* infer context fix place error.

* update

* update

* init

* move CUDAStream to phi

* fix compilation

* merge develop

* add stream_owned_ member

* split cuda_stream.h

* fix cpu compile

* fix constructor

* fix bug

* fix windows compile

* fix inference test_levit

* fix windows tests

* multi stream support handle lazy init.

* support eigen lazy init

* update

* fix ci problem

* compile into one static library

* fix xpu compile

* fix xpu compile

* fix inference compile

* fix inference compile

* add custom test

* revert one file

* clean glog header in public header

* move marco pos

* add some kernel(csr*dense->csr, dense*dense->csr) of SparseTensor matmul

* fix CI

* fix CI

* fix comment

* fix comment

Use int64_t in GetGpuLaunchConfig1D and ElementwiseKernel as index type to support large tensor. (#43506)

* Change some data type from int to int64_t in GetGpuLaunchConfig1D to support large tensor.

* Use int64_t in ElementwiseKernel as index type to support large tensor.

* use GpuMemcpy and GpuMemset

* sparse convert kernel support double dispatch by indices dtype

* cudaMemcpyKind->gpuMemcpyKind

* refine enforce code

* refine enforce code

* fix compile failed

* fix infrt failed

* add context pool unittests

* fix timeout

* polish details

* change option pos

* add dll decl for wndows

* fix pre-commit error

* move dll_decl and export DeviceContext

* replace lost dll_decl.h

* fix_paddle_numel_check

* fix_paddle_numel_check

* Using the maximum workspace_size of all alogirhms to limit the workspace size in exhaustive search mode.

* Use the system cudaMalloc and cudaFree to allocate workspace during searching.

* Enable switch of two kind of workspace setting methods.
Co-authored-by: NLiu Yiqun <liuyiqun01@baidu.com>

* add interploate cpu kernel

* fix nullptr bug

* add interpolate gpu kernel

* fix unit test error

* remove raw kernels

* add cuda kernel impl

* add infermeta

* recover accidentally deleted kernels in interpolate op

* fix grad x_grad name error

* remove interpolate_v2_op.h

* rm unused codes

* fix xpu build error

* fix build error

* fix namespace error

* add register header for nup

* fix infermeta error

* modify by review

* add the missing args in test_trt_convert_nearest_interp_v2

* add maximum limit for grid of reduce, elementwise and gather

* add {} after if

* infrt add trt engine

* fix register

* file generate

* fix ci error

* fix conflict

* add copyright

* update

* update

* update

* update engine name

* refactor trt code

* update

* update

* update

* update

* fix conflict

* update

* fix compile with cuda

* fix gpu conetxt callback

* fix gpu callback

* fix callback early destruct problem

* Added cuBlasLtHandle_t to device context.

* Added fused_gemm_epilogue op.

1. Added fused_gemm_epilogue op to leverage cuBlastLt Epilogue.
2. Support fusion Act(X*Y + bias), X'dims >=2 and Y'dims shoule be 2.
2. Act currently only be supported ReLU. (Will add GeLU in the future).

* Added UT to fused_gemm_epilogue op.

* Added LinearAct Pattern

1. Added LinearAct into graph_pattern_detector.* to define (2.)'s
pattern.
2. LinearAct is used to detect act(element_add(matmul_v2(x, w), bias)).
3. act currently only support ReLU (Will support GeLU in the future).

* Added FuseGemmEpiloguePass

1, Added FuseGemmEpiloguePass to handle nn.Linear + Act{ReLU}
fusion (GeLU will be supported in the future).
2. Only support matmul_v2 from nn.Linear.

* Added pybind to BuildStrageter.fuse_gemm_epilogue_.

* Added UT for fuse_gemm_epilogue_pass.

* GeLU support and EpilogueSingleton

1. Added GeLU support to fused_gemm_epilogue op.
2. Added EpilogueSingleton to cache auxiliary pointer.
3. Added related UTs.

* Rename cublaslt_epilogue_opto gemm_epilogue_op.*.

* Added both train and infer pattern to LinearAct.

