News for hackers. Stuff that matters.
Re:Intel CEO: 'The entire industry is motivated to eliminate... by shihab on Thursday December 14, 2023 @03:23PM (#38646746)
But you can't support Pytorch without a proper foundation in place. They don't need to support zillion _end_ libraries, sure, but they do need to have at least a very good set of standard libraries, equivalent of Cublas, Curand etc.

And they don't. My work recently had me working with rocRAND (Rocm's answer to Curand). It was frankly pretty bad- the design, performance (50% slower in places that don't make any sense because generating random numbers is not exact that complicated), and documentation (God it was awful).

Now, that's a small slice of the larger pie. But imagine if this trend continues for other libraries.

Re:Intel CEO: 'The entire industry is motivated to eliminate... by singhrac on Thursday December 14, 2023 @04:11PM (#38647311)
Generating random numbers is a bit complicated! I wrote some of the samplers in Pytorch (probably replaced by now) and some of the underlying pseudo-random algorithms that work correctly in parallel are not exactly easy... running the same PRNG with the same seed on all your cores will produce the same result, which is probably NOT what you want from your API.

But, to be honest, it's not that hard either. I'm surprised their API is 2x slower, Philox is 10 years old now and I don't think there's a licensing fee?

Re:Intel CEO: 'The entire industry is motivated to eliminate... by eternityforest on Thursday December 14, 2023 @06:19PM (#38648811)
I wonder if the next generation chips are going to just have a dedicated hardware RNG per-core if that's an issue?
Re:Intel CEO: 'The entire industry is motivated to eliminate... by lelanthran on Friday December 15, 2023 @09:03AM (#38654247)
Why bother?

It's not the generation that matters so much, it's the gathering of entropy, which comes from peripherals and not possible to generate on-die.

If you don't need cryptographically secure randomness, you still want the entropy for generating the seeds per thread/die/chip.

Re:Intel CEO: 'The entire industry is motivated to eliminate... by eternityforest on Friday December 15, 2023 @02:00PM (#38657356)
It absolutely is possible to generate entropy on-die, assuming you actually want entropy and not just a unique value that gets XORed with the seed, so you can still have repeatable seeds.

Pretty much every chip has an RNG which can be as simple as just a single free running oscillator you sample

Re:Intel CEO: 'The entire industry is motivated to eliminate... by lelanthran on Friday December 15, 2023 @04:38PM (#38659144)
> Pretty much every chip has an RNG which can be as simple as just a single free running oscillator you sample

Every chip may have some sort of noise to sample, but they are nowhere near good sources of entropy.

Entropy is not a binary thing (you either have it or don't), it's a spectrum and entropy gathered on-die is poor entropy.

Look, I concede that my knowledge on this subject is a bit dated, but the last time I checked there were no good sources of entropy on-die for any chip in wide use. All cryptographically secure RNGs depend on a peripheral to grab noise from the environment to mix into the entropy pool.

A free-running oscillator is a very poor source of entropy.

Re:Intel CEO: 'The entire industry is motivated to eliminate... by eternityforest on Friday December 15, 2023 @05:13PM (#38659470)
For non-cryptographic applications, a PRNG like xorshift reseeded by a few bits from an oscillator might be enough.

As I understand it, the reason they don't use on-chip RNGs by themselves isn't due to lack of entropy, it's because people don't trust them not to put a backdoor on the chips or to have some kind of bug.

Intel has https://en.m.wikipedia.org/wiki/RDRAND but almost all chips seem to now have some kind of RNG.

Re:Intel CEO: 'The entire industry is motivated to eliminate... by shihab on Thursday December 14, 2023 @07:20PM (#38649428)
> Generating random numbers is a bit complicated!

I know! I just wrote a whole paper and published a library on this!

But really, perhaps not as much as many from outside might think. The core of a Philox implementation can be around 50 lines of C++ [1], with all the bells and whistles maybe around 300-400. That implementation's performance equals CuRAND's , sometimes even surpasses it! (the API is designed to avoid maintaining any rng states on device memory, something curand forces you to do).

