ComputerGuru

8 ms·

301.
by ComputerGuru 10mo ago
Hardware-accelerated SHA support has a patchy history. I wrote an article some years ago about the prevalence of SHA instructions in x86 in x86_64 CPUs [0], like the current mess we see now with AVX-512, Intel invented something useful then
302.
by ComputerGuru 10mo ago
Thanks for citing your sources, I think your source may be out if date, though! The “free init time hack” was killed in August (unless I’m missing something - never used it myself). https://aws.amazon.com/blogs/compute&
303.
by ComputerGuru 10mo ago
Sounds like a more performant and cleaner solution than TFA.
304.
by ComputerGuru 10mo ago
Good opportunity to plug this folklore legend: https://neosmart.net/forums/threads/an-extended-history-of-t...
305.
by ComputerGuru 10mo ago
Two suggestions: UX on mobile isn’t great. It wasn’t obvious to me where the second model output was and I was thrown off even more so because the option to vote for model 1 output was presented without ever even seeing model two output. Se
306.
by ComputerGuru 10mo ago
“Eye on the left” is different from “the left eye”. First can be ambiguous, second really isn’t.
307.
by ComputerGuru 10mo ago
> its worth a few seconds of extra risk There really isn’t that much increase; when there’s another driver then you both have the combined the light output of both headlights, coming from two different directions.
308.
by ComputerGuru 10mo ago
Interesting thanks for sharing. I can see how that makes sense for switching apps but imho for switching windows of the same app that benefit is negated since the thumbnail (without intense scrutiny) is generally too similar between windows
309.
by ComputerGuru 10mo ago
Nice to see another contender in this space. If OP is here, can you comment on runtime sandboxing and interop support? Can I selectively disable (or intercept) certain features like network support?
310.
by ComputerGuru 10mo ago
Not exactly alt-tab but it’s a ui-less immediate switcher (snappy af, zero latency) to switch between windows of the same app with alt-backtick (next to escape), originally a macOS feature: https://neosmart.net/EasySwitch&#x
311.
by ComputerGuru 10mo ago
I wrote hddrand to write random data and optionally read it back to verify integrity. https://github.com/mqudsi/hddrand
312.
by ComputerGuru 10mo ago
There is only a single sentence about how this works with US-resident data and foreign intelligence or LE apparatus. I'm under no illusion about the willingness of most corporations to hand over everything from subscriber info to conte
313.
by ComputerGuru 11mo ago
My issue with it is that it requires a lot of duplication between it and a traditional rdbms; you can’t use it alone because it doesn’t offer filtering without a search vector (i.e. what some vendors call a scroll function).
314.
by ComputerGuru 11mo ago
I had only heard positive things about pgvector but when you Google comparisons with leading vector dbs you keep getting seo slop from Tiger Data pushing pgvector with very suspicious benchmarks that turned me off it altogether instead ht
315.
by ComputerGuru 11mo ago
Building and linking llvm sucks.
316.
by ComputerGuru 11mo ago
It's not throwing any information away because it can be faithfully reconstructed (via an admittedly arduous process), therefore no entropy has been lost (if you consider the sum of both "input bytes" and "knowledge of u
317.
by ComputerGuru 11mo ago
Most style guides would call that an error, em dash should be used without surrounding spaces (while an en dash requires them). The only publication I know that has (recently?) eschewed that advice is WaPo. If the idea was to make it more v
318.
by ComputerGuru 11mo ago
Money
319.
by ComputerGuru 11mo ago
Yup. The beauty of it is that the underlying ai accelerator/hardware is completely abstracted away. There’s a CoreML ONNX execution provider, though I haven’t used it. No more fighting with hardcoded cuda:0 everywhere. The only pain po
320.
by ComputerGuru 11mo ago
uv doesn’t fundamentally solve the issues. It didn’t invent venv or pip. What fundamentally solves the issue is to use an onnx version of the model.
321.
by ComputerGuru 11mo ago
No. It depends on the object.
322.
by ComputerGuru 11mo ago
dxdiag.exe
323.
by ComputerGuru 11mo ago
We can see right through you.
324.
by ComputerGuru 11mo ago
Wow, I used this when you first posted a public link to it and I see it’s come a long way since; I should definitely check it out again! Congrats on the success thus far!
325.
by ComputerGuru 11mo ago
New AOT C# is nice, but not fully doable with the most common dependencies. It addresses a lot of the old issues (size, bloat, startup latency, etc)
326.
by ComputerGuru 11mo ago
Are these custom boards or just mini pcie network cards you could swap out?
327.
by ComputerGuru 11mo ago
Unless they’re now doing iPhone-level parts locking, getting the right keyboard would be an eBay purchase away.
328.
by ComputerGuru 11mo ago
Probably a better link: https://msrc.microsoft.com/update-guide/vulnerability/CVE-20...
329.
by ComputerGuru 11mo ago
Each H100 can do 60 TFLOPS of f32 operations, while a single RTX 3080 can do roughly half that (just under 30). So complete back-of-the-envelope answer would be 16x as long (since nanochat is targeting four hours with 8xH100) 64 hours isn’t
330.
by ComputerGuru 11mo ago
If you’re not writing/modifying the model itself but only training, fine tuning, and inferencing, ONNX now supports these with basically any backend execution provider without needing to get into dependency version hell.

More ›