1. Added support of fwd graph with grap_ops linking to LinearAct.
2. Added related changes to fuse_gemm_epilogue_pass for above
modification.

* Changed CUDA requirement from 11.4 to 11.6 for fuse_gemm_epilogue_pass.

* Added identity activation support to gemm_epilogue_op.

* Added Linear Fusion (matmul_v2 + ele_add)

1. Added matmul_v2 + ele_add pattern to LinearActPattern.
2. Added matmul_v2 + ele_add support to fuse_gemm_epilogue_pass.

* Rename gemm_epilogue_op.* to fused_gemm_epilogue_op.*

* Add fused_gemm_epilogue_grad op.

1. Added fused_gemm_epilogue_grad to support backward epilogue fusion.

* Add UTs to fused_gemm_epilogue_grad_op.

* Change attribute name in fused_gemm_epilogue_grad_op for clearing.

* Allow DX and DBias be dispensable to fused_gemm_epilogue_grad op.

* Added ElementwiseAdd+Matmul+Act graph pattern detection.

* Fuse backward of Linear( Act(x))

1. Added backward fusion pass to Linear( Act(x)).
2. Added backward fusion pass to Linear(x).

* Added UTs to backward fusion of Linear(Act(x)).

* Complete document of arguments to fused_gemm_epilogue_op.

* Made arguments of some functions pass by reference.

* Modify code with review comments.

1. Made arguments of some function pass by reference.
2. Removed redundant code.
3. Followed Google code style to change code.

* Made 'const' code style be consistent

* Fixed random seed of python UTs.

* Set Compiling constrains to cuBlasLt

1. Require CUDA 11.6+
2. Remove fuse_gemm_epilogue related tests when CUDA < 11.6.

* Code Reivew from Paddle

1. Changed arguments name is_first_gemm to without_x_gradient for
clearing.
2. Applied PADDLE_THROW in fused_gemm_epilogue_op.

* Remove EpilogueSingleton

1. Applied ReserveSpace to replace Epilogue for passing auxiliary
pointers between FWD and BWD.

* Fix a logical error and enhance UTs.

1. Added act op count checking in UTs.
2. Fix issue to fuse backward or ReLU(Linear(X)).
3. TODO: solve GELU fusion issues.

* Fix Linear and GeLU fusion issues.

1. Modified graph_detech_pattern to fit with both linear wiht gelu or
relu.
2. Modified data range in Uts to allow negative values.

* Removed fused_gemm_epilogue_op.h.

* Rename namespace pten to phi.

* Rename name of arguments in fused_gemm_epilogue_op

1. bias -> Bias.
2. out -> Out.
3. reserve_space -> ReserveSpace.

* Change EpiloguePassActivationCache as local variable.

1. Removed singleton in EpiloguePassActivationCache.
2. Made EpiloguePassActivationCache as an argument to each pass
functions.

* support kps backend and compile

* resolve conflict

* fix kps backend trans

* test in xpu2 device

* remove dummy kernel

* rename pten_utils to phi_utils

* rename pten_utils target

* rename Pten to Phi

* replace pten with phi

* resolve conflict

* support cudnn kernel moving

* polish cmake rules

* add unittest for coverage

* remove orig kernel

* remove softmax cudnn kernel

* fix softmax test failed

* fix npu func error

* resolve conflict

* rename gpu dnn kernels

* fix name rule error

* fix compile error

* update fp16 namespace

* rename pten to phi

* fix infrt compile failed

* resolve conflict

* change CUDA implementaion of bernoulli OP

* fix CI

* change Vector to std::vector and provide MixVector class as a helper wrapper class

* solve the multi-gpu hang problem

* remove the duplicate template instantialize

* Copy vector to cpu

* add CopyToCPU

* xxx

* final version: fix the problem of all reduce

* remove mixvector dependence

* fix

* merge

* fix code

* fix by CI

* rename pten dir to phi

* rename namespace to phi

* rename infrt pten dir to phi

* resolve conflict

* rename pten to phi in cmake

* revert all infrt change

* change needed files

* fix infrt failed

* fix inference failed