MCPcopy Create free account

hub / github.com/NVIDIA/cutlass / functions

Functions19,553 in github.com/NVIDIA/cutlass

⨍ Functions 19,553◇ Types & classes 10,397↳ Endpoints 26
↓ 18 callersMethodfind
include/cute/container/type_list.hpp:55
↓ 18 callersFunctionget
( input: Layout, mode: Any = ..., *, loc: Optional[ir.Location] = None, ip: Optional[ir.In
python/CuTeDSL/cutlass/cute/core.py:2405
↓ 18 callersFunctionget_gmem_layout
include/cutlass/detail/collective/mixed_input_utils.hpp:527
↓ 18 callersFunctionget_logical_ptr
include/cutlass/detail/collective/mixed_input_utils.hpp:506
↓ 18 callersFunctioninitialize_tensor
test/unit/gemm/device/gemm_testbed_3x_ptr_array.hpp:208
↓ 18 callersFunctionis_integer
Check if an object is static integer or dynamic integer
python/CuTeDSL/cutlass/cute/typing.py:698
↓ 18 callersMethodis_source_needed
include/cutlass/epilogue/collective/default_epilogue.hpp:148
↓ 18 callersFunctionload_with_pointer_offset
Loads a fragment from memory at the location pointed to by the iterator. (scalar loads)
include/cutlass/gemm/warp/mma_simt_tile_iterator.h:472
↓ 18 callersFunctionmake_thread_cooperative_group
(size: int)
examples/python/CuTeDSL/cute/blackwell/kernel/attention/fmha/fmha.py:104
↓ 18 callersFunctionmax_common_layout
include/cute/layout.hpp:1382
↓ 18 callersFunctionop_tag
Dispatches `op` to the appropriate *Tag dictionary depending on whether `op` is an ActivationOp or FunctionalOp. This is useful for cases in
python/cutlass_cppgen/backend/library.py:166
↓ 18 callersMethodpermute
Permute self.layout according to indices
python/cutlass_cppgen/backend/evt/ir/tensor.py:131
↓ 18 callersMethodsm90_mma_major_mode
(self)
python/CuTeDSL/cutlass/utils/layout.py:27
↓ 18 callersMethodstore
( self, value: Union[Numeric, cutlass_arith.ArithValue, int, float, bool], *,
python/CuTeDSL/cutlass/cute/core.py:1593
↓ 18 callersFunctionsynclog_emit_tma_load
include/cutlass/arch/synclog.hpp:778
↓ 18 callersFunctiontile_shape
include/cute/atom/mma_atom.hpp:626
↓ 17 callersMethodaccumulator_type
(self)
python/cutlass_library/gemm_operation.py:146
↓ 17 callersMethodadd
Elementwise add operator (1-by-2)
include/cutlass/matrix.h:276
↓ 17 callersFunctionagent_sync
Syncs all threads within an agent.
python/CuTeDSL/cutlass/pipeline/helpers.py:888
↓ 17 callersMethodcan_implement
test/unit/transform/device/sm90_sparse_gemm_compressor_legacy.hpp:300
↓ 17 callersFunctioncanonical_warp_group_idx
Returns a warp-uniform value indicating the canonical warp group index of the calling threads. Threads within the warp must be converged.
include/cutlass/cutlass.h:148
↓ 17 callersMethodcheck_cmd_line_flag
* Checks whether a flag "--<flag>" is present in the commandline */
tools/util/include/cutlass/util/command_line.h:106
↓ 17 callersFunctioncomposition
( lhs: Layout, rhs: Union[Layout, Shape, Tile], *, loc: Optional[ir.Location] = None, ip:
python/CuTeDSL/cutlass/cute/core.py:4375
↓ 17 callersFunctioncreate_and_permute_tensor
( shape, permute_order, dtype, dt_or_a=0, dynamic_modes=None, ref_tensor=None )
examples/python/CuTeDSL/cute/blackwell/kernel/attention/mamba2_ssd/mamba2_ssd.py:3322
↓ 17 callersFunctionfast_max
include/cutlass/fast_math.h:775
↓ 17 callersFunctionfind
include/cute/container/tuple.hpp:254
↓ 17 callersFunctionflatten
(t)
python/pycute/int_tuple.py:51
↓ 17 callersFunctionfold
include/cute/algorithm/tuple_algorithms.hpp:387
↓ 17 callersMethodget_trip_count
Calculate the number of trips needed for the current block. The trip count depends on the mask type and the block coordinates.
