2022-04-17 02:17:57 <KipIngram   > Well, I've chased this damn device driver all night - no luck.
2022-04-17 02:18:18 <KipIngram   > Along the way, though, I got the entire Chapel flow for using the GPU working as well.
2022-04-17 02:18:47 <KipIngram   > Built an application and everything, but when I try to run it I hit the same wall - "Platform not found."
2022-04-17 02:19:28 <KipIngram   > This is one of those things that will be just as simple as pie when I know what to do - one little command to stick one little rinky dink piece of software in the right place.
2022-04-17 09:52:18 <lagash      > KipIngram: you mean https://linux-hardware.org ?
2022-04-17 10:55:51 <KipIngram   > I do.  Sorry.
2022-04-17 10:58:23 <KipIngram   > No reply at all yet though.
2022-04-17 11:28:14 <TCZ         > Wesołych Świąt
2022-04-17 11:53:05 <KipIngram   > Happy Easter, TCZ!
2022-04-17 12:19:23 <KipIngram   > So, there is an open-source project called ZLuda, which is supposed to be a drop-in replacement for CUDA, and purportedly supports Intel graphics.  Apparently Nvidia and Intel got into some sort of "corporation tiff" and Nvidia refused to provide Intel support in CUDA.  But ZLuda is announced to be "buggy all over the place," and some time in the recent past the maintainer "resigned" "for privacy reasons,"
2022-04-17 12:19:25 <KipIngram   > (whatever that means).  So the project is currently headless.
2022-04-17 12:19:58 <KipIngram   > So OpenCL looks like my only viable path forward, unless I decided to try to tackle the ZLuda code myself.  That's probably a bigger job than I want to undertake.
2022-04-17 12:21:05 <KipIngram   > At least with the SSE/AVX the instsructions are right there ready for me to execute - no "OS wrangling required."
2022-04-17 13:07:41 <KipIngram   > So that ForwardCom instruction set looks pretty well thought through.
2022-04-17 13:08:01 <KipIngram   > Not "Forthish," in that it has lots of operand specification capacity, but still well-organized.
2022-04-17 13:10:22 <eris[m]     > what's forwardcom?
2022-04-17 15:56:22 <KipIngram   > https://www.forwardcom.info/
2022-04-17 15:56:29 <KipIngram   > I linked it a day or two ago.
2022-04-17 15:56:39 <KipIngram   > "Open source instruction set."
2022-04-17 15:56:47 <KipIngram   > Neither RISC nor CISC - some of both.
2022-04-17 16:04:41 <Zarutian_HTC> I have a bit of a confession to make, I am nearly done porting the inner interpreter of zeForth (an eForth variant) to IBM/370
2022-04-17 16:05:05 <travisb_    > so you can run it under Hercules?
2022-04-17 16:05:19 <Zarutian_HTC> probably
2022-04-17 16:06:25 <Zarutian_HTC> but I am aiming to implement an kernel compatible to KeyKOS
2022-04-17 16:07:45 <Zarutian_HTC> to do that I need to implement I/O for BSC and DASDs (IBM terms)
2022-04-17 16:09:10 <Zarutian_HTC> I have been leafing through IBM/370 PoOP (Principles of OPeration)
2022-04-17 16:09:49 <Zarutian_HTC> some of the stuff is just wtf
2022-04-17 16:10:59 <Zarutian_HTC> the MOVE instructions in particular
2022-04-17 16:11:40 <Zarutian_HTC> they are nothing like MOV on other architectures
2022-04-17 16:12:24 <Zarutian_HTC> basically they are variants of Forth CMOVE and co if you squint
2022-04-17 16:13:25 <Zarutian_HTC> but IBM did invent something like interrupt vectors though
2022-04-17 16:15:34 <Zarutian_HTC> I am still wraping me head around how IO is done with channels
2022-04-17 16:16:26 <Zarutian_HTC> I am sloting it into DMA but with command lists
2022-04-17 16:16:37 <Zarutian_HTC> concept wise
2022-04-17 16:23:05 <KipIngram   > Wow - there are still System 370's around?
2022-04-17 16:23:25 <KipIngram   > Isn't that like... 50-ish years old?
