[Pharo-project] new Cog VMs uploaded

Philippe Marschall kustos at gmx.net
Wed May 4 21:19:10 CEST 2011


On 03.05.2011 23:40, Eliot Miranda wrote:
> 
> 
> On Tue, May 3, 2011 at 11:57 AM, Philippe Marschall <kustos at gmx.net
> <mailto:kustos at gmx.net>> wrote:
> 
>     On 03.05.2011 19:51, Eliot Miranda wrote:
>     >
>     >
>     > On Sun, May 1, 2011 at 4:38 AM, Philippe Marschall
>     <kustos at gmx.net
>     <mailto:kustos at gmx.net>
>     > <mailto:kustos at gmx.net
>     <mailto:kustos at gmx.net>>> wrote:
>     >
>     >     On 29.04.2011 19:03, Eliot Miranda wrote:
>     >     >
>     >     >
>     >     > On Fri, Apr 29, 2011 at 9:04 AM, Philippe Marschall
>     >     <kustos at gmx.net
>     <mailto:kustos at gmx.net>
>     >     <mailto:kustos at gmx.net
>     <mailto:kustos at gmx.net>>
>     >     > <mailto:kustos at gmx.net
>     <mailto:kustos at gmx.net>
>     >     <mailto:kustos at gmx.net
>     <mailto:kustos at gmx.net>>>> wrote:
>     >     >
>     >     >     On 27.04.2011 19 <tel:27.04.2011%2019> <tel:27.04.2011%2019>
>     >     <tel:27.04.2011%2019>:23, Eliot Miranda wrote:
>     >     >     > Philippe,
>     >     >     >
>     >     >     >     can you please profile before and after?  This could
>     >     very well
>     >     >     be to
>     >     >     > do with the additional cost of shallowCopy for
>     contexts, but
>     >     it is
>     >     >     > inevitable that a correct implementation is going to
>     be more
>     >     expensive
>     >     >     > than the simple block-copy.  The new implementation can be
>     >     optimized,
>     >     >     > but we need to profile first to be sure we're tilting
>     at the
>     >     relevant
>     >     >     > windmill.
>     >     >
>     >     >
>     >     >     What kind of profiling do you have in mind, MessageTally?
>     >     >
>     >     >
>     >     > Yes, that would be fine.  Thanks!
>     >
>     >     OK, there you go. As you'll see the runtime is about the same,
>     that's
>     >     because we do more iterations in the same time. Also note the
>     time we're
>     >     spending in WriteStream >> #nextPutAll: with interestingly has a
>     >     different growing behavior than WriteStream >> #nextPut: (only
>     25% vs
>     >     100%). I already raised the issue once but people argued that the
>     >     current code is prefect.
>     >
>     >
>     > Hmmm.  GC behaviour is very different.  Also requestContext Also shows
>     > up in the 2382 profiles but not the 2370 traces.  You'll need to dig a
>     > little deeper to account for these differences.  I guess they
>     could well
>     > be to do with the new (correct) MethodContext copying behaviour
>     > retaining more state and hence stressing the GC more.
> 
>     What approach do you suggest should I take?
> 
> 
> I have no idea.

You understand that when I hear these words from a VM guru I get a bit
worried about the challenges ahead of me.

> You could try and break down the various activities and
> benchmark them separately I suppose.  You could look at the profiles
> you're already generating in higher resolution to see where the time is
> going.  You could also try and extract the SystemProfiler from OpenQwaq,
> which will run on Cog, and is better than MessageTally.  Andreas wrote
> it to get around some of MessageTally's limitations.  It should
> hopefully show primitives quite accurately.

I'll see what I can do.

Cheers
Philippe




More information about the Pharo-project mailing list