Will Killian's Avatar

Will Killian

@willkill07.bsky.social

HPC / Runtimes / Programming Models / C++ The views and opinions expressed in this account are those of my own and do not represent those of my employer, NVIDIA.

165 Followers  |  69 Following  |  37 Posts  |  Joined: 23.11.2023  |  2.1461

Latest posts by willkill07.bsky.social on Bluesky

Matt Godbolt’s blog

This December, I'll be posting an article & video each day until Christmas in the Advent of Compiler Optimisations! #AoCO2025
Each day we'll explore a fun optimisation in C or C++; some low-level, x86 or ARM-specific, some high-level. Hope you'll join me!
YT: youtube.com/mattgodbolt
Blog: xania.org

13.11.2025 19:51 β€” πŸ‘ 50    πŸ” 13    πŸ’¬ 2    πŸ“Œ 0
Preview
Sponsor @hpc-social on GitHub Sponsors The hpc-social resources provide an open platform for community engagement, including a variety of open source work on code and resources hosted on the hpc.social domain and associated community ca...

P.S.: If you'd like to support our open-source projects including our jobs board, Slack/Discord/Mastodon, community map, weekly online chat, podcasts, community syndicated blog, discussion area, surveys, please click on one of the links below: github.com/sponsors/hpc... or github.com/sponsors/hpc...

11.11.2025 23:32 β€” πŸ‘ 1    πŸ” 2    πŸ’¬ 0    πŸ“Œ 0
Preview
Welcome to HPC.social! A welcome to our high performance computing community!

Dear all #HPC and #AI/#Quantum/#RCD and #Supercomputing folks and friends: It's just 3 years since we launched the survey.hpc.social/results/hpc-... that led to the creation of the multi-faceted set of resources of the hpc.social website and tools. This is just to say THANK YOU and see you at #SC25!

11.11.2025 23:19 β€” πŸ‘ 2    πŸ” 2    πŸ’¬ 0    πŸ“Œ 0

Fun fact: prices of DDR5 are at a near-high right now and over 50% more expensive than a few months ago :(

08.11.2025 02:02 β€” πŸ‘ 1    πŸ” 0    πŸ’¬ 1    πŸ“Œ 0
Preview
Accelerating Federal Permitting of Data Center Infrastructure By the authority vested in me as President by the Constitution and the laws of the United States of America, it is hereby ordered: Sectionβ€― 1. Β Policy and

@glennklockwood.com ah, there is this:

www.whitehouse.gov/presidential...

28.10.2025 22:39 β€” πŸ‘ 0    πŸ” 0    πŸ’¬ 1    πŸ“Œ 0
Preview
Energy Department Announces New Public-Private Partnership Model, Two Supercomputers, to Accelerate American Dominance in Science and Technology The U.S. Department of Energy today announced two new AMD-accelerated artificial intelligence supercomputers at Oak Ridge National Laboratory, one of which will be built at record speeds thanks to a n...

I just get pointed to the mention of DOE’s new public-private partnership model. But I’m failing to find out more information. Same thing with the new 100K GPU NVIDIA system at ALCF.

www.energy.gov/articles/ene....

28.10.2025 20:13 β€” πŸ‘ 0    πŸ” 0    πŸ’¬ 1    πŸ“Œ 0

I thought Lux was the OCI partner system. Discovery is OLCF-6 and doesn’t include Oracle.

28.10.2025 20:06 β€” πŸ‘ 0    πŸ” 0    πŸ’¬ 0    πŸ“Œ 0

I’ve found every thunderbolt cable that I’ve used has a clearly marked ⚑️logo on the ends.

Granted, my sample size is 3 and I believe all came from Apple.

20.08.2025 02:47 β€” πŸ‘ 1    πŸ” 0    πŸ’¬ 0    πŸ“Œ 0
Preview
Ignite Your SC25 Journey: HPC Crash Course & Student Programming β€’ SC25 Designed to ignite curiosity, spark innovation, and fuel lasting connections, Student Programming can empower your week at SC25.

#SC25 will open with a 2-day #HPC Crash Course based on OLCF training at ORNL.

The course is free with Student Program registration, but spots are limited.

sc25.supercomputing.org/2025/08/igni...

19.08.2025 18:44 β€” πŸ‘ 3    πŸ” 2    πŸ’¬ 0    πŸ“Œ 0

It’s a wonderful game… So many expansions and high replay value, too!