> running the same PRNG with the same seed on all your cores will produce the same result

You're right. Solution here is to utilize multiple generator objects, one per thread, ensuring each produces statistically independent random streams. Some good algorithms (Philox for example), allow you to use any set of unique values as seeds for your threads (e.g. thread id).

[1] https://github.com/msu-sparta/OpenRAND/blob/main/include/ope...

Re:Intel CEO: 'The entire industry is motivated to eliminate... by carterschonwald on Thursday December 14, 2023 @08:34PM (#38650021)
Cool! I’ll have a lookseee. I’ve my own experiments in this space.
Re:Intel CEO: 'The entire industry is motivated to eliminate... by paulmd on Thursday December 14, 2023 @11:28PM (#38650995)
for GPGPU, the better approach is CBRNG like random123.

https://github.com/DEShawResearch/random123

if you accept the principles of encryption, then the bits of the output of crypt(key, message) should be totally uncorrelated to the output of crypt(key, message+1). and this requires no state other than knowing the key and the position in the sequence.

moreover, you can then define the key in relation to your actual data. the mental shift from what you're talking about is that in this model, a PRNG isn't something that belongs to the executing thread. every element can get its own PRNG and keystream. And if you use a contextually-meaningful value for the element key, then you already "know" the key from your existing data. And this significantly improves determinism of the simulation etc because PRNG output is tied to the simulation state, not which thread it happens to be scheduled on.

(note that the property of cryptographic non-correlation is NOT guaranteed across keystreams - (key, counter) is NOT guaranteed to be uncorrelated to (key+1, counter), because that's not how encryption usually is used. with a decent crypto, it should still be very good, but, it's not guaranteed to be attack-resistant/etc. so notionally if you use a different key index for every element, element N isn't guaranteed to be uncorrelated to element N+1 at the same place in the keystream. If this is really important then maybe you want to pass your array indexes through a key-spreading function etc.)

there are several benefits to doing it like this. first off obviously you get a keystream for each element of interest. but also there is no real state per-thread either - the key can be determined by looking at the element, but generating a new value doesn't change the key/keystream. so there is nothing to store and update, and you can have arbitrary numbers of generators used at any given time. Also, since this computation is purely mathematical/"pure function", it doesn't really consume any memory-bandwidth to speak of, and since computation time is usually not the limiting element in GPGPU simulations this effectively makes RNG usage "free". my experience is that this increases performance vs CuRand, even while using less VRAM, even just directly porting the "1 thread = 1 generator" idiom.

Also, by storing "epoch numbers" (each iteration of the sim, etc), or calculating this based on predictions of PRNG consumption ("each iteration uses at most 16 random numbers"), you can fast-forward or rewind the PRNG to arbitrary times, and you can use this to lookahead or lookback on previous events from the keystream, meaning it serves as a massively potent form of compression as well. Why store data in memory and use up your precious VRAM, when you could simply recompute it on-demand from the original part of the original keystream used to generate it in the first place? (assuming proper "object ownership" ofc!) And this actually is pretty much free in performance terms, since it's a "pure function" based on the function parameters, and the GPGPU almost certainly has an excess of computation available.

In the extreme case, you should be able to theoretically "walk" huge parts of the keystream and find specific events you need, even if there is no other reference to what happened at that particular time in the past. Like why not just walk through parts of the keystream until you find the event that matches your target criteria? Remember since this is basically pure math, it's generated on-demand by mathing it out, it's pretty much free, and computation is cheap compared to cache/memory or notarizing.

(ie this is a weird form of "inverted-index searching", analogous to Elastic/Solr's transformers and how this allows a large number of individual transformers (which do their own searching/indexing for each query, which will be generally unindexable operations like fulltext etc) to listen to a single IO stream as blocks are broadcast from the disk in big sequential streaming batches. Instead of SSD batch reads you'd be aiming for computation batch reads from a long range within a keystream.)

