touisteur

Moreover, informing about the actual realities of export restrictions doesn't seem useless or condescending (I'm not commenting about the actual article here, which... seems hacking in technical savvy ... to be generous - but news about the topic in general).

Supply chains are complex, rife for corruption at many levels and giving some understanding how it happens seems a worthy news item for me, especially to understand the effectiveness of such sanctions, and as an educated citizen, either clamor for tighter supply-chain control, or be OK with the status quo, or ask for less control or for alternatives.

Another way to see this is NVIDIA invested heavily, heavily in making a language and its compiler useable (for some level of useable.......) on enthusiast-affordable and poor-scrappy-researcher-affordable hardware on a complex weird-to-program-from-asm architecture and provided good-enough performance (and showcasing lots of very good optimizations) on standard HPC libraries and a bit of HW/SW co-design and kept at it for a long long time... with some level of backwards compatibility too. The recipe is here.

I don't know about Iraq, but residential buildings are routinely bombed on Ukraine, big cities, not sure one can draw parallels here. Spending nights and even being schooled in bomb shelters will take its toll, but from meeting many Ukrainian refugees, many will make light of it, very often wary of complaining while their friends, family, countrypeople, still in-country are still waking up to sirens and explosions.

The Ukranian utilities people are relentless, resourceful heroes, keeping power on, under the conditions they're facing.

Wondering who actually pays public price on those AMD chips in big-OEM servers. Not saying the AmpereOne isn't also discount-able too, but these public prices always feel like signalling, to me, more than a reference to the actual price. Or maybe I'm lucky...

Modern low-latency GCs never reach 1ms on all the workloads I've put them through. Mind you I don't GC terabytes of RAM so who knows what happens there.

That and radar... mode S can transmit data packets, uplink and downlink. Never seen it in use but it's in the European Mode S standard so any compliant radar has tested and qualified its use... data rate is very low but enough for some chat.

Yeah they've refrained from going past 2 on all Zen stuff and here we are. What a thing.

With the B100 somehow announced to have lower scalar FP64 throughput than the H100 (did they remove the DP tensor cores ?), one will have to rely on Ozaki schemes (dgemm with int8 tensor cores) and lots of the recent body of work on mixed-precision linear algebra show there's a lot of computing power to be harnessed from Tensor Cores. One of the problems of HPC now is a level of ossification of some codebases (or the lack of availability of porting/coding/optimizing people). You shouldn't have to rewrite everything every 5 years but the hardware constructors go where they go and we still haven't found the right level of abstraction to avoid big porting efforts.

Most of the toolchain got hidden behind openvino and there was no hardware released for years. Keembay was 'next year' for years. I have some code for DSP using it that I can't use anymore. Has Intel actually released new shave cores, with an actual dev environment ? I'm curious.

Next level of this thread is with formal proof (e.g. SPARK). Left to the reader to complete the paragraph.

SHAVE from MOVIDIUS was fun, before Intel bought them out.

Ozaki has been doing fp64 matrix-multiplication using int8 tensor cores

https://arxiv.org/html/2306.11975v4

Interesting AF.

Then I don't understand why they stopped before they even released a second gen, and instead went for nebulous promises of falcon shores, weird archs or maybe some sprinkled Gaudi, or maybe something in some 'soon' future. The Arc boards weren't bad and at least you could start porting to SYCL/oneapi for a good reason (aka there's hardware available...).

I was working some years ago on 'happy path fuzzing', trying to find heuristics to guide through code avoiding all error handling, runtime checks. Never got better results than afl-go or other targeted fuzzing, but you have to know what's your happy path.

Also tried to use previous-version or previous-previous-version coverage ('precise' through gcov, or intel processor trace, or sampled perf traces, down until poor-man's-profiler samples) coupled with program repair tools, and... never managed to jump from fun small toy examples to actual 100+kloc applications. Maybe one day.

I've always wondered how people actually get the profiles for Profile-Guided-Optimization. Unit tests probably won't actuate high-performance paths. You'd need a set of performance-stress tests. Is there a write-up on how everyone does it in the wild ?

Time (and the market ?) will tell whether all the people clamoring for NVIDIA alternatives actually put their money on it (understanding that NVIDIA's headstart is a long-term heavy investment on software too: compilers, libraires, and of course hardware/software co-design). I still can't fathom how Intel thought Arc and/or Ponte Vecchio would pay for themselves on day one.

