24 Commits

Author	SHA1	Message	Date
Chao Liu	8732ea04fb	tweaked params for direct conv; added a dummy winograd	2018-11-24 03:55:36 -06:00
Chao Liu	dbffe05a98	add host winograd 3x3 conv	2018-11-21 13:21:34 -06:00
Chao Liu	a21b0d27a5	refactor	2018-11-20 10:51:28 -06:00
Chao Liu	6790b8f3cc	rename	2018-11-20 10:43:37 -06:00
Chao Liu	d2a488ddec	hand tuned params	2018-11-20 10:34:16 -06:00
Chao Liu	c587726190	faster: output skip LDS	2018-11-16 06:08:11 -06:00
Chao Liu	5096a157da	refactor	2018-11-16 01:45:24 -06:00
Chao Liu	73480fee36	refactor	2018-11-15 23:53:23 -06:00
Chao Liu	adf4b173b3	refactor	2018-11-15 23:22:06 -06:00
Chao Liu	99d05ba77f	refactor	2018-11-15 22:41:01 -06:00
Chao Liu	c8d0356a34	faster	2018-11-15 18:54:57 -06:00
Chao Liu	1812666a47	improved blockwise_tensor_op	2018-11-14 08:55:45 -06:00
Chao Liu	ff31af2227	refactor	2018-11-13 15:46:19 -06:00
Chao Liu	0404f777b4	refactor	2018-11-13 15:22:41 -06:00
Chao Liu	08c7f74391	add 2nd version of blockwise_tensor_op	2018-11-07 18:35:54 -06:00
Chao Liu	5d2cafcb24	clean up	2018-11-07 11:26:13 -06:00
Chao Liu	84b36bc18d	sucess cuda run	2018-11-04 14:55:55 -06:00
Chao Liu	6a45afba95	conv: update tensorDesc calculation	2018-11-04 04:55:37 -06:00
Chao Liu	1b648f2f42	use constant tensor descriptor	2018-11-04 04:08:51 -06:00
Chao Liu	9657baec32	initial direct conv correct run	2018-11-02 00:25:21 -05:00
Chao Liu	dfa0213942	convolution: init cuda run	2018-10-30 11:12:21 -05:00
Chao Liu	2f2cf35bf4	initial cuda build	2018-10-22 11:51:10 -05:00
Chao Liu	d51b81588f	cpu direct conv	2018-10-19 01:26:21 -05:00
Chao Liu	fc98757acd	start adding convolution	2018-10-08 22:49:58 -05:00