Anyway I don't know how much that maps to your particular use-case but that's the best advice I can give. Procedural generation using a rewindable, element-specific keystream is a very potent form of compression, and very cheap. But, even if all you are doing is just avoiding having to store a bunch of CuRand instances in VRAM... that's still an enormous win even if you directly port your existing application to simply use the globalThreadId like it was a CuRand stateful instance being loaded/saved back to VRAM. Like I said, my experience is that because you're changing mutation to computation, this runs faster and also uses less VRAM, it is both smaller and better and probably also statistically better randomness (especially if you choose the "hard" algorithms instead of the "optimized" versions like threefish instead of threefry etc).

That is the reason why you shouldn't do the "just download random numbers", as a sibling comment mentions (probably a joke) - that consumes VRAM, or at least system memory (and pcie bandwidth). and you know what's usually way more available as a resource in most GPGPU applications than VRAM or PCIe bandwidth? pure ALU/FPU computation time.

buddy, everyone has random numbers, they come with the fucking xbox. ;)

Re:Intel CEO: 'The entire industry is motivated to eliminate... by paulmd on Friday December 15, 2023 @12:24AM (#38651277)
thinking this through a little bit, you are launching a series of gradient-descent work tasks, right? taskId is your counter value, weightIdx is your key value (RNG stream). That's how I'd port that. Ideally you want to define some maximum PRNG usage for each stage of the program, which allows you to establish fixed offsets from the epoch value for a given event. Divide your keystream in whatever advantageous way, based on (highly-compressible) epoch counters and event offsets from that value.

in practice, assuming a gradient-descent event needs a lot of random numbers, having one keystream for a single GD event might be too much and that's where key-spreading comes in. if you take the "weightIdx W at GradientDescentIdx G" as the key, you can have a whole global keystream-space for that descent stage. And the key-spreading-function lets you go between your composite key and a practical one.

https://en.wikipedia.org/wiki/Key_derivation_function

(again, like threefry, there is notionally no need for this to be cryptographically secure in most cases, as long as it spreads in ways that your CBRNG crypto algorithm can tolerate without bit-correlation. there is no need to do 2 million rounds here either etc. You should actually pick reasonable parameters here for fast performance, but good enough keyspreading for your needs.)

I've been out of this for a long time, I've been told I'm out of date before and GPGPUs might not behave exactly this way anymore, so please just take it in the spirit it's offered, can't guarantee this is right but I've specifically gazed into the abyss the CuRand situation a decade ago and this was what I managed to come up with. I do feel your pain on the stateful RNG situation, managing state per-execution-thread is awful and destroys simulation reproducibility, and managing a PRNG context for each possible element is often infeasible. What a waste of VRAM and bandwidth and mutation/cache etc.

And I think that cryptographic/pseudo-cryptographic PRNG models are frankly just a much better horse to hook your wagon to than scientific/academic ones, even apart from all the other advantages. Like there's just not any way mersenne twister or w/e is better than threefish, sorry academia

--

edit: Real-world sim programs are usually very low-intensity and have effectively unlimited amounts of compute to spare, they just ride on bandwidth (sort/search or sort/prefix-scan/search algorithms with global scope building blocks often work well).

And tbh that's why tensor is so amazing, it's super effective at math intensity and computational focus, and that's what GPUs do well, augmented by things like sparse models etc. Make your random not-math task into dense or sparse (but optimized) GPGPU math, plus you get a solution (reasonable optimum) to an intractible problem in realtime. The experienced salesman usually finds a reasonable optimum, but we pay him in GEMM/BLAS/Tensor compute time instead of dollars.

Sort/search or sort/prefix-sum/search often works really well in deterministic programs too. Do you ever have a "myGroup[groupIdx].addObj(objIdx) stage? that's a sort and prefix-sum operation right there, and both of those ops run super well on GPGPU.

