[m-rev.] For review: Implement support for threadscope profiling.

Paul Bone pbone at csse.unimelb.edu.au
Fri Dec 4 09:20:17 AEDT 2009


On Thu, Dec 03, 2009 at 05:29:20PM +1100, Peter Wang wrote:
> On 2009-12-03, Paul Bone <pbone at csse.unimelb.edu.au> wrote:
> > 
> > This has been commited with the revised changelog:
> 
> It occurred to me that MR_PROFILE_PARALLEL_EXECUTION_SUPPORT should
> imply MR_LL_PARALLEL_CONJ anyway, shouldn't it?

I havn't quite decided - this is bad.

Some of this stuff might be applicable to grades that don't use parallel
conjunction.

At some time I'd like to make parallel conjunction work with GCD for the hlc
backend.  (Maybe we can get another student to do that).  At which point
MR_HL_PARALLEL_CONJ should be defined and MR_PROFILE_PARALLEL_EXECUTION_SUPPORT
might be applicable to both hlc.par and asm_fast.par grades.  This will
probably mean that we won't have a parallel grade that doesn't support parallel
conjunction.

I want to talk to Zoltan about making a new grade with this parallelism
profiling stuff in it.  Maybe it should be called par_prof and one can have
asm_fast.gc.par and asm_fast.gc.par_prof.  Mostly I want to discuss the pros
and cons of this and also learn how to create a new grade.

-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 489 bytes
Desc: Digital signature
URL: <http://lists.mercurylang.org/archives/reviews/attachments/20091204/22fe90ff/attachment.sig>


More information about the reviews mailing list