2022-07-22 10:52:17 <KipIngram   > Well, still trouble with Python opencl.  Every example I found and tried failed in the same kind of way - it may be that things have changed a little and my exanples were old.  It does seem to be basically working.  I just need to learn a little more about what I'm doing rather than just run code blind.
2022-07-22 10:52:24 <KipIngram   > PyOpenGL seemed to work fine, though.
2022-07-22 10:54:52 <KipIngram   > The step that fails on opencl is almost the last one - where it tries to hand off the little GPU kernel program it has created to the GPU.
2022-07-22 10:55:26 <KipIngram   > And it's a silly language error - it just reports that some required arguments are missing.  So maybe if I actually learn what I'm doing I'll see how to correct it.
2022-07-22 10:55:50 <KipIngram   > I was just hoping to get to see something run all the way through, but no joy.
2022-07-22 11:02:43 <KipIngram   > Eventually I'd like to figure out a nice Forth interface to GPU computing, but I really have no idea what that would look like yet.  Any of you guys ever see that done?
2022-07-22 12:18:22 <imode       > KipIngram: yes!
2022-07-22 12:18:34 <imode       > https://forthsalon.appspot.com/
2022-07-22 12:54:10 <KipIngram   > Oh, thta's kind of interesting.
2022-07-22 12:56:51 <KipIngram   > So those things are run using the GPU?  Is there any info on the internals?
2022-07-22 12:57:46 <KipIngram   > Well, I just drove in and picked up my new work computer.  I got an invite a week and a half ago or so to "refresh" - looks like they're starting that program back up.
2022-07-22 12:58:10 <KipIngram   > I chose a 14" Macbook Pro.  It'll be one with the new chip in it, which will my first exposure to that.
2022-07-22 12:58:23 <KipIngram   > That's an ARM architecture, right?
2022-07-22 12:59:15 <KipIngram   > When I first got the one I'm using  now, the "transition" (which I'd dreaded ahead of time) turned out to be remarkably easy - someone had thought the process through really well.
2022-07-22 12:59:21 <KipIngram   > I hope it's still that easy.
2022-07-22 12:59:39 <KipIngram   > Before it can get on the IBM network it has to have all the right Naziware installed and running.
2022-07-22 13:00:04 <KipIngram   > But it was almost a one-double-click process last time.
2022-07-22 13:02:26 <KipIngram   > They told me I could buy the old one if I wanted to, for about $270.  I haven't researched whether that's a good price or not.  It's keyboard has started getting a little flaky, but it might just need cleaning.
2022-07-22 13:02:41 <KipIngram   > Old one's a 13" Air.
2022-07-22 13:03:39 <MrMobius    > I would have bought the one from my old work if they offered. it was a small company so no restrictions on what you could put on it. i had a lot of hobby things on there
2022-07-22 13:04:17 <MrMobius    > I bought essentially the same exact thing when I left but the much less expensive model
2022-07-22 13:08:57 <imode       > I don't know about the internals, but view-source should tell you something.
2022-07-22 13:09:58 <imode       > here's the JS. view-source:https://forthsalon.appspot.com/static/haiku.js
2022-07-22 13:11:37 <imode       > looks like it generates a shader from your code.
2022-07-22 13:11:41 <imode       > cool.
2022-07-22 13:11:58 <KipIngram   > I used to do hobby things on the Air too, but then they decided to cripple the USB ports to keep people from "stealing IP."
2022-07-22 13:12:13 <KipIngram   > Meanwhile, they left the ability to burn DVDs via USB in place.  :-|
2022-07-22 13:12:32 <KipIngram   > But, I suppose someone got to report to his executive that "action had been taken."
2022-07-22 13:12:50 <KipIngram   > I think that's behind a whole lot of things those guys do.
2022-07-22 13:17:10 <KipIngram   > Oh, I didn't know you could just stick "view-source" on the address bar like that.  That's pretty cool.
2022-07-22 13:18:03 <imode       > yep, it's a special resource type that your browser understands.
2022-07-22 13:20:01 <thrig       > w3m: Can't load view-source:https://...
2022-07-22 13:20:17 <KipIngram   > I'd always done it via the menus.
2022-07-22 13:53:07 <KipIngram   > I asked about this pyopencl error on the #opencl channel last night.  So far, though... silence.
2022-07-22 13:53:13 <KipIngram   > No activity at all.
2022-07-22 13:53:51 <KipIngram   > I really do think it could just be a compatibility problem.  I've got opencl 3.0, and it's only recently come out.
2022-07-22 13:54:15 <KipIngram   > So if I write something using the 3.0 documentation, it may run fine.
2022-07-22 13:56:14 <KipIngram   > Oh, on the boot sector Forth I may switch the TOS element from cx to bx.
2022-07-22 13:56:49 <KipIngram   > In the old 16-bit code you can do things with bx that you can't do with cx, specifically you can use it as an address.  That simplifies @ and !.
2022-07-22 13:57:26 <KipIngram   > I ought to go through and use aliases for everything like I did i my 64-bit system.
2022-07-22 14:50:01 <KipIngram   > Actually, TOS == cx may be best.  bx is the only unused register that can hold an address - there are a number of places I'd have to save it in order to use it if it's TOS.
2022-07-22 14:50:19 <KipIngram   > I think it would gobble up most of the aforementioned benefits.
2022-07-22 14:51:02 <KipIngram   > And also for >> and << it's nice to already have the shift count in cl, where I need it.
2022-07-22 22:15:17 <remexre     > does anyone here have a design they've liked to get CASE-like control flow with quotations?
2022-07-22 22:56:22 <remexre     > (the obvious one with 2n+1 quotations just feels gross to implement, possibly in part because I don't have PICK)
2022-07-22 22:57:21 <imode       > 2n+1 quotations?
2022-07-22 22:57:52 <remexre     > { cond1 } { body1 } { cond2 } { body2 } { if-none-match } 2 CASE-FOR-QUOTATIONS
2022-07-22 22:58:04 <imode       > ah.
2022-07-22 22:58:12 <remexre     > where CASE-FOR-QUOTATIONS does a loop on n
2022-07-22 22:58:21 <imode       > makes sense.
2022-07-22 22:58:24 <remexre     > and stashes the quotations in the r-stack or wherever