Reading an ASPLOS paper where the author claim a ~115% reduction in Miss per Thousands Instruction, over the baseline.

I don't think percentage work like that.

(I feel this suggest the baseline is 2.15x their optimised result, but the reduction in that case in 115/215 not 1.15)

Should I do something about this ?

#asplos #academia

On Sunday we're hosting LATTE ☕, our workshop on Languages and Tools for Accelerator design! We have 15 super cool papers and plenty of time for discussion!

You can attend virtually via zoom or in person with #asplos in Rotterdam https://capra.cs.cornell.edu/latte25/

Also, we have a cozy little Matrix chat for the community over at https://matrix.to/#/#latte-chat:fossi-foundation.org

Are you going to be in Rotterdam on Monday, e.g, for #ASPLOS or #EuroSys?

Come to our Workshop on User-Scheduable Languages https://usl-workshop.github.io/
USLs, as popularized by Halide, provide experts with great control over how to optimize their program for diverse hardware targets and have found use in many modern compilation frameworks.

We will have talks from Mary Hall, Thomas Koehler, Carl-Johan Seger, Alex Aiken, Alex Reinking, and Alex Zinenko.