Zig's New Relationship with LLVM • Hacker News Board

Re: Zig's New Relationship with LLVM

#291

Post by MaulingMonkey » Tue, Sep 29, 2020, 5:20 PM UTC

ndesaulniers wrote: post #254

MaulingMonkey wrote: post #55

Earlier quoted context omitted.

Gold is much better than a 30% speedup IME. I went from >60 seconds per (config x arch x platform x near-leaf binary) combination with bfd, to 6x faster / Additionally, I'll note that gold's incremental linking is in-place binary patching, from what I can tell, and potentially another order-of-magnitude speedup even on your existing C projects - slide 11+ certainly appears to be describing such, and the caveats + res…

Have you tried LLD; LLD beat the crap out of Gold IME.

Not yet on a large enough project that I'd be able to tell. It's on my radar for the next time I find myself in that position though!

Re: Zig's New Relationship with LLVM

#292

Post by pron » Tue, Sep 29, 2020, 9:03 PM UTC

atombender wrote: post #283

pron wrote: post #105

Earlier quoted context omitted.

You are correct, that's still a WIP, but the goal is to have "safe Zig" (i.e. Zig with safety checks on) catch all undefined behaviour modulo explicitly marked unsafe casts (which correspond to unsafe Rust code).

By "catch", do you mean at compile time?

Some at compile-time, and some at runtime.

Re: Zig's New Relationship with LLVM

#293

Post by Guvante » Wed, Sep 30, 2020, 12:23 AM UTC

steveklabnik wrote: post #289

Guvante wrote: post #288

Earlier quoted context omitted.

Yeah rather than supporting backporting things like that they instead decided that every feature should have a lint that shows you how to update your edition. Basically the idea was rather than making it as easy as possible to stay on the oldest version they focused that effort on making it as easy as possible to update once you decided that was what you wanted to do.

We do backport all kinds of things, and we also the vast majority of things land on all editions. This one is tricky for a few reasons that don't apply to most things.

Fair sorry to paint too broad a stroke. Should have made the fact that I was implying "everything" there...

Especially after NLL landed on 2015 ><

Re: Zig's New Relationship with LLVM

#294

Post by steveklabnik » Wed, Sep 30, 2020, 1:55 AM UTC

Guvante wrote: post #293

steveklabnik wrote: post #289

Earlier quoted context omitted.

We do backport all kinds of things, and we also the vast majority of things land on all editions. This one is tricky for a few reasons that don't apply to most things.

Fair sorry to paint too broad a stroke. Should have made the fact that I was implying "everything" there... Especially after NLL landed on 2015 ><

It’s all good :)

Re: Zig's New Relationship with LLVM

#295

Post by butterisgood » Tue, Oct 06, 2020, 12:10 AM UTC

Arelius wrote: post #224

butterisgood wrote: post #29

I’ve often wanted to try out Zig at work but the dependency on the latest LLVM meant doing that first. Building LLVM from scratch on our platform takes more resources (including time) than I normally have available (LLDB needs like 6GB of RAM to link?) So, I welcome this news!

Even if your build system isn't set up for it, you could potentially do a cross-compiled build for just LLVM your target architecture, then just rely on those binaries?

That’s a really interesting idea... Got a new spare time project to try out!