Groups | Search | Server Info | Keyboard shortcuts | Login | Register [http] [https] [nntp] [nntps]
Groups > sci.physics.relativity > #671581 > unrolled thread
| Started by | Mild Shock <janburse@fastmail.fm> |
|---|---|
| First post | 2026-07-22 21:00 +0200 |
| Last post | 2026-07-29 17:05 +0200 |
| Articles | 9 on this page of 89 — 10 participants |
Back to article view | Back to sci.physics.relativity
The Wuhan Virus that destroyed Python [ggml Manifesto] Mild Shock <janburse@fastmail.fm> - 2026-07-22 21:00 +0200
Deadlock Exorcism: Switch from Push to Pull [A pi-calculus Specification of Prolog] (Re: The Wuhan Virus that destroyed Python [ggml Manifesto]) Mild Shock <janburse@fastmail.fm> - 2026-07-23 00:23 +0200
Why do you even need a mpmc queue? [Thunder Kittens] (Re: Deadlock Exorcism: Switch from Push to Pull) Mild Shock <janburse@fastmail.fm> - 2026-07-23 08:43 +0200
Trivial balancing example for (int i=0; i<global_id; i++) (Re: Why do you even need a mpmc queue? [Thunder Kittens]) Mild Shock <janburse@fastmail.fm> - 2026-07-23 08:57 +0200
Enqueue/dequeue need not be fast and can spinn ["fairness" questions] (Was: Trivial balancing example for (int i=0; i<global_id; i++)) Mild Shock <janburse@fastmail.fm> - 2026-07-23 09:11 +0200
The Pixel Phone AI Experiment Song (Re: Enqueue/dequeue need not be fast and can spinn ["fairness" questions] ) Mild Shock <janburse@fastmail.fm> - 2026-07-23 09:21 +0200
Re: Why do you even need a mpmc queue? [Thunder Kittens] (Re: Deadlock Exorcism: Switch from Push to Pull) Ross Finlayson <ross.a.finlayson@gmail.com> - 2026-07-23 08:24 -0700
Potential Python Recovery: Free Threading [3.13 release] (Was: The Wuhan Virus that destroyed Python [ggml Manifesto]) Mild Shock <janburse@fastmail.fm> - 2026-07-23 10:19 +0200
Re: Potential Python Recovery: Free Threading [3.13 release] (Was: The Wuhan Virus that destroyed Python [ggml Manifesto]) Ross Valikhanov <kavna@rl.ru> - 2026-07-23 16:01 +0000
Re: The Wuhan Virus that destroyed Python [ggml Manifesto] Ramon Dubenkov <omd@nnk.ru> - 2026-07-23 13:38 +0000
The things XILINX braught to the AMD table (Was: The Wuhan Virus that destroyed Python [ggml Manifesto]) Mild Shock <janburse@fastmail.fm> - 2026-07-23 18:47 +0200
NIVIDIA evacuated its Chinese market [Tau Scaling] (Was: The things XILINX braught to the AMD table) Mild Shock <janburse@fastmail.fm> - 2026-07-23 19:11 +0200
NVIDIA evacuated its Chinese market [Tau Scaling] (Re: The things XILINX braught to the AMD table) Mild Shock <janburse@fastmail.fm> - 2026-07-23 19:12 +0200
Re: NVIDIA evacuated its Chinese market [Tau Scaling] (Re: The things XILINX braught to the AMD table) Lane W <cactus_DAC@yahoo.com> - 2026-07-23 11:22 -0600
Micro penis mother sung arias (Was: NVIDIA evacuated its Chinese market [Tau Scaling]) Mild Shock <janburse@fastmail.fm> - 2026-07-24 14:38 +0200
Re: Micro penis mother sung arias (Was: NVIDIA evacuated its Chinese market [Tau Scaling]) Lane W <cactus_DAC@yahoo.com> - 2026-07-24 07:15 -0600
Micro penis brain is in constant hiatus (Was: Micro penis mother sung arias) Mild Shock <janburse@fastmail.fm> - 2026-07-24 15:24 +0200
Re: Micro penis brain is in constant hiatus (Was: Micro penis mother sung arias) Mild Shock <janburse@fastmail.fm> - 2026-07-24 15:36 +0200
Ignoramus or Ignorabimus: I don't care (π-WAM) (Re: Micro penis brain is in constant hiatus) Mild Shock <janburse@fastmail.fm> - 2026-07-24 15:38 +0200
Re: Ignoramus or Ignorabimus: I don't care (π-WAM) (Re: Micro penis brain is in constant hiatus) Lane W <cactus_DAC@yahoo.com> - 2026-07-24 08:31 -0600
You are a moron, brainless putin payed (Was: Ignoramus or Ignorabimus: I don't care (π-WAM)) Mild Shock <janburse@fastmail.fm> - 2026-07-24 18:01 +0200
Re: You are a moron, brainless putin payed (Was: Ignoramus or Ignorabimus: I don't care (π-WAM)) Lane W <cactus_DAC@yahoo.com> - 2026-07-24 10:27 -0600
Yeah keep reading my posts, uninspired fool (Was: You are a moron, brainless putin payed) Mild Shock <janburse@fastmail.fm> - 2026-07-24 19:45 +0200
Re: Yeah keep reading my posts, uninspired fool (Was: You are a moron, brainless putin payed) Lane W <cactus_DAC@yahoo.com> - 2026-07-24 12:11 -0600
LoL (Was: Yeah keep reading my posts, uninspired fool ) Mild Shock <janburse@fastmail.fm> - 2026-07-24 20:12 +0200
Re: LoL (Was: Yeah keep reading my posts, uninspired fool ) Lane W <cactus_DAC@yahoo.com> - 2026-07-24 12:53 -0600
Out of the blue accusation span 15 days [Empirical USENET study] (Was: Ignoramus or Ignorabimus: I don't care (π-WAM)) Mild Shock <janburse@fastmail.fm> - 2026-07-24 20:26 +0200
A brain desease of 20 days [Rossy Boy] (Re: Ignoramus or Ignorabimus: I don't care (π-WAM)) Mild Shock <janburse@fastmail.fm> - 2026-07-29 18:40 +0200
Re: A brain desease of 20 days [Rossy Boy] (Re: Ignoramus or Ignorabimus: I don't care (π-WAM)) Mantra Mahonov <hnaam@aat.ru> - 2026-07-29 21:15 +0000
I didn't use a Ryzen Halo, whats wrong with you? (Was: A brain desease of 20 days [Rossy Boy]) Mild Shock <janburse@fastmail.fm> - 2026-07-29 23:24 +0200