20.08.2025 02:43 β€” πŸ‘ 2    πŸ” 0    πŸ’¬ 0    πŸ“Œ 0

I’m increasingly convinced by an argument I saw recently that LLMs are effectively power tools (apologies for not remembering the cite!):

- Incredibly useful for experts in a field that are able to correct for their errors

- Really dangerous for layfolk who don’t know what they don’t know

12.08.2025 02:13 β€” πŸ‘ 5    πŸ” 2    πŸ’¬ 3    πŸ“Œ 0
August 1, 2025
AMERICAN ECONOMIC ASSOCIATION
Committee on Economic Statistics and Committee on Government Relations


Statement from the American Economic Association on the
Dismissal of the BLS Commissioner

Leaders of the American Economic Association express their grave concern over the dismissal of the Commissioner of the Bureau of Labor Statistics (BLS) earlier today.
The independence of the federal statistical agencies is essential to the proper functioning of a modern economy. Accurate, timely, and impartial statistics are the foundation upon which households, businesses, and policymakers make critical decisions. Undermining the independence or credibility of these agencies threatens the integrity of the information that markets, institutions, and the public rely on every day.
Measuring the vast and dynamic U.S. economy in real time is inherently challenging. It is standard practice for statistical estimates to be revised as more complete and higher-quality data become available. These revisions reflect the commitment of statistical agencies to accuracy, transparency, and methodological rigor-not failure or bias.
The BLS has long had a well-deserved reputation for professional excellence and nonpartisan integrity.
Safeguarding this tradition is vital for the continued health of the U.S. economy and public trust in our institutions.
We call upon elected officials to respect and preserve the independence of the nation's statistical infrastructure.

Lawrence Katz
President, American Economic Association
Katharine Abraham
President-Elect, American Economic Association
Karen Dynan
Chair, American Economic Association Committee on Economic Statistics
Kenneth Troske
Chair, American Economic Association Committee on Government Relations

August 1, 2025 AMERICAN ECONOMIC ASSOCIATION Committee on Economic Statistics and Committee on Government Relations Statement from the American Economic Association on the Dismissal of the BLS Commissioner Leaders of the American Economic Association express their grave concern over the dismissal of the Commissioner of the Bureau of Labor Statistics (BLS) earlier today. The independence of the federal statistical agencies is essential to the proper functioning of a modern economy. Accurate, timely, and impartial statistics are the foundation upon which households, businesses, and policymakers make critical decisions. Undermining the independence or credibility of these agencies threatens the integrity of the information that markets, institutions, and the public rely on every day. Measuring the vast and dynamic U.S. economy in real time is inherently challenging. It is standard practice for statistical estimates to be revised as more complete and higher-quality data become available. These revisions reflect the commitment of statistical agencies to accuracy, transparency, and methodological rigor-not failure or bias. The BLS has long had a well-deserved reputation for professional excellence and nonpartisan integrity. Safeguarding this tradition is vital for the continued health of the U.S. economy and public trust in our institutions. We call upon elected officials to respect and preserve the independence of the nation's statistical infrastructure. Lawrence Katz President, American Economic Association Katharine Abraham President-Elect, American Economic Association Karen Dynan Chair, American Economic Association Committee on Economic Statistics Kenneth Troske Chair, American Economic Association Committee on Government Relations

Statement from the largest economics association about the BLS firing

As context: AEA approximately never makes such public statements

This is a big deal

02.08.2025 20:13 β€” πŸ‘ 5114    πŸ” 2132    πŸ’¬ 119    πŸ“Œ 114

πŸ’₯Spack v1.0 is out!πŸ’₯

This is a huge milestone. We reworked the core to add compiler dependencies, and we're introducing a stable package API.

πŸš€1.0 also adds concurrent builds, better includes, and much more -- read it all in the release notes!

github.com/spack/spack/...

20.07.2025 10:45 β€” πŸ‘ 41    πŸ” 16    πŸ’¬ 0    πŸ“Œ 5

Why are they called "relaxed" atomics, but whenever I have to use them, I become very stressed?

10.07.2025 15:36 β€” πŸ‘ 52    πŸ” 5    πŸ’¬ 6    πŸ“Œ 0
NSF, NASA and NIH budgets per year, inflation adjusted from 2000-2025 along with the proposed cuts. NSF includes research component only. Massive cuts across all sectors, well below support spanning 25 years.

NSF, NASA and NIH budgets per year, inflation adjusted from 2000-2025 along with the proposed cuts. NSF includes research component only. Massive cuts across all sectors, well below support spanning 25 years.

The President's Budget request as released yesterday will gut scientific research. Why should you care?
1) Science is fundamentally a jobs program. Many 100,000s are employed to do science and work for you, the US taxpayer.