examples/python/CuTeDSL/utils/fmha_helpers.py:396
↓ 17 callersFunctionis_numpy_tensor
(inp)
python/cutlass_cppgen/utils/datatypes.py:77
↓ 17 callersMethodis_source_needed
Determine if the source is needed. May return false if
include/cutlass/epilogue/threadblock/epilogue_with_broadcast.h:122
↓ 17 callersFunctionisnan
include/cutlass/half.h:505
↓ 17 callersFunctionlaunch
( A: cute.Tensor, B: cute.Tensor, x: cute.Int32, y: cute.Int32, C: cute.Tensor, D: cut
examples/python/CuTeDSL/dsl_tutorials/jax/cutlass_call_basic.py:59
↓ 17 callersMethodload_init
include/cutlass/gemm/collective/sm100_mma_warpspecialized.hpp:498
↓ 17 callersFunctionmake_cotiled_copy
include/cute/atom/copy_atom.hpp:526
↓ 17 callersFunctionmax_common_vector
include/cute/layout.hpp:1409
↓ 17 callersFunctionnorm
include/cutlass/complex.h:479
↓ 17 callersMethodreduce
( self, dSK_tiled_mma: cute.TiledMma, tdQtdQ: cute.Tensor, tma_atom_dQ_acc: cu
examples/python/CuTeDSL/cute/blackwell/kernel/attention/fmha/fmha_bwd.py:2193
↓ 17 callersFunctionreference_gemm
examples/cute/tutorial/blackwell/example_utils.hpp:40
↓ 17 callersFunctionrewrite_tensor_shape
Rewrite tensor shape while keeping the same stride and iterator. This is primarily for debug friendliness - shows the actual expert's shape
examples/python/CuTeDSL/cute/blackwell/kernel/moe/moe_utils.py:234
↓ 17 callersMethodrun
Executes one test
test/unit/gemm/device/gemm_testbed_3x.hpp:3127
↓ 17 callersFunctionstatic
( value: Any, *, loc: Optional[ir.Location] = None, ip: Optional[ir.InsertionPoint] = None, )
python/CuTeDSL/cutlass/cute/core.py:2240
↓ 17 callersFunctionsum
(x, dim)
python/cutlass_cppgen/epilogue/evt_ops.py:49
↓ 17 callersMethodwidth
(self)
python/CuTeDSL/cutlass/cute/typing.py:87
↓ 17 callersMethodwrite
(self, *args)
examples/41_fused_multi_head_attention/piped_subprocess.py:73
↓ 17 callersFunctionx
(bx)
docs/jquery.js:23
↓ 16 callersFunctionTensorCopy
tools/util/include/cutlass/util/reference/host/tensor_copy.h:118
↓ 16 callersFunctionabs
include/cute/numeric/math.hpp:65
↓ 16 callersMethodaccum_ref
examples/41_fused_multi_head_attention/gemm/mma_from_smem.h:117
↓ 16 callersMethodargument_index
Returns the index of an argument by name
tools/profiler/src/problem_space.cpp:506
↓ 16 callersFunctionatomicAdd
( dst_ptr: Pointer, val: Int32, *, loc: Optional[ir.Location] = None, ip: Optional[ir.Inse
python/CuTeDSL/cutlass/utils/distributed.py:64
↓ 16 callersFunctionbF
(b0,b1)
docs/jquery.js:16
↓ 16 callersMethodclear
Efficient clear method
include/cutlass/array.h:360
↓ 16 callersMethodcontains
Determines whether a location is within a tensor
include/cutlass/tensor_view.h:190
↓ 16 callersFunctioncreate_and_permute_tensor
( shape, permute_order, dtype, min_val=-2, max_val=2, is_dynam
examples/python/CuTeDSL/cute/blackwell/kernel/attention/fmha/fmha_bwd.py:2838
↓ 16 callersFunctiondevice_cc
Returns the compute capability of the device with ID `device`. :param device: ID of the device to query :type device: int :return:
python/cutlass_cppgen/backend/utils/device.py:69
↓ 16 callersFunctionepilogue
(efc_config, C, D, alpha, beta, X, x_factor, Y)
examples/python/CuTeDSL/cute/blackwell/efc/custom_epilogue_dense_gemm.py:450
↓ 16 callersMethodget_block_coord
examples/77_blackwell_fmha/kernel/fmha_tile_scheduler.hpp:73
↓ 16 callersMethodget_desc_ptr
Get the raw gmem pointer to a specific TMA descriptor in workspace. :param tensor_name: Name identifying which tensor's descriptor
examples/python/CuTeDSL/cute/blackwell/kernel/moe/moe_utils.py:362
↓ 16 callersFunctionget_work_k_tile_start
include/cutlass/gemm/kernel/sm90_tile_scheduler_stream_k.hpp:698
↓ 16 callersMethodis_host_problem_shape_available
include/cutlass/gemm/group_array_problem_shape.hpp:125
↓ 16 callersMethodis_valid
examples/77_blackwell_fmha/kernel/fmha_tile_scheduler.hpp:68
↓ 16 callersFunctionload_with_byte_offset
Loads a fragment from memory with additional logical offset
include/cutlass/gemm/warp/mma_complex_tensor_op_tile_iterator_sm80.h:253
↓ 16 callersFunctionmake_ordered_layout
include/cute/layout.hpp:423
↓ 16 callersFunctionmake_utccp_copy
include/cute/atom/copy_traits_sm100.hpp:3876
↓ 16 callersFunctionpipeline_check_is_producer
include/cutlass/pipeline/sm90_pipeline.hpp:64
↓ 16 callersFunctionpipeline_init_arrive_relaxed
Used to guarantee that the Pipeline init is visible to all producers and consumer threadblocks in the cluster
include/cutlass/pipeline/sm90_pipeline.hpp:1375
↓ 16 callersMethodprocedural_name
(self)
python/cutlass_library/gemm_operation.py:361
↓ 16 callersMethodstride
Returns the stride of the layout
include/cutlass/layout/tensor_op_multiplicand_sm80.h:127
↓ 16 callersFunctionthr_size
include/cute/atom/mma_atom.hpp:644
↓ 16 callersFunctiony
(t,e)
python/docs/_static/clipboard.min.js:7
↓ 15 callersFunctionRematerializeBlockIdxZ
Helper to rematerialize block Idx. Reduces register liveness.