ASML stocks are plunging, bye bye dutchies (Was: NVIDIA evacuated its Chinese market [Tau Scaling]) Mild Shock <janburse@fastmail.fm> - 2026-07-28 14:17 +0200
Little Data Center on Your Palm [AI Laptops for 500 USD] (Was: The Wuhan Virus that destroyed Python [ggml Manifesto] Mild Shock <janburse@fastmail.fm> - 2026-07-24 17:58 +0200
2008: 4 Blades + Tesla S1070 versus 2026: 1 AI Laptop (Re: Little Data Center on Your Palm [AI Laptops for 500 USD]) Mild Shock <janburse@fastmail.fm> - 2026-07-24 18:16 +0200
Re: Little Data Center on Your Palm [AI Laptops for 500 USD] (Was: The Wuhan Virus that destroyed Python [ggml Manifesto] Bradford Babkoff <ffb@odbb.ru> - 2026-07-24 18:05 +0000
LoL (Was: Little Data Center on Your Palm [AI Laptops for 500 USD]) Mild Shock <janburse@fastmail.fm> - 2026-07-24 20:11 +0200
Hurry the blue bus doesnt stop indefinitely (Was: The Wuhan Virus that destroyed Python [ggml Manifesto]) Mild Shock <janburse@fastmail.fm> - 2026-07-24 20:36 +0200
Not SIMD, a MIMD design for NVIDIA Volta (Re: Hurry the blue bus doesnt stop indefinitely) Mild Shock <janburse@fastmail.fm> - 2026-07-24 20:57 +0200
Could take 3-4 months find machine / browser (Was Not SIMD, a MIMD design for NVIDIA Volta) Mild Shock <janburse@fastmail.fm> - 2026-07-24 21:15 +0200
The Koan of pi-WAM queues [FORTRAN-S] (Was: Not SIMD, a MIMD design for NVIDIA Volta) Mild Shock <janburse@fastmail.fm> - 2026-07-26 19:52 +0200
The turbo capping of AI Laptops (Re: The Koan of pi-WAM queues [FORTRAN-S]) Mild Shock <janburse@fastmail.fm> - 2026-07-26 20:01 +0200
Re: The Koan of pi-WAM queues [FORTRAN-S] (Was: Not SIMD, a MIMD design for NVIDIA Volta) Ross Finlayson <ross.a.finlayson@gmail.com> - 2026-07-26 20:33 -0700
Why forget Bulgarians, never on my mind (Re: The Koan of pi-WAM queues [FORTRAN-S] (Was: Not SIMD, a MIMD design for NVIDIA Volta) Mild Shock <janburse@fastmail.fm> - 2026-07-27 09:14 +0200
miniTriton CUDA is an alternative to torch variants (Was: Why forget Bulgarians, never on my mind) Mild Shock <janburse@fastmail.fm> - 2026-07-27 09:40 +0200
Andrej Karpathy original gangster of Budget Laptop (Was: miniTriton CUDA is an alternative to torch variants) Mild Shock <janburse@fastmail.fm> - 2026-07-27 09:51 +0200
Re: Andrej Karpathy original gangster of Budget Laptop (Was: miniTriton CUDA is an alternative to torch variants) Ross Finlayson <ross.a.finlayson@gmail.com> - 2026-07-27 01:41 -0700
Re: Why forget Bulgarians, never on my mind (Re: The Koan of pi-WAM queues [FORTRAN-S] (Was: Not SIMD, a MIMD design for NVIDIA Volta) Ross Finlayson <ross.a.finlayson@gmail.com> - 2026-07-27 01:38 -0700
The evolution of hardware and GPT-2 training (Was: Why forget Bulgarians, never on my mind) Mild Shock <janburse@fastmail.fm> - 2026-07-27 10:56 +0200
How speed up π-WAM with vector operations (Was: The evolution of hardware and GPT-2 training) Mild Shock <janburse@fastmail.fm> - 2026-07-27 11:08 +0200
AI accelerator extend from GPU to CPU [Zero Copying] (Was: How speed up π-WAM with vector operations) Mild Shock <janburse@fastmail.fm> - 2026-07-27 11:21 +0200
The invention of vector and matrix registers [NVIDIA Volta] (Was: AI accelerator extend from GPU to CPU [Zero Copying]) Mild Shock <janburse@fastmail.fm> - 2026-07-27 13:20 +0200
Maybe they should have named it NVIDIA Einstein [Rossy Boy Toe Sucking] (Re: The invention of vector and matrix registers [NVIDIA Volta] (Was: AI accelerator extend from GPU to CPU [Zero Copying]) Mild Shock <janburse@fastmail.fm> - 2026-07-27 17:14 +0200
π-WAM is not adding decimals, it is removing decimals (Re: The invention of vector and matrix registers [NVIDIA Volta]) Mild Shock <janburse@fastmail.fm> - 2026-07-27 18:36 +0200
In Budget Laptops the TOPS come with low energy footprint (Was: π-WAM is not adding decimals, it is removing decimals) Mild Shock <janburse@fastmail.fm> - 2026-07-27 18:44 +0200
Java picky concerning JIT-ing [Luckier with C++/C or FORTRAN compilers?] (Re: The Koan of pi-WAM queues [FORTRAN-S]) Mild Shock <janburse@fastmail.fm> - 2026-07-29 12:55 +0200
Potato Computer owner impressed by Ukraine Tech [Rossy Boys Brother?] (Re: Hurry the blue bus doesnt stop indefinitely) Mild Shock <janburse@fastmail.fm> - 2026-07-27 16:57 +0200
Got it. Or are you too stupid? [New Usenet Mantra] (Re: The Wuhan Virus that destroyed Python [ggml Manifesto]) Mild Shock <janburse@fastmail.fm> - 2026-07-27 19:00 +0200
Re: Got it. Or are you too stupid? [New Usenet Mantra] (Re: The Wuhan Virus that destroyed Python [ggml Manifesto]) Kim Baitchorov <bvhkoc@bmc.ru> - 2026-07-27 22:38 +0000
Clueless about MIMD as usual [Flynn's Taxonomy] (Was: Rossy Boy is neither Einstein nor Zweistein) Mild Shock <janburse@fastmail.fm> - 2026-07-28 11:29 +0200
confused rossy boy is confused (Re: Clueless about MIMD as usual [Flynn's Taxonomy]) Mild Shock <janburse@fastmail.fm> - 2026-07-29 11:20 +0200
Gemini, DeepSeek, OpenAI more clever than rossy boy (Re: confused rossy boy is confused) Mild Shock <janburse@fastmail.fm> - 2026-07-29 11:21 +0200
In AI Acceleration nobody cares about CivetWeb (Re: Gemini, DeepSeek, OpenAI more clever than rossy boy) Mild Shock <janburse@fastmail.fm> - 2026-07-29 11:29 +0200
