highway VS JDK

The bf16 dot instruction replaces 6 instructions: https://github.com/google/highway/blob/master/hwy/ops/x86_12...

[0] for those interested in Highway.

It's also mentioned in [1], which starts off

> Today we're sharing open source code that can sort arrays of numbers about ten times as fast as the C++ std::sort, and outperforms state of the art architecture-specific algorithms, while being portable across all modern CPU architectures. Below we discuss how we achieved this.

[0] https://github.com/google/highway

[1] https://opensource.googleblog.com/2022/06/Vectorized%20and%2..., which has an associated paper at https://arxiv.org/pdf/2205.05982.pdf.

Thanks so much!

Everyone working on this self-selected into contributing, so I think of it less as my team than ... a team?

Specifically want to call out: Jan Wassenberg (author of https://github.com/google/highway) and I started gemma.cpp as a small project just a few months ago + Phil Culliton, Dan Zheng, and Paul Chang + of course the GDM Gemma team.

C++ users can enjoy Highway [1].

[1] https://github.com/google/highway/

At that point it is better to have some kind of DSL that should not be in the main language, because it would target a much lower level than a typical program. The best effort I've seen in this scene was Google's Highway [1] (not to be confused with HighwayHash) and I even once attempted to recreate it in Rust, but it is still distanced from my ideal.

[1] https://github.com/google/highway

Interesting, thanks for sharing :)

At the time we open-sourced Highway, the standardization process had already started and there were some discussions.

I'm curious why stdlib is the only path you see to default? Compare the activity level of https://github.com/VcDevel/std-simd vs https://github.com/google/highway. As to open-source usage, after years of std::experimental, I see <200 search hits [1], vs >400 for Highway [2], even after excluding several library users.

But that aside, I'm not convinced standardization is the best path for a SIMD library. We and external users extend Highway on a weekly basis as new use cases arise. What if we deferred those changes to 3-monthly meetings, or had to wait for one meeting per WD, CD, (FCD), DIS, (FDIS) stage before it's standardized? Standardization seems more useful for rarely-changing things.

1: https://sourcegraph.com/search?q=context:global+std::experim...

2: https://sourcegraph.com/search?q=context:global+HWY_NAMESPAC...

Thanks for the link. We were previously using GFNI for bit reversal and 8-bit shifts, and I just extended that to our 8-bit BroadcastSignBit (https://github.com/google/highway/pull/1784).

You could study Google's Highway library [1].

[1] https://github.com/google/highway

Runtime feature detection need not be rare nor hard, it's a few dozen lines of boilerplate. You can even write your code just once: see https://github.com/google/highway#examples.

I felt out of the loop, thinking that Zero VM was some kind of new distro for OpenJDK but chasing <https://packages.debian.org/sid/openjdk-22-jre-zero#:~:text=...> to <https://sources.debian.org/src/openjdk-11/11.0.23%2B9-1/debi...> lead me to https://github.com/openjdk/jdk/tree/jdk-22-ga/src/hotspot/cp...

It seems that it's a specific CPU target for the Hotspot JIT for non-mainstream architectures (or for research purposes, as I saw mentioned once)

Completely gutted from the OpenJDK, last I checked. See here for the culprit PR: https://github.com/openjdk/jdk/pull/18688

> Yes, they're changing one aspect of signal handler use to work around this problem. They're not stopping the use of signal handlers in general. Hotspot continues to use signals for efficiency in general. See https://github.com/openjdk/jdk/blob/9059727df135dc90311bd476...

This whole thread is about SIGSEGV, and specifically their SIGSEGV handling. However, catching normal signals is not about efficiency.

Some of their exception handling is still odd: There is no reason for a program that receives SIGILL to ever attempt continuing. But others is fine, like catching SIGFPE to just forward an exception to the calling code.

(Sure, you could construct an argument to say that this is for efficiency if you considered the alternative to be implementing floating point in software so that all exceptions exist in user-space, but hardware floating point is the norm and such alternative would be wholly unreasonable.)

> The wonderful thing about choosing not to care about facts is having whatever opinions you want.

I appreciate the irony of you making such statement, proudly thinking that your opinion equals fact, and therefore any other opinion is not.

This discussion is nothing but subjective opinion vs. subjective opinion. Facts are (hopefully, as I can only speak for myself) inputs to both our opinions, but no opinion about "good" or "bad", "nasty" or not can ever be objective. Objective code quality does not exist.

I remember talking to Brendan about the PreserveFramePointer patch during my first months at Netflix in 2015. As of JDK 21, unfortunately it is no longer a general purpose solution for the JVM, because it prevents a fast path being taken for stack thawing for virtual threads: https://github.com/openjdk/jdk/blob/d32ce65781c1d7815a69ceac...

Map doesn't implement the Collection interface.

https://github.com/openjdk/jdk/blob/master/src/java.base/sha...