18 Commits

Author	SHA1	Message	Date
Po Yen Chen	c6eac9746f	Fix type errors in composes<>	2024-04-09 13:18:17 +00:00
rocking	83b8a99018	Merge branch 'ck_tile/refactor' into ck_tile/elementwise	2024-04-09 19:45:43 +08:00
carlushuang	89a75a97fa	fix some bug in group-mode masking and codegen. update README	2024-04-09 19:01:25 +00:00
Po Yen Chen	ecc64bce12	Generalize the composes<> template	2024-04-09 10:14:56 +00:00
Po Yen Chen	6ed739f913	Fix wrong value produced by saturating	2024-04-09 09:27:58 +00:00
Po Yen Chen	5d0ebdbfe4	Re-use already-existing scales<> functor template	2024-04-09 08:06:38 +00:00
Po Yen Chen	db0d7c6a99	Use conditional_t<> to simplify code	2024-04-09 06:52:54 +00:00
rocking	525b89e538	1. codgen the f8 api and kernel ... 2. f8 host code	2024-04-08 21:36:23 +00:00
rocking	5860f3134a	Merge branch 'ck_tile/refactor' into ck_tile/elementwise	2024-04-09 02:37:42 +08:00
Po Yen Chen	e49498f616	Set fp8 rounding error for check_err()	2024-04-08 12:39:37 +00:00
rocking	5c3fdeb0b8	Remove f8 pipeline, we should share the same pipeline even in f8	2024-04-08 09:56:23 +00:00
rocking	f7d81364f3	To prevent compiler issue, remove the elementwise function we have not used.	2024-04-08 09:44:21 +00:00
carlushuang	42ebffe822	1).support receipe in generate.py 2).use simplified mask type 3).change left/right to pass into karg	2024-04-07 23:30:34 +00:00
rocking	286c74468d	Add element function to fmha api	2024-03-29 18:05:36 -04:00
carlushuang	0e7df1999f	wip fix	2024-03-06 14:31:36 +00:00
carlushuang	a67473fff8	now can build	2024-03-04 20:45:51 +00:00
carlushuang	112d521b09	fix xx	2024-03-03 23:48:31 +00:00
carlushuang	f69356b1d7	add code	2024-02-28 22:57:19 +00:00