therealcamino

They weren't "changed" from random because it's not trained, it's a fake model with random weights, is what that post is implying.

This would be interesting to read if not for the LLM prose style.

If it requires physically contiguous RAM to work, then it's not really participating in the full virtual memory system, really. It would be using an exception to it, that can be accommodated to some extent by the OS, but not sitting in demand-paged storage like the rest of the system.

Maybe. The use cases have always been around FP arithmetic over arrays, because that's what is easy to parallelize. I saw a standalone systolic array box attached to a MicroVAX circa 1990. Lots of PIM approaches in the mid-90s, too, but mostly what survived from that era are shared memory NUMA multiprocessors and using GPUs for general purpose computing.

What does "watch your arguments" even mean? It's just "Dont say anything I disagree with" with a threatening tone.

Thanks for this. RISC-V brings out the armchair critics for some reason.

Uh, pretty hard?

I love how there's a section called "The SuperML Take" when the whole thing is obviously written start to finish by Claude or GPT.

I'll save you a click: it'd be bad if you can no longer choose what underlying model to use in Cursor after the SpaceX acquisition.

Not worth the time to read it.

Well, VHDL was heavily based on Ada.

That's certainly an interesting data point, but it was a 90 minute programming contest, not peer-reviewed research.

They weren't exotic, they just weren't part of your writing style

The reason "--" autocorrects to an em dash in practically any word processing software (not talking about browsers) is that that's the accepted way to type it on a typewriter. And you don't need to go into any system settings to enable it. It came in around when things like Smart Quotes came in.

Complexity of the fab processes is isn't what the parent was talking about. They're talking about the major changes in the relationship between fabless semiconductor companies and commercial foundries.

The complexity of actual fabrication was always, and still is, entirely within the foundry. But in the early days of that model, designs could be more easily handed off at the logical level, leaving the physical design to back end companies, which makes designs much more portable between foundries. (The publisher analogy.) What's changed is that the complexity of physical design has exploded, and you can't make the handoff at nearly as high a level, and there is much more work that depends directly on the specific process you are targeting. Much more work at the physical level falls to the fabless semi companies. So it is much more work to retarget a design to a different foundry or process.

The site looks nice, but I think you'd benefit from having some beta testers who are not familiar with HW design already. I mention a few issues below, which I hope you take as constructive suggestions of how to improve the site, and not just as criticism.

It feels a bit like you are expecting completely inexperienced users, but the site doesn't quite anticipate their needs. Loading on mobile was not legible (which is fine, it's an early version.) Switching to the desktop version, the instruction for the initial design to "set" a bit is unclear. What is the function you want the user to implement? It's underspecified.

Further, there needs to be an explanation of what a testbench is, before you present a "Run" button, as well as an explanation as to what stimui the TB will present, and even the entire idea of Verilog simulation. It would be good to have an opportunity to see the testbench code and what the site expects the correct output to be; I didn't see an option for that. New users may not understand waveforms at all without an explanation.

The AI component is fine, but it feels a bit like all of the educational aspect has been delegated to the user to ask questions of the AI. A user who is inexperienced would not even know to use the word "testbench" is to frame their questions. I would suggest some careful thought as to who exactly your target audience is, and specifying early on what prior knowledge you expect them to have.