--- Log opened Wed May 25 00:00:33 2022 00:01 -!- muurkha [~kragen@adjuvant.canonical.org] has joined #hplusroadmap 00:20 < lkcl> nmz787, there's a sphinx plugin for converting back again. ascii-art --> xdot i think 00:40 < lkcl> kanzure, hwt syntax looks near-identical to nmigen, with some weirdness around assignment and comparisons 02:53 -!- darsie [~darsie@84-113-55-200.cable.dynamic.surfer.at] has joined #hplusroadmap 03:42 -!- Molly_Lucy [~Molly_Luc@user/Molly-Lucy/x-8688804] has quit [Read error: Connection reset by peer] 04:46 -!- yashgaroth [~ffffffff@2601:5c4:c780:6aa0::93] has joined #hplusroadmap 06:22 < lsneff> a year from now is going to be a pretty awful time to graduate 06:22 < lsneff> Picked my timing pretty badly 06:37 -!- mirage33591 [~mirage335@64.79.52.86] has quit [Quit: Client closed] 06:43 < nsh> pretty awfullest time *so far* 06:47 < kanzure> lsneff: between maaku and myself we have all sorts of employers that would take you in a heartbeat. 06:47 < kanzure> but this won't necessarily last forever 06:51 < lsneff> I’m not really worried for myself, but my friends and family 06:52 < lsneff> My friends are all graduating around the same time, but most of them aren’t engineers 07:17 < kanzure> https://www.prnewswire.com/news-releases/sosvs-indiebio-and-genome-project-write-partner-to-fund-and-advise-startups-301553921.html 07:18 < muurkha> it may be a difficult time specifically for engineers though 07:19 < muurkha> when the dot-bomb fell in 02000 and 02001, there were a lot of extremely competent programmers who spent several months involuntarily unemployed 07:21 < muurkha> imagine Elon Musk lays off several thousand Twitter employees and all the startups are in Survival Panic Don't Hire Anybody Mode 07:22 -!- Molly_Lucy [~Molly_Luc@user/Molly-Lucy/x-8688804] has joined #hplusroadmap 07:23 < muurkha> and at the same time more and more companies are willing to hire remote, and not willing to pay employees a premium for living in the US, so graduates of Rose-Hulman are competing for jobs with graduates of the IITs 07:23 < muurkha> it could be a pretty tight job market out there for US programmers for a few years 07:24 < muurkha> and it won't affect firefighters, middle managers, nurses, machinists, and truck drivers in the same way 07:26 < lsneff> I’ve got a couple film student friends haha 07:26 < lsneff> That’s a good point though, a lot of engineering aren’t really necessary 07:27 < muurkha> yeah, Hollywood probably isn't going to be affected by potential Twitter layoffs, and your gaffer and best boy can't WFH. but eager Ukrainian film students can do your sound mixing and editing 07:27 < muurkha> I still think it's a good bet to be building the systems that run the new society, particularly in a discipline where the most important barrier to entry is your personal ability rather than your certifications or the family you come from; and the salaries in the IT sector I think back that up 07:27 < muurkha> but there may be a few years where it's hard to find a job at all 07:27 < lsneff> I bet Twitter needs a 10th of engineers it has, that kind of thing 07:27 < lsneff> Yes 07:28 < muurkha> nothing is *necessary* except dying 07:28 < muurkha> (and we're working on that!) 07:28 < muurkha> but engineering can be very *profitable* 07:28 < lsneff> necessary has many different meanings 07:29 < muurkha> did you see the YC email to startups that's been circulating? 07:29 < lsneff> I saw it on HN I believe 07:32 < muurkha> a hacker friend of mine spent a couple of years during the dot-bomb working as a massage therapist 07:33 < lsneff> That’s unfortunate 07:33 < muurkha> I don't know, I kind of wish I could do that 07:34 < lsneff> You could 07:34 < muurkha> I'm not as cute and female and nonthreatening as she is 07:34 < kanzure> "How to break md5 and other hash functions" http://merlot.usc.edu/csac-f06/papers/Wang05a.pdf 07:35 < muurkha> and I think her massage skills were already better than mine at the time 07:35 < muurkha> but that matters at a later stage of the sales pipeline 07:37 < lsneff> Yes, I suppose it does haha 07:56 < juri_> during dot-bomb, i worked at a dollar general. 