apparent memory increase

apparent memory increase

[Tom Tromey]

You may have seen this warning from the memory consumption tester:

http://gcc.gnu.org/ml/gcc-regression/2008-05/msg00041.html

... related to the recent identifier GC patch.

I looked into this a little.  My theory is that this is an artifact of
how the tester collects its data.  In particular I suspect the tester
was not including the identifier data in its memory stats -- but now,
because identifiers are allocated via the GC, they are included.

If this theory is in error, let me know and I will investigate some
more.  In this case, it would be helpful to have the script that
generates the report.

Tom

Re: apparent memory increase

[Richard Guenther]

On Thu, May 22, 2008 at 1:04 AM, Tom Tromey <tromey@redhat.com> wrote:
> You may have seen this warning from the memory consumption tester:
>
> http://gcc.gnu.org/ml/gcc-regression/2008-05/msg00041.html
>
> ... related to the recent identifier GC patch.
>
> I looked into this a little.  My theory is that this is an artifact of
> how the tester collects its data.  In particular I suspect the tester
> was not including the identifier data in its memory stats -- but now,
> because identifiers are allocated via the GC, they are included.
>
> If this theory is in error, let me know and I will investigate some
> more.  In this case, it would be helpful to have the script that
> generates the report.

For big testcases I actually see a consistend reduction in peak
overall memory usage
(the number if you would look at 'top').  For small testcases I can
see both ups and downs,
probably because of changes in the number and points of GC collections
(but I guess
that is expected).

Richard.

Re: apparent memory increase

[Tom Tromey]

>>>>> "Richard" == Richard Guenther <richard.guenther@gmail.com> writes:

Richard> For big testcases I actually see a consistend reduction in
Richard> peak overall memory usage (the number if you would look at
Richard> 'top').  For small testcases I can see both ups and downs,
Richard> probably because of changes in the number and points of GC
Richard> collections (but I guess that is expected).

If you are looking with top, I suppose noise for smaller test cases
could come from differences in the overhead imposed by the GC versus
an obstack.  Also the new scheme wastes some amount of memory per
identifier -- on average, half the minimum allocation size.

Tom

Re: apparent memory increase

[Jan Hubicka]

> You may have seen this warning from the memory consumption tester:
> 
> http://gcc.gnu.org/ml/gcc-regression/2008-05/msg00041.html
> 
> ... related to the recent identifier GC patch.
> 
> I looked into this a little.  My theory is that this is an artifact of
> how the tester collects its data.  In particular I suspect the tester
> was not including the identifier data in its memory stats -- but now,
> because identifiers are allocated via the GC, they are included.

This is my understanding too, that is why I did not reply to that mail.
SBRK memory didn't increased, so we just moved stuff into GGC from
memory we didn't track at all.

Honza
> 
> If this theory is in error, let me know and I will investigate some
> more.  In this case, it would be helpful to have the script that
> generates the report.
> 
> Tom