Paul Bone
2016-12-23 12:27:49 UTC
Hi,
I'm doing some heap utilisation profiling and noticed that in .gcd grades
allocations (cells) in the heap appear to be larger than in a regular gc
grade. But when I edit boehm_gc/Makefile.direct and comment out the
-DNO_DEBUGGING line (to get a collector with debugging support in a regular
"gc" grade) this doesn't happen.
As I understand the only things that should increase the size of heap cells
are term size profiling and memory attribution profiling. Why would this
also happen in a grade such as hlc.par.gcd? I've searched and cannot find
what might be causing this.
Any ideas?
What I'm trying to do is make the gc_dump code i boehm_gc available. It is
made available when the NO_DEBUGGING macro is *not* defined, which is
controlled by boehm_gc/Makefile.direct, it is defined by default and not
defined in gcd grades. I can comment this out manually as I have done, but
that's less convenient because then anyone else who will use my scripts
(which I will publish) will need to do the same and rebuild Mercury.
Thanks.
I'm doing some heap utilisation profiling and noticed that in .gcd grades
allocations (cells) in the heap appear to be larger than in a regular gc
grade. But when I edit boehm_gc/Makefile.direct and comment out the
-DNO_DEBUGGING line (to get a collector with debugging support in a regular
"gc" grade) this doesn't happen.
As I understand the only things that should increase the size of heap cells
are term size profiling and memory attribution profiling. Why would this
also happen in a grade such as hlc.par.gcd? I've searched and cannot find
what might be causing this.
Any ideas?
What I'm trying to do is make the gc_dump code i boehm_gc available. It is
made available when the NO_DEBUGGING macro is *not* defined, which is
controlled by boehm_gc/Makefile.direct, it is defined by default and not
defined in gcd grades. I can comment this out manually as I have done, but
that's less convenient because then anyone else who will use my scripts
(which I will publish) will need to do the same and rebuild Mercury.
Thanks.
--
Paul Bone
http://paul.bone.id.au
Paul Bone
http://paul.bone.id.au