[PATCH] [Core] Add parallel infrastructure to lld.

[Alex Rosenberg]

We're in the process of putting together a large internal buildbot farm that will include tests almost exactly like the ones Sean has listed.

I'd love for lab to also be doing some LNTs here.

I'd also suggest some benchmarking of small resulting executables, as per Sean's point about parallel_sort.

Alex

On Apr 9, 2013, at 5:54 PM, Sean Silva wrote:

> 
>  On a side note: now that it seems that LLD is starting to get into a phase of its life where it is being optimized, we *really* should have bots that test for performance regressions. Otherwise, we have no real baseline and can't tell whether things are actually getting faster or not.
> 
>  We probably want at least:
>  - Release build of LLVM+Clang
>  - Debug build of LLVM+Clang
>  - Release build of Chromium (or some other gargantuan project)
>  - Debug build of Chromium
> 
>  on linux, mac, windows. Especially now that system dependent parallel stuff is being added, it is important to have a common baseline for meaningful progress to be made (commits benefitting one platform may adversely affect another). It shouldn't take much computing power since the link step does not take very long (compared to compiling), so it would be sufficient to just stash a "golden" set of object files and run just the actual link step.
> 
> http://llvm-reviews.chandlerc.com/D649
> _______________________________________________
> llvm-commits mailing list
> llvm-commits at cs.uiuc.edu
> http://lists.cs.uiuc.edu/mailman/listinfo/llvm-commits