Run with minimum HTTPS and .mjs type (Re: In AI Acceleration nobody cares about CivetWeb) Mild Shock <janburse@fastmail.fm> - 2026-07-29 11:49 +0200
Your strictness is your problem , not mine [See WebLLM] (Re: Run with minimum HTTPS and .mjs type) Mild Shock <janburse@fastmail.fm> - 2026-07-29 11:51 +0200
Re: In AI Acceleration nobody cares about CivetWeb (Re: Gemini, DeepSeek, OpenAI more clever than rossy boy) Lane W <cactus_DAC@yahoo.com> - 2026-07-29 07:06 -0600
Re: In AI Acceleration nobody cares about CivetWeb (Re: Gemini, DeepSeek, OpenAI more clever than rossy boy) Ross Finlayson <ross.a.finlayson@gmail.com> - 2026-07-29 07:20 -0700
You are still chewing on SIMD. LoL (Re: Clueless about MIMD as usual [Flynn's Taxonomy]) Mild Shock <janburse@fastmail.fm> - 2026-07-29 17:14 +0200
Hurry Rossy Boy, the blue bus is waiting (Re: You are still chewing on SIMD. LoL) Mild Shock <janburse@fastmail.fm> - 2026-07-29 17:54 +0200
Look how they advertized CUDA and logical threads (Re: Hurry Rossy Boy, the blue bus is waiting) Mild Shock <janburse@fastmail.fm> - 2026-07-29 17:56 +0200
Forget any arithmetization of product FSA (Re: Look how they advertized CUDA and logical threads) Mild Shock <janburse@fastmail.fm> - 2026-07-29 17:57 +0200
Re: You are still chewing on SIMD. LoL (Re: Clueless about MIMD as usual [Flynn's Taxonomy]) Ross Finlayson <ross.a.finlayson@gmail.com> - 2026-07-29 10:48 -0700
Rossy Boys tears could cool a data center [pi-WAM Interleaved Synchronized Emulator] (Was: You are still chewing on SIMD. LoL) Mild Shock <janburse@fastmail.fm> - 2026-07-29 20:03 +0200
I don't use Rust, you are crazy [Jump off a bridge, idiot] (Re: Rossy Boys tears could cool a data center [pi-WAM Interleaved Synchronized Emulator]) Mild Shock <janburse@fastmail.fm> - 2026-07-29 20:24 +0200
Re: I don't use Rust, you are crazy [Jump off a bridge, idiot] (Re: Rossy Boys tears could cool a data center [pi-WAM Interleaved Synchronized Emulator]) Ross Finlayson <ross.a.finlayson@gmail.com> - 2026-07-29 12:27 -0700
Re: I don't use Rust, you are crazy [Jump off a bridge, idiot] (Re: Rossy Boys tears could cool a data center [pi-WAM Interleaved Synchronized Emulator]) Ross Finlayson <ross.a.finlayson@gmail.com> - 2026-07-29 13:40 -0700
Hack ecosystem ignorance paired with paranoia [Nand to Tetris] (Re: Rossy Boys tears could cool a data center) Mild Shock <janburse@fastmail.fm> - 2026-07-29 22:51 +0200
A funny Q16.16 experiment with Hack (Re: Hack ecosystem ignorance paired with paranoia [Nand to Tetris]) Mild Shock <janburse@fastmail.fm> - 2026-07-29 23:12 +0200
Re: A funny Q16.16 experiment with Hack (Re: Hack ecosystem ignorance paired with paranoia [Nand to Tetris]) Alexey Bessonov <nexes@ebxsvs.ru> - 2026-07-29 21:24 +0000
Summer Challenge: libSQL = Prolog+Modes [VDBE versus π-WAM] (Re: A funny Q16.16 experiment with Hack) Mild Shock <janburse@fastmail.fm> - 2026-07-30 11:26 +0200
I wrote Hack VM for π-WAM from scratch [4 Months total JavaScript, Python and Java] (Re: Hack ecosystem ignorance paired with paranoia [Nand to Tetris]) Mild Shock <janburse@fastmail.fm> - 2026-07-30 19:35 +0200
Re: I wrote Hack VM for π-WAM from scratch [4 Months total JavaScript, Python and Java] (Re: Hack ecosystem ignorance paired with paranoia [Nand to Tetris]) Mild Shock <janburse@fastmail.fm> - 2026-07-30 19:49 +0200
For WebGPU I first had SIMD in mind (Re: I wrote Hack VM for π-WAM from scratch) Mild Shock <janburse@fastmail.fm> - 2026-07-30 19:50 +0200
Corr.: 4 Months --> 4 Weeks (Re: For WebGPU I first had SIMD in mind) Mild Shock <janburse@fastmail.fm> - 2026-07-30 20:05 +0200
MIPS is a big Huffman mess [But Hack could do it] (Re: I wrote Hack VM for π-WAM from scratch) Mild Shock <janburse@fastmail.fm> - 2026-07-30 22:30 +0200
Not declarative with PHI (Φ) nodes (Re: MIPS is a big Huffman mess [But Hack could do it]) Mild Shock <janburse@fastmail.fm> - 2026-07-30 22:42 +0200
Quo Vadis: Extend investigations to WebNN (Re: I wrote Hack VM for π-WAM from scratch) Mild Shock <janburse@fastmail.fm> - 2026-07-31 20:45 +0200
Re: Quo Vadis: Extend investigations to WebNN (Re: I wrote Hack VM for π-WAM from scratch) Jereb Pohlebaev <obje@bbvoeoaa.ru> - 2026-07-31 20:23 +0000
Lamas in a cradle and Lamas on the edge [Red Pyjama] (Was: The Wuhan Virus that destroyed Python [ggml Manifesto]) Mild Shock <janburse@fastmail.fm> - 2026-07-29 13:02 +0200
AI Accelerators and ISO Prolog multi-threading (Was: Lamas in a cradle and Lamas on the edge [Red Pyjama]) Mild Shock <janburse@fastmail.fm> - 2026-07-29 17:01 +0200
Actor/Erlang is dead, no Thread and Mailbox conflation [golang channels] (Was: AI Accelerators and ISO Prolog multi-threading) Mild Shock <janburse@fastmail.fm> - 2026-07-29 17:05 +0200
Page 5 of 5 — ← Prev page 1 2 3 4 [5]
| From | Mild Shock <janburse@fastmail.fm> |
|---|---|
| Date | 2026-07-30 19:50 +0200 |
| Subject | For WebGPU I first had SIMD in mind (Re: I wrote Hack VM for π-WAM from scratch) |
| Message-ID | <114g2tn$m1rd$3@solani.org> |
| In reply to | #671751 |
Hi,
The op-codes are all uniform, have the
same sub fields. Already Z-80 CPU differs here.
