to_mma_32x8_layout_sr()
  • shared_16x16_to_mma_32x8_layout_trans()
  • shared_16x16_to_mma_32x8_smoothlayout()
  • shared_16x32_to_mma_32x16_layout()
  • shared_16x32_to_mma_32x16_smoothlayout()
  • shared_32x16_to_mma_32x16_layout()
  • shared_32x16_to_mma_32x16_smoothlayout()
• tilelang.intrinsics.mma_macro_generator module
  • INT4TensorCoreIntrinEmitter
    • INT4TensorCoreIntrinEmitter.mma()
  • INT4TensorCoreIntrinEmitterWithLadderTransform
    • INT4TensorCoreIntrinEmitterWithLadderTransform.mma()
  • TensorCoreIntrinEmitter
    • TensorCoreIntrinEmitter.M_DIM
    • TensorCoreIntrinEmitter.N_DIM
    • TensorCoreIntrinEmitter.WARP_SIZE
    • TensorCoreIntrinEmitter.dtype_abbrv
    • TensorCoreIntrinEmitter.extract_thread_binding()
    • TensorCoreIntrinEmitter.get_store_index_map()
    • TensorCoreIntrinEmitter.is_m_first
    • TensorCoreIntrinEmitter.ldmatrix_a()
    • TensorCoreIntrinEmitter.ldmatrix_b()
    • TensorCoreIntrinEmitter.make_mma_load_layout()
    • TensorCoreIntrinEmitter.make_mma_store_layout()
    • TensorCoreIntrinEmitter.mma()
    • TensorCoreIntrinEmitter.stmatrix()
  • TensorCoreIntrinEmitterWithLadderTransform
    • TensorCoreIntrinEmitterWithLadderTransform.ldmatrix_a()
    • TensorCoreIntrinEmitterWithLadderTransform.ldmatrix_b()
    • TensorCoreIntrinEmitterWithLadderTransform.mma()
• tilelang.intrinsics.utils module
  • get_ldmatrix_offset()
  • get_mma_micro_size()
  • index_to_coordinates()
  • mfma_store_index_map()
  • mma_store_index_map()
  • shared_16x16_to_mma_32x8_layout()
  • shared_16x32_to_mma_32x16_layout()
  • shared_32x16_to_mma_32x16_layout()

Module contents