Competition brings out the best on gpumode. 🔥
10billiontokens has written the best QR factor on Blackwell as of today. Few more days left for the competition as well
Hoping the competitors are making use of the parallelisms available for the upper triangular matrix T like constructing it in logarithmic number of steps.
dhu.randhar is beating Cuda Colonels’s forces and currently ahead by 101us.
a cool username I must say.
No Digg Deeper questions have been answered for this story yet.
This is resolved given GPU mode competition is now raising the bar for GPU kernel performance for LinAlg.
https://www.gpumode.com/news/linear-algebra-kernels-age-of-research
There are more kernels to be written
One major regret I have on the Scalable Shampoo paper is giving up on what now is called SOAP approach jn Appendix B due to lack of access to gpu / good eigh / qr on tpu.
Competition brings out the best on gpumode. 🔥
10billiontokens has written the best QR factor on Blackwell as of today. Few more days left for the competition as well
Hoping the competitors are making use of the parallelisms available for the upper triangular matrix T like constructing it in logarithmic number of steps.
dhu.randhar is beating Cuda Colonels’s forces and currently ahead by 101us.
a cool username I must say.
@_arohan_ Just need to cut 293us and it would be best number to get 🥇
@_arohan_