optimize phi::IntArray(common::vectorize()) in kernels/gpudnn #20635

Job	Run time
Clone-linux / Clone Paddle	4m 26s
Mac-CPU / Build and test	18m 3s
build docker images / Build docker	2m 27s
Linux-CPU / Build and test	47m 49s
Distribute-stable-build / Build	1h 14m 50s
PR-CI-SOT / Check bypass / Check bypass	4s
Linux-XPU / Check bypass / Check bypass	3s
Linux-DCU / Check bypass / Check bypass	3s
Linux-XPU / Build	2h 18m 7s
PR-CI-SOT / Build and Test	1h 12m 49s
Linux-DCU / Build	1h 33m 17s
Linux-NPU / Check bypass / Check bypass	5s
Linux-IXUCA / Check bypass / Check bypass	5s
Linux-NPU / Test	11m 26s
Linux-IXUCA / Test	2m 45s
Linux-DCU / Test	1h 32m 14s
Linux-XPU / Test	31m 41s
Clone-linux / Clone Paddle	4m 26s
Mac-CPU / Build and test	18m 3s
Linux-CPU / Build and test	47m 49s
Linux-DCU / Check bypass / Check bypass	3s
build docker images / Build docker	2m 27s
Linux-XPU / Check bypass / Check bypass	3s
Distribute-stable-build / Build	1h 14m 50s
PR-CI-SOT / Check bypass / Check bypass	4s
Linux-DCU / Build	1h 33m 17s
PR-CI-SOT / Build and Test	1h 12m 49s
Linux-NPU / Check bypass / Check bypass	5s
Linux-XPU / Build	2h 18m 7s
Linux-IXUCA / Check bypass / Check bypass	5s
Linux-IXUCA / Test	2m 21s
Linux-XPU / Test	31m 41s
Linux-NPU / Test	11m 26s
Linux-DCU / Test	1h 32m 14s
	19h 40m 4s

Provide feedback