2022-04-17 16:26:12 <Zarutian_HTC> 55 years
2022-04-17 16:27:25 <Zarutian_HTC> but here is a thing, lots of programs and software systems run on top of IBM/370, even though the hw is now mostly emulated
2022-04-17 16:28:56 <KipIngram   > Yeah, I was just commenting yesterday on how legacy software has "steered" hardware development.
2022-04-17 16:29:06 <Zarutian_HTC> long life of software that
2022-04-17 16:29:37 <Zarutian_HTC> indeed hence my interest in BitGrid
2022-04-17 16:29:43 <KipIngram   > I can't help wondering what we might have now that we don't, for having done things that way.
2022-04-17 16:29:53 <KipIngram   > Yes.
2022-04-17 16:30:27 <Zarutian_HTC> note that comp sci is STILL rather young field
2022-04-17 16:30:28 <KipIngram   > What if we'd committed to high core count as soon as our semiconductor capability was good enough to start creating them?
2022-04-17 16:30:51 <KipIngram   > And just re-written the software as necessitated by that?
2022-04-17 16:31:18 <Zarutian_HTC> or kept FlowBasedProgramming informed by the unit machines era
2022-04-17 16:31:22 <KipIngram   > I was noting yesterday that we did the same thing again when the interest in virtualization developed.
2022-04-17 16:31:35 <KipIngram   > Added another layer of hardware, to give hypervisors a place to live.
2022-04-17 16:31:53 <KipIngram   > Since the OS's had gobbled up ring 0.
2022-04-17 16:32:17 <Zarutian_HTC> IBM/360, 370, 390 all that were designed to be easily virtualizable
2022-04-17 16:33:28 <Zarutian_HTC> never jived with intels ring terminology, supervisor/problem state is easier to think about
2022-04-17 16:34:35 <KipIngram   > Notice that we mostly just used two of the rings.
2022-04-17 16:37:45 <Zarutian_HTC> ring 0 and 4 iirc
2022-04-17 16:39:30 <Zarutian_HTC> I think doing FBP ontop of BitGrid (or fpga/cpld inspired device with BitGrid blocks, buses and routing boxes) would be quite interesting
2022-04-17 16:39:34 <KipIngram   > Yes - the extremes.  Maybe that made sense, given that at the time no one knew what the next need would be.
2022-04-17 16:39:57 <KipIngram   > I can see why "the OS" needed to be at the highest level - it needed to be able to do anything.
2022-04-17 16:40:17 <KipIngram   > I assume someone just thought it was more likely that we'd think of something to put between OS and user than "below user."
2022-04-17 16:41:03 <KipIngram   > But then it turned out what was needed was to downgrade the OS, while at the same time letting them "think" they were still in charge.
2022-04-17 16:41:25 <Zarutian_HTC> or just the IBM PC inventor thought that they were good enough for supervisor and problem mode substitutions
2022-04-17 16:41:39 <KipIngram   > So, Management Engine.
2022-04-17 16:42:18 <Zarutian_HTC> this is what 3*0 were built for, you can have as many layers of VMs as you like
2022-04-17 16:42:34 <KipIngram   > So much of what CPU designers have done has been devoted to "propping up" a use model that developed decades ago.
2022-04-17 16:43:29 <KipIngram   > It's kind of like all the work the Army Corp of Engineers has done to keep the Mississippi River running where it's "supposed to."
2022-04-17 16:43:42 <Zarutian_HTC> there is a reason why I refer to x86 code (even 64 bit) as legacy bytecode
2022-04-17 16:43:47 <KipIngram   > "Supposed to," because we've gotten used to it.
2022-04-17 16:43:56 <KipIngram   > Oh, sure.
2022-04-17 16:44:04 <KipIngram   > We don't really program the hardware, even in assembly.
2022-04-17 16:44:14 <KipIngram   > We program a vm created by the microcode.
2022-04-17 16:44:49 <Zarutian_HTC> often leaky ones that per spectre and intel meltdown
2022-04-17 16:44:57 <KipIngram   > Exactly.
