[HN Gopher] CUDA-L2: Surpassing cuBLAS Performance for Matrix Mu...
___________________________________________________________________
CUDA-L2: Surpassing cuBLAS Performance for Matrix Multiplication
Through RL
Author : dzign
Score : 40 points
Date : 2025-12-04 21:04 UTC (1 hours ago)
(HTM) web link (github.com)
(TXT) w3m dump (github.com)
| stonogo wrote:
| Am I reading this wrong, or does this only support FP16 inputs,
| and compares its performance against an FP32 solver?
| bgwalter wrote:
| -4 -4 -4 -4 -4
| krapht wrote:
| This is a standard which few kernels will ever meet. I'd say
| requiring a numerical proof is the same as requiring no proof
| at all - because it won't ever happen unless you're validating
| silicon or something equally expensive.
| Q6T46nT668w6i3m wrote:
| I guess it depends on your definition of proof but I'd say
| the reasoning and justifications sections of a TOMS article
| qualifies and that's a standard nearly every popular library
| meets.
| j2kun wrote:
| They claim the algorithm "discovered" the new techniques, but the
| methods described in section 5 do not seem all that novel to me.
| It smells like it could be "laundering" the literature [1] and
| reshuffling existing techniques. This is not inherently a bad
| thing, but I would hope that if it is borrowing existing
| techniques, the appropriate citation would eventually make it
| into this paper.
|
| [1]: https://www.argmin.net/p/lore-laundering-machines
| AlexCoventry wrote:
| In the future, we will all be Jurgen Schmidhuber. :-)
| alyxya wrote:
| There generally aren't new techniques when optimizing something
| ubiquitous. Instead, there are a lot of ways to apply existing
| techniques to create new and better results. Most ideas are
| built on top of the same foundational principles.
| Q6T46nT668w6i3m wrote:
| You're not kidding. I just looked. There isn't anything novel
| in that section. I assumed from the description they found
| novel methods but this is standard GPU Gems advice.
| alyxya wrote:
| The chart confused me because I expected to see performance
| numbers of CUDA-L2 compared to the others, but instead it shows a
| chart showing the speedup percentage of CUDA-L2 over the others.
| In some sense, the bar chart effectively inverts the performance
| of torch.matmul and cuBLAS with how much percentage it shows. 0%
| on the bar chart would only mean equal performance.
___________________________________________________________________
(page generated 2025-12-04 23:00 UTC)