800x Speed Boost on Nvidia GPUs
scmp.comhttp://cjcm.ijournals.net.cn/jslxxb/ch/reader/view_abstract....
Appears to be this, though different title: https://www.sciencedirect.com/science/article/pii/S095579972...
I wonder if this is also a CUDA-bypass, PTX optimization that led to the 10x performance gain by Deepseek: https://xyzlabs.substack.com/p/deepseeks-latest-shocker-who-...
Which GPUs? I don't have paid access to the article so I can't read much of anything worthwhile.
it doesnt say https://archive.md/Hq7ms
Well that is highly disappointing. I was hoping to gain insight into what they were doing so I could see how they managed to speed up an algorithm by 800x.