239 Commits

This Branch

This Branch

All Branches

Author	SHA1	Message	Date
Chao Liu	81497a93a0	reorginze files	2019-06-11 23:49:51 -05:00
Chao Liu	88b77181aa	rename files, added header guard, added namespace	2019-06-11 14:35:59 -05:00
Chao Liu	05e046654c	remove .hip extension	2019-06-11 09:12:41 -05:00
Chao Liu	a68b16a5d9	refactor	2019-06-06 19:49:40 -05:00
Chao Liu	0a386c46a9	use more constexpr for Array	2019-06-06 19:26:08 -05:00
Chao Liu	08c692433e	fix build issue	2019-06-06 17:13:15 -05:00
Chao Liu	7a89684f92	refactor	2019-06-06 16:50:35 -05:00
Chao Liu	eafdabba77	fix build issue	2019-06-04 20:11:04 -05:00
Chao Liu	709f13a6d7	use more constexpr	2019-06-04 20:00:48 -05:00
Chao Liu	498e71b098	try using more constexpr	2019-06-04 17:02:49 -05:00
Chao Liu	917d7a2b1d	use vectorized read and write for threadwise generic tensor copy	2019-06-03 18:58:01 -05:00
Chao Liu	33c38e2a23	tuned	2019-06-03 10:50:25 -05:00
Chao Liu	7cdb665a50	refactor	2019-06-01 10:48:48 -05:00
Chao Liu	97ba755f2f	refactor	2019-05-31 22:28:32 -05:00
Chao Liu	8d4607403e	refactor	2019-05-31 09:59:00 -05:00
Chao Liu	3a6044aa84	refactor	2019-05-31 01:25:14 -05:00
Chao Liu	9ef124cc15	fix build issue	2019-05-30 21:29:08 -05:00
Chao Liu	ad5db169b6	added lds doubl buffer for implicit gemm v4 (nchw, kcyx)	2019-05-30 20:55:21 -05:00
Chao Liu	cb78cc74b7	added implicit gemm v4 (nchw, kcyx)	2019-05-30 20:28:01 -05:00
Chao Liu	b2439ec9dd	adding implicit gemm v4 (nchw, kcyx)	2019-05-30 17:50:49 -05:00
Chao Liu	0a2657312e	adding implicit gemm v4 (nchw, kcyx)	2019-05-30 00:07:39 -05:00
Chao Liu	8c385cf5cf	implicit gemm v3 added lds double buffer	2019-05-24 17:17:22 -05:00
Chao Liu	e17b495dff	refactor	2019-05-24 16:43:29 -05:00
Chao Liu	8fcf3f1e68	added implicit GEMM v3	2019-05-23 23:54:31 -05:00
Chao Liu	8ce1480430	refactor	2019-05-23 22:53:35 -05:00
Chao Liu	1cc683a3a3	adding implicit gemm v3	2019-05-23 22:10:40 -05:00
Chao Liu	8a4b59785b	adding implicit gemm v3	2019-05-22 19:39:56 -05:00
Chao Liu	2a48812edb	behavior has changed (better and worse), figuring out why	2019-05-21 16:43:56 -05:00
Chao Liu	acd7082fe1	adding ConstantMergedTensorDescriptor, refactering ConstantTensorDescriptor, Sequence	2019-05-21 16:17:58 -05:00
Chao Liu	cd29b09a82	refactor	2019-05-19 11:41:00 -05:00
Chao Liu	a6b95c393b	rework sequence	2019-05-18 23:21:02 -05:00
Chao Liu	df73287b82	rework sequence	2019-05-17 14:56:39 -05:00
Chao Liu	33b5a8556b	adding implicit gemm v3	2019-05-16 22:23:18 -05:00
Chao Liu	5e5c27a63b	adding implicit gemm v3	2019-05-16 13:22:40 -05:00
Chao Liu	b7d052459d	adding implicit gemm v3	2019-05-15 09:58:17 -05:00
Chao Liu	4957d5a399	refactored	2019-05-02 14:49:20 -05:00
Chao Liu	4a99f54c31	remove dead code	2019-05-02 11:09:42 -05:00
Chao Liu	d2436a58c4	v1r3 nchw*cyxk=nkhw lds double buffer	2019-04-27 15:06:15 -05:00
Chao Liu	63cdc6d2a4	fix v1r3 output reorder bug	2019-04-27 00:58:49 -05:00
Chao Liu	c138e2126d	nchw*cyxk*nkhw on AMD	2019-04-26 18:03:55 -05:00
Jing Zhang	49d5af1002	ds_read_offset	2019-04-26 15:55:26 -05:00
Chao Liu	3ce77700b6	debugging ds_read asm	2019-04-26 15:34:55 -05:00
Chao Liu	b93d2e1b57	fix batch gemm asm bug	2019-04-26 14:40:19 -05:00
Chao Liu	46a0aec185	trying gemm asm	2019-04-26 10:20:02 -05:00
Chao Liu	2603bb0fe3	tuning on vega 20	2019-04-25 17:28:59 -05:00
Chao Liu	a903146427	implicit gemm v1r3 nchw_cyxk_nkhw	2019-04-25 15:14:39 -05:00
Chao Liu	569ad66e2a	added implicit gemm v1r3 lds_double_buffer NCHW * CYXK = KNHW, reworked static functionals	2019-04-23 17:51:14 -05:00
Chao Liu	87d8740bf5	added lds double buffer (on C dimension) for implicit gemm v1r3, as a result, it should achieve 90% of peak for all filter sizes, on CHWN format	2019-04-19 17:49:25 -05:00
Chao Liu	6d066ede00	added implicit gemm v1r3, refactored decomposition of wei tensor (loop over y, x first, and C second) to allow easy lds double buffer on C	2019-04-19 16:46:29 -05:00
Chao Liu	5ce19234a4	added GridwiseConvolutionImplicitGemm_v1r2_nchw_cyxk_khwn	2019-04-19 14:22:02 -05:00