Re: VIRT_ADDR_VARIES

[Paul Eggert <eggert@cs.ucla.edu>]

On 11/08/11 00:56, Andreas Schwab wrote:
> Paul Eggert <eggert@cs.ucla.edu> writes:
> 
>> Compiling without VIRT_ADDR_VARIES speeds up aset by 44% on my host
> 
> I cannot reproduce that.  The difference I get is lost in the noise.

This appears to be weirdness on my end -- my host gives reproducible
results when I use those executables, but when I build them again
I get results showing that VIRT_ADDR_VARIES speeds things up, which
can't be right.  I'll try to get to the bottom of this but
for now please ignore my claim about measurable performance difference.

>> Sorry, I don't know the context.  What's the bug here?
> 
> $ valgrind src/temacs 
> [...]
> Segmentation fault

Is that a segmentation fault in valgrind itself, or in temacs?

I don't see why running temacs under valgrind would affect
VIRT_ADDR_VARIES; surely valgrind doesn't change the pure array's
address.

But you're right: if setting VIRT_ADDR_VARIES has no measurable
performance difference and if it lets people debug better, that's
a pretty strong argument for having it always on.