vighneshiyer

6 ms·

1.
by vighneshiyer 1y ago
Here is the relevant bit from their whitepaper ( https://storage.googleapis.com/deepmind-media/DeepMind.com/B... ): > AlphaEvolve was able to find a simple code rewrite (within an arithmetic unit within the matmu
2.
by vighneshiyer 2y ago
Existing mixed-placement algorithms depend on hyperparameters, heuristics, and initial states / randomness. If afforded more compute resources, they can explore a much wider space and in theory come up with better solutions. Some algor
3.
by vighneshiyer 2y ago
Cadence in particular has been quite receptive to allowing academics and researchers to benchmark new algorithms against their tools. They have also been quite permissive with letting people publish TCL scripts for their tools ( https:&#x2F
4.
by vighneshiyer 2y ago
So I'm not sure what Google is referring to here. As you can see in the ISPD paper ( https://vlsicad.ucsd.edu/Publications/Conferences/396/c396.p... ) on page 5, they openly compare Cadence CMP with AutoDM
5.
by vighneshiyer 2y ago
> One key argument in the rebuttal against the ISPD article is that the resources used in their comparison were significantly smaller. To me, this point alone seems sufficient to question the validity of the ISPD work's conclusions.
6.
by vighneshiyer 2y ago
You are correct. For commercial use, the GPUs used for training and fine-tuning aren't a problem financially. However, if we wanted to rigorously benchmark AlphaChip against simulated annealing or other floorplanning algorithms, we hav
7.
by vighneshiyer 2y ago
I have published an addendum to an article I wrote about AlphaChip ( https://vighneshiyer.com/misc/ml-for-placement/ ) at the very bottom that addresses this rebuttal from Google and the AlphaChip algorithm in gener
8.
by vighneshiyer 2y ago
This work from Google (original Nature paper: https://www.nature.com/articles/s41586-021-03544-w ) has been credibly criticized by several researchers in the EDA CAD discipline. These papers are of interest: - A rebutta