HackerTrans
TopNewTrendsCommentsPastAskShowJobs

charles_irl

no profile record

Submissions

Routing for serverless servers with Pingora, Envoy, and Spanner

modal.com
11 points·by charles_irl·16 giorni fa·0 comments

Speculation Is All You Need

modal.com
12 points·by charles_irl·21 giorni fa·0 comments

Cutting inference cold starts by 40x with LP, FUSE, C/R, and CUDA-checkpoint

modal.com
91 points·by charles_irl·2 mesi fa·18 comments

How to Achieve Serverless GPUs

modal.com
8 points·by charles_irl·2 mesi fa·0 comments

Three types of LLM workloads and how to serve them

modal.com
75 points·by charles_irl·6 mesi fa·5 comments

Host overhead is killing your inference efficiency

modal.com
3 points·by charles_irl·8 mesi fa·0 comments

Quantized Float Exposed

quant.exposed
2 points·by charles_irl·8 mesi fa·1 comments

Against SQL (2021)

scattered-thoughts.net
82 points·by charles_irl·9 mesi fa·77 comments

Length-extension attacks are still a thing

00f.net
2 points·by charles_irl·9 mesi fa·1 comments

The future of Python web services looks GIL-free

blog.baro.dev
3 points·by charles_irl·9 mesi fa·0 comments

Lexical differential highlighting instead of syntax highlighting

wordsandbuttons.online
2 points·by charles_irl·9 mesi fa·0 comments

CReact – JSX for the Cloud

github.com
1 points·by charles_irl·9 mesi fa·0 comments

QUIC and the end of TCP sockets

codemia.io
62 points·by charles_irl·9 mesi fa·82 comments

In C++ modules globally unique module names seem to be unavoidable

nibblestew.blogspot.com
2 points·by charles_irl·9 mesi fa·0 comments

Stupid jj Tricks

andre.arko.net
3 points·by charles_irl·9 mesi fa·0 comments

We reverse-engineered Flash Attention 4

modal.com
5 points·by charles_irl·10 mesi fa·0 comments

A Tour of eBPF in the Linux Kernel: Observability, Security and Networking

lucavall.in
2 points·by charles_irl·10 mesi fa·0 comments

Categorical Foundations for Cute Layouts

research.colfax-intl.com
39 points·by charles_irl·10 mesi fa·6 comments

Pocket Casts, You Altered the Deal, So I Will Alter Your App

blog.matthewbrunelle.com
12 points·by charles_irl·10 mesi fa·3 comments

Modal Notebooks: How we built a cloud GPU notebook that boots in seconds

modal.com
4 points·by charles_irl·10 mesi fa·0 comments

comments

charles_irl
·2 mesi fa·discuss
Very cool!

> Unless you've actually watched tokens stream at those rates, the numbers are hard to internalize. This is the rendering.

I built something similar recently, for the same reason: https://modal.com/llm-almanac/token-timing-simulator.

I like that the output rendering is closer to typical UIs -- syntax highlighting in code mode, tool calls, dim-italic reasoning.

One feature mine has that the author, or anyone else who vibe codes their own version after seeing this, might like to steal is modeling the distribution of output latencies. My implementation is hacky (log-normal roughyl estimated from p50, p90, and p99 values), but still, when you set those to realistic values, it recreates the "jitter" you see in many LLM UIs.

antirez is right that generation tok/s isn't flat as a function of context length, which is a weakness of both simulators.
charles_irl
·2 mesi fa·discuss
To clarify: we do content-based hashing, and when we say "shared bytes aren’t guaranteed to be in the exact same container image layer", what we mean is that

FROM some/image RUN pip install torch==2.7.1

and

FROM another/image RUN pip install torch==2.7.1

will produce images with very high overlap in contents, which will be shared by a content-based cache, but those images' final layers are disjoint from the perspective of a layerwise cache.
charles_irl
·2 mesi fa·discuss
You're absolutely right!
charles_irl
·2 mesi fa·discuss
Yep! That should start in ten seconds or so -- about a second per gigabyte of weights, plus a second to start the container and a few seconds to load the memory snapshot.

There are a few limitations with snapshotting, e.g. it generally fails when using multiple GPUs, which we document here: https://modal.com/docs/guide/memory-snapshots.
charles_irl
·2 mesi fa·discuss
Cutting latencies by 40x! Unfortunately couldn't fit the whole title in the character limit :<
charles_irl
·6 mesi fa·discuss
> Ed Zitron also called out the business model of GPU-as-a-service middleman companies like modal deeply unsustainable, and I also don't see how they can make a profit if they are only reselling public clouds.

You got a link for that? I work on Modal and would be interested in seeing the argument!

We think building a proper software layer for multitenant demand aggregation on top of the public clouds is sufficient value-add to be a sustainable business (cf DBRX and Snowflake).
charles_irl
·6 mesi fa·discuss
Sorry to lead with a bunch of jargon! Wanted to make it obvious that we'd give concrete recommendations instead of palaver.

The technical terms there are later explained and diagrammed, and the recommendations derived from something close to first principles (e.g. roofline analysis).
charles_irl
·6 mesi fa·discuss
oof ty, willfix
charles_irl
·8 mesi fa·discuss
Inspired by https://float.exposed, which was on the front page recently, I put together this visualizer for lower precision/quantized floating point numbers -- specifically, all of the formats in the Open Compute Project's Microscaling Formats Spec (https://www.opencompute.org/documents/ocp-microscaling-forma...).
charles_irl
·10 mesi fa·discuss
Thanks! I think computers are fun and I want reading about them to be fun too.

I was also reminded of HazyResearch's MegaKernels. Didn't want to distract from the main thrust of the post, but definitely think that's a promising approach.
charles_irl
·10 mesi fa·discuss
Hey, one of the authors here!

Reductively, software engineering means taking an idea and mapping it into code. So one form of "reverse" engineering would be taking the code and extracting the ideas. That's what we did here.

Because the source is public, there's quite a lot to work with from the start -- the warp specializations are named and there are helpful comments in many places.

But for many components, we didn't have much. Maybe the clearest case of "reverse engineering" explained in the post is with the cubic approximation for the rational part of the exponentiation. That required staring at some inline assembly and doing math.
charles_irl
·10 mesi fa·discuss
Cool paper! The authors use the fact that the M1 chip supports both ARM's weaker memory consistency model and x86's total order to investigate the performance hit from using the latter, ceteris paribus.

They see an average of 10% degradation on SPEC and show some synthetic benchmarks with a 2x hit.