Another difference to a Z-80 CPU is that
their instruction stream was 8-bit, instructions
can 1, 2, 3 or 4 byte long. On the other
hand in my Hack VM all instructions are
one 32-bit chunk. The porting of a first
prototype that I already had, to WebGPU / WGSL
only took like 1-2 hours. The execution
of Hack VM is very simple, version 1.0,
for a single shader:
fn run() {
var pc : i32 = 0;
var accu : i32 = 0;
while (pc < i32(arrayLength(&code))) {
var instr : i32 = code[pc];
pc += 1;
var value : i32 = run_get(instr);
accu = run_fun(instr, accu, value);
run_set(instr, accu);
pc += run_jump(instr, accu);
}
}
https://github.com/Jean-Luc-Picard-2021/gigabudget/blob/b8946e891be774c40522267ab17062d32b023e7a/course/example63/boot.mjs#L176-L187
I first though this will be perfect for
SIMD. Until I learnt that modern GPUs have
anyway MIMD. Hell Yeah, thats much better!
Bye
Mild Shock schrieb:
> Hi,
>
> > Just downloading some other person's code
>
> I didn't do that, I wrote Hack VM for pi-WAM
> from scratch, over the last 4 weeks. I came
> back from holidays on end of June 2026, and now
>
> we have end of July 2026. But its only possible
> because the instruction set is very smal, like
> ca. 8 functions and ca. 8 modes and ca. 8 conditions,
>
> so its ca. 8 x 8 x 8 = 512 opcodes, each has an
> A parameter and a D parameter simultaneously.
> It has currently the following CPU backends:
>
> - Now supports interleaved synchronous emulation.
> - Now supports warp parallelism via Java platform threads.
> - Now supports warp parallelism via Python system threads.
> - Now supports warp parallelism via JavaScript worker threads.
> - Note: For Python free threads are not yet fully tested.
> - Note: For JavaScript web workers are not yet fully tested.
>
> https://www.dogelog.ch/typtab/doclet/book/14_install/05_notes22/110_224.html
>
>
> But frankly I came to encounter Hack not from
> the usual university curriculum web resources,
> but indirectly through a post about a Prolog
>
> emulation of Hack, using constrained horn clauses (CHC):
>
> Verifying Nand2Tetris Assembly
> https://www.philipzucker.com/nand2tetris-chc/
>
> The binary encoding is currently that the functions,
> modes and conditions eat up a nibble (4-bit), in
> total 12-bit, which I use then 10-bit for A parameter
>
> and 10-bit for D parameter. I used AI freemium, Codex
> by ChatGPT from within IntelliJ to do some fragment
> code translations automatically from Java to JavaScript
>
> or from JavaScript to Python.
>
> Have Fun!
>
> Bye
>
> Mild Shock schrieb:
>> Hi,
>>
>> > Who exactly is the thief? Does this person
>> > have stats in the Rogue class in dungeons
>> > and dragons?
>>
>> The conspiracy theory of a stealing of Torso VDBE,
>> by Rossy Boy, is probably a result of complete
>> ignorance of the Hack ecosystem.
>>
>> Hack is a very popular computer science project,
>> with a couple of subprojects in hardware and
>> software. It goes also by the name Nand to Tetris,
>>
>> and is programming language agnositic. You can do
>> Hack experiments in any programming language, be
>> it BASIC, ADA or Rust. Nobody cares.
>>
>> The gist are projects like here, first to
>> educate yourself about Hack:
>>
>> https://www.nand2tetris.org/course
>>
>> And then to use Hack in different contexts:
>>
>> https://www.nand2tetris.org/copy-of-talks
>>
>> For didactic purposes, I used Hack for my WebGPU
>> experiment. I didn't even take a look at Torso
>> VDBE, why should I? Hack is nicely documented,
>>
>> has even a book, and fusing the two 16-bit
>> instruction types A and D, into a single 32-bit
>> instruction stream, is nowhere patented.
>>
>> Bye
>>
>> Mild Shock schrieb:
>>> Hi,
>>>
>>> Rossy Boys tears could cool a data center,
>>> he thinks there exists no literature about
>>> serial algorithms of parallel stuff, and
>>>
>>> he also thinks normal forms lead to optimizing
>>> something. LoL, what a utter bullshit. I did
>>> alreay a serial implementation of a parallel
>>>
>>> simulation of my pi-WAM. Just lookup the literature
>>> about pi-calulus. I published it a few days ago,
>>> its part of 2.2.4 released already:
>>>
>>> Parallel π-WAM: An Interleaved Synchronous Emulator
>>> https://medium.com/2989/0196089e143a
>>>
>>> Whats your point, Rossy Boy? Except you post pretend
>>> nonsense not knowing what you are doing?
>>>
>>> Bye
>>>
>>> Ross Finlayson schrieb:
>>>> No, troll, these are serial algorithms their optimized forms.
>>>>
>>>> Normal sorts of forms, ....
>>>>
>>>>
>>>> Yeah, everybody already figured out "interpreters" and
>>>> "programs" and "spawning".
>>>>
>>>> Go spawn yourself.
>>>>
>>>>
>>>
>>
>
[toc] | [prev] | [next] | [standalone]
| From | Mild Shock <janburse@fastmail.fm> |
|---|---|
| Date | 2026-07-30 20:05 +0200 |
| Subject | Corr.: 4 Months --> 4 Weeks (Re: For WebGPU I first had SIMD in mind) |
| Message-ID | <114g3ol$m2d8$2@solani.org> |
| In reply to | #671753 |
Hi,
There is a typo here:
Subject: I wrote Hack VM for π-WAM from scratch
[4 Months total JavaScript, Python and Java]
It should say:
Subject: I wrote Hack VM for π-WAM from scratch
[4 Weeks total JavaScript, Python and Java]
Sorry!
Bye
Mild Shock schrieb:
> Hi,
>
> The op-codes are all uniform, have the
> same sub fields. Already Z-80 CPU differs here.