include/cutlass/gemm/threadblock/index_remat.h:78
↓ 15 callersMethod_checkCudaErrors
(self, result: Any)
python/CuTeDSL/cutlass/utils/hardware_info.py:158
↓ 15 callersFunctioncheck_cuda_status
include/cutlass/experimental/distributed/device/detail.hpp:46
↓ 15 callersFunctioncluster_wait
include/cute/arch/cluster_sm90.hpp:66
↓ 15 callersFunctioncomposition
include/cute/swizzle_layout.hpp:308
↓ 15 callersMethodconditional_producer_try_acquire
( self, intra1_acc_producer_state, intra1_acc_pipeline, C )
examples/python/CuTeDSL/cute/blackwell/kernel/attention/mamba2_ssd/mamba2_ssd.py:2925
↓ 15 callersFunctionconstruct
include/cute/algorithm/tuple_algorithms.hpp:626
↓ 15 callersFunctioncopy_aligned
include/cute/algorithm/copy.hpp:331
↓ 15 callersMethoddecode
( self, # MMA blk_tile_shd: cute.Tile, mma_tile_mnk: cute.Tile, tiled_
examples/python/CuTeDSL/cute/blackwell/kernel/attention/mixed_input_fmha/mixed_input_fmha_decode.py:502
↓ 15 callersFunctiongcd
include/cute/int_tuple.hpp:174
↓ 15 callersFunctiongemm_zero_acc
examples/77_blackwell_fmha/collective/fmha_common.hpp:56
↓ 15 callersMethodget_k_tile_iterator
include/cutlass/gemm/kernel/sm100_tile_scheduler.hpp:480
↓ 15 callersFunctionget_valid_schedules
(tile_description, cuda_version, is_aligned, data_types, layout, instantiation_level,
python/cutlass_library/sm90_utils.py:443
↓ 15 callersMethodhas_node
Check if the node is in the graph
python/cutlass_cppgen/backend/evt/ir/dag_ir.py:118
↓ 15 callersMethodi8
Create an i8 constant with the given value.
python/CuTeDSL/cutlass/base_dsl/tvm_ffi_builder/mlir_builder.py:161
↓ 15 callersMethodinit_state
examples/88_hopper_fmha/collective/fmha_collective_load.hpp:104
↓ 15 callersMethodinit_tensormap_from_atom
( self, copy_atom: cute.CopyAtom, dst_ptr: cute.Pointer, warp_id: int,
python/CuTeDSL/cutlass/utils/tensormap_manager.py:87
↓ 15 callersFunctionis_torch_tensor
(inp)
python/cutlass_cppgen/utils/datatypes.py:197
↓ 15 callersFunctionlayout
include/cute/tensor_impl.hpp:523
↓ 15 callersMethodload_dltensor
Lazily load the DLTensorWrapper. This function loads the DLTensorWrapper when needed, avoiding overhead in the critical path of calli
python/CuTeDSL/cutlass/cute/runtime.py:195
↓ 15 callersFunctionload_with_byte_offset
Loads a fragment from memory with additional logical offset
include/cutlass/gemm/warp/scale_bias_tile_iterator.h:258
↓ 15 callersFunctionload_with_pointer_offset
Loads a fragment from memory with additional logical offset
include/cutlass/gemm/warp/mma_tensor_op_tile_iterator.h:419
↓ 15 callersFunctionp
(by,bw,bv)
docs/jquery.js:23
↓ 15 callersFunctionprint
include/cute/atom/copy_atom.hpp:619
↓ 15 callersFunctionstore_with_pointer_offset
Stores a fragment to memory with additional pointer offset
include/cutlass/gemm/warp/mma_tensor_op_tile_iterator.h:4636
↓ 15 callersMethodto_llvm_ptr
Get the LLVM pointer representation of this pointer. (Used by internal API to propagate loc and ip) :param loc: Source location for
python/CuTeDSL/cutlass/cute/core.py:1665
↓ 15 callersFunctiontodevice
Pass the host_data to device memory
python/cutlass_cppgen/backend/memory_manager.py:92
↓ 15 callersMethodupdate
(self, shape)
python/cutlass_cppgen/backend/evt/ir/layout_nodes.py:65
↓ 15 callersMethodvisit
include/cutlass/epilogue/threadblock/fusion/visitor_store.hpp:321
← previousnext →601–700 of 19,553, ranked by callers