On Sunday, 5 May 2013 at 01:55:36 UTC, bearophile wrote:
> Jonathan M Davis:
>
>> So, it's
>> probably only a matter of time before we have a more precise garbage collector
>> in druntime (which will seriously reduce the false pointers problem, though I
>> can't remember if it completely eliminates it).
>
> Probably it doesn't use things like a shadow stack, so it doesn't trace stack pointers well. So probably it's not fully precise.
>

It is conservative with stack and unions. So it isn't fully precise. I do think the tradeoff that is made here is the right one.

On Sat, May 4, 2013 at 12:41 AM, Rob T <alanb@ucora.com> wrote:

> I have significantly reduced CG performance issues (up to 3x faster) with strategically placed disabling and enabling of the GC, so it may be that the problem can be solved using carefully placed GC.disable and GC.enable calls at points in the compilation process. This is not the best solution, but it could be a quick temporary fix until the GC problem is dealt with directly.


Hello Rob

I am totally stuck with my compile time parsing memory issues at this point. I have 8GB RAM on my machine and the compiler goes out of memory even when parsing a small block. So I am re-opening this old thread. Is it possible for you to share your work with me so that I can experiment a little bit.

Thanks and Regards
- Puneet