Andrew Zonenberg on Nostr: nprofile1q…ydpy5 Welcome to the very exclusive club of people who have got so fed ...
nprofile1qy2hwumn8ghj7un9d3shjtnddaehgu3wwp6kyqpq7wkkj73azdfr23hmgd07fn0h39xvy0w72garahk5p8hg8jv8dx5snydpy5 (nprofile…dpy5) Welcome to the very exclusive club of people who have got so fed up with the performance of some piece of commercial software that they optimized it themself, with only the binary.
In my case it was the matrix solver inner loop of Sonnet, which at the time was only using SSE and not AVX (and had some other tweaks possible, like refactoring some bounds checks done each loop iteration to be static before the loop entry)
In my case it was the matrix solver inner loop of Sonnet, which at the time was only using SSE and not AVX (and had some other tweaks possible, like refactoring some bounds checks done each loop iteration to be static before the loop entry)