07:57 < muurkha> what was that like? 08:37 -!- superkuh [~superkuh@user/superkuh] has quit [Quit: putting in new hardware] 09:12 < nmz787> lsneff: why do you think a year from now will be terrible for graduating??? 09:13 < nmz787> Intel just said they are facing a shortage of workers due to other companies demanding, I guess 09:13 < nmz787> and unbanning previously ban-fired people 09:13 < nmz787> (for example) 09:15 < muurkha> nmz787: did you see the YC email to startups that's been circulating? 09:15 < nmz787> https://www.oregonlive.com/silicon-forest/2022/05/intel-facing-labor-shortage-revokes-no-rehire-rule-from-prior-layoffs.html 09:16 < nmz787> muurkha: nah I don't read HN but once or twice a year 09:16 < muurkha> surely better for your mental health 09:16 < nmz787> people have been talking about a recession, but I guess I'm in some weird niche where things seem relatively OK 09:17 < muurkha> .t https://nitter.net/refsrc/status/1527238287471292417 09:17 < saxo> URLError: (title:73) 09:17 < muurkha> .t https://nitter.net/refsrc/status/1527238287471292417 09:17 < saxo> URLError: (title:73) 09:17 < muurkha> "YC's message to founders:" 09:19 < nmz787> that message doesn't scare me or sound any alarms :/ 09:19 < nmz787> I've been working under those terms for my entreprenurial thrusts the whole time 09:22 < muurkha> right, but they mean that VC-funded startups are going to be hiring a lot less 09:22 < nmz787> idk, seems like small potatoes 09:23 < nmz787> maybe I just have my head in the sand/clouds/forest 09:23 < muurkha> I think VC-funded startups account for a significant fraction of the tech jobs in Silicon Valley and an even larger fraction of the job *openings* 09:24 < nmz787> but are they *interesting* tech jobs, or just stupid cat pic websites and questionable applications of AI? 09:24 < muurkha> but less so in hardware, less so outside Silicon Valley 09:25 < nmz787> personally I can't see myself ever working for the majority of tech companies doing anything with software these days 09:25 < muurkha> a lot of them are pretty trivial 09:25 < nmz787> too boring in terms of product, innovation at fundamental levels, etc 09:26 < muurkha> here are the companies that email is addressing: https://techcrunch.com/2021/08/31/here-are-all-the-companies-from-y-combinators-summer-2021-demo-day-part-1/ 09:28 < nmz787> if these companies die, they clearly had a bad idea? 09:29 < muurkha> not necessarily, their ideas are only a small factor in their success 09:29 -!- spaceangel [~spaceange@ip-78-102-216-202.net.upcbroadband.cz] has joined #hplusroadmap 09:32 < nmz787> well regardless of how cynical I am in the morning, I think interesting people will have interesting opportunities 09:33 < nmz787> "AI powered video game coach" geez... what happened to just playing a game all day until the sun goes down and the only light on in the house is from your TV/monitor? 09:35 < nmz787> way too many of these companies to finish reading this 09:36 < muurkha> oil/gas well control software, seaweed-based atmospheric carbon capture, Filipino ghost kitchen food delivery, banking with higher-yield checking, Southeast Asian restaurant management software to claw back customers from food delivery apps, a debit card with "cash back" in the form of stocks instead of cash, high school math coding games, a password manager for startups, no-code payment scripting 09:36 < muurkha> for Southeast Asian SMBs, early-stage ovarian-cancer blood tests, another online store builder, NLP search engine, digital bookkeeping for Vietnam, space trash removal, automated hospital-patient communications over WhatsApp, opening private equity to unqualified investors, something or other vague about online stores, IoT patient monitoring for nurses, in-store inventory with barcodes, a 09:36 < muurkha> marketplace for coding courses, something about mental health and wellness, medical device purchase optimization for surgery centers, by-the-minute coworking spaces, plus another couple hundred 09:37 < muurkha> and another