Reduce inference library size and compile time

[chalsliu]

PR types

Others

PR changes

Others

Description

Reduce inference library size and compile time.

[paddle-bot]

你的PR提交成功，感谢你对开源项目的贡献!
请关注后续CI自动化测试结果，详情请参考Paddle-CI手册。
Your PR has been submitted. Thanks for your contribution!
Please wait for the result of CI firstly. See Paddle CI Manual for details.

[paddle-bot]

❌ The PR is not created using PR's template. You can refer to this Demo.
Please use PR's template, it helps save our maintainers' time so that more developers get helped.

[jiweibo]

LGTM

[chalsliu]

使用方式：

1、编译前执行：python tools/remove_grad_op_and_kernel.py --only_kernel
2、cmake 参数加上： -DREDUCE_INFERENCE_LIB_SIZE=ON

[chalsliu]

效果：

1. with third_party pre-compiled
2. 90并发编译

whl_so_SIZE -> size of libpaddle.so
infer_so_SIZE -> size of libpaddle_inference.so

REDUCE	WITH_PYTHON	STRIP	ARCH_NUM	whl_so_SIZE	infer_so_SIZE
OFF	ON	ON	2 (sm_80 sm_86)	960M	924M
ON	ON	ON	2 (sm_80 sm_86)	678M (-282M) ✅	643M (-281M) ✅
OFF	OFF	ON	2 (sm_80 sm_86)	NA	922M
ON	OFF	ON	2 (sm_80 sm_86)	NA	639M (-283M) ✅
OFF	ON	ON	5 (sm_61 sm_70 sm_75 sm_80 sm_86)	2.0G	2.0G
ON	ON	ON	5 (sm_61 sm_70 sm_75 sm_80 sm_86)	1.4G (-613M) ✅	1.4G (-613M) ✅

[luotao1]

@qingqing01 @chalsliu @jiweibo 请问 tools/infer_prune_patches/ 还需要维护么？