Optimize performance of depthwise_conv_fwd

[ZzSean]

PR types

Performance optimization

PR changes

OPs

Describe

Optimize performance of depthwise_conv_fwd

• Method: Reduce modulo calculations and other redundant calculations

• Result:

config	pytorch	paddle dev	paddle this PR	speedup
input[2048, 1024, 4, 4] filter[1024, 1, 4, 4] stride=1 pad=0 dilation=1	0.4124ms	1.9699ms	0.375ms	5.25x

[paddle-bot[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.