similar number the next day 09:37 < muurkha> out of those I'd say that about a third sound like they're technically interesting 09:37 < nmz787> yeah, most seem like duplicates of existing things, or are silly gimmicks 09:37 < nmz787> I was thinking more like 10% 09:37 < nmz787> but that's close enough to 30% 09:37 < muurkha> well, my list there probably isn't a representative sample 09:37 < muurkha> it's the first companies on the list 09:38 < nmz787> I've been reading it for a while 09:38 < muurkha> I listed 22 09:39 < nmz787> in other more exciting news, "git add -p " is just what I've wanted many times for several years and I just learned that it's so easy and intuitive that I am stupid for not learning it a decade ago 09:39 < muurkha> out of those I think probably 6 or 7 probably have interesting technical challenges 09:39 < nmz787> https://git-scm.com/book/en/v2/Git-Tools-Interactive-Staging#_staging_patches 09:39 < muurkha> yeah, git add -p is good 09:39 < muurkha> the version in magit is better if you like emacs or even can possibly tolerate emacs 09:40 < muurkha> magit is worth running emacs for even if you don't use emacs for anything else 09:40 < nmz787> I've avoided learning emacs for 25 years or so... I don't know if I want to try now 09:42 < muurkha> you don't have to learn emacs to run magit in it 09:42 < nmz787> .wik magit 09:42 < saxo> "Magit is an interface to the Git version control system / (a Git Client), / implemented as a GNU Emacs package / written in Elisp. / It is made available through the MELPA package repository, / on which it is the most-downloaded non-library package, with almost three million [...]" - https://en.wikipedia.org/wiki/Magit 09:42 < nmz787> eh, why do I need a new interface to git? 09:43 < nmz787> the interactive staging works just as I'd expect 09:43 < muurkha> well, among other things because it has a much better way to do interactive staging 09:43 < muurkha> it allows you to expand, collapse, stage, and unstage chunks 09:44 < muurkha> admittedly part of the reason it's comfortable or me is that I can use Emacs as an editor so the ability to click through a chunk and get to the relevant position in the file, then edit it, is a useful feature 09:45 < muurkha> and also I can easily use text search to search through the staging buffer with ^S and ^R 09:45 < muurkha> what editor do you use for programming? 09:47 < nmz787> usually pycharm and/or sublime text 09:48 < muurkha> the magit-status buffer is like git status except that you can expand or collapse file diffs or diff chunks with tab 09:48 < muurkha> hmm, I don't know what the sublime and pycharm keybindings are like 09:48 < nmz787> I'm ok with VI but that's mostly just for quick one line edits 09:49 < muurkha> when you're looking at a diff in magit-status you can stage it with s or unstage it with u 09:49 < muurkha> or revert it with k 09:50 < muurkha> whereas with git's normal interface, when you're looking at a diff, you have to exit out of there and fire up git add -p or whatever 09:50 < muurkha> and I don't even know how to selectively unstage diff chunks, or revert them to the committed version, with the git CLI 09:52 < muurkha> I was thinking that maybe someone might have implemented PyCharm or Sublime keybindings in Emacs, since there are multiple implementations of vi keybindings, some of which are highly regarded 09:54 < muurkha> but I'm not seeing anything 09:59 < muurkha> anyway magit is, like, the UI for Git that you would design if you thought about what would be the ideal set of available user interactions 10:08 < muurkha> the git log is a hypertext list of commit "pages" with a display of the graph. each commit displays the diffs, and you can expand and collapse the diffs to an arbitrary level of detail. they have hyperlinks to their parents (though not their children, for that you have to use the log buffer) and to the branches they're on, and to the files being changed, so you can click through a diff to open the 10:09 < muurkha> old version of the file. You can easily text-search through an individual commit. the status buffer has the similar features I mentioned above, and also you can push, pull, commit, or switch to the log with two keystrokes 10:09 -!