code loading performance
Locked
 
|
Hi all,
Currently loading GToolkit takes quite some time (arount 18 minutes at the best of times) we were wondering if there’s been any attempts at improving code loading times in Metacello/Monticello. Or mabye there are some ideas on how one might start doing something like this? Cheers, George |
Re: code loading performance
|
|
Hi Georges,
There have been quite some improvements those last weeks on the performance of loading classes and methods. But we are still waiting for the https://github.com/pharo-project/pharo/pull/5292 to be integrated. And you have to encapsulate the loading code into: SourceFiles deferFlushDuring: [...] and use the latest pharo 8.0 image. You can give a try with this and tell us how it goes! Cheers, Vincent -----Original Message----- From: Pharo-dev On Behalf Of George Ganea Sent: Friday, 6 December 2019 17:08 To: [hidden email] Cc: Chis Vasile Andrei <[hidden email]> Subject: [Pharo-dev] code loading performance Hi all, Currently loading GToolkit takes quite some time (arount 18 minutes at the best of times) we were wondering if there’s been any attempts at improving code loading times in Metacello/Monticello. Or mabye there are some ideas on how one might start doing something like this? Cheers, George |
Re: code loading performance
|
|
In reply to this post by George Ganea
Yes, some improvements are needed.
There are many steps: - downloading the code (networking) - loading the code (file io) - compiling the code - writing to the changes file - writing to the epicea log - system notifications - reactions to those system notifications BTW, doing a cleanup for production is also too slow, so I am guessing most time is lost in the last 4 items. > On 6 Dec 2019, at 17:07, George Ganea <[hidden email]> wrote: > > Hi all, > > Currently loading GToolkit takes quite some time (arount 18 minutes at the best of times) we were wondering if there’s been any attempts at improving code loading times in Metacello/Monticello. Or mabye there are some ideas on how one might start doing something like this? > > Cheers, > George |
|
|
In reply to this post by George Ganea
Hi Vincent,
I will definitely try it out and report back. Thank you! George
Hi Georges, There have been quite some improvements those last weeks on the performance of loading classes and methods. But we are still waiting for the https://github.com/pharo-project/pharo/pull/5292 to be integrated. And you have to encapsulate the loading code into: SourceFiles deferFlushDuring: [...] and use the latest pharo 8.0 image. You can give a try with this and tell us how it goes! Cheers, Vincent -----Original Message----- From: Pharo-dev On Behalf Of George Ganea Sent: Friday, 6 December 2019 17:08 To: pharo-dev at lists.pharo.org Cc: Chis Vasile Andrei <dedal18 at yahoo.com> Subject: [Pharo-dev] code loading performance Hi all, Currently loading GToolkit takes quite some time (arount 18 minutes at the best of times) we were wondering if there’s been any attempts at improving code loading times in Metacello/Monticello. Or mabye there are some ideas on how one might start doing something like this? Cheers, George |
|
|
In reply to this post by Sven Van Caekenberghe-2
Thank you Sven,
we’re already skipping writing to the epicea log, I’m now wondering can we also skip the notifications or not really? Cheers, George > On 6 Dec 2019, at 19:01, Sven Van Caekenberghe <[hidden email]> wrote: > > Yes, some improvements are needed. > > There are many steps: > > - downloading the code (networking) > - loading the code (file io) > - compiling the code > - writing to the changes file > - writing to the epicea log > - system notifications > - reactions to those system notifications > > BTW, doing a cleanup for production is also too slow, so I am guessing most time is lost in the last 4 items. > >> On 6 Dec 2019, at 17:07, George Ganea <[hidden email]> wrote: >> >> Hi all, >> >> Currently loading GToolkit takes quite some time (arount 18 minutes at the best of times) we were wondering if there’s been any attempts at improving code loading times in Metacello/Monticello. Or mabye there are some ideas on how one might start doing something like this? >> >> Cheers, >> George > > |
|
|
In reply to this post by George Ganea
On Fri, Dec 6, 2019 at 5:08 PM George Ganea <[hidden email]> wrote: Hi all, Could be even worse when you dl from low-bandwith countries like in Africa. While waiting a better solution, it possible to have build images available from PharoLauncher ? Regards, -- Serge Stinckwic h Int. Research Unit on Modelling/Simulation of Complex Systems (UMMISCO) Sorbonne University (SU) French National Research Institute for Sustainable Development (IRD) U niversity of Yaoundé I, Cameroon
|
|
|
In reply to this post by George Ganea
Hi Vincent, Unfortunately we did not see an improvement, and even worse, it seems to break one of our builds that is downstream from GToolkit. We got all sorts of errors that seemed to be related to incomplete method code. Two examples below. Syntax Error on line 2: ''']'' expected' groupedByIntervalType: aTemporalIntervalType fromTimestamp: startTimestamp toTimestamp: endTimestamp We’ve now switched our CI to a fixed version image that is prior to this merge. Even though the code we’re loading is in a private repo, please let me know if there is anything I can do to help debug this. Thank you, George
|
|
|
Maybe the best is to revert for now
|
«
Return to Pharo Smalltalk Developers
|
1 view|%1 views
| Free forum by Nabble | Edit this page |