shaklee3

45 ms·

601.
by shaklee3 6y ago
It's happened to me on a GPU. I had ecc disabled for a speed increase, and got a reproducible bit error.
602.
by shaklee3 6y ago
His books are excellent, especially the latest bpf one. Highly recommended.
603.
by shaklee3 6y ago
For what it's worth, your name alone has become quite famous in the performance world, and that's due to your blog posts and books. I would be a bit shocked if an interviewer for that field didn't know you.
604.
by shaklee3 6y ago
In that case, why would you try to develop/compile remotely? Just target a different architecture.
605.
by shaklee3 6y ago
The problem with power 9, and likely 10, is that it was far too expensive and hard to get. Even two years after it was announced you had to talk to IBM salespeople, and the server was about 50% more expensive than an equivalent Intel. For a
606.
by shaklee3 6y ago
All the time. You'll frequently see front page titles like "My company (YC19) Hiring Engineers"
607.
by shaklee3 6y ago
Are you on wsl1 or 2?
608.
by shaklee3 6y ago
Sorry, that was not my intent. I was pointing out that Nvidia has made significant contributions already to the c++ standard, so this is not the first thing they've done.
609.
by shaklee3 6y ago
You're right, sorry. Mixing up terms.
610.
by shaklee3 6y ago
They have about the same amount of verboseness in my experience.
611.
by shaklee3 6y ago
You have been able to do that for a long time with UVA.
612.
by shaklee3 6y ago
Nvidia has had many members on the c++ standards committee for a while.
613.
by shaklee3 6y ago
Gpudirect is only on the data center cards afaik.
614.
by shaklee3 6y ago
This should have the 2019 tag in the title
615.
by shaklee3 6y ago
When I can run a TPU in my own data center, there is. Until then it precludes a lot of applications.
616.
by shaklee3 6y ago
Let's just leave this one alone then. I can't argue with someone who claims anything is possible, yet absolutely nobody seems to be doing what you're referring to (except you). A100 now tops all MLPerf benchmarks, and the una
617.
by shaklee3 6y ago
I gave you an example of something you can't do, which is an overlap-save FFT, and you ignored that completely. Please implement it, or show me any example of someone implementing any custom FFT that's not a simple, standard, batc
618.
by shaklee3 6y ago
Well, there you go. For one TensorFlow is not a generic framework like cuda is, so you lose a whole bunch of the configurability you have with cuda. So, for example, even though there is an FFT raw function, there doesn't appear to be
619.
by shaklee3 6y ago
Please show me the API where I can write a generic function on a TPU. I'm talking about writing something like a custom reduction or a peak search, not offloading a tensor flow model. I'll make it easier for you, directly from Goo
620.
by shaklee3 6y ago
A TPU is a chip you cannot program. It's purpose built and can't run the fraction of the type of workloads that a GPU can.
621.
by shaklee3 6y ago
There's a lot of truth in it, but for a slightly different reason. Almost everyone I work with knows or writes C regularly. They're usually very senior people, and not once have I ever heard them talk about a blog, let alone write
622.
by shaklee3 6y ago
I can definitely takes the cake for an easy to use syntax for doing all kinds of linear algebra. It's only able to do that because the extensive operator overloading that C++ allows. However, don't want that for you for a speed. U
623.
by shaklee3 6y ago
Totally agree. Look at something like asyncio and how many things have been improved or deprecated since 3.5. it makes it really hard to keep up. Sure, if you want to stick to basic python it's certainly easier to learn than C++, but t
624.
by shaklee3 6y ago
I think the two major c++ documentation websites are far better than the official Python documentation. They have more examples, the different variants are clear across which version of the language you are using, and I never need to go els
625.
by shaklee3 6y ago
Excellent way to phrase it. For lock of a better way of saying it, you don't know what you don't know. You can become comfortable with certain things, but the pace at which stuff moves in this industry makes you stale if you don&#
626.
by shaklee3 6y ago
Of course I researched the decision outside of HN. But just the superficial exposure you have to so many different things that you wouldn't have by just browsing around is invaluable. And just by hearing something like a name can take
627.
by shaklee3 6y ago
I have the complete opposite feeling. As long as you filter yourself fairly well on which topics you read, HN can be immensely useful. You can learn about new things that can help you a lot in your career, that you otherwise may have missed
628.
by shaklee3 6y ago
Do you know if this applies to epyc as well?
629.
by shaklee3 6y ago
Huh? Mellanox (now part of Nvidia) has a massive data center presence, and it's growing their NICs are leaps and bounds above Intel's.
630.
by shaklee3 6y ago
https://www.google.com/amp/s/www.history.com/.amp/news/6-thi...

More ›