- superkuh [~superkuh@user/superkuh] has joined #hplusroadmap 10:12 < muurkha> and there are features I never use for things like reflogs, cherry-picking, stashing, notes, rebasing, and so on. for those I use the git CLI because I don't use them often enough to make it seem appealing to justify spending time on learning a more efficient way to do them 10:13 < muurkha> but the everyday stuff of looking at diffs, interactively staging, committing, pushing, and pulling, is just head and shoulders better than gitk or the CLI 10:13 < muurkha> I haven't tried GitHub's proprietary GUI so I can't compare it 10:29 < kanzure> .tw https://twitter.com/lkcl/status/1529499538486329344 10:29 < saxo> @ChristopherA @Crossbar @tropicsquare @supranational @proxy @risc_v chris we're designing Draft bitmanip ops for Power ISA including Galois Field (3 types, GF2, GF(Prime), GF(2^n), and some bigint as well, i'd be interested in your thoughts https://libre-soc.org/openpower/sv/biginteger/analysis/ https://libre-soc.org/openpower/sv/bitmanip/ (@lkcl, in reply to tw:1529482350651793413) 10:42 < lkcl> the way the vectorised bigint ops work is, we assume "scalar operation equals element operation", so SVP64 provides the concept of hardware-looping "for free" 10:42 < lkcl> but because the Power ISA has add-with-carry, if you chain a lot of adds-with-carrys together, that's, duh, a Vector BigInt Add! 10:43 < kanzure> listening to sam talk about his disintermediation design for derivatives markets https://www.wirestream.tv/customer/cftc/2022/05-25/ 10:45 < muurkha> lkcl: it looks "for free" from the software side 10:45 < lkcl> then it becomes a matter for the hardware to go, "oh you wanted this batch of adds-with-carrys? i can hand that to a massive-wide ALU for you" 10:45 < lkcl> muurkha, basically, yes. 10:46 < lkcl> no pissing about "oh god how do i work out how to jam this into 4-wide or 8-wide or whatever-wide SIMD" 10:46 < muurkha> but as you know, from the hardware side, it's not "for free"; it means the hardware needs to implement a massive lookahead carry tree, or run at a lower clock frequency, or have a latency of several cycles for a vector operation in at least the add-with-carry cae 10:46 < muurkha> *case 10:47 < lkcl> that's the hardware-designer's problem :) 10:47 < lkcl> and it's not that hard. 10:47 < muurkha> well, multiple cycles of extra latency isn't the hardware designer's problem 10:47 < lkcl> if you look up how people do carry-propagation on 64-bit numbers to create bigint in software, you can duplicate that in hardware 10:47 < muurkha> it's the user's problem 10:48 < muurkha> and if you want to limit the extra latency to that case, you have to distinguish between vector add-with-carry and vector add 10:48 < muurkha> and all this adds verification effort 10:48 < lkcl> let's say worst-case that you have to have a 3-stage pipeline 10:48 < muurkha> why would that be the worst case? 10:48 < lkcl> * massive suite of multiple 64-bit SIMD adds 10:49 < lkcl> * analysis of each individual 64-bit add to calculate the carry-propagation 10:49 < lkcl> * add one to each 64-bit SIMD add where it was needed (as determined by carry-prop) 10:49 < muurkha> I know propagation delay in CMOS is a little tricky because it depends on fanout and the amount of capacitive load, but how many propagation delays are your clock cycles today? 10:50 < lkcl> 5 ghz has typically been around 17 to 19 gates 10:50 < lkcl> but let's say you have to have a 3-stage pipeline, there 10:50 < muurkha> a full adder is already 5 NAND/NORs I think? 10:51 < muurkha> and then you need an additional gate for every factor of 4 that your vector grows 10:51 < muurkha> okay, so 3 stages is probably pretty reasonable 10:51 < lkcl> well, that's not a "problem" in that OoO CPUs have been dealing with pipeline latency for several decades, now 10:51 < lkcl> it's pretty normal to have a 64-bit add in a single clock cycle 10:51 < muurkha> yeah 10:52 < muurkha> or, more amazingly, a multiply-accumulate 10:52 < lkcl> there's a fascinating piece of research making it public on how to do add/sub/cmp optimisation 10:52 < lkcl> yeah the two best-known algorithms there are Dadda and Wallace 10:52 < muurkha> thanks! 