https://tinyvision.ai/
09:40:31	 From Chris (@UncensoerdNews) : $35 + $5 for pods + $2.80 for shipping from china.
09:40:45	 From Chris (@UncensoerdNews) : Can you provide a link to the code, easier to see?
09:48:55	 From Don Golding : Code will be here, this is new code, need to update GIT: https://github.com/angelus9/AI-Robotics/tree/main
09:49:24	 From Don Golding : use this branch: https://github.com/angelus9/AI-Robotics/tree/CORE-I-Forth-Computer-for-TinyVision-V3-Board
09:51:49	 From Chris (@UncensoerdNews) : Here is an invite to the Tiny Vision Discord Server Forth channel. 
https://discord.gg/ZyhFF8AN
09:53:40	 From Don Golding : thanks Chris
09:54:01	 From Chris (@UncensoerdNews) : Thank you for porting to the pico-ice.
09:54:47	 From Don Golding : it is their tinyvision.ai board, should work on that one too.
09:55:18	 From Don Golding : this one: https://tinyvision.ai/products/upduino-v3-1
09:55:45	 From Richard Hair : Nice!
09:55:48	 From Chris (@UncensoerdNews) : I prefer the pico-ice to the upduino.  Has a better clock, and an RP2040.
09:56:20	 From Don Golding : it is all source, works on any FPGA.
09:57:12	 From Chris (@UncensoerdNews) : Ah yes, but the rest of the eco system matters.  Pico-Ice has a much richer sdk.
10:09:02	 From Travis Bemann : does the Pico-Ice have a Free toolchain that runs under Linux?
10:09:58	 From Chris (@UncensoerdNews) : Yes.  Yosys is the open source Verilog synthesis tool chain for linux, pico-ice and lots of other boards.
10:11:43	 From Travis Bemann : I think I saw 'Yosys' mentioned when playing around with the iCEBreaker and Mecrisp-Ice
10:13:40	 From Chris (@UncensoerdNews) : Correct.  The open source synthesis require an open source description of the fpga, which exists for the Lattice ICe40 chips used on the icebreaker, ULX3S and lots of other boards.
10:14:55	 From Chris (@UncensoerdNews) : When mentioning luts good to mention the board as well.
10:15:55	 From Chris (@UncensoerdNews) : Great question.
10:16:55	 From Chris (@UncensoerdNews) : May I answer that question please?
10:17:36	 From Chris (@UncensoerdNews) : Let me do so here. 
The MicroCore answers that question.  Every time he ran into a performance limitation, he added an opcode.
10:17:59	 From Chris (@UncensoerdNews) : Adding peripherals is a great use case.
10:18:09	 From Demitri Peynado : Reacted to "this one: https://..." with 👍
10:18:36	 From Chris (@UncensoerdNews) : There are many verilog libraries for I/O.
10:19:30	 From Chris (@UncensoerdNews) : And also FPGAs have very large numbers of I/O ports.
10:20:01	 From Chris (@UncensoerdNews) : @ZeptoForth have you run into a need for additional ports?
10:22:41	 From Travis Bemann : I haven't personally run into a need for additional ports myself, but on the RP2040 there is a kind of asymmetry where if you need more than, say, 2 SPI ports you have to use the PIO's, which is a pain if you want all your SPI ports to operate in the same fashion
10:23:14	 From Chris (@UncensoerdNews) : Reacted to "I haven't personally..." with 👍
10:23:32	 From Travis Bemann : and if you're using, say, an STM32F407 you're really out of luck because there's no PIO-like peripherals
10:24:27	 From Travis Bemann : (even though the STM32F407 in theory often has far more peripherals than the RP2040, one is limited by the exact package for the chip you're using)
10:25:00	 From Siva : What is the cell size?
10:26:14	 From Chris (@UncensoerdNews) : @Travis you asked about the tool chains. 

Here is the Unofficial Mecrisp Ice Documentation.  https://mecrisp-ice.readthedocs.io/en/latest/
10:26:31	 From Siva : Are arithmetic operations defined as core words?
