extradoc / minute / pypy-sync-2005-09-22.txt

pypy-sync developer meeting 22th September 

Time & location: 1pm (30 minutes) at #pypy-sync 


         Samuele Pedroni
         Anders Lehmann (minutes/moderation)
         Anders Chrigström
         Christian Tismer 
         Holger Krekel
         Eric van Riet Paap
         Michael Hudson
         Adrien Di Mascio

Regular Topics 

- activity reports (3 prepared lines of info). 
  All Attendees submitted activity reports (see `IRC-Log`_ 
  at the end and 'LAST/NEXT/BLOCKERS' entries in particular)

- resolve conflicts/blockers
  No conflicts were discovered (which we could address).

Topics of the week

1. translate_pypy_new

We should try to make translate_pypy_new be the default way of translating pypy this week (ale and pedronis). There are wishes as to be able to specify options in the target, that translate_pypy_new should propagate to the translation process.

Better naming and better choice of default values should be done first.

Batch mode is another wish

2. A pypy scrap-book ?

It was agreed that it was a nice idea, and that we should setup a file for 
that purpose. (ale will do that)

3. Candidates for Refactoring

Two candidates were mentioned. The backendoptimizations and the that we should separate inlining and malloc removal. Both topics was deemed as hard and as good topics for a sprint.

Next pypy-sync meeting

Scheduled for next Thursday, Sept. 29nd 1300h CET, conducted by Anders Lehmann.


Anders closes the meeting at 13:34pm.

.. _`IRC-log`:


