Hacker News Reader

Comment by iLoveOncall | Hacker News Reader

I feel for the maintainer, but looking at the performance chart in the readme, I think any engineer will have a hard time getting their company to approve paying a license for such a small improvement over LLVM ld.

My big-tech company and my team don't care that build time for a small part of our application is 5 minutes (not a clean build, that would be an hour), so I know they wouldn't pay to save 10 seconds. And linking is not even all of the time to build your application.

From a financial point of view, the fact that the author is in Singapore and has a goal of $10K a month doesn't help.

View on HN

If an engineer at a FAANG company reduced their global link times by even 10% they'd be promoted into early retirement.

Mold is an order of magnitude faster.

The guys that improved dependency download time by up to 80% at my company (Amazon) sure as hell are enjoying the 3% raise they got like everyone else.

I agree with you that Mold is an order of magnitude faster than the competition (sometimes). However in that case the absolute value is more representative than the relative number. Your linking time going from 10 seconds to 1 or 2 seconds is only a tiny improvement.

Nobody only spent 10s on linking. For any reasonable sized binary (in the range of 100 to 200 MiB) it is somewhere around 50s to 60s before mold. People regularly links 1GiB binary for living during their development cycle. If you use small binary or primarily using Go, probably not the target audience of mold.

Otherwise I actually agree mold is amazing but have limited commercialization potential unless author expands the scope (like RAD Game Tools) / take some VC money (much like Emerge Tools)

10s is an extreme link time with lld. lld can link clang in 6s. Mold cuts that in half, but it's still true that you save 3s, not a minute.

I disagree. Our lld link times for the Linux build of our project is over 3 minutes, with LTO disabled.

I sort of want to hear about this project but also I'm sort of afraid to find out. How long does it take with mold?

I work in video games, and all the game projects I've worked on for the last 10 years or so have been multi minute link times. We predominantly iterate on windows with MSVC, and it's usually "server" binaries that are produced for linux.

Afraid I don't know how long it takes with mold, sorry!

If that were true, the author would have retired, because they were a Google engineer when they wrote lld, which is bonkers fast compared to gold, Google's previous linker.

Early retirement and, to use rui's word choice, "earn a comfortable income" are not inconsistent.

But to your credit it does sound that rui saved Google (vastly) more money than they ever paid him.

There are hundreds of people in platforms, language tools, ads, search, and probably everywhere in the company that have saved Google way more than they ever got paid. The thing to understand is that Google's scale is not your scale, as an employee. You make a change that saves the company a million dollars per month. The number is large because the company is large, not because you're amazing.

I think a single employee netting even a single digit percentage in a year's worth of effort in improved performance for all of Google is an astonishing accomplishment. Trivializing a "million dollars per month" improvement just says so much about how little programmers here realize their worth to technology companies, and intellectualize ways to justify it.

Incredibly wrong take.

Yeah, businesses are often cheapskates, and a couple seconds’ worth of speedup aren’t worth any money, especially if you’re removing 10-20 seconds from a multi-hour build pipeline running on dedicated build boxes.

Reply on news.ycombinator.com