|
|
Subscribe / Log in / New account

Soller: Real hardware breakthroughs, and focusing on rustc

Soller: Real hardware breakthroughs, and focusing on rustc

Posted Dec 4, 2019 11:39 UTC (Wed) by jezuch (subscriber, #52988)
In reply to: Soller: Real hardware breakthroughs, and focusing on rustc by Elv13
Parent article: Soller: Real hardware breakthroughs, and focusing on rustc

That's a lot of dead code. Could it be (also?) deduplication of functions independently included into object files and/or of identical template specializations?


to post comments

Soller: Real hardware breakthroughs, and focusing on rustc

Posted Dec 4, 2019 17:49 UTC (Wed) by Elv13 (subscriber, #106198) [Link]

Well, of course it deduplicates more. That's one of LTO main advantage. But overall, this kind of gain is to be expected. With such a good overview of what can and cannot be executed, you can dead-code-elliminate down to the `if` themselves. Then simplify the code further. In that specific case, it was pulling half of Qt and KF5. Yet, it didn't *really* use much of that. In most case, the module/kf5-frameworks were pulled indirectly by each other for a single class/function, which wasn't even relevant to the part of the library I was actually using. This isn't just common. With high level softwares, this is pretty much unavoidable. The Gnome stack isn't different. However, I think LTO has less effect on it because the runtime gobject-introspection will (maybe?) keep some dead code alive. With Python, it is absurdly full of dead code and there is no way around it.


Copyright © 2025, Eklektix, Inc.
Comments and public postings are copyrighted by their creators.
Linux is a registered trademark of Linus Torvalds