MCPcopy Create free account

hub / github.com/bytedance/flux / functions

Functions784 in github.com/bytedance/flux

⨍ Functions 784◇ Types & classes 309
↓ 106 callersMethoddata_ptr
src/reduce_scatter/reduce_scatter_kernel.hpp:474
↓ 64 callersMethodget
include/flux/flux.h:667
↓ 46 callersFunctionto_cutlass_element
include/flux/cuda/cuda_common.h:93
↓ 28 callersFunctionunify_type
include/flux/flux.h:1039
↓ 25 callersFunctionprint_per_kernel
include/flux/cuda/cuda_common_device.hpp:74
↓ 25 callersFunctionptr_offset
src/all_gather/ths_op/all_gather_gemm_kernel_crossnode.cc:64
↓ 24 callersMethodend
src/reduce_scatter/sm90_epilogue_evt.hpp:225
↓ 16 callersMethodget_op
include/flux/op_registry.h:239
↓ 16 callersMethodstr
include/flux/op_registry.h:85
↓ 13 callersFunctionflux_barrier_all_on_stream
src/ths_op/flux_shm.cc:195
↓ 13 callersFunctionget_arch
()
python/flux/util.py:37
↓ 12 callersFunctionget_int_from_env
src/cuda/utils.cc:34
↓ 12 callersFunctionptr_offset
src/all_gather/ths_op/all_gather_gemm_kernel.cc:71
↓ 11 callersFunctionfrom_torch_dtype
src/ths_op/ths_op.cc:43
↓ 11 callersFunctionget_bool_from_env
src/cuda/utils.cc:49
↓ 11 callersFunctionnanosleep
include/flux/cuda/cuda_common_device.hpp:161
↓ 11 callersMethodreset
include/flux/utils.h:98
↓ 10 callersMethodgemm_device
///////////////////// CRTP functions /////////////////////
include/flux/cuda/gemm_impls/gemm_v3_impl.hpp:64
↓ 10 callersFunctionmake_gemm_v3_meta
include/flux/gemm_meta.h:173
↓ 10 callersMethodnew_group
(self, ranks)
python/flux/dist_utils.py:64
↓ 10 callersMethodrun
src/reduce_scatter/reduce_scatter_kernel.hpp:2097
↓ 8 callersMethodevt
include/flux/cuda/gemm_impls/gemm_v2_impl.hpp:101
↓ 8 callersFunctionflux_create_tensor_list
src/ths_op/flux_shm.cc:182
↓ 8 callersFunctiontuple_transform
include/flux/flux.h:226
↓ 8 callersFunctionunify_type
include/flux/gemm_hparams.h:28
↓ 7 callersMethodadd
src/ths_op/ths_op.cc:194
↓ 7 callersFunctionmake_gemm_v3_hparams
include/flux/gemm_hparams.h:95
↓ 7 callersMethodrun
src/all_gather/ths_op/transfers.hpp:68
↓ 7 callersMethodsync
src/reduce_scatter/epilogue_vectorized_reduce_scatter.hpp:191
↓ 7 callersFunctionto_gemm_v3_hparams
include/flux/gemm_hparams.h:102
↓ 6 callersMethodGemmOnly
src/comm_none/ths_op/gemm_only.cc:213
↓ 6 callersMethodalignment_c
include/flux/cuda/gemm_impls/gemm_v2_impl.hpp:97
↓ 6 callersFunctionensure_nvml_init
src/cuda/cuda_common.cc:23
↓ 6 callersMethodget_tile_offset
Obtains the calling threadblock's tiled coordinates for the given tile index
src/all_gather/sm80_all_gather_gemm_threadblock_swizzle.hpp:102
↓ 6 callersMethodstart
Start the timer for a given stream (defaults to the default stream)
include/flux/cuda/cuda_common.h:182
↓ 6 callersMethodstop
Stop the timer
include/flux/cuda/cuda_common.h:189
↓ 5 callersFunctioncheck_nvshmem_init
pynvshmem/src/functions.cpp:43
↓ 5 callersFunctioncopy_continous_kernel
src/reduce_scatter/reduce_scatter_kernel.hpp:257
↓ 5 callersMethodelapsed_millis
include/flux/utils.h:103
↓ 5 callersFunctionget_bool_from_env_debug
src/all_gather/ths_op/all_gather_gemm_kernel_crossnode.cc:68
↓ 5 callersMethodget_stride_b
include/flux/cuda/gemm_impls/gemm_v2_impl.hpp:364
↓ 5 callersMethodget_workspace_size
src/reduce_scatter/gemmk_visitor_load.hpp:82
↓ 5 callersMethodregister_one
src/ths_op/ths_op.cc:263
↓ 5 callersFunctionto_make_expr
include/flux/flux.h:814
↓ 5 callersFunctiontuple_cartesian_product
include/flux/flux.h:321
↓ 5 callersFunctiontuple_slice
include/flux/flux.h:163
↓ 4 callersMethodAGKernel
src/all_gather/ths_op/all_gather_gemm_kernel.cc:132
↓ 4 callersFunctionatomic_load_acquire_sys
include/flux/cuda/cuda_common_device.hpp:110
↓ 4 callersFunctionatomic_load_sys
include/flux/cuda/cuda_common_device.hpp:105
↓ 4 callersMethoddefault_collective_epilogue
include/flux/cuda/gemm_impls/gemm_v3_impl.hpp:50
↓ 4 callersMethoddefault_kernel_params
include/flux/cuda/gemm_impls/gemm_v2_impl.hpp:213
↓ 4 callersMethodforward