    <aleale> Hi everyone, my watch says 1 o'clock. Have anyone heard from pedronis, ludal, stakkars and arigo
    <aleale> Should we ait for them ?
    <hpk> not sure
    <adim> aleale: Hi
    <arre> Samuele is not in the office.
    <hpk> armin flew to goetheborg today 
    <adim> ludal won't be here today
    <hpk> let's start the meeting anyway, i'd say
    <aleale> Ok 
    <aleale> Welcome 
    <mwh> hello
    <aleale> Lets start with what we have prepared
    <aleale> Prev: a little on translate_pypy_new
    <aleale> Next: translate_pypy_new ?
    <aleale> Blockers: -
    <arre> PREV: Getting the astcompiler is shape
    <arre> NEXT: Even more work on the astcompiler
    <arre> BLOCKERS: None
    <adim> LAST: astcompiler
    <adim> NEXT: none
    <adim> BLOCKERS: none
    <mwh> last: dealt with UoB stuff
    <mwh> next: continue to attempt to resolve UoB situation
    <mwh> blockers: UoB administrative delays
    ??? stakkars [] has joined #pypy-sync
    <hpk> last: EU issues, py lib/test stuff
    <hpk> next: relaxing, pytest related things
    <hpk> blockers: -
    <aleale> welcome stakkars, we just doing the round of prepared stuff
    <stakkars> ok
    <stakkars> DONE: optimization and benchmarking
    <stakkars> NEXT: more of this, splitting generated code
    <stakkars> BLOCK: None
    <aleale> ericvp: Do you have something?
    <hpk> probably he is not actually here at the moment
    <aleale> The only block seems to be mwh's wrestling with UoB bureaucraci
    <aleale> I dont know if we can help, anyone ?
    <mwh> not really
    ??? bertf [] has joined #pypy-sync
    <mwh> we talked about it in the consortium meeting yesterday
    <aleale> yes
    <mwh> if the uob doesn't work out, it looks like i'll join armin at hhud
    <mwh> (administratively more than physically :)
    <aleale> Hi bertf. we are just about moving to the weekly topics
    <aleale> Ok, mwh good luck
    <bertf> yea, sorry, was in the wrong channel. Hi
    <mwh> aleale: thanks
    <mwh> let's not waste time on this one today :)
    <aleale> The weekly topics are:
    <aleale>        1. translate_pypy_new
    <aleale>        2. A pypy scrap-book ?
    <aleale>        3. Candidates for Refactoring
    <hpk> bertf: hi bert 
    <ericvrp> sorry, was asleep. will paste now
    <ericvrp> last: transformation for faster exception handling in llvm
    <ericvrp> next: nightly cronjob benchmarking of pypy-c and pypy-llvm
    <ericvrp> blockers: -
    <aleale> first I would like if we could give a hint on in what direction we are heading with translate_pypy_new
    <aleale> Thanks ericvp
    <aleale> But since armin and pedronis are not here we better defer it to #pypy ?
    <mwh> can someone quickly summarize the point of translate_pypy_new ?
    <stakkars> yes.  I heared Samuele saying he would help with this
    <mwh> is it just to tidy up the various translation target/option/... stuff
    <mwh> ?
    <hpk> mwh: yes
    <mwh> right, thanks
    <aleale> It was/is an attempt to clean up te mess and maybe add some flexibility, I think
    <aleale> Defer ?
    <hpk> yes, the idea is a more unified model for options regarding our pypy entry points
    <hpk> aleale: can one currently use translate_pypy_new already? 
    ? hpk/#pypy-sync hasn't tried so far
    <aleale> I have tried it on targetrichards and standalone. seems to work
    <aleale> Actually I think all the checked in versions have work (with different naming of options)
    <hpk> it probably makes sense to switch it in sometime next week, after some more people have tried to use it 
    <stakkars> I was a bit puzzled about new option names. I know the old ones by heart. that made me lazy to try it.
    <ericvrp> It is working, some options do the opposite of what you expect and default values need to be choosen better
    <ericvrp> but yes, it is working
    <aleale> the naming had to change due to optparse - maybe we dont want to use that
    <aleale> Hi pedronis
    ??? pedronis [] has joined #pypy-sync
    <aleale> Hi pedronis
    ? pedronis/#pypy-sync sorry
    <stakkars> ha
    <hpk> no, we should use optparse 
    <aleale> Ok, we were discussing the state of translate_pypy_new
    <hpk> but keep the option meanings and default values as close as sensible to the current script 
    <hpk> like Eric said
    <aleale> sure
    <stakkars> what was the expected changeof the refactoring?
    <pedronis> there's an issue about that
    <aleale> less mess/ more flexibility
    <stakkars> just better layout, or adding features easily?
    <stakkars> sorry, I didn't look last time. Is it considered ready?
    <pedronis> well, one thing we may want is pass --usemodulus to targertpypy.
    <hpk> yip
    <pedronis> other is to run it unattended with a failed/successed exit value and writing to logs (maybe)
    <hpk> yip, but we can add that 
    <hpk> i think it's good to head for actually using it some time soon
    <stakkars> a complete batch mode: compile with options, run tests, do next one.
    <aleale> Ok, I'll conclude that we want to have translate_pypy_new brougth into a usable state soon
    <pedronis> at the moment if used with -d option on targetrichads it crashed
    <stakkars> --usemodules is indeed a thing where I'd like to have the defaults in the target. It would be nice if the
               translate_pypy was able to read the options which a target has and to support them
    <aleale> and that we want to add more features.
    <stakkars> a way to specify specific things in a target without cluttering translate
    <pedronis> well, but also to take them from the command line
    <pedronis> target specific options
    <aleale> Who have time to look at it ( I wont be able to do anything before tuesday)?
    <stakkars> I want to provide them in the commandline, for the target, without defining them in translate, necessarily
    <pedronis> I can look at it a bit
    <aleale> I have prepared for target specific options (add a dictionary called "options" in the target)
    <stakkars> as an aside: I'd like to renew the idea of saving state to a file which can be postprocessed, right before the
    <aleale> I think we should go to the next topic (8 mins left)
    <mwh> a scrap book sounds like a nice idea, but is there that much to go in it?
    <aleale> mwh: what to you mean?
    <mwh> maybe in misunderstood the idea then
    <mwh>       ^I
    <aleale> I think we should start thinking about it and collecting stuff.
    <aleale> Then we can later decide how to present it
    <hpk> well, just start a .txt file 
    <mwh> makes sense
    <mwh> i certainly don't know where to find all our conference presentations
    <aleale> what about pictures and video?
    <aleale> should it be a directory instead
    <hpk> we don't have much in that area (apart from the sprint pictures i took)
    <hpk> they can be referenced along with the sprint reports 
    <aleale> ok I wil start the scrapbook then
    <aleale> topic 3 : Candidates for refactoring
    <mwh> well, the backendoptimizations are still pretty scary
    <mwh> don't know if that's fixable, though
    <aleale> It seems that we have enough tasks for the next week so I think we can postpone it ?
    <stakkars> we should try to seperate inlining and malloc removal
    <aleale> mwh: do you have time to look at that ?
    <mwh> not really
    <aleale> stakkars: will that be done as part of the current optimisations
    <stakkars> I hope so. But it is difficult.
    <aleale> Ok the candidates will be recorded in the minutes. Anyway the time is up
    <mwh> might be a better sprint topic
    ??? mwh [] has left #pypy-sync ["ERC Version 5.0 (CVS) $Revision:
              1.771 $ (IRC client for Emacs)"]
    <aleale> Have we choesen a moderator for next week ?
    <stakkars> yes, the list is completely incomplete :)
    <hpk> yes, who is doing it next week? 
    <aleale> Any takers?
    <aleale> Ok I can do it again ?
    <stakkars> I think of a number, you guess it. who is closes takes it
    ? stakkars/#pypy-sync thinks this didn't work
    <aleale> I would like to add : "Incredible job You've done with optimisations. Great job"
    <hpk> stakkars, aleale: what about the two of you doing it the next two times? 
    <hpk> after that we already have the paris sprint and can plan further
    <aleale> I will do it next week . See you all then
    <stakkars> no problem
    <hpk> ok, great.
    <aleale> and we are 4 minutes late
    <stakkars> no idea why exactly we, but it's fine
    ??? stakkars [] has left #pypy-sync []
    <adim> see you
    <hpk> see you
    ??? adim [] has left #pypy-sync []
    <aleale> bye. I have to prepare my sons 18 birthday now - so the minutes will first be done tomorrow 
    <ericvrp> goodbye
    <bertf> bye
     [01:40pm][aleale] [#pypy-sync(+ns)]                                                                                      
     [Lag  0] [O/1 N/6 I/0 V/0 F/0]                                                                                [U:a:S:b:h]