tilelang package

Subpackages

• tilelang.autotuner package
  • Module contents
    • AutoTuner
      • AutoTuner.from_kernel()
      • AutoTuner.run()
      • AutoTuner.set_compile_args()
    • AutotuneResult
      • AutotuneResult.latency
      • AutotuneResult.config
      • AutotuneResult.ref_latency
      • AutotuneResult.libcode
      • AutotuneResult.func
      • AutotuneResult.kernel
      • AutotuneResult.config
      • AutotuneResult.func
      • AutotuneResult.kernel
      • AutotuneResult.latency
      • AutotuneResult.libcode
      • AutotuneResult.ref_latency
    • JITContext
      • JITContext.out_idx
      • JITContext.ref_prog
      • JITContext.supply_prog
      • JITContext.rtol
      • JITContext.atol
      • JITContext.max_mismatched_ratio
      • JITContext.skip_check
      • JITContext.cache_input_tensors
      • JITContext.profiler
      • JITContext.target
      • JITContext.atol
      • JITContext.cache_input_tensors
      • JITContext.max_mismatched_ratio
      • JITContext.out_idx
      • JITContext.profiler
      • JITContext.ref_prog
      • JITContext.rtol
      • JITContext.skip_check
      • JITContext.supply_prog
      • JITContext.target
    • autotune()
    • check_tensor_list_compatibility()
    • jit()
• tilelang.cache package
  • Submodules
    • tilelang.cache.kernel_cache module
      • KernelCache
  • Module contents
    • cached()
    • clear_cache()
• tilelang.carver package
  • Subpackages
    • tilelang.carver.arch package
      • Subpackages
      • Submodules
      • Module contents
    • tilelang.carver.roller package
      • Subpackages
      • Submodules
      • Module contents
    • tilelang.carver.template package
      • Submodules
      • Module contents
  • Submodules
    • tilelang.carver.analysis module
      • BlockInfo
      • IterInfo
      • check_func_with_dynamic()
      • collect_block_iter_vars_used_in_access_region()
      • collect_vars_used_in_prim_expr()
      • detect_dominant_read()
      • find_var_from_func()
      • get_coalesced_veclen()
      • get_max_shared_memory_per_block()
      • get_max_threads_per_block()
      • get_reduction_blocks()
      • get_root_block()
      • is_broadcast_epilogue()
      • normalize_prim_func()
    • tilelang.carver.common_schedules module
      • get_block()
      • get_output_blocks()
      • try_inline()
      • try_inline_contiguous_spatial()
    • tilelang.carver.matmul_analysis module
      • IterKind
      • IterTrait
      • auto_inline_consumer_chain()
      • auto_inline_consumers()
      • auto_inline_producers()
      • collect_vars_from_expr()
      • detect_iter_traits()
      • find_arg_idx_from_buffer_chain()
      • find_first_similar_buffer()
      • find_first_similar_region()
      • find_last_producer_from_buffer()
      • get_dequantize_block()
      • get_in_out_dtypes()
      • get_index_map()
      • get_ladder_stage3_map()
      • get_propagate_map()
      • get_tensorized_func_and_tags()
      • inline_transpose_block()
      • is_identity_block()
      • is_identity_or_transpose_block()
      • is_transpose_block()
      • layout_propagate_chain()
      • make_iter_fusion_index_map()
      • normalize_to_matmul()
    • tilelang.carver.utils module
      • get_rasterization_code()
      • get_roller_hints_from_func()
      • get_roller_hints_from_output_nodes()
      • retrieve_func_from_module()
  • Module contents
• tilelang.common package
  • Submodules
    • tilelang.common.transform_kind module
      • TransformKind
  • Module contents
• tilelang.contrib package
  • Submodules
    • tilelang.contrib.cc module
      • create_executable()
      • create_shared()
      • create_staticlib()
      • cross_compiler()
      • get_cc()