> Another difference to a Z-80 CPU is that
>
> their instruction stream was 8-bit, instructions
> can 1, 2, 3 or 4 byte long. On the other
> hand in my Hack VM all instructions are
>
> one 32-bit chunk. The porting of a first
> prototype that I already had, to WebGPU / WGSL
> only took like 1-2 hours. The execution
>
> of Hack VM is very simple, version 1.0,
> for a single shader:
>
> fn run() {
> var pc : i32 = 0;
> var accu : i32 = 0;
> while (pc < i32(arrayLength(&code))) {
> var instr : i32 = code[pc];
> pc += 1;
> var value : i32 = run_get(instr);
> accu = run_fun(instr, accu, value);
> run_set(instr, accu);
> pc += run_jump(instr, accu);
> }
> }
>
> https://github.com/Jean-Luc-Picard-2021/gigabudget/blob/b8946e891be774c40522267ab17062d32b023e7a/course/example63/boot.mjs#L176-L187
>
>
> I first though this will be perfect for
> SIMD. Until I learnt that modern GPUs have
> anyway MIMD. Hell Yeah, thats much better!
>
> Bye
>
> Mild Shock schrieb:
>> Hi,
>>
>> > Just downloading some other person's code
>>
>> I didn't do that, I wrote Hack VM for pi-WAM
>> from scratch, over the last 4 weeks. I came
>> back from holidays on end of June 2026, and now
>>
>> we have end of July 2026. But its only possible
>> because the instruction set is very smal, like
>> ca. 8 functions and ca. 8 modes and ca. 8 conditions,
>>
>> so its ca. 8 x 8 x 8 = 512 opcodes, each has an
>> A parameter and a D parameter simultaneously.
>> It has currently the following CPU backends:
>>
>> - Now supports interleaved synchronous emulation.
>> - Now supports warp parallelism via Java platform threads.
>> - Now supports warp parallelism via Python system threads.
>> - Now supports warp parallelism via JavaScript worker threads.
>> - Note: For Python free threads are not yet fully tested.
>> - Note: For JavaScript web workers are not yet fully tested.
>>
>> https://www.dogelog.ch/typtab/doclet/book/14_install/05_notes22/110_224.html
>>
>>
>> But frankly I came to encounter Hack not from
>> the usual university curriculum web resources,
>> but indirectly through a post about a Prolog
>>
>> emulation of Hack, using constrained horn clauses (CHC):
>>
>> Verifying Nand2Tetris Assembly
>> https://www.philipzucker.com/nand2tetris-chc/
>>
>> The binary encoding is currently that the functions,
>> modes and conditions eat up a nibble (4-bit), in
>> total 12-bit, which I use then 10-bit for A parameter
>>
>> and 10-bit for D parameter. I used AI freemium, Codex
>> by ChatGPT from within IntelliJ to do some fragment
>> code translations automatically from Java to JavaScript
>>
>> or from JavaScript to Python.
>>
>> Have Fun!
>>
>> Bye
[toc] | [prev] | [next] | [standalone]
| From | Mild Shock <janburse@fastmail.fm> |
|---|---|
| Date | 2026-07-30 22:30 +0200 |
| Subject | MIPS is a big Huffman mess [But Hack could do it] (Re: I wrote Hack VM for π-WAM from scratch) |
| Message-ID | <114gc9m$mmiv$2@solani.org> |
| In reply to | #671751 |
Hi, There is Prolog compiler which spits out Hack. From there on your are free to develop and/or use any Hack realization that goes from abstract to concrete. You could replace the CPU backends that realize a Hack VM by MIPS. Shouldn't be difficult. Basically I refused to think in Huffman Coding (*) while designing Hack VM. On the other hand the MIPS architecture looks like a big Huffman mess. Already its initial design has 3 instructions types: Type format (bits) R opcode(6) rs(5) rt(5) rd(5) shamt(5) funct(6) I opcode(6) rs(5) rt(5) imme(16) J opcode(6) addr(26) While my Hack has only 1 instruction type, when binary encoded for Hack VM, the currently used design looks as follows: Type format (bits) AD opcode(4) mode(4) cond(4) imme(10) addr(10) But since its an abstract machine, nothing prevents you from translating Hack code into MIPS before executing it. In has far you have to distinguish Hack, which is specified in Prolog. And Hack VM which is a virtual machine, with the above instruction packing. And which has currently a JavaScript runtime, a Python runtime and a Java runtime. Bye (*) https://en.wikipedia.org/wiki/Huffman_coding Mild Shock schrieb: > Hi, > > > Just downloading some other person's code > > I didn't do that, I wrote Hack VM for pi-WAM > from scratch, over the last 4 weeks. I came > back from holidays on end of June 2026, and now > > we have end of July 2026. But its only possible > because the instruction set is very smal, like > ca. 8 functions and ca. 8 modes and ca. 8 conditions, > > so its ca. 8 x 8 x 8 = 512 opcodes, each has an > A parameter and a D parameter simultaneously. > It has currently the following CPU backends: > > - Now supports interleaved synchronous emulation. > - Now supports warp parallelism via Java platform threads. > - Now supports warp parallelism via Python system threads. > - Now supports warp parallelism via JavaScript worker threads. > - Note: For Python free threads are not yet fully tested. > - Note: For JavaScript web workers are not yet fully tested. > > https://www.dogelog.ch/typtab/doclet/book/14_install/05_notes22/110_224.html > > > But frankly I came to encounter Hack not from > the usual university curriculum web resources, > but indirectly through a post about a Prolog > > emulation of Hack, using constrained horn clauses (CHC): > > Verifying Nand2Tetris Assembly > https://www.philipzucker.com/nand2tetris-chc/ > > The binary encoding is currently that the functions, > modes and conditions eat up a nibble (4-bit), in > total 12-bit, which I use then 10-bit for A parameter > > and 10-bit for D parameter. I used AI freemium, Codex > by ChatGPT from within IntelliJ to do some fragment > code translations automatically from Java to JavaScript > > or from JavaScript to Python. > > Have Fun! > > Bye > > Mild Shock schrieb: >> Hi, >> >> > Who exactly is the thief? Does this person >> > have stats in the Rogue class in dungeons >> > and dragons? >> >> The conspiracy theory of a stealing of Torso VDBE, >> by Rossy Boy, is probably a result of complete >> ignorance of the Hack ecosystem. >> >> Hack is a very popular computer science project, >> with a couple of subprojects in hardware and >> software. It goes also by the name Nand to Tetris, >> >> and is programming language agnositic. You can do >> Hack experiments in any programming language, be >> it BASIC, ADA or Rust. Nobody cares. >> >> The gist are projects like here, first to >> educate yourself about Hack: >> >> https://www.nand2tetris.org/course >> >> And then to use Hack in different