(self, input: torch.Tensor)
python/flux/ag_gemm.py:201
↓ 4 callersMethodforward
src/comm_none/ths_op/gemm_only.cc:252
↓ 4 callersMethodgemm_kernel
src/comm_none/gemm_v3_comm_none.hpp:60
↓ 4 callersMethodget_all_prof_results
src/ths_op/ths_op.cc:143
↓ 4 callersMethodget_code
src/ths_op/ths_op.cc:124
↓ 4 callersMethodget_workspace_size
Returns the workspace size (in bytes) needed for these parameters
src/all_gather/sm80_all_gather_gemm.hpp:424
↓ 4 callersFunctionis_fp8_dtype
(dtype: torch.dtype)
python/flux/util.py:125
↓ 4 callersFunctionpretty_format
(node)
pynvshmem/tools/autogen_pybind.py:76
↓ 4 callersFunctionprint_if
include/flux/cuda/cuda_common_device.hpp:40
↓ 4 callersMethodreset_signals
src/all_gather/ths_op/all_gather_gemm_kernel.cc:674
↓ 4 callersFunctionsizeof_dtype
include/flux/flux.h:592
↓ 4 callersFunctionto_cutlass_layout_c
include/flux/cuda/cuda_common.h:149
↓ 4 callersFunctionto_gemm_shape
include/flux/cuda/gemm_impls/gemm_v2_impl.hpp:62
↓ 4 callersFunctiontuple_filter
include/flux/flux.h:271
↓ 3 callersFunctionGetLaunchProp
src/reduce_scatter/reduce_scatter_kernel.hpp:1830
↓ 3 callersFunction_preload_libs
(libname)
python/flux/cpp_mod.py:30
↓ 3 callersFunctionatomic_add_release_sys
include/flux/cuda/cuda_common_device.hpp:143
↓ 3 callersMethodcopy_local
src/all_gather/ths_op/all_gather_gemm_kernel.cc:687
↓ 3 callersFunctioncudaipc_create_tensor_list
src/ths_op/flux_shm.cc:99
↓ 3 callersMethoddefault_collective_mma
include/flux/cuda/gemm_impls/gemm_v3_impl.hpp:39
↓ 3 callersMethodgather
src/all_gather/ths_op/all_gather_gemm_kernel.cc:667
↓ 3 callersMethodget_barrier_workspace_size
Get the workspace size needed for barrier
src/reduce_scatter/epilogue_evt.hpp:176
↓ 3 callersFunctionget_from_torch_classes
(name: str)
python/flux/cpp_mod.py:80
↓ 3 callersFunctionget_gpu_device_name
why not std::string? flux/th_op is compiled with -D_GLIBCXX_USE_CXX11_ABI=0 but flux/cuda is not
src/cuda/cuda_common.cc:32
↓ 3 callersMethodget_hparams
include/flux/op_registry.h:183
↓ 3 callersMethodget_tile_offset
Obtains the calling threadblock's tiled coordinates for the given tile index
src/reduce_scatter/tile_scheduler/threadblock_swizzle.hpp:77
↓ 3 callersMethodinitialize
src/reduce_scatter/reduce_scatter_kernel.hpp:1951
↓ 3 callersFunctionis_fp8_dtype
include/flux/flux.h:586
↓ 3 callersFunctionis_fp8_torch_dtype
src/ths_op/ths_op.cc:65
↓ 3 callersFunctionis_grouped_gemm_impl
include/flux/flux.h:580
↓ 3 callersMethodkernel_params
src/reduce_scatter/gemm_v2_reduce_scatter.hpp:237
↓ 3 callersFunctionmake_space_meta_hparams_pair
include/flux/gemm_hparams.h:450
↓ 3 callersFunctionnvshmem_create_tensor_list
src/ths_op/flux_shm.cc:63
↓ 3 callersMethodoutput_tile_thread_map
include/flux/cuda/gemm_impls/gemm_v2_impl.hpp:153
↓ 3 callersMethodrecord_best
src/ths_op/ths_op.cc:207
↓ 3 callersMethodreduce
src/reduce_scatter/sm90_reduce_scatter_utils.hpp:307
↓ 3 callersFunctionsub_barrier_sync
include/flux/cuda/cuda_common_device.hpp:171
↓ 3 callersFunctionto_cutlass_layout_a
include/flux/cuda/cuda_common.h:127
↓ 3 callersFunctionto_cutlass_layout_b
include/flux/cuda/cuda_common.h:137
↓ 3 callersFunctionto_gemm_shape
include/flux/cuda/gemm_impls/gemm_grouped_impl.hpp:45
↓ 3 callersFunctionto_gemm_v3_meta
include/flux/gemm_meta.h:178
↓ 3 callersFunctiontuple_enumerate
include/flux/flux.h:405
↓ 3 callersMethodvisit_hparams
include/flux/op_registry.h:268
↓ 3 callersFunctionwait_value
src/all_gather/ths_op/all_gather_gemm_kernel_crossnode.cc:77
↓ 3 callersFunctionwrite_value
src/all_gather/ths_op/all_gather_gemm_kernel_crossnode.cc:82
↓ 2 callersMethodGemmRS
src/reduce_scatter/ths_op/gemm_reduce_scatter.cc:210
↓ 2 callersMethodProfilingContext
src/ths_op/ths_op.cc:117
↓ 2 callersMethod_is_cudaStream_argument
(self, type, name)
pynvshmem/tools/autogen_pybind.py:211
↓ 2 callersMethod_is_type_pointer
(self, type, name)
pynvshmem/tools/autogen_pybind.py:221
next →1–100 of 784, ranked by callers