A collection of DIPs

via Digitalmars-d digitalmars-d at puremagic.com
Wed Sep 9 08:09:21 PDT 2015


On Wednesday, 9 September 2015 at 14:00:52 UTC, Brandon Ragland 
wrote:
> Seriously, the fact that the GC has gone un-noticied for so 
> long is a HUGE turn off for just about ANY would be C like

It has not gone un-noticed, that's how @nogc came about. So 
people are aware that it is not suitable for interactive use.

The key focus seems to be to rewrite Phobos so that you don't 
need a GC there at all. That approach I know nothing about...

Unfortunately, how to do C++-style memory management is still not 
obvious to me either. And I am also not happy with the 
library-allocator approach since that will make it more difficult 
for the compiler to optimize memory allocations. I think memory 
handling should be part of the language, not a runtime or library 
issue.

So I very much hope for a language spec that takes a more modern 
approach to address memory management so that we can get 
something that is better than  C++ for interactive applications.

> D could SERIOUSLY use a rewrite or three of the garbage 
> collector.

You actually have to change the language semantics if you want a 
faster garbage collector. You can only scan so many cache-lines 
and unfortunately scanning is limited by total memory that can 
contain pointers and not the size of the GC heap itself... So 
either you need to group pointers on cache-lines and use advanced 
static analysis, limit GC memory access or take the toll of a 
concurrent GC (which would set back progress a couple of years).



More information about the Digitalmars-d mailing list