Re:Intel CEO: 'The entire industry is motivated to eliminate... by lumost on Thursday December 14, 2023 @03:33PM (#38646864)
Folks also underestimate how complex these libraries are. There are dozens of projects to make BLAS alternatives which give up after ~3-6 months when they realize that this project will take years to be successful.
Re:Intel CEO: 'The entire industry is motivated to eliminate... by eternityforest on Thursday December 14, 2023 @06:21PM (#38648832)
How does that work? Why not pick up where the previous team left off instead of everyone starting new ones? Or are they all targeting different backends and hardware?
Re:Intel CEO: 'The entire industry is motivated to eliminate... by haltist on Thursday December 14, 2023 @06:43PM (#38649052)
It's a compiler problem and there is no money in compilers [1]. If someone made an intermediate representation for AI graphs and then wrote a compiler from that intermediate format into whatever backend was the deployment target then they might be able to charge money for support and bug fixes but that would be it. It's not the kind of business anyone wants to be in so there is no good intermediate format and compiler that is platform agnostic.

1: https://tinygrad.org/

Re:Intel CEO: 'The entire industry is motivated to eliminate... by whatshisface on Thursday December 14, 2023 @11:09PM (#38650895)
JAX is a compiler.
Re:Intel CEO: 'The entire industry is motivated to eliminate... by haltist on Friday December 15, 2023 @12:07AM (#38651193)
So are TensorFlow and PyTorch. All AI/ML frameworks have to translate high-level tensor programs into executable artifacts for the given hardware.
Re:Intel CEO: 'The entire industry is motivated to eliminate... by nradov on Thursday December 14, 2023 @06:18PM (#38648794)
Instead of generating pseudorandom numbers you can just download files of them.

https://archive.random.org/

Re:Intel CEO: 'The entire industry is motivated to eliminate... by hcrean on Thursday December 14, 2023 @09:16PM (#38650372)
Or you could just re-use the same number; no one can prove it is not random.

https://xkcd.com/221/

Re:Intel CEO: 'The entire industry is motivated to eliminate... by slavik81 on Thursday December 14, 2023 @07:57PM (#38649714)
If you haven't already, please consider filing issues on the rocrand GitHub repo for the problems you encountered. The rocrand library is being actively developed and your feedback would be valuable for guiding improvements.
Re:Intel CEO: 'The entire industry is motivated to eliminate... by shihab on Friday December 15, 2023 @08:00PM (#38660770)
Appreciate it, will do.
Re:Intel CEO: 'The entire industry is motivated to eliminate... by jart on Friday December 15, 2023 @02:49AM (#38651966)
I honestly don't see why it's so hard. On my project we wrote our own gemm kernels from scratch so llama.cpp didn't need to depend on cublas anymore. Only took a few days and a few hundred lines of code. We had to trade away 5% performance.
Re:Intel CEO: 'The entire industry is motivated to eliminate... by soulbadguy on Friday December 15, 2023 @04:00AM (#38652362)
For a given set kernels, and a limited set of architectures, the problem is relatively easy.

But covering all the important kernels acros all the crazy architecture out there and with relatively good performance and numerical accuracy ... Much harder

Re:Intel CEO: 'The entire industry is motivated to eliminate... by rightbyte on Friday December 15, 2023 @04:13AM (#38652417)
> generating random numbers

You can't bench implementations of random numbers against each other purely on execution speed.

A better algorithm (better statistical properties) will be slower.

Re:Intel CEO: 'The entire industry is motivated to eliminate... by codetrotter on Friday December 15, 2023 @05:32AM (#38652813)
I have the fastest random number generator in the world. And it works in parallel too!

https://i.stack.imgur.com/gFZCK.jpg

Re:Intel CEO: 'The entire industry is motivated to eliminate... by shihab on Friday December 15, 2023 @02:51PM (#38658015)
Yeah. In this instance, I was talking about the same algorithm (phillox), the difference is purely in implementation.