erichkeane wrote:
> > > > At the moment it looks between 1-2%, would there be an acceptable
> > > > increase, or does the artifact size make it just completely untenable?
> > >
> > >
> > > A 1-2% increase in startup time would be unproblematic -- what I'm
> > > actually seeing in local tests is a ~20% increase in wall-time when
> > > running clang on `int main() { return 0; }` (which should be about
> > > representative of `./configure` use cases...) with a dylib release build.
> >
> >
> > Interesting, I am not seeing the 20% increase, but I'm deep into trying to
> > mess with CMake to minimize the startup time. I'd be interested in seeing
> > if the patch I'm working on makes a difference if you'd be able/willing to
> > test it.
>
> I did this one: #232409
>
> It reduces the overhead at load time to a reasonable amount as far as I can
> tell, but I'd love to see if it satisifies your concerns (regarding run-time).
Actually, I THINK I have helped the libclang-cpp.so size significantly. WITH
that other patch:
No ClangiR: 86.4mb
WITH 232409 patch: 127.6mb (~47.7% increase)
WITHOUT it: 159.5mb (~84.6% increase)
So I at least picked up some? Probably not enough for Fedora I suspect, but
hopefully it gets us significantly 'closer'.
https://github.com/llvm/llvm-project/pull/221291
_______________________________________________
cfe-commits mailing list
[email protected]
https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
