* update xccl lib & use native Reduce in dygraph

* minor

* modify name

* merge develop

* fix param

* fix exp gen bug

* fix sum_grad

* comment

* [KUNLUN] add op: maxpool_with_index

* use DeviceContext::Alloc() instead of DenseTensor::mutable_data()

* fix file format

* solve clip unittest failure

* minor fix

* Revert "solve clip unittest failure" since the issue is fixed
in #49535

This reverts commit 1127adc66e79afe35ac3c00bb34e6aaa7cd7d78b.

* align with xdnn on the definition of mask in max_pool_with_index

* minor

* support PaddlePaddle Backend on Triton

* fix test cases

* fix Codestyle

* add test case

* add test case

* extract fuse pass logic to header file

* adjust namespaces

* Update paddle/fluid/framework/ir/mkldnn/activation_onednn_fuse_pass.h

update date
Co-authored-by: NTomasz Socha <tomasz.socha@intel.com>

* add inline remove static
Co-authored-by: NTomasz Socha <tomasz.socha@intel.com>

* fix cast issue

* add ut

* Add the cumsum 0d tensor

* xpu and cpu judge the 0d  tensor

* change to 2022 to 2023 in new commit

* fix the reverse logic

* revert to use default XPU stream for computing

XPUContext now has a null stream by default. If you want to use a separate stream
 (e.g. in async collective communication), you should create a dedicated XPUContext
and invoke its XPUContext::CreateStream()

* minor

* add test for composite with dy2st

* add more log

* refine munmap freq for ref_cnt_mmap_allocator

* add shm reuse logic

* fix compile bug

* fix compile bug

* fix bug of file refcount

* fix compile bug

* fix compile bug

* refine code for delete shm case

* polish code

* refine shm cache pool size setting logic

* set buffer is 2

* refine shm cache size logic

* refine max shm cache

* refine shm cache size

* reshape_transpose_matmul_pass_tester rewritten

* matmul_transpose_reshape_pass_tester rewritten

* mkldnn to onednn

* new exe supports CUDA Graph

* fix

* fix

* fix

* fix FLAGS_use_stream_safe_cuda_allocator in unittest

* insert output of coalesce_tensor op to skip_gc_var

* fix

* proto type of composite grad in paddle

* proto type of composite grad in paddle

* refactor composite api with phi

* fix compile error

* support static graph code-gen for squeeze op

* generate static graph code of unsqueeze

* refine op name

* fix compile error

* add extra output in op_compat

* remove debug log

* fix clang compile error

* support prim switch flag

* support prim switch flag

* fix dygraph error

* merge develop

* add code_gen

* add necessary files without codegen

* fix code_gen bug

* add deps

* modify igmnore

* add ignore

* delete std cout

* add composite logic for backward.py

* add tanh first order grad composite

* support enable_prim flag for static graph

* throw expection when both GrapOpMaker and GradCompOpMaker not been registered

* reorganize the directory of prim api tests

* fix windows error

* add eager_utils

* add eager_utils

* modify code gen

* add composite parse

* add unittest for get_grad_op_desc

* code optimize

* fix static test on windows

* support generate static graph code for imag and real op

* fix windows compile error in test_static_prim

* merge develop

* disable test eager in inference

* prim code gen

* disable eager compile in inference

* origin_yaml codegen success

* rm other file

* rm gitignore file

* code_style

* add eager test

* code_style

* clear #

* merge develop

* clear #

* remove useless files

* modify static test

* support bool flag from singlton

* merge develop

* recover git ignore

* fix conflict

* clear prim_gen

* recover git ignore for generated op

* parse_yaml success

* fix test compile error

* remove some tests

* add python test

* code_style

* revert parse_utils+ clear prim_gen

* fix some name issue

* add composite code gen

* modify backward yaml

* fix static composite grad maker code gen

* remove addtional files

* add some static funcs unit test

* fix some bugs

* fix composite grad maker register code gen

* optimize some functions

* modify gen cmake

* add more api gen

* add header

* modify static

* add static expand unsqueeze

* comments

* modify compopmaker

* revert

* modify gen name
Co-authored-by: NJiabinYang <360788950@qq.com>
Co-authored-by: Nzyfncg <zhangyunfei07@baidu.com>
Co-authored-by: Ncxxly <chenxx_id@163.com>
Co-authored-by: Ncharles-hit <wanghao107@baidu.com>

