menaerus

39 ms·

781.
by menaerus 2y ago
Under what circumstances can you cut down your LOADS and STORE from and to main memory by an order of magnitude without observing major improvements in algorithm runtime that is memory-bound?
782.
by menaerus 2y ago
I skimmed through the paper real quickly. There's no performance data on inference speedups in the paper. Only the benchmarks relevant for training. They also, interestingly, don't compare against the flash-attention. Flash-attent
783.
by menaerus 2y ago
> It's usually the case that memory access is also slower than matrix multiplications so this will be faster. Burning flops to save memory movement. I haven't read this paper (yet) but isn't this the case that mostly appli
784.
by menaerus 2y ago
AST representation of 6k LoC could probably fit for the most of its part into the L3 cache of modern CPUs. C and C++ build process is mostly a memory-bound problem.
785.
by menaerus 2y ago
> If performance of o3 doesn't meet expectations, there'll be plenty of people making excuses for it I agree and I can definitely see that happening but it is also not impossible, given the incentive and impact of this technolo
786.
by menaerus 2y ago
Perhaps what he meant is that the public will be able to benchmark the model themselves by throwing different difficulty math problems at it and not necessarily the FrontierMath benchmark. It should become pretty obvious if they were faking
787.
by menaerus 2y ago
"The grant covers personal living, material, and coaching expenses over 12 months, allowing founders to focus on developing their founding idea. While graduates receive personal funding of 2.500€, students can receive 1.000€ a month, a
788.
by menaerus 2y ago
"Here is what I want from a future debugger when working on a hard to debug problem: It should be able to show me the source code It should be able to show me the IR It should be able to show me disassembly I want to be able to step th
789.
by menaerus 2y ago
Thanks for the link. A holdout set which is yet to be used to verify the 25% claim. He also says that he doesn't believe that OpenAI would self-sabotage themselves by tricking the internal benchmarking performance since this will get e
790.
by menaerus 2y ago
I don't have anything to like or not to like. I'm not a user of ZFS filesystem. I'm just dismissing your invalid argumentation. Percona's data is nothing about the scale for reasons I already mentioned.
791.
by menaerus 2y ago
I understand but I'm just curious to learn under what conditions.
792.
by menaerus 2y ago
Refuting the "it doesn't scale" argument with a data from a blog that showcases a single workload (TPC-C) with 200G+10tables dataset (small to medium) at 2vCPU (wtf) machine with 16 connections (no thread pool so overprovisio
793.
by menaerus 2y ago
I'd be curious if you measured 50x on a single core implementation or is the algorithm distributed to multiple cores? I ask because you say that the results are similar to Zen4 so this would sorta imply that you run and measure single-
794.
by menaerus 2y ago
You're going a long way not to address reasonably simple questions I had. You're very combative for no obvious reason - I think I had my arguments laid out in the most objective form I could but unfortunately you seem to be very t
795.
by menaerus 2y ago
To prove that a change isn't harmful is still a hypothesis.
796.
by menaerus 2y ago
How do you run thousands of VM's on 192 physical cores? That must be assuming that large percentage of VM's are idling most of the time.
797.
by menaerus 2y ago
Plus assuming that your codebase is clangd parseable. Many aren't out of the box. Plus assuming that you're not reading/modifying code on a (remote) machine where you don't have access to IDEs but simple editors only. Ot
798.
by menaerus 2y ago
Sorry, that's exactly what you said and the reason why we are having this discussion in the first place. I am guilty of being too patient with trolls such as yourself. If you're not a troll, then you're clueless or detached f
799.
by menaerus 2y ago
> They do not. The chip has 105MB L3 cache and they tested on 128MB of memory. This exceeds the size of L3 cache and thus, it is not a proper test of L3 cache. First, you claimed that there was no L3 BW test. Now, I am not even sure if y
800.
by menaerus 2y ago
Parent comment I am responding to is arguing that CPU caches are not that relevant because the CPU for bigger workloads is anyways bottlenecked by the system memory BW. And thus, AVX-512 is irrelevant because it can only provide compute boo
801.
by menaerus 2y ago
> Your 642 GB/s figure should be for a single Golden Cove core Correct. > That is also significantly below your 642 GB/sec prediction. Not exactly the prediction. It's an extract from one of the Chips and Cheese article
802.
by menaerus 2y ago
Not sure I understood you. You think that AVX-512 workload and store-load BW are irrelevant because main system memory (RAM) cannot keep up with the speed of CPU caches?
803.
by menaerus 2y ago
I mean you wouldn't run a database on a GB10 device or cluster of them thereof. GH200 is another story, however, the potential improvement wrt the databases-in-GPUs still falls short of the question if there are enough workloads that a
804.
by menaerus 2y ago
Ok, fair enough, I didn't explain myself very well. What I more specifically meant is that AMD up until zen5 could not (1) drive 2x AVX-512 computations (2) handle 2x AVX-512 memory loads + 1x AVX-512 memory store in the same
805.
by menaerus 2y ago
> AVX512 was that strategy’s zenith and it was a disaster for them since they had to cut clock speeds when AVX-512 operations ran while AMD was able to implement them without any apparent loss in clock speed. AMD up until zen 5 didn&#x27
806.
by menaerus 2y ago
> So, typical database operations, massively parallel in nature like join or filter, would run about that faster. Given workload A how much of the total runtime JOIN or FILTER would take in contrast to the storage engine layer for exampl
807.
by menaerus 2y ago
Recent report says there are 1M paying customers. At ~30USD for 12 months this is ~3.6B of revenue which kinda matches their reported figures. So to break even at their ~5B costs assuming that they need no further major investment in infras
808.
by menaerus 2y ago
Because under general case you cannot read from concurrent hashmap unless you make sure that write is not taking place. And to do that we need locks, either exclusive (mutex) or more fine grained locks such as rw-lock.
809.
by menaerus 2y ago
Yes, I mean if you have a hashmap for read-only workload where insertion is not going to take place after initial build-up then there isn't much to do.
810.
by menaerus 2y ago
Concurrent reads are also an issue if you hold an exclusive lock.

More ›