Re: more question about porting/gcc

[Andrew Colvin <apc@abcj.demon.co.uk>]

> Content-Type: TEXT/PLAIN; charset=US-ASCII
> 
> On Fri, 20 Mar 1998, Jo Even Skarstein wrote:
> 
> > I can understand that a process is given the largest available block
> > at startup, but is it possible to modify Malloc() so it uses the
> > *smallest* possible block? Theoretically this is just as bad (if not
> > worse) as using the largest block, but if possible it should be tested
> > and see if things improves. 
> 
> But that would be even worse, since it's impossible to prevent some
> fragmentation, and so programs would always get the smallest possible
> amount of memory. If there was a 10MB block and a 100KB block, the
> programs would get only 100K! Really the amount of memory required by a
> program would need to be known before the program is loaded, then the most
> suitably sized block could be used.
> 
I must agree with this. The only lasting solution (where we won't be 
revisiting this again in the future) is to have paged memory - maybe 
we can fix the memory protection in the process and improve the 
virtual memory. I am willing to help with this but openly admit that 
I do not have the expertise.  Do we still have the people out there 
that have, and can spare, the resources.

This thread started about mshrink'ing in time.  Did your starting 
processes have their fastload bit set?  If not, and you have a lot of 
memory, you might as well go and make yourself a cup of tea.  I ran 
into this problem when I got outside and tried running it with mint 
1.08 (I think). My hard disk thrashed until my whole swap partition 
had been zero'd (well a large portion of it anyway).  Admittedly this 
is a lot slower than memory but I did realise pdq that I had to set 
my "bits" appropriately.

Andrew.