31.05.2025 15:30 β€” πŸ‘ 224    πŸ” 124    πŸ’¬ 5    πŸ“Œ 10
Preview
Announcing the 44 Accepted SC25 Workshops β€’ SC25 SC25 has accepted 33 workshops to which you may submit a paper for possible publication. Eleven workshops are symposium-style.

Announcing the 44 Accepted #SC25 Workshops

o Workshops with Proceedings

o Symposium-style Workshops

sc25.supercomputing.org/2025/05/anno...

#HPC

20.05.2025 19:13 β€” πŸ‘ 5    πŸ” 6    πŸ’¬ 2    πŸ“Œ 0

I want them to be successful because I think something else is needed. But running large apps at scale is always gonna be a bottleneck for any non von Neumann architecture.

06.05.2025 16:35 β€” πŸ‘ 1    πŸ” 0    πŸ’¬ 1    πŸ“Œ 0

- SambaNova’s RDU dataflow execution vs Next Silicon’s Mill execution are very, very different (a priori static vs dynamic)
- For Next Silicon, unlikely/non-parallel can execute on device via RISC β€” which is something that Samba Nova cannot handle at all(?)
- SW stack is completely different

06.05.2025 16:33 β€” πŸ‘ 1    πŸ” 0    πŸ’¬ 0    πŸ“Œ 0

When will they learnnnnnn

02.05.2025 13:55 β€” πŸ‘ 3    πŸ” 0    πŸ’¬ 1    πŸ“Œ 0
A picture of the cover of Sy Brand’s new book titled β€œBuilding a Debugger β€” Write a native x64 debugger from scratch”

A picture of the cover of Sy Brand’s new book titled β€œBuilding a Debugger β€” Write a native x64 debugger from scratch”

@tartanllama.xyz Look at what came in the post yesterday!

09.04.2025 14:30 β€” πŸ‘ 1    πŸ” 0    πŸ’¬ 1    πŸ“Œ 0

repost to upset a C++ programmer:

int main() {
int a = 0xE + 1; // compiles
int b = 0xE+1; // doesn't
}

18.03.2025 15:13 β€” πŸ‘ 146    πŸ” 33    πŸ’¬ 13    πŸ“Œ 2

Yeah, wow that’s incredibly annoying.

I am able to do it with only one additional template instantiation though!

godbolt.org/z/h97xGnq3a

12.03.2025 03:00 β€” πŸ‘ 1    πŸ” 0    πŸ’¬ 0    πŸ“Œ 0

At the end of the day, everything is a data movement problem.

06.03.2025 03:49 β€” πŸ‘ 0    πŸ” 0    πŸ’¬ 0    πŸ“Œ 0
Preview
In This Club, You Must 'Earn the Exa' There have been some recent press releases and headlines with the phrase β€œAI Exascale” in them. Other than flaunting the word exascale or even zettascale, these stories do not provide […]

Don't make me point to the sign

www.hpcwire.com/2024/10/17/i...

07.01.2025 17:17 β€” πŸ‘ 10    πŸ” 2    πŸ’¬ 0    πŸ“Œ 0

The stat and AoC nerd in me wonders how many earned multiples of 50 on Christmas Day over all 10 years.

27.12.2024 13:53 β€” πŸ‘ 0    πŸ” 0    πŸ’¬ 0    πŸ“Œ 0
An edited version of a Spotify Wrapped image that says "Your year in C++". It says I had 7.8k segfaults, that puts me in the top 0.1% of bad programmers worldwide. At the bottom is the C++ logo and a disclaimer that says "This is not real please don't get angry at me"

An edited version of a Spotify Wrapped image that says "Your year in C++". It says I had 7.8k segfaults, that puts me in the top 0.1% of bad programmers worldwide. At the bottom is the C++ logo and a disclaimer that says "This is not real please don't get angry at me"

Damn, I didn't expect C++ Wrapped to go so hard this year

04.12.2024 16:40 β€” πŸ‘ 539    πŸ” 82    πŸ’¬ 12    πŸ“Œ 4

Do you have any papers referencing how passing an overload sets could/would work?

02.12.2024 03:56 β€” πŸ‘ 1    πŸ” 0    πŸ’¬ 1    πŸ“Œ 0

Btw the below is why I find mixed precision #HPC so interesting. There’s potentially *huge* performance upsides, but if you get it wrong, it’s *all* wrong.

29.11.2024 16:40 β€” πŸ‘ 11    πŸ” 2    πŸ’¬ 2    πŸ“Œ 0

You mean beamer? πŸ€”

27.11.2024 19:43 β€” πŸ‘ 2    πŸ” 0    πŸ’¬ 0    πŸ“Œ 0

@willkill07 is following 20 prominent accounts