Skip to content

Nhwc fwd #79

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Jump to bottom
Open
wants to merge 44 commits into
base: master
Choose a base branch
from
Open

Nhwc fwd #79

wants to merge 44 commits into from

Conversation

@carlushuang
Copy link
Collaborator

@carlushuang carlushuang commented Jan 29, 2021

[WIP]

carlushuang and others added 30 commits January 14, 2021 14:39
@carlushuang
add 1st version of nhwc
6a94b64
@carlushuang
fixup following part of nhwc
4731118
@carlushuang
Merge remote-tracking branch 'origin/master' into nhwc_fwd
bec0823
@carlushuang
host build
6fa9c74
@carlushuang
fix crash
1b3df13
@carlushuang
remove several code
b72e0a1
@carlushuang
k_pack for smem
cacc8ef
@carlushuang
more add
88a3c36
@carlushuang
update code
05219e4
@carlushuang
support kpack in main loop
a782aaa
@carlushuang
tiny fix kpack
c9c97cb
@carlushuang
nxe=0 now works!
ad0b413
@carlushuang
fix a bug in fwd validation
d23bee5
@carlushuang
fix a bug in non 1x1 case
2647213
@carlushuang
add magic div
801261f
@carlushuang
fwd support split batch (#78)
8afe37a 
* fwd support split batch

* remove confusing assert

* fix several >4G address type
@carlushuang
split 4G support
5bb8587
@carlushuang
fix type
fc4e716
@carlushuang
Merge remote-tracking branch 'origin/master' into nhwc_fwd
17f1e52
@carlushuang
fix size
425ddbc
@carlushuang
Merge remote-tracking branch 'origin/master' into nhwc_fwd
f381b0c
@carlushuang
remove a useless check
ded5a91
@carlushuang
pad gemm_n
20d501b
@carlushuang
update nhwc
24d72e7
@carlushuang
refactor some code
1c62a04
@carlushuang
further update
8edb357
@carlushuang
update code
ba43174
@carlushuang
enable magic div by default
830c21d
@carlushuang
fix a bug in magic number calculation
7cbc45b
@carlushuang
update small change
6b2ddb2
carlushuang and others added 14 commits March 11, 2021 18:44
@carlushuang
inference prototype for gfx1030
c40dc9b
@carlushuang
gfx1030 kernel asm
6359fe5
@carlushuang
pretty asm
6813498
@carlushuang
tiny fix
6b05999
@carlushuang
256x16 tile size
f533622
@carlushuang
proper global inst
de7c670
@carlushuang
remove limitation
798c0a4
@carlushuang
Merge remote-tracking branch 'origin/nhwc_fwd' into nhwc_fwd
3d62902
@carlushuang
add create_base_args
4d57876
@carlushuang @shaojiewang
Nhwc fwd gemmk split (#88)
d756706 
* add raw code to pta case

* fix bug for non-pta case

* add some tunables

* update all configs

* remove redundant print

* refactor driver

Co-authored-by: shaojiewang <wsjmessi@163.com>
@carlushuang
fix a bug in find max gks
cf45e48
@carlushuang
add missing configs
2cba95c
@carlushuang
Nhwc inference test gfx1030 (#90)
2dc86a8 
* 256x8

* 512x8

* optimize double global prefetch

* 512x8x8

* 1024x8

* 512x4

* 256x4

* 128x4, 384x4

* 256x8x8

* int8x4

* 512x8 int8

* update 256x8x16, 512x16x8

* 1024x16x8

* 256x4
@carlushuang
ignore print in wrw driver
ae7cef3
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Reviewers

No reviews

Assignees

No one assigned

Labels

None yet

Projects

None yet

Milestone

No milestone

Development

Successfully merging this pull request may close these issues.

2 participants

@carlushuang