      • get_cplus_compiler()
      • get_global_symbol_section_map()
      • get_target_by_dump_machine()
    • tilelang.contrib.dlpack module
      • convert_func()
      • to_pytorch_func()
    • tilelang.contrib.hipcc module
      • compile_hip()
    • tilelang.contrib.nvcc module
      • callback_libdevice_path()
      • compile_cuda()
      • find_cuda_path()
      • get_cuda_version()
      • have_cudagraph()
      • have_fp16()
      • have_int8()
      • have_tensorcore()
      • parse_compute_version()
  • Module contents
• tilelang.engine package
  • Submodules
    • tilelang.engine.callback module
      • register_cuda_postproc()
      • register_cuda_postproc_callback()
      • register_hip_postproc()
      • register_hip_postproc_callback()
    • tilelang.engine.lower module
      • canon_target_host()
      • device_codegen()
      • device_codegen_without_compile()
      • extrac_params()
      • get_device_call()
      • get_host_call()
      • has_device_kernel_launch()
      • host_codegen()
      • is_cpu_device_backend()
      • is_device_call()
      • is_device_call_c_device()
      • lower()
    • tilelang.engine.param module
      • CompiledArtifact
      • KernelParam
    • tilelang.engine.phase module
      • LowerAndLegalize()
      • OptimizeForTarget()
  • Module contents
• 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
    • 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
      • INT4TensorCoreIntrinEmitterWithLadderTransform
      • TensorCoreIntrinEmitter
      • TensorCoreIntrinEmitterWithLadderTransform
    • 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
• tilelang.jit package
  • Subpackages
    • tilelang.jit.adapter package
      • Subpackages
      • Submodules
      • Module contents
  • Submodules
    • tilelang.jit.env module
    • tilelang.jit.kernel module
      • JITKernel
  • Module contents
    • compile()
    • jit()
• tilelang.language package
  • Subpackages
    • tilelang.language.tir package
      • Submodules
      • Module contents
  • Submodules
    • tilelang.language.allocate module
      • alloc_fragment()
      • alloc_local()
      • alloc_shared()
      • alloc_var()
    • tilelang.language.builtin module
      • CreateListofMBarrierOp()
      • CreateTMADescriptorOp()
      • FenceProxyAsyncOp()
      • GetMBarrierOp()
      • MBarrierExpectTX()
      • MBarrierWaitParity()
      • NoSetMaxNReg()
      • SetMaxNReg()
      • TMALoadOp()
      • TMAStoreArrive()
      • TMAStoreWait()
      • WaitWgmma()
    • tilelang.language.copy module
      • buffer_load_to_tile_region()
      • buffer_region_to_tile_region()
      • buffer_to_tile_region()
      • c2d_im2col()
      • copy()
      • region()
    • tilelang.language.customize module
      • atomic_add()
      • atomic_addx2()
      • clamp()
      • dp4a()
      • reshape()
      • view()
    • tilelang.language.fill module
      • clear()
      • fill()
    • tilelang.language.frame module
      • FrameStack
      • LetFrame
      • get_let_value()
      • has_let_value()
    • tilelang.language.gemm module
      • gemm()
    • tilelang.language.kernel module
      • FrameStack
      • Kernel()
      • KernelLaunchFrame
      • get_block_binding()
      • get_block_bindings()
      • get_thread_binding()
      • get_thread_bindings()
    • tilelang.language.memscope module
    • tilelang.language.parallel module
      • Parallel()
    • tilelang.language.pipeline module
      • Pipelined()
    • tilelang.language.print module
      • print()
    • tilelang.language.proxy module
      • BaseTensorProxy
      • BufferProxy
      • FragmentBufferProxy
      • LocalBufferProxy
      • SharedBufferProxy
      • TensorProxy
