* [CustomDevice] add custom device api

* update

* update

* test=document_fix

* update

* update

* add  examples

* Support stream priority for standalone executor

* Fix compile error

* Fix compile error

* Fix compile error

* Fix compile error

* Fix compile error

* [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

* 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>

* add reduce_sum_int64 and reduce_sum_int8 xpu kernels

* [PHI] add clip grad kernel with support type float32 and int32

* [PHI unittest] add clip_grad unit test

* adapt code to clang-format

* update xpu api output with clip_grad api

* remove int8 support of reduce_sum xpu kernel since it can not pass unit tests

* adapt license date, add code for XPUDataType convertion

* add int8 support of reduce_sum

* add reduce_sum unit tests for dtype int64, int8, and add more test cases

* update license date

* remove buggy bitwise and, or and xor xpu kernels, refine bitwise not xpu kernel

* change license date

* 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

* 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

* kunlun add support for c_concat and c_split

* replace mutable_data() and ShareDataWith()

* fix xpu unittest issue: zero_dim_tensor

* deal with leftout issue introduced by #49470

xpu2_op_list.cc. test=kunlun

* rename kernel

* delete sig

* modify code according comment

* fix ci bugs

* add stack grad kernel optimization

* add basic optimization kernel for stack_grad_kernel

* optimization of stack_grad_kernel for last dim stack and change code format with pre-commit

* migrate shaple sgd, split,sign xpu kernels to phi, test=kunlun

* fix dlrm throughput problem, test=kunlun

* add xpu einsum, fill_diagonal, and diagonal kernels, test=kunlun

* add bitwise and, bitwise not, bitwise or and bitwise xor

* correct typo

* migrate shaple sgd, split,sign xpu kernels to phi, test=kunlun

* fix dlrm throughput problem, test=kunlun

* add warp transducer code

* move dropout_impl from fluid to phi

* move cuda_graph_with_memory_pool from fluid to phi

* update namespace

* remove cuad_graph in fluid

* fix mac-build

* fix bugs

* correct CodeStyle

* fix mac-build

* fix mutable_data

* fix stl include

* fix copy param

* nullptr bugfix for XPU pg mode

Also a few kernels is added to xpu whitelist

* increase error msg length

* fix codestyle

* add double complex<float> complex<double> dtype support for syevj_batched

* fix use_syevj flag for precision loss when input dtype of syevj_batch is complex128 in some case

* optimize eigh in different case

* fix missing ; bug

* fix use_syevj bug

* fix use_cusolver_syevj_batched flag