contexts: >> >> https://www.nand2tetris.org/copy-of-talks >> >> For didactic purposes, I used Hack for my WebGPU >> experiment. I didn't even take a look at Torso >> VDBE, why should I? Hack is nicely documented, >> >> has even a book, and fusing the two 16-bit >> instruction types A and D, into a single 32-bit >> instruction stream, is nowhere patented. >> >> Bye >> >> Mild Shock schrieb: >>> Hi, >>> >>> Rossy Boys tears could cool a data center, >>> he thinks there exists no literature about >>> serial algorithms of parallel stuff, and >>> >>> he also thinks normal forms lead to optimizing >>> something. LoL, what a utter bullshit. I did >>> alreay a serial implementation of a parallel >>> >>> simulation of my pi-WAM. Just lookup the literature >>> about pi-calulus. I published it a few days ago, >>> its part of 2.2.4 released already: >>> >>> Parallel π-WAM: An Interleaved Synchronous Emulator >>> https://medium.com/2989/0196089e143a >>> >>> Whats your point, Rossy Boy? Except you post pretend >>> nonsense not knowing what you are doing? >>> >>> Bye >>> >>> Ross Finlayson schrieb: >>>> No, troll, these are serial algorithms their optimized forms. >>>> >>>> Normal sorts of forms, .... >>>> >>>> >>>> Yeah, everybody already figured out "interpreters" and >>>> "programs" and "spawning". >>>> >>>> Go spawn yourself. >>>> >>>> >>> >> >
[toc] | [prev] | [next] | [standalone]
| From | Mild Shock <janburse@fastmail.fm> |
|---|---|
| Date | 2026-07-30 22:42 +0200 |
| Subject | Not declarative with PHI (Φ) nodes (Re: MIPS is a big Huffman mess [But Hack could do it]) |
| Message-ID | <114gcut$mn2q$4@solani.org> |
| In reply to | #671755 |
Hi, Another choice for naming Hack, would be to call it an intermediate format. But this is typically used here: The intermediate representation, or IR for short, is an in-memory data structure that represents executable code. https://www.llvmpy.org/llvmpy-doc/dev/doc/llvm_concepts.html#ssa-form-and-phi-nodes So I still like the term abstract machine, as already used in the past by David H. D. Warren for the famous, and in my opinion infamous: Warren Abstract Machine 1983 https://en.wikipedia.org/wiki/Warren_Abstract_Machine Maybe you can take the term abstract machine as a hint that it is more lower level, and more imperative. Not something highlevel, that is easily malleable. But abstract also captures the notion that there is still a level further down, making it concrete. And you find many Prolog systems that did just that, they compile WAM into a further instruction stream, like x86 or whatever, for binary compiled code, that is not interpreted WAM. Bye Mild Shock schrieb: > Hi, > > There is Prolog compiler which spits out Hack. > From there on your are free to develop > and/or use any Hack realization that goes > > from abstract to concrete. You could > replace the CPU backends that realize > a Hack VM by MIPS. Shouldn't be difficult. > > Basically I refused to think in Huffman > Coding (*) while designing Hack VM. On the > other hand the MIPS architecture looks > > like a big Huffman mess. Already its > initial design has 3 instructions types: > > Type format (bits) > R opcode(6) rs(5) rt(5) rd(5) shamt(5) funct(6) > I opcode(6) rs(5) rt(5) imme(16) > J opcode(6) addr(26) > > While my Hack has only 1 instruction > type, when binary encoded for Hack VM, > the currently used design looks as follows: > > Type format (bits) > AD opcode(4) mode(4) cond(4) imme(10) addr(10) > > But since its an abstract machine, nothing > prevents you from translating Hack code > into MIPS before executing it. > > In has far you have to distinguish Hack, > which is specified in Prolog. And Hack VM > which is a virtual machine, with the above > > instruction packing. And which has currently > a JavaScript runtime, a Python runtime > and a Java runtime. > > Bye > > (*) > https://en.wikipedia.org/wiki/Huffman_coding >
[toc] | [prev] | [next] | [standalone]
| From | Mild Shock <janburse@fastmail.fm> |
|---|---|
| Date | 2026-07-31 20:45 +0200 |
| Subject | Quo Vadis: Extend investigations to WebNN (Re: I wrote Hack VM for π-WAM from scratch) |
| Message-ID | <114iqfe$obtb$2@solani.org> |
| In reply to | #671751 |
Hi, Since we have a good flow, and since NPUs share the same system memory, and possibly a lot of other traits as well with the GPU in libary(edge/furryhaze), we just developed. The idea here is to do first some off Dogelog experiments and then create a library that provides npu_exec/2 for pi-WAM code, the analogue to gpu_exec/2. A name suggestion would be: - edge/ironpaw.p The new Prolog library The NPU will be clearly underutilized when only doing scalar, not sure whether this is even permitted. But in the long run it is planned that pi-WAM will have vector and matrix traits anyways. Here is an example goal can be run with matrix and quantization traits: ?- [X,Y] ins 0..3, Z is X*2+Y*3+4, T is X*3-Y*2-1 These traits will demand some CPU, GPU and NPU translation. If we keep these traits simple, we might indeed arrive at concrete realization from the same abstract machine LoL, ironpaw the little brother of ironfist. Bye Mild Shock schrieb: > Hi, > > > Just downloading some other person's code > > I didn't do that, I wrote Hack VM for pi-WAM > from scratch, over the last 4 weeks. I came > back from holidays on end of June 2026, and now > > we have end of July 2026. But its only possible > because the instruction set is very smal, like > ca. 8 functions and ca. 8 modes and ca. 8 conditions, > > so its ca. 8 x 8 x 8 = 512 opcodes, each has an > A parameter and a D parameter simultaneously. > It has currently the following CPU backends: > > - Now supports interleaved synchronous emulation. > - Now supports warp parallelism via Java platform threads. > - Now supports warp parallelism via Python system threads. > - Now supports warp parallelism via JavaScript worker threads. > - Note: For Python free threads are not yet fully tested. > - Note: For JavaScript web workers are not yet fully tested. > > https://www.dogelog.ch/typtab/doclet/book/14_install/05_notes22/110_224.html > > > But frankly I came to encounter Hack not from > the usual university curriculum web resources, > but indirectly through a post about a Prolog > > emulation of Hack, using constrained horn clauses (CHC): > > Verifying Nand2Tetris Assembly > https://www.philipzucker.com/nand2tetris-chc/ > > The binary encoding is currently that the functions, > modes and conditions eat up a nibble (4-bit), in > total 12-bit, which