      • make_tensor()
      • ptr()
    • tilelang.language.reduce module
      • reduce()
      • reduce_absmax()
      • reduce_abssum()
      • reduce_max()
      • reduce_min()
      • reduce_sum()
  • Module contents
    • annotate_layout()
    • import_source()
    • symbolic()
    • use_swizzle()
• tilelang.layout package
  • Submodules
    • tilelang.layout.fragment module
      • Fragment
      • make_swizzled_layout()
    • tilelang.layout.layout module
      • Layout
    • tilelang.layout.swizzle module
      • make_swizzled_layout()
  • Module contents
• tilelang.math package
  • Module contents
    • cdiv()
    • next_power_of_2()
• tilelang.primitives package
  • Subpackages
    • tilelang.primitives.gemm package
      • Submodules
      • Module contents
  • Module contents
• tilelang.profiler package
  • Submodules
    • tilelang.profiler.bench module
      • do_bench()
  • Module contents
    • Profiler
      • Profiler.params
      • Profiler.result_idx
      • Profiler.supply_type
      • Profiler.adapter
      • Profiler.adapter
      • Profiler.assert_allclose()
      • Profiler.assert_consistent()
      • Profiler.determine_profiler()
      • Profiler.do_bench()
      • Profiler.func
      • Profiler.params
      • Profiler.result_idx
      • Profiler.run_once()
      • Profiler.supply_type
      • Profiler.with_default_adapter()
• tilelang.testing package
  • Module contents
    • main()
    • set_random_seed()
• tilelang.tools package
  • Submodules
    • tilelang.tools.Analyzer module
      • AnalysisResult
      • Analyzer
    • tilelang.tools.plot_layout module
      • plot_layout()
  • Module contents
• tilelang.transform package
  • Submodules
    • tilelang.transform.simplify module
      • Simplify()
      • apply_simplify()
      • simplify_prim_func()
  • Module contents
    • AnnotateDeviceRegions()
    • ClusterPlanning()
    • FrontendLegalize()
    • IfStmtBinding()
    • InjectFenceProxy()
    • InjectPTXAsyncCopy()
    • InjectSoftwarePipeline()
    • LayoutInference()
    • LegalizeSafeMemoryAccess()
    • LegalizeVectorizedLoop()
    • LoopVectorizeDynamic()
    • LowerDeviceStorageAccessInfo()
    • LowerHopperIntrin()
    • LowerTileOp()
    • MakePackedAPI()
    • MergeIfStmt()
    • MultiVersionBuffer()
    • PipelinePlanning()
    • RewriteWgmmaSync()
    • ThreadPartialSync()
    • ThreadSync()
    • VectorizeLoop()
    • WarpSpecialized()
    • WarpSpecializedPipeline()
• tilelang.utils package
  • Submodules
    • tilelang.utils.deprecated module
      • deprecated()
    • tilelang.utils.language module
      • array_reduce()
      • is_fragment()
      • is_global()
      • is_local()
      • is_shared()
      • is_shared_dynamic()
      • retrieve_func_from_module()
    • tilelang.utils.target module
      • check_cuda_availability()
      • check_hip_availability()
      • determine_target()
    • tilelang.utils.tensor module
      • TensorSupplyType
      • adapt_torch2tvm()
      • get_tensor_supply()
      • map_torch_type()
      • torch_assert_close()
  • Module contents

Submodules

• tilelang.env module
  • disable_cache()
  • enable_cache()
  • is_cache_enabled()

Module contents

class tilelang.TqdmLoggingHandler(level=0)

Bases: Handler

Custom logging handler that directs log output to tqdm progress bar to avoid interference.

emit(record)

Emit a log record. Messages are written to tqdm to ensure output in progress bars isn’t corrupted.

tilelang.set_log_level(level)

Set the logging level for the module’s logger.

Parameters:

• level (str or int) – Can be the string name of the level (e.g., ‘INFO’) or the actual level (e.g., logging.INFO).

• OPTIONS – ‘DEBUG’, ‘INFO’, ‘WARNING’, ‘ERROR’, ‘CRITICAL’