<fossy>OriansJ: I would be too concerned about performance rn it is sufficiently fast in 99% of cases <OriansJ>I figured out an O(1) solution for the application of macros; the downside is that it is 8times slower <OriansJ>who knew hashing really cut down the processing time <OriansJ>it went from 8times slower to 10times faster <OriansJ>(possibly more as 1 second is down to .1 second) <OriansJ>with probably the worst hash possible <OriansJ>with a 65536 entry table for 13K distinct symbols <OriansJ>So it could probably be even faster if I can get the collision rate down <xentrac>heh, my paper address book uses the first and third letter of people's name as the hash <xentrac>lumped into categories: ab, cd, ef, gh, ij, etc., to fit the hash table on the page <OriansJ>i = (i << 5)+ i + s[0]; would probably give a better spread <xentrac>yeah but it's harder to do in my head <xentrac>you can always count on bob to go too far <OriansJ>xentrac: did you have your meeting with the Bobs yet? ;-p <OriansJ>for added fun converting @9 to 0009 is now done only exactly once (instead of as many times as @9 appears) <OriansJ>and when it is built via GCC it is so fast, it doesn't even show up in htop when running the M2-Planet tests <OriansJ>and commit is up for review (I admit it is probably ugly) <OriansJ>I'll profile hex2 tomorrow and figure out what is slow there but first sleep <OriansJ>I spent way too much sleeptime on programming today. Good night <OriansJ>hopefully with a speed upgrade in hex2 and a release of mescc-tools; janneke will be able to cut his bootstrap time by half <OriansJ>oh thank god; finally after a week. guix pull and guix package -i emacs finally worked <OriansJ>(yes I was forced to use mg to do development this week) <OriansJ>oh tree-undo mode, org-mode and magit how I missed you <vagrantc>couldn't you have just used an old generation? <OriansJ>hence my frustrations going from a fresh tarball to successfully doing a single guix pull <OriansJ>if it wasn't for efraim's help and his missing tarballs collection; I'd still be stuck at the unable to successfully guix pull <efraim>I gathered a couple while trying to revive the mips64el port for Guix, but the 1GB of ram on the board wasn't enough <efraim>most of them I grabbed from bayfront <OriansJ>sadly 4GB might be too small for builds soon. (ghc really can't tolerate that little) <OriansJ>It is weird, I remember when 8MB was alot of RAM <efraim>the mips board has 1GB and I gave it 4GB of swap, all on an SSD. I forget if it exhausted itself compiling guix or guile, but I only made it as far as about 0.15 <OriansJ>efraim: I'd say guile probably as it really doesn't like less than 2GB of RAM for some reason <OriansJ>possibly due to how it handles garbage collecion; resulting in rather bad page thrashing even with an SSD <OriansJ>builds in general have gotten way less [computer resource] efficient for questionable improvments in developer efficiency <OriansJ>rain1: they have gotten alot cheaper over the last few years but they still have a price premium over spinning disks; ironically all of the biggests disks are SSD only (100TB+ Drives) and spinning disks are now just a fraction of the size of SSDs (<20TB) <xentrac>hopefully they will reincarnate soon ***ChanServ sets mode: +o rekado