I use then 10-bit for A parameter > > and 10-bit for D parameter. I used AI freemium, Codex > by ChatGPT from within IntelliJ to do some fragment > code translations automatically from Java to JavaScript > > or from JavaScript to Python. > > Have Fun! > > Bye > > Mild Shock schrieb: >> Hi, >> >> > Who exactly is the thief? Does this person >> > have stats in the Rogue class in dungeons >> > and dragons? >> >> The conspiracy theory of a stealing of Torso VDBE, >> by Rossy Boy, is probably a result of complete >> ignorance of the Hack ecosystem. >> >> Hack is a very popular computer science project, >> with a couple of subprojects in hardware and >> software. It goes also by the name Nand to Tetris, >> >> and is programming language agnositic. You can do >> Hack experiments in any programming language, be >> it BASIC, ADA or Rust. Nobody cares. >> >> The gist are projects like here, first to >> educate yourself about Hack: >> >> https://www.nand2tetris.org/course >> >> And then to use Hack in different contexts: >> >> https://www.nand2tetris.org/copy-of-talks >> >> For didactic purposes, I used Hack for my WebGPU >> experiment. I didn't even take a look at Torso >> VDBE, why should I? Hack is nicely documented, >> >> has even a book, and fusing the two 16-bit >> instruction types A and D, into a single 32-bit >> instruction stream, is nowhere patented. >> >> Bye >> >> Mild Shock schrieb: >>> Hi, >>> >>> Rossy Boys tears could cool a data center, >>> he thinks there exists no literature about >>> serial algorithms of parallel stuff, and >>> >>> he also thinks normal forms lead to optimizing >>> something. LoL, what a utter bullshit. I did >>> alreay a serial implementation of a parallel >>> >>> simulation of my pi-WAM. Just lookup the literature >>> about pi-calulus. I published it a few days ago, >>> its part of 2.2.4 released already: >>> >>> Parallel π-WAM: An Interleaved Synchronous Emulator >>> https://medium.com/2989/0196089e143a >>> >>> Whats your point, Rossy Boy? Except you post pretend >>> nonsense not knowing what you are doing? >>> >>> Bye >>> >>> Ross Finlayson schrieb: >>>> No, troll, these are serial algorithms their optimized forms. >>>> >>>> Normal sorts of forms, .... >>>> >>>> >>>> Yeah, everybody already figured out "interpreters" and >>>> "programs" and "spawning". >>>> >>>> Go spawn yourself. >>>> >>>> >>> >> >
[toc] | [prev] | [next] | [standalone]
| From | Jereb Pohlebaev <obje@bbvoeoaa.ru> |
|---|---|
| Date | 2026-07-31 20:23 +0000 |
| Subject | Re: Quo Vadis: Extend investigations to WebNN (Re: I wrote Hack VM for π-WAM from scratch) |
| Message-ID | <114j08g$3mn0p$1@news.nntp4.net> |
| In reply to | #671759 |
Mild Shock wrote:
> The idea here is to do first some off Dogelog experiments and then
> create a library that provides npu_exec/2 for pi-WAM code, the analogue
> to gpu_exec/2. A name suggestion would be:
>
> - edge/ironpaw.p
hanoi(N) :-
move(N, left, center, right).
move(1, Source, Target, _) :-
format("Move disk from ~w to ~w~n", [Source, Target]).
move(N, Source, Target, Auxiliary) :-
N > 1,
M is N - 1,
move(M, Source, Auxiliary, Target),
move(1, Source, Target, _),
move(M, Auxiliary, Target, Source).
[toc] | [prev] | [next] | [standalone]
| From | Mild Shock <janburse@fastmail.fm> |
|---|---|
| Date | 2026-07-29 13:02 +0200 |
| Subject | Lamas in a cradle and Lamas on the edge [Red Pyjama] (Was: The Wuhan Virus that destroyed Python [ggml Manifesto]) |
| Message-ID | <114cmkf$jmhc$1@solani.org> |
| In reply to | #671581 |
Hi, Why does this Lama have a red pyjama. Oh, its a baby Lama. Its still in the cradle and needs some training: RedPajama-Data-v2 https://github.com/togethercomputer/RedPajama-Data But then Andrej Karpathy recently showed GPT-2 training on rented GPUs for less than 100 USD in less then 2 hours. So where do these grown up Lamas go. Well Georgi Gerganov prefered C++/C when he shouted Llama Llama Red Pyjama. But you also find WebLLM, wrapping the underlying C++/C GPU interface via the W3C standard WebGPU / WGSL, with JavaScript: In-Browser LLM Inference Engine https://webllm.mlc.ai/ My experience with WebLLM 6 months ago on an iPad Pro 2024, still a little early stage performance and robustness. But hey hardware of AI mobile iGPUs is still evolving, and AI laptop, AI smartphones and AI tablets, will soon feature Chinese hardware such some new Kirin AI in 2027. Bye Mild Shock schrieb: > Hi, > > Remember when first all local AI was Python > and PyTorch APIs. And then suddently people strated > using bare metal C/C++ Code. Here is the story: > > How it started: > > GPT-J or GPT-J-6B is an open-source large > language model (LLM) developed by EleutherAI > in 2021. As the name suggests, it is a > generative pre-trained transformer model > designed to produce human-like text that > continues from a prompt. > https://www.eleuther.ai/ > > How it was going [Georgi Gerganov]: > > So a few days later comes out the LLaMA, I do > some calculations and I figure out “Okay, 65 > billion parameters. You probably need about > 40 gigs of RAM, with 4-bit quantization. So > this can run on a MacBook. Why not do it?” > > Why I was able to do it so quickly - basically, > for all that I saw it’s pretty much GPT-J architecture > with some modifications, like some extra memorization > layers. It’s minor changes. Basically, again, the > existing code for the GPT-J, I just simply > modified it there, it happened pretty quickly. > https://changelog.com/podcast/532 > > Georgi Gerganov, Bulgarian, now with Hugging > Face, ggml-cann also running on Chinese AI chips. > ggml Manifesto https://github.com/ggml-org/ggml > > Bye >
[toc] | [prev] | [next] | [standalone]
| From | Mild Shock <janburse@fastmail.fm> |
|---|---|
| Date | 2026-07-29 17:01 +0200 |
| Subject | AI Accelerators and ISO Prolog multi-threading (Was: Lamas in a cradle and Lamas on the edge [Red Pyjama]) |
| Message-ID | <114d4kg$kfj1$1@solani.org> |
| In reply to | #671725 |
