16 Commits (a6ff5240f519380257f206fbc9c7f720fff4badc)

Author	SHA1	Message	Date
Liu Yiqun	3fefee8a06	Use scalar implementation instead of neon implementation to avoid out of range memory access in the tail conv3x3.	7 years ago
Liu Yiqun	9e524fce1b	Remove the generation of executable, such as paddle_trainer, when setting WITH_C_API.	8 years ago
hedaoyuan	0146e8bfc5	Merge pull request #3843 from Xreki/build_android_clang ... Enable the use of clang to cross-compile for Android	8 years ago
Liu Yiqun	3360e9cdb8	Change the definition of vmlaq_laneq_f32 from template function to macro.	8 years ago
Liu Yiqun	c54c7d91a0	Use template to deliver const argument instead, to remove the compiling error "argument to __builtin_neon_vgetq_lane_f32 must be a constant integer".	8 years ago
hedaoyuan	90bf4f60ae	Add stride support 2 for NeonDepthwiseConvTranspose.	8 years ago
hedaoyuan	840104c99a	Add NeonDepthwiseConvTransposeFunction.	8 years ago
hedaoyuan	40d47fae95	[Refine code]Move class Padding into the NeonDepthwiseConv.h.	8 years ago
hedaoyuan	4b6b7251c1	Refine NeonDepthwiseConv.	8 years ago
hedaoyuan	f7e75a03cf	Refine the neon depthwise convolution code(separate the Function and kernel).	8 years ago
hedaoyuan	5df384d67f	Remove NeonDepthwiseConv.h	8 years ago
hedaoyuan	227fdfb65d	Refine NeonDepthwiseConvFunction.	8 years ago
hedaoyuan	f00c4112d2	Neon depthwise conv with filterSize = 4 and stride = 2.	8 years ago
hedaoyuan	6dcff9a444	Neon depthwise conv with filterSize = 3 and stride = 2.	8 years ago
hedaoyuan	b7885b087b	Add DepthwiseConvKernel for filter size is 4.	8 years ago
hedaoyuan	0dffe68ca9	Add NeonDepthwiseConvFunction.	8 years ago