10:53 < lkcl> Dadda i think is designed so that each carry-propagation from one of the half/full adder phases goes into the *back* of the queue of the ones to be carried out later 10:53 < lkcl> *specifically* so that any "propagation delay" (which you validly mention) is mitigated 10:54 < muurkha> https://en.wikipedia.org/wiki/Dadda_multiplier 10:54 < muurkha> .t 10:54 < saxo> "The Dadda multiplier is a hardware binary multiplier design invented by computer scientist Luigi Dadda in 1965. It uses a selection of full and half adders to sum the partial products in stages (the Dadda tree or Dadda reduction) until two numbers are left." - https://en.wikipedia.org/wiki/Dadda_multiplier 10:54 < lkcl> so you basically subdivide a 64x64 multiply down into a batch of 64 8x8 multiplies 10:54 < lkcl> and then cascade-add all of them together exactly as you would in Long Multiply / Vedic Multiply 12:02 -!- CryptoDavid [uid14990@id-14990.uxbridge.irccloud.com] has joined #hplusroadmap 13:11 -!- faceface [~faceface@user/faceface] has joined #hplusroadmap 13:43 < kanzure> https://csrc.nist.gov/csrc/media/events/workshop-on-elliptic-curve-cryptography-standards/documents/presentations/session6-bos-joppe.pdf 15:17 -!- spaceangel [~spaceange@ip-78-102-216-202.net.upcbroadband.cz] has quit [Remote host closed the connection] 15:31 < ^ditto> [freenode] guys . its wrong room for hplusroadmap ? cuz its room was crowded before 15:33 < ^ditto> [freenode] heyyy 15:52 -!- helleshin [~talinck@108-225-123-172.lightspeed.cntmoh.sbcglobal.net] has joined #hplusroadmap 15:52 -!- helleshin [~talinck@108-225-123-172.lightspeed.cntmoh.sbcglobal.net] has quit [Client Quit] 15:52 -!- helleshin [~talinck@108-225-123-172.lightspeed.cntmoh.sbcglobal.net] has joined #hplusroadmap 18:20 -!- darsie [~darsie@84-113-55-200.cable.dynamic.surfer.at] has quit [Ping timeout: 255 seconds] 18:59 -!- yashgaroth [~ffffffff@2601:5c4:c780:6aa0::93] has quit [Quit: Leaving] 19:00 -!- CryptoDavid [uid14990@id-14990.uxbridge.irccloud.com] has quit [Quit: Connection closed for inactivity] 19:10 < fenn> http://www.astranis.com SDR powered internet microsats with $250 million in funding 19:22 < superkuh> I guess they'll have done the link budgets but there are physical limits to gain proportional to the size of your antenna(s) aperture. 19:23 < superkuh> GEO and micro aren't good partners in that respect. 19:24 < superkuh> I guess the cell site antennas will have to be proportionally larger. 20:13 < fenn> QOA: Cocoa-free chocolate developed through precision fermentation https://qoacompany.com/ 20:20 < fenn> "Safer Management helps public schools track attendance, an important metric for funding — and, of course, education. It’s a modern system of QR codes and facial recognition that reliably tracks who’s in class." 20:20 < fenn> well that sounds positively dystopian 20:21 < fenn> prometheus: if you're still there, we're on libera now 20:25 < fenn> sigh.. "the NoCode movement (or No-Code movement) refers to a growing collection of tools and techniques which allow non-technical people to build apps, websites, games and other types of software without having to learn to code or worry about servers, hosting and other software problems." 20:25 < fenn> "software creation should be accessible to anyone, not just those who have already learnt to code or studied computer science." 20:27 < fenn> call me when your shit breaks, my rates are reasonable 20:28 < fenn> it's like the web dev kids just discovered dynamically linked libraries 20:35 < maaku> lsneff: I graduated in 2008. welcome to the club! 20:36 < maaku> muurkha: this downturn is affecting way more than just tech 21:17 -!- mirage335 [~mirage335@64.79.52.86] has joined #hplusroadmap 22:05 < muurkha> maaku: yeah, I'm sure 22:06 < muurkha> fenn: or Excel --- Log closed Thu May 26 00:00:34 2022