News for hackers. Stuff that matters.

A comment on AMD's CDNA 3 Compute Architecture. Parent comment

Re:AMD's CDNA 3 Compute Architecture by treprinum on Sunday December 17, 2023 @05:05PM (#38676938)
The issue is that RDNA and CDNA are different so when an enthusiast makes a fast RDNA code it doesn't mean it would work well on CDNA and vice-versa. Not sure why AMD had to go this route, only high-end pros will write software for CDNA and they won't get any mindshare.
[ Parent | on HN ]
Re:AMD's CDNA 3 Compute Architecture by JonChesterfield on Sunday December 17, 2023 @05:52PM (#38677238)
What do you have in mind for code that works noticeably better on one than the other?
[ Parent | on HN ]
Re:AMD's CDNA 3 Compute Architecture by jacoblambda on Sunday December 17, 2023 @10:54PM (#38679087)
That's not accurate at all.

Someone writing for the rocm platform will be writing HIP and then HIPCC will compile that down into the actual runtime targeting the given architecture. HIP is pretty platform agnostic so very little device specific optimization tends to go into idiomatic HIP.

[ Parent | on HN ]
Re:AMD's CDNA 3 Compute Architecture by treprinum on Monday December 18, 2023 @05:20AM (#38680883)
There are many low-level details in CUDA to get it working fast on a given GPU architecture that compilers don't address at all.
[ Parent | on HN ]