12 Commits (3d5e3a910faaec30b15eb694d5af7ec5ef435746)

Author SHA1 Message Date
zihaomu 840b1d5c94 add depthwise add fuse 2 years ago
zihaomu 82616eec41 fix possible segmentation fault error in winograd on x86 2 years ago
zihaomu 71765858dc fix invalid memory access 2 years ago
zihaomu 71c6339af0 remove old convolution branch, and optimize conv3d and conv1d. 2 years ago
Zihao Mu 0a650b573b
Merge pull request #22840 from zihaomu:optimze_conv_memory_usage 2 years ago
Zihao Mu cee8c86b6e fixed bug at winograd of SIMD128 and more robust code. 2 years ago
Zihao Mu 0fa43e3aac Optimize the winograd futher more. 3 years ago
Zihao Mu 1e2ceca4df add enableWinograd API for Net. 3 years ago
Zihao Mu 2cd7e17b65 replace v_add with + 3 years ago
Zihao Mu bb64db98d8
Further optimization of Conv2D, fused Conv_Add_Activation, bring latest code from ficus OpConv.fx. (#22401) 3 years ago
Zihao Mu 59b870a87a
Merge pull request #21910 from zihaomu:fast_conv_ARM 3 years ago