Skip to content
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.

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

Create rav1e glue code for dav1d 0.6.0 new ASM #2195

Closed
4 tasks
shssoichiro opened this issue Mar 6, 2020 · 5 comments · Fixed by #2377
Closed
4 tasks

Create rav1e glue code for dav1d 0.6.0 new ASM #2195

shssoichiro opened this issue Mar 6, 2020 · 5 comments · Fixed by #2377
Labels
SIMD Architecture-specific SIMD optimization speed performance

Comments

@shssoichiro
Copy link
Collaborator

shssoichiro commented Mar 6, 2020

Follow-up to #2152

  • Use prep_bilin/prep_8tap AVX-512 asm
  • Use HBD put_8tap/prep_8tap/prep_bilin NEON asm
  • Use 10-bit SGR NEON asm (specifically DOES NOT WORK with 12-bit)
  • Use LBD/HBD cdef NEON asm (can we? isn't there a mismatch between our cdef and dav1d's? same for AVX-512 cdef...)
@shssoichiro shssoichiro added SIMD Architecture-specific SIMD optimization speed performance labels Mar 6, 2020
@lu-zero
Copy link
Collaborator

lu-zero commented Mar 8, 2020

I prepared already medek/nasm-rs#28

@shssoichiro
Copy link
Collaborator Author

Seems to be no response from the maintainer there. I hate to say we might want to fork it... but should we fork it?

@tdaede
Copy link
Collaborator

tdaede commented Apr 4, 2020

We've forked it before. I guess we can fork it again.

@lu-zero
Copy link
Collaborator

lu-zero commented Apr 4, 2020

Or we can write the check in build.rs directly. Once I'm fully awake I can send a pr if nobody beats me at it.

@shssoichiro
Copy link
Collaborator Author

#2223

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
SIMD Architecture-specific SIMD optimization speed performance
Projects
None yet
Development

Successfully merging a pull request may close this issue.

3 participants