2023-06-23 10:43:03 <veltas      > Division by invariant integers using multiplication https://gmplib.org/~tege/divcnst-pldi94.pdf
2023-06-23 10:46:53 <veltas      > Looking back through my git logs, why am I so productive on Thursdays?
2023-06-23 10:48:01 <veltas      > Apparently this corroborates with research on the more and less productive days of the week
2023-06-23 11:01:09 <KipIngram   > That's in Hacker's Delight.
2023-06-23 11:01:20 <KipIngram   > I currently use it to do modulo 255 in my block buffer hash.
2023-06-23 11:01:48 <KipIngram   > You're looking forward to the weekend and feeling good about its imminence.
2023-06-23 11:01:59 <KipIngram   > Got hump day behind you.
2023-06-23 11:03:13 <thrig       > Thursday, I don't care about you
2023-06-23 15:07:22 <gordonjcp   > veltas: multiply by reciprocal?
2023-06-23 16:10:02 <unjust      > does anyone here have a copy of: 'MOUSE, A Language for Microcomputers'?
2023-06-23 16:44:55 <drakonis    > tried libgen?
2023-06-23 16:46:44 <drakonis    > ah
2023-06-23 16:46:48 <drakonis    > you might find this in archive.org
2023-06-23 16:47:05 <drakonis    > i think they were importing byte magazine scans into the archive a while back
2023-06-23 16:47:44 <drakonis    > i'm considering calling my forth implementation "metaforth" but is it a bit on the nose?
2023-06-23 16:48:45 <drakonis    > unjust: its indeed in archive.org
2023-06-23 16:48:47 <drakonis    > https://archive.org/details/BYTE_Vol_04-07_1979-07_Automating_Eclipses
2023-06-23 16:49:09 <drakonis    > https://archive.org/details/byte-magazine-1979-07-rescan
2023-06-23 16:50:14 <drakonis    > https://archive.org/details/byte-magazine-1980-06 has corrections
2023-06-23 16:51:01 <drakonis    > the first url seems to be the good one? i'm having a bit of trouble figuring out which one is the washed out copy that looks worse
2023-06-23 16:51:22 <drakonis    > ah, nevermind that, the first url is the latest scan
2023-06-23 17:11:55 <unjust      > drakonis: thanks for the byte article, but i'm looking for the book written by the same author
2023-06-23 17:14:18 <drakonis    > the title makes it difficult to find a digital version
2023-06-23 17:15:08 <unjust      > agreed, and apparently the publisher folded soon after it was published (according to the author)
2023-06-23 17:15:29 <drakonis    > plus its extremely obscure
2023-06-23 17:15:34 <drakonis    > so its even harder to find anything
2023-06-23 17:18:02 <unjust      > it's an interesting language though, and given its similarity to forth i thought this channel might be worth asking in. (http://www.massmind.org/techref/language/mouse.htm and https://www.youtube.com/watch?v=EaunbXlDUWI sum it well enough though)
2023-06-23 17:24:54 <drakonis    > fair enough
2023-06-23 18:34:47 <KipIngram   > You know, I don't understand why people get all bent out of shape over someone posting a meme picture in the Dresden Reddit.  We do expect posts to be "topical," but in the Meme category I think a fairly loose net is appropriate.  My reasoning is usually that if someone thought it was Dresden related, someone else might too.
2023-06-23 18:34:57 <KipIngram   > Some folks are just so uptight over it all.
2023-06-23 18:40:28 <KipIngram   > gordonjcp: What veltas was talking about was avoiding the division instruction.  If you don't know what number you're going to need to divide by, you can't.  But if you have a particular case, where you know in advance what you'll need to divide by, then there's a trick you can play to get the right answer via multiplication.
2023-06-23 18:40:39 <KipIngram   > And no, it's not reciprocal - we're talking about integers here.
2023-06-23 18:40:50 <KipIngram   > Integers modulo 2^N, where N is your word size.
2023-06-23 18:41:57 <KipIngram   > The case I used it in was needing to know mod 255 of a block number - that's how I choose which disk buffer to house it in.
2023-06-23 18:42:14 <KipIngram   > For various reasons I only had 255 buffers, not 256.
2023-06-23 18:42:40 <KipIngram   > I figured out the number I'd need to multiply by and did that with a series of shifts and adds.
2023-06-23 18:51:05 <Zarutian_iPad> so 127 - 7 >> ?
2023-06-23 18:57:12 <KipIngram   > The number I needed to multiply by was 0x1010101.
2023-06-23 18:58:08 <KipIngram   > And after multiplying there were a few other cheap steps.
2023-06-23 18:58:54 <KipIngram   > Actually I see that I didn't do shifts and adds - I did use imul.
2023-06-23 19:02:41 <KipIngram   > gordonjcp: This is all rooted somehow in the math of finite fields.