<mark_weaver>I'd like to see the debate between Gustafson and Kahan. <mark_weaver>fwiw, although I appreciate the elegance of the Unum representation, I'm very skeptical that solves any deep problems with inexact arithmetic, and it seems that Kahan disagreed as well. <DeeEff[m]>I suppose I would have to see the talk myself, but at a glance they seem to correct a significant number of shortcomings with inexactness. <DeeEff[m]>The idea would be that with unums, nothing is inexact. <DeeEff[m]>The data parallelism and bit operation lookups seem to offer an advantage that floats cannot offer. Basically hardware backed unum solutions would be able to save power / computing time. <DeeEff[m]>Well, not possible in the sense of an exact decimal (think pi) but possible in the sense that things are represented by intervals, not single values. <DeeEff[m]>So your interval width corresponds directly to your significance. <mark_weaver>sorry, I don't have time to discuss this further now. <mark_weaver>but if you're serious about this, feel free to dig in and implement it in Guile :) <mark_weaver>Guile supports user-defined numeric types through GOOPS. <DeeEff[m]>I was actually thinking of giving it a shot in C, then porting to CHICKEN, but I have more projects than time nowadays :) <ijp>but newer than you'd think <daviid>no gmail account no git send-mail ? <jeffers-media>wingo: interested in learning about compilers and thought contributing to some of the compiler tasks mentioned in your blog would be a fun practical suppliment <jeffers-media>since that post was in february, are there any updates regarding those tasks? ***thomas_ is now known as adverb
<zv>actually, if anyone can answer this question i would be really thankful <zv>ijp said that modules need not span an entire file <zv>my sicp4.scm implementation is in *exactly* one file, i am now about to make a VERY large change that I want to be able to turn on/off with a single variable or function call <ijp>that's a very broad question <jeffers-media>anyone happen to know the nick of the Robin Templeton, the main developer of guile-emacs? <ijp>they might still go by bipt, but I'm not sure <wingo>jeffers-media: no updates, contributions welcome :) i would start by mailing a plan to the list, if you want feedback, or if you prefer to just hack that's fine <wingo>re: elisp, there was a recent rebase <ijp>the "linking multiple modules in one file" is a problem I'm going to end up running into sooner or later <wingo>it would be really nice to solve <wingo>i don't quite know how you would go about loading (foo bar) but that module is actually bundled into (foo) <wingo>like module foo/bar is in foo.go <wingo>or maybe even (bar) is in foo.go <wingo>puzzling over this 2.0 peval bug <civodul>2.0 peval bugs? how's that possible? :-) <wingo>somehow it doesn't exist in master and i also want to know why that is :) <wingo>it was fixed in master but not back-ported <wingo>civodul: do you know the right incantation to forward the netbsd gnulib time-internal.h bug to gnulib? <wingo>istr it's possible to do that with bug tags or something, dunno <civodul>wingo: i think it's "reassign NNN PACKAGE" <amz3`>I am kind of stuck in my project(s) does anyone needs help for something? <amz3`>well, I found a nice workaround! <amz3`>I implemented a missing procedure from the bindings and make use of it <amz3`>so it's not really a work around <amz3`>the gist of it is that one can use `cursor-search' in place of `cursor-search-near' for exact match which is what I need <galex|713>amz3`: what are the kind of “bad choices” of mongodb? <amz3`>I mean that they do bad strategic choices for themself, like doing analytics using pgsql <amz3`>which seems like a good thing when you want to do analytics but not a good thing for mongodb which primary competitor is pgsql <galex|713>I thought nosql and sql were for different purposes? <galex|713>(yet a friend said me pgsql was better than mongodb, dunno why, probably performance) <paroneayea>(random-state-from-platform), it says it's not necessarily safe for high-security applications <paroneayea>I'm guessing if it succeds at getting info from /dev/urandom it is <lfam>paroneayea: I wondered the same thing when I read it <lfam>I did ask here but I don't remember if there were any replies <amz3`>paroneayea: I wanted to let you know, that I've hitting dustycloud with my search engine, let me know if there is a problem