I'm your Man has been rarely covered but someone did a good job of it https://youtu.be/qH06n-5-B70

There's also "A Thousand Kisses Deep" and its rare cover by Jackson Browne https://youtu.be/w_mV4tdw2wY

And "Dance Me To The End of Love" which has also been beautifully covered.

I'm also very fond of Chelsea Hotel (more touching if you get the context) and Boogie Street.

Most of Cohen isn't very accessible to me as I miss many cultural markers, but they still very often feel appropriate (pop-up) in times of melancholy...

Err. It still does by default change your name (if you're a woman). But you can ask and keep your 'maiden' (ugh) name or have both. It gets a bit trickier for kids' family names...

It's a matter of degrees. Absolute trust is a rare thing, but people have given examples of relative trust. Your car won't break down and you can trust it with your kids' life, almost never challenging its trustworthiness, but still you can do checkups or inspections, because some of the bult-in redundancies might be strained. Trusting aircraft but still doing inspections. Trusting your colleagues to do their best but still doing reviews because every fucks up once in a while.

The idea of trusting a next-token-predictor (jesting here) is akin to trusting your System 1 - there's a degree to find where you force yourself to enable System 2 and correct biases.

Yes of course. Although my two DTs and tapering off the bottle gave me a new appreciation to this movie. Reality can be far worse than Aronofsky's worst.

Isn't it part of the plot of Requiem for a Dream ?

Heavy use of unifdef there. Following their progress was amazing from an 'let's own this thing now' perspective, seeing the size of code changes, how they decided which things had to go.

I would have been bragging too, were I removing ifdefs by the shovelload every day for weeks.

As for zen3 and zen4, it's been a pleasure to work with good vectorization perf, and it seems they've been doing avx512 'right', at least for HPC I know.

Add to that the 12 channels of DDR per socket and the quite generous cache, and once you've swallowed the need to buy 24 sticks for 2 sockets, these things are beautiful beasts, at what Intel has taught us, a very reasonable price (ask you OEM for prices, public prices on CPUs are nuts).

Not really any paradigm shift since the introduction of Tensor Cores in NVIDIA archs. Anything Ampere or Lovelace, will do to teach yourself CUDA up to the crazy optimization techniques and the worst libraries that warp the mind. You'll only miss on HBM which allows you to cheat on memory bandwidth, amount of VRAM (teach yourself on smaller models...), double precision perf and double precision tensor cores (go for an A30 then and not sure they'll keep them - either the x30 bin, or DP tensor cores - ever since "DGEMM on Integer Matrix Multiplication Unit" - https://arxiv.org/html/2306.11975v4 ). FP4, DPX, TMA, GPUDirect are nice but you must be pretty far out already for them to be mandatory...

As for everyone getting the same thing wrong being unlikely, seeing how much 'inspiration' there is in the space, especially anything helping the race to the bottom on price... it's not unseen, as in a whole generation of motherboards caps blowing or leaking, bad TPM implementations... not saying your overall point is wrong but this part I don't know

But GP's explanation still holds. The PSU will give the reference voltage, while being less and less efficient. Unable to deliver proper voltage (because it becomes less efficient with time), will not deliver any, until you remove the drives and it can deliver it? Did I misunderstand something ?

Now I'm thinking of these HPE over-over-over-specced PSUs I run and y'all got me worried.

Very common for me to write shell scripts that generate shell scripts to run.

Instead of creating gnuplot scripts to loaded, sometimes a good sequence of echo "set terminal png 800,600; plot..." | gnuplot > my.png

or a very long imagemagick drawing command or a simple svg generated then converted to png then assembled (thank montage), can do the job.

The whole advantage of generating intermediate shell scripts, is to replay one part specifically to debug a small part of a script pipeline.

And it worked mainly because they were a drop-in for Intel processors. Which was and is an amazing feat. I and most people could and can run anything compiled (but avx512 stuff back then on zen1 and 2 ?) without a hitch. And it was still a huge uphill battle and Intel let it happen, what with their bungling of the 10nm process.

I don't see how the same can work here. HIP isn't it right now (every time I try, anyway).

It might rather be because of who is (was) doing the HF of RLFH, or where most big moat builders went https://amp.theguardian.com/technology/2024/apr/16/techscape...