2022-04-17 16:45:00 <Zarutian_HTC> at that*
2022-04-17 16:45:17 <KipIngram   > Cache guys and speculative execution guys weren't sufficiently familiar with what one another were doing.
2022-04-17 16:45:35 <KipIngram   > Jesus - that sentence has got to have a grammar error in it somewhere.
2022-04-17 16:46:31 <KipIngram   > Speculative execution guys failed to roll back all the side effects of abandoned instructions.
2022-04-17 16:46:50 <KipIngram   > And someone out there was smart.
2022-04-17 16:46:59 <Zarutian_HTC> do you know what I thought when I first read about Intel SGX? damn what a shitty way to half implement GlobalPlatform secure processor (used in smartcards like sim, conax, and others)
2022-04-17 16:47:14 <KipIngram   > I remember when I first caught on to what was going on there (Spectre/Meltdown) - I was damn impressed.
2022-04-17 16:48:13 <KipIngram   > Well, among numerous other things, Intel is a bureaucracy, and bureaucracies aren't legendary for their excellence.
2022-04-17 16:48:59 <KipIngram   > Half the time (or more) there are about 50 different agendas in play across a big corporation at any given time.
2022-04-17 16:49:25 <KipIngram   > Divisions fighting with one another for tech turf, etc.
2022-04-17 16:49:26 <Zarutian_HTC> when speculative out of order came on the scene I recall asking a cs guy about it and if it could possibly be exploited, thence when spectre and meltdown came, I just thought finally someone exploited that
2022-04-17 16:49:49 <KipIngram   > You were paying better attention than I was.
2022-04-17 16:50:07 <KipIngram   > I left the digital field for a chunk of time in there, and worked in pulse power research.
2022-04-17 16:50:20 <KipIngram   > And when I came back in the mid 1990's, it tooks some catching up for sure.
2022-04-17 16:50:45 <KipIngram   > That missed period was like from 1987-1996, and a hell of a lot happened during that time.
2022-04-17 16:51:01 <KipIngram   > I was off playing Buck Rogers.
2022-04-17 16:51:06 <KipIngram   > Railguns and so forth.
2022-04-17 16:51:13 <Zarutian_HTC> naah, I just like to point out flaws possible hacks just from descriptions of hw/sw things
2022-04-17 16:52:10 <Zarutian_HTC> I was just born two years prior to the start of that year range of yours
2022-04-17 16:52:37 <KipIngram   > :-)  Junior.
2022-04-17 16:52:53 <KipIngram   > I have no complaints, though - I've had a hell of a fun career.
2022-04-17 16:53:01 <Zarutian_HTC> I have attacked rough draft designs this way for years
2022-04-17 16:53:02 <KipIngram   > Got to touch on quite a few really interesting technologies.
2022-04-17 16:53:40 <KipIngram   > Also that flip flop during that period was critical to meeting my wife, and it's hard for me to imagine not having had her around.
2022-04-17 16:54:23 <KipIngram   > I had a pathetically bad first marrage, and did NOT intend to get married again, but changed my mind when I met her - that was a good move.
2022-04-17 16:54:30 <Zarutian_HTC> that is quite nice indeed!
2022-04-17 16:54:36 <KipIngram   > s/marrage/marriage/
2022-04-17 16:55:07 <KipIngram   > We just celebrated our 25th a coupld of months ago - it's been a good quarter century.
2022-04-17 16:55:18 <KipIngram   > Gad, I can't type today.
2022-04-17 16:55:35 <Zarutian_HTC> my attacking this way has gotten few folks to really think through their designs and implications
2022-04-17 16:56:02 <KipIngram   > so, I don't know if you were on to catch my current struggle - I'm trying to lay hands on the right driver to connect up OpenCL with my Intel UHD graphics hardware.
2022-04-17 16:56:23 <KipIngram   > Got a couple of GPU computing toolchains all ready to go, except for that one link in the chain.
2022-04-17 16:57:57 <Zarutian_HTC> sorry, I mostly dealt with mechatronics. Aside frome the low level algo dabling in gfx I have no idea where to finds these kind of drivers
2022-04-17 16:58:00 <KipIngram   > On of those chains involves Python; if I can get that one running I think I'll be able to do GPU computing from Jupyter Notebook.
