IRC channel logs
2022-08-03.log
back to list of logs
<wingo>got a generational collection working in whippet, in a way that allows (but does not require) conservative root-finding as in bdw-gc <civodul>wingo: wo0t! sounds like it's getting serious :-) <drakonis>so, how exactly would i proceed with transitioning kernel to scheme? <dthompson>hmmm guile doesn't seem to optimize away multiplications by zero. <dthompson>I ran a little test: ,x (lambda (x) (+ (* x 0) x)) <dthompson>I guess guile could only optimize it away if it knew that x was a number. <dthompson>this even more contrived example generates bytecode that eliminates the math: ,x (lambda (x) (let ((y 3)) (+ (* y 0) y))) <dthompson>so nvm, guile will do the right thing if it knows <lilyp>tbf you can do operator overloading with goops <lilyp>so i could define my own record type for which (* x 0) yields 1 <dthompson>lilyp: yeah but at compile time guile knows if it's dealing with primitive math operators or not <dthompson>there's lots of math optimizations that are happening that wouldn't be possible otherwise *dthompson goes to find the channel logs <dthompson>oh wow! nice indeed! a new gc is very exciting <ArneBab>Sidenote: Some weeks ago I offered that I could review and merge simple compatibility patches if I could get commit access but didn’t see replies. My goal is to avoid frustration of contributors who provide boring-and-simple-and-useful improvements. Did I miss something? Whom should I ask to move forward? <civodul>ArneBab: howdy! i didn't see your earlier request, but what you suggest sounds good to me <civodul>like you write, merging the boring-and-simple-and-useful patches quickly enough would be great for everyone <civodul>so unless wingo disagrees, let's do that :-) <civodul>and of course you're welcome to ask for guidance on IRC, esp. at the beginning <ArneBab>civodul: sounds good! Thank you! (and high five for persistence! :-) — I know how easy it is to miss stuff myself …)