10:26:34	 From Travis Bemann : @Chris (@UncensoerdNews) Yeah, I've seen that
10:26:48	 From Don Golding : yes
10:27:01	 From Don Golding : one line of code
10:28:10	 From Travis Bemann : umm shouldn't _minus be the other way around?
10:28:13	 From Chris (@UncensoerdNews) : @Travis, are there any functions too slow on the RP2040?  Something that could be done faster/better on the FPGA.  Multiplys come to mind.
10:28:16	 From Siva : Cell size is 32 bits?
10:29:32	 From Richard Hair : Nice presentation - processor designed around target hardware environment...
10:29:33	 From Travis Bemann : @Chris (@UncensoerdNews) creating arbitrary GPIO output sequences that are not limited to just what you can do with PWM or the PIO's
10:58:17	 From Chris (@UncensoerdNews) : https://en.wikipedia.org/wiki/Cheney%27s_algorithm
11:06:06	 From Chris (@UncensoerdNews) : He is very quiet.
11:06:42	 From Kevin Appert : IDK, seems normal to me
11:07:55	 From Brad Nelson : FWIW I think the biggest utility with non-moving collectors is interoperability with non-GCed languages (they hate it when you move stuff out from under them). But using Cheney's is great if you don't have that constraint (and 50% is at least a constant factor.
11:16:27	 From Brad Nelson : Whoops, bumped the record stop button on accident...
11:16:46	 From Brad Nelson : have resumed
11:17:03	 From Richard Hair : Auto-filtering content :)
11:26:16	 From Travis Bemann : @Brad Nelson even zeptoscript has the "it doesn't like memory being moved out from under it" problem because zeptoforth doesn't expect memory to move; I solved that problem with a word named ENSURE, which guarantees that a certain number of bytes will be available in the heap without GC occurring (by triggering an early GC if necessary, and raising an out-of-memory exception if not enough space is available after the GC), so in conditions where I need memory to not move, I call ENSURE so I can allocate a given amount of memory without any movement occurring
11:26:51	 From Chris (@UncensoerdNews) : Reacted to "@Brad Nelson even ze..." with 👍
11:36:38	 From Brad Nelson : Reacted to "@Brad Nelson even ze..." with 👍
11:36:40	 From Travis Bemann : you can also get support in #Mecrisp in irc.hackint.org
11:42:20	 From Chris (@UncensoerdNews) : Great Video , how did you create it.
11:43:40	 From Brad Nelson : Several pics out of Gemini + Adobe Premier
11:43:59	 From Travis Bemann : I figured you were harnessing the power of AI :D
11:44:18	 From Travis Bemann : it just seemed so AI-riffic
11:45:56	 From Dave Jaffe : Vintage Computer Festival Aug 2nd & 3rd at Compter History Museum in Mountain VIew, CA
11:52:15	 From Travis Bemann : a key thing to remember is the difference between simplicity in external API's and simplicity in internals
11:52:25	 From Brad Nelson : yeah
11:52:41	 From Brad Nelson : I do think we suffer from a lack of gateways in.
11:52:44	 From Travis Bemann : a complex implementation can easily be hidden behind a simple API, and a simple implementation can saddle the user with a complex external interface
11:52:54	 From Brad Nelson : yes^
11:53:38	 From Travis Bemann : to me it is better to have a complex implementation with a simple interface than a simple implementation with a complex interface in most cases
11:54:44	 From Brad Nelson : I think aspirationally you can have both, but not sure reality lives up to that.
11:55:47	 From Brad Nelson : Simple often means "the same as something else I already know", so it is a challenge.
11:56:43	 From Bob Armstrong : Why should either be complex ?
A reason to go directly from the interactive simplicity at the machine level to the interactive simplicity at the ` math level is the short path to reliability .
11:57:11	 From Travis Bemann : ideally everything would be simple
11:57:19	 From Brad Nelson : mm
12:01:12	 From Travis Bemann : another major advantage of forth-based softcore systems is on-the-fly reconfiguration
12:01:23	 From Travis Bemann : that's what Matthias Koch uses Mecrisp Ice for in his day job