projects / platform / core / ml / nnfw.git / commit
? search:
summary | shortlog | log | commit | commitdiff | tree
(parent: 8d2222f) | patch
[NNFW] optimized Conv on CPU (#4964)
authorАндрей Шедько/AI Tools Lab /SRR/Engineer/삼성전자 <a.shedko@samsung.com>
Wed, 10 Apr 2019 03:54:57 +0000 (06:54 +0300)
committer박세희/On-Device Lab(SR)/Principal Engineer/삼성전자 <saehie.park@samsung.com>
Wed, 10 Apr 2019 03:54:57 +0000 (12:54 +0900)
commitda96d8f8219abf6b5b2d59905c1cee1b0a0fbc81
tree528cff322daf0ecfec12c0778c79886a1bde1461tree | snapshot
parent8d2222f661cb7fcdc7c54aeec8012c61b3335b77commit | diff
[NNFW] optimized Conv on CPU (#4964)

Moved condition and offset calculation out of the innermost loop in cpu
convolution kernel. Achieved 2+x speed up on mobilenet on neurun.

Signed-off-by: Andrei Shedko <a.shedko@samsung.com>
libs/cker/include/cker/operation/Conv.h diff | blob | history
Domain: Machine Learning / ML Framework;