Hi, Usual question: > Why implement both pre-emptive threading AND cooperative tasks/engines? I had implemented the ISO proposal in formerly Jekejeke Prolog, you find the ISO proposal here: ISO/IEC DTR 13211–5:2007 Prolog multi-threading support https://logtalk.org/plstd/threads.pdf But the ISO proposal doesn't match modern WebGPU APIs, where your logical threads can live remotely in a dedicated GPU in the VRAM there, and where you would have launch parameters that say: Hey please run 4096 compute shaders for me, that have independet thread state. Using cooperative multi-tasking as the orchestrator works well. Bye Mild Shock schrieb: > Hi, > > Why does this Lama have a red pyjama. > Oh, its a baby Lama. Its still in the cradle > and needs some training: > > RedPajama-Data-v2 > https://github.com/togethercomputer/RedPajama-Data > > But then Andrej Karpathy recently showed > GPT-2 training on rented GPUs for less > than 100 USD in less then 2 hours. > > So where do these grown up Lamas go. > Well Georgi Gerganov prefered C++/C > when he shouted Llama Llama Red Pyjama. > > But you also find WebLLM, wrapping the > underlying C++/C GPU interface via the > W3C standard WebGPU / WGSL, with JavaScript: > > In-Browser LLM Inference Engine > https://webllm.mlc.ai/ > > My experience with WebLLM 6 months > ago on an iPad Pro 2024, still a little early > stage performance and robustness. > > But hey hardware of AI mobile iGPUs is > still evolving, and AI laptop, AI smartphones > and AI tablets, will soon feature Chinese > > hardware such some new Kirin AI in 2027. > > Bye > > Mild Shock schrieb: >> Hi, >> >> Remember when first all local AI was Python >> and PyTorch APIs. And then suddently people strated >> using bare metal C/C++ Code. Here is the story: >> >> How it started: >> >> GPT-J or GPT-J-6B is an open-source large >> language model (LLM) developed by EleutherAI >> in 2021. As the name suggests, it is a >> generative pre-trained transformer model >> designed to produce human-like text that >> continues from a prompt. >> https://www.eleuther.ai/ >> >> How it was going [Georgi Gerganov]: >> >> So a few days later comes out the LLaMA, I do >> some calculations and I figure out “Okay, 65 >> billion parameters. You probably need about >> 40 gigs of RAM, with 4-bit quantization. So >> this can run on a MacBook. Why not do it?” >> >> Why I was able to do it so quickly - basically, >> for all that I saw it’s pretty much GPT-J architecture >> with some modifications, like some extra memorization >> layers. It’s minor changes. Basically, again, the >> existing code for the GPT-J, I just simply >> modified it there, it happened pretty quickly. >> https://changelog.com/podcast/532 >> >> Georgi Gerganov, Bulgarian, now with Hugging >> Face, ggml-cann also running on Chinese AI chips. >> ggml Manifesto https://github.com/ggml-org/ggml >> >> Bye >> >
[toc] | [prev] | [next] | [standalone]
| From | Mild Shock <janburse@fastmail.fm> |
|---|---|
| Date | 2026-07-29 17:05 +0200 |
| Subject | Actor/Erlang is dead, no Thread and Mailbox conflation [golang channels] (Was: AI Accelerators and ISO Prolog multi-threading) |
| Message-ID | <114d4rl$kfj1$4@solani.org> |
| In reply to | #671729 |
Hi, Mostlikely for high performance computing à la, the Actor/Erlang model is dead, they might rely on MPMC (Multiple Producer, Multiple Consumer) queue entities separate from the threads. The ISO Prolog multi-threading support had also such threads. But besides that was also Actor/Erlang leaning in practice, like SWI, where threads have some default queues. So an actor is basically a Thread and Mailbox conflation. While a MPMC queue is a kind of separate Mailbox, where multiple "actors" can read from and write from. A kind of localized Linda Tuple store. Which Programming language did adopted the non-Actor pi-calculus model? Right golang with its channels. Bye Mild Shock schrieb: > Hi, > > Usual question: > > > Why implement both pre-emptive threading > AND cooperative tasks/engines? > > I had implemented the ISO proposal in formerly Jekejeke > Prolog, you find the ISO proposal here: > > ISO/IEC DTR 13211–5:2007 > Prolog multi-threading support > https://logtalk.org/plstd/threads.pdf > > But the ISO proposal doesn't match modern WebGPU APIs, > where your logical threads can live remotely in a dedicated GPU > in the VRAM there, and where you would have launch > > parameters that say: Hey please run 4096 compute > shaders for me, that have independet thread state. Using > cooperative multi-tasking as the orchestrator works well. > > Bye > > Mild Shock schrieb: >> Hi, >> >> Why does this Lama have a red pyjama. >> Oh, its a baby Lama. Its still in the cradle >> and needs some training: >> >> RedPajama-Data-v2 >> https://github.com/togethercomputer/RedPajama-Data >> >> But then Andrej Karpathy recently showed >> GPT-2 training on rented GPUs for less >> than 100 USD in less then 2 hours. >> >> So where do these grown up Lamas go. >> Well Georgi Gerganov prefered C++/C >> when he shouted Llama Llama Red Pyjama. >> >> But you also find WebLLM, wrapping the >> underlying C++/C GPU interface via the >> W3C standard WebGPU / WGSL, with JavaScript: >> >> In-Browser LLM Inference Engine >> https://webllm.mlc.ai/ >> >> My experience with WebLLM 6 months >> ago on an iPad Pro 2024, still a little early >> stage performance and robustness. >> >> But hey hardware of AI mobile iGPUs is >> still evolving, and AI laptop, AI smartphones >> and AI tablets, will soon feature Chinese >> >> hardware such some new Kirin AI in 2027. >> >> Bye >> >> Mild Shock schrieb: >>> Hi, >>> >>> Remember when first all local AI was Python >>> and PyTorch APIs. And then suddently people strated >>> using bare metal C/C++ Code. Here is the story: >>> >>> How it started: >>> >>> GPT-J or GPT-J-6B is an open-source large >>> language model (LLM) developed by EleutherAI >>> in 2021. As the name suggests, it is a >>> generative pre-trained transformer model >>> designed to produce human-like text that >>> continues from a prompt. >>> https://www.eleuther.ai/ >>> >>> How it was going [Georgi Gerganov]: >>> >>> So a few days later comes out the LLaMA, I do >>> some calculations and I figure out “Okay, 65 >>> billion parameters. You probably need about >>> 40 gigs of RAM, with 4-bit quantization. So >>> this can run on a MacBook. Why not do it?” >>> >>> Why I was able to do it so quickly - basically, >>> for all that I saw it’s pretty much GPT-J architecture >>> with some modifications, like some extra memorization >>> layers. It’s minor changes. Basically, again, the >>> existing code for the GPT-J, I just simply >>> modified it there, it happened pretty quickly. >>> https://changelog.com/podcast/532 >>> >>> Georgi Gerganov, Bulgarian, now with Hugging >>> Face, ggml-cann also running on Chinese AI chips. >>> ggml Manifesto https://github.com/ggml-org/ggml >>> >>> Bye >>> >> >
[toc] | [prev] | [standalone]
Page 5 of 5 — ← Prev page 1 2 3 4 [5]
Back to top | Article view | sci.physics.relativity
csiph-web