* memory holder optimize

Change-Id: Ic91af8ac6f2853336d28a9fbbc5e8d0c57b5d05e

* memory holder optimize

Change-Id: I2fd1c14ecc17f5d5ce88b87890381ea801e6367f

* fix large scale memory holder

Change-Id: Ief0992b02b00220e16c72cc637a56e7b5788140f

* fix large scale memory holder

Change-Id: I910142a3952ead643a5604f8f80955f3e6efe655

* fix op version checker of pass bug

* fix code style

* update  pass version

* zero_copy_tensor supports int8_t

* register instance norm, test=op_version

Co-authored-by: Nzhoushunjie <zhoushunjie@baidu.com>

* Register op version for linspace,test=op_version

* Register op version for linspace,test=op_version

* Register op version for linspace,test=op_version

* Register op version for linspace,test=op_version

* Register op version for linspace,test=op_version

* test=develop, add op_register_version for roll_op

* dot op support complex types

* matmul support complex types

* add test case

* matmul broadcast gradient support complex

* move conjFunctor to complex_functor.h

* fix rank_attention, test=op_version

* Add operator checkpoints for new attributes.

* Fix adding subsequent checkpoint to quantize op.

* Add mkldnn nearest_interp and bilinear_interp op
* don't run mkldnn interpolate in default
* add interpolate_mkldnn_pass

* add the support the op version check for matmul, test=op_version

* add the op version check for the elementwise ops, test=op_version

* add the support check for elementwise_ops, test=op_version

* fix momentum op register

* upgrade instance_norm, test=op_version

* fix

* add debug code

* add place info

* fix compile problem

* add place for output

change the elementwise ops version check, test=op_version

Add moving_average_abs_max_scale op_register_version

fix ut:trt_resnext_test, trt_quant_int8_yolov3_r50_test, test_trt_dynamic_shape_ernie, test_trt_dynamic_shape_ernie_fp16_ser_deser, trt_cascade_rcnn_test (#29977)

register op version for conv2d_transpose, conv3d_transpose and depthwise_conv2d_transpose, test=op_version (#29937)

test=develop
test=op_version

* map matmul/squeeze2+matmul/reshape2+matmul to mul

PADDLE_RETRY_CUDA_SUCCESS used wrong sleep time so it can cause timeout in unittest. This PR fixed it.

After we searched the doc in https://pubs.opengroup.org/onlinepubs/7908799/xsh/unistd.h.html, the time unit of sleep in unistd.h takes "seconds", usleep takes "microseconds", Sleep in windows.h takes "milliseconds".

* add gru op_register_version; test=op_version;

* Update fc,mul version;test=op_version;

* flush denormals to zero, test=develop

* add comments, test=develop

* add op version for fake_quant and fake_dequant ops, test=op_version, test=develop

* fix a bug in reset_tensor_array, test=develop

* ci coverage, test=develop