2022-04-17 16:58:50 <KipIngram   > Yeah, they're hard to find, because they have to be EXACTLY RIGHT for your particular setup.
2022-04-17 16:59:00 <KipIngram   > I've found all the generic pieces now, I think.
2022-04-17 17:02:06 <Zarutian_HTC> Jypter Notebooks? I have not looked at those yet
2022-04-17 17:06:57 <KipIngram   > Jupyter Notebook is quite cool.  It runs an engine that provides a localhost webserver - you then browse to localhost:8888 with your browser, and you can interact with it there.  It's sort of a fusion of Python, loose format spreadsheet cells, markup, and so on - you can use it very much like a "scientific lab notebook" that intelligently runs all your code and presents your results in whatever way you decide
2022-04-17 17:07:00 <KipIngram   > you like.
2022-04-17 17:07:33 <KipIngram   > https://jupyter.org/
2022-04-17 17:08:00 <KipIngram   > It's currently "in play" in a pretty major way in university research labs and so on.
2022-04-17 17:08:14 <KipIngram   > You can share these notebooks, people can collaborate around them, etc.
2022-04-17 17:16:45 <KipIngram   > So, just having that tool capable of doing some gpu "heavy lifting" - that's a goal.  But I also just want to learn more about it to see if there's any interesting ways to wire Forth to it.
2022-04-17 17:17:31 <KipIngram   > It is, after all, far and away my computer's strongest compute capability.
2022-04-17 17:17:43 <KipIngram   > Probably by an order of magnitude at least, maybe more.
2022-04-17 17:18:03 <KipIngram   > Though it is just a 13" notebook, so it's not going to have the beefiest GPU.
2022-04-17 17:47:37 <Zarutian_HTC> so like a calve level or some such?
2022-04-17 17:51:13 <KipIngram   > I don't grok "calve level."
2022-04-17 18:07:32 <Zarutian_HTC> think about the word "beefy" for a bit
2022-04-17 18:18:49 <KipIngram   > Oh.  Haha - got it.
2022-04-17 18:19:06 <KipIngram   > Sorry about that - I'm in the middle of trying to do my taxes - distracted a little I guess.
2022-04-17 18:19:36 <KipIngram   > Somehow ##forth is a lot more entertaining than the IRS...
2022-04-17 18:23:11 <Zarutian_HTC> oh, right. Where I live doing taxes does not take long. Basically there is a web portal where you can do your returns. Most of the stuff is prefilled due to information filed at the tax authority by others about matter pertaining to you
2022-04-17 18:25:49 <Zarutian_HTC> and that stuff, like payroll can be easily exported out of the companys accounting system and imported into tax authorities system when the accountant files taxes for the company
2022-04-17 18:28:20 <Zarutian_HTC> why is it so sensible? because both everybody hated sending in the paper ones and the tax authority folks hated checking for errors
2022-04-17 19:20:03 <KipIngram   > I'm done now - I use TurboTax and it usually goes pretty well.  And the past few years haven't been painful - we seem to have our throughout the year witholding set sensibly.
2022-04-17 19:46:42 <travisb_    > the gov't should simply make their own tax-submission web application that will put quicken out of business
2022-04-17 19:47:42 <travisb_    > s/quicken/intuit
2022-04-17 19:48:05 <travisb_    > a friend of mine works with some people in the US gov't who're suing intuit because of misleading advertising
2022-04-17 19:49:34 <travisb_    > intuit markets "free" turbotax, with just a momentary, hardly readable disclaimer in their ads, which lets the user enter their personal information and like to only then tell the user that they have to pay for the paid version
2022-04-17 20:02:36 <KipIngram   > Yeah.  The other thing they do is offer to subtract their fee from your refund, if you are getting a refund, only they don't make it clear that that carries an extra $40 processing fee.
2022-04-17 20:02:49 <KipIngram   > I'd rather pull out my credit card and save the $40.
2022-04-17 20:03:16 <KipIngram   > They are really sneaky about a number of things, and very aggressive about marketing additional services to you.
2022-04-17 20:03:47 <KipIngram   > I've kind of gotten used to being on the lookout for it over the years.