Re: PATCH: new GC implementation

[Richard Henderson <rth at cygnus dot com>]

On Wed, Sep 22, 1999 at 10:44:47PM +0200, Andi Kleen wrote:
> I could imagine that it would be a maintenance nightmare though:
> 
> If it was hardcoded there might be the problem that the compiler will
> stop working on a future OS version with changed memory map. 

There's an easy portable way to find a huge contiguous hunk of vma.
You mmap a gigantic hunk of anonymous memory, then get rid of it.

It's kind of a shame unix doesn't have a reserve-vm-area syscall,
but (if necessary) the same effect can be had by leaving that huge
area mapped with PROT_NONE.  It probably won't be necessary though;
I'll have to think about a good solution here...

> > ... for targets with 128-bit types -- ia32 will soon be one of
> > those, in fact.
> 
> For long double ? 

For SSE.

> Sorry I was wrong. I would keep the MAP_ANONYMOUS bit for now,
> until /dev/zero is fixed to not use zeromap_page_range anymore.

We should certainly use MAP_ANONYMOUS when it's available.


r~