* change feed_op to phi kernel

* fix ci bugs

* fix build bugs

* fix ci bugs

* fix compile bugs

* fix ci bugs

* perfect code

* perfect comment code

* fix install bugs

* modify code according comment

* remove visitor in feed_op

* modify according comment

* perfect code according comment

* add infershape

* fix py3 bugs

* fix getexpected kernel type

* fix getexpected kernel type

* fix ci bugs

* add registry for custom device

* fix py3 bugs

* fix floating point error

* fix py3 test bugs

* Support call backward() without params in dy2st

* support elementwise base func

* fix compiling error and add test

* support vjp for div using comp

* remove additional change

* fix dy2st error with magic num

* fix dy magic num

* another magic

* another magic

* another magic

* add skip rename strategy

* support add vjp

* support add with new axis cal

* support sub vjp

* [prim] add multiply vjp rules

* [prim] add multiply vjp rules

* [prim] fix no infershape with composite in _append_backward_ops

* [prim] add expand vjp rule

* [prim] add exp vjp rule

* uncomment infer shape for reshape/sum static prim api

* [prim] fix tanh nullptr error

* remove some print message

* fix magic number in run_program relative tests @JiaBinYang

* [prim] add expand,multiply,exp vjp rules

* fix only support single direction reduce error

* infer reduce dims using out dims
Co-authored-by: NJiabinYang <360788950@qq.com>

* support the 'data_transform' for generating static graph ops

* reset 'pow' code

* change the 'GetKernelTypeForVar'

* Update warpctc for cuda-12

* Deprecate cudaProfilerInitialize for CUDA > 11

* Deprecate CUSPARSE_MV_ALG_DEFAULT for CUDA_VERSION >= 11040

* Add the missing thrust header

* add outvar name for nvtx mark

* nly network created with kEXPLICIT_BATCH can setsetMaxBatchSize

* [CINN]Switch cinn GIT_TAG from v0.2 into develop

* fix branch name

* specify commit

* disable unittest

* disable unittest

* add trt_support_nhwc_pass

* fix ctr_double_accessor.h

* fix graph_brpc_client.h non-const reference to pointer

* fix common_table.h

* fix graph_py_service.cc, server.cc, server.h

This reverts commit 4d5265b8.

* add gpu_cpu_map_matmul_to_mul_pass to kGpuLowerPrecisionPasses

* disable fc_elementwise_layernorm_fuse_pass in mixed precision

* polish static grad op maker gen

* fix some bugs

* fix static code gen

* solve conflict

* modify composite grad maker name

1 update xccl lib
2 when using comm_ctx, the allocator should be set manually.

* support elementwise base func

* fix compiling error and add test

* remove additional param

* support vjp for div using comp

* remove additional change

* fix dy2st error with magic num

* fix dy magic num

* another magic

* another magic

* add more test

* fix windows problem

* another magic

* fix windows compile

* invoke ci

* add skip rename strategy

* support add vjp

* fix test_tanh

* support add with new axis cal

* fix resnet and some test

* add composite log

* support sub vjp

* enhance_tests

* support more dtype for full

* add fmha_flashattention oss plugin

* update trt engine to set in/out data type

* update

* Update engine.cc

* Update engine.cc

* update

* set engine output type before freeze the network

* update

* update trt autoscan ut

* update

* update ut

* fix equal bug, update ut

* fix cast and equal ut

* update cast ut using TRT < 8.4

* set datatype from scope

* check output var is nullptr

* Update op_converter.h

* update tensorrt_engine_op_test ut

* update

* clear ProcessGroupCustom manually

* fix bug

* fix bug

* move destroy ProcessGroup to ProcessGroupIdMap

* enable destroy to all device

* remove unused comments

* change to internal api

* Update process_group.cc

* Update process_group.cc

* update get_device to custom

* add custom_device api

* rm is_compiled_with_custom_device from framework

* add todo comments