Chris Muller uploaded a new version of Tools to project The Inbox:
http://source.squeak.org/inbox/Tools-cmm.733.mcz ==================== Summary ==================== Name: Tools-cmm.733 Author: cmm Time: 6 December 2016, 3:50:06.783172 pm UUID: 46663c9b-de5d-4b65-ae5e-def988146a88 Ancestors: Tools-nice.732 Smalltalk garbageCollect can sometimes take many minutes, and there's no need to do it when opening a ProcessBrowser. =============== Diff against Tools-nice.732 =============== Item was changed: ----- Method: ProcessBrowser class>>open (in category 'instance creation') ----- open - "ProcessBrowser open" - "Create and schedule a ProcessBrowser." - Smalltalk garbageCollect. ^ToolBuilder open: self new! |
On 06.12.2016, at 22:50, [hidden email] wrote: > Chris Muller uploaded a new version of Tools to project The Inbox: > http://source.squeak.org/inbox/Tools-cmm.733.mcz > > ==================== Summary ==================== > > Name: Tools-cmm.733 > Author: cmm > Time: 6 December 2016, 3:50:06.783172 pm > UUID: 46663c9b-de5d-4b65-ae5e-def988146a88 > Ancestors: Tools-nice.732 > > Smalltalk garbageCollect can sometimes take many minutes, what? > and there's no need to do it when opening a ProcessBrowser. > > =============== Diff against Tools-nice.732 =============== > > Item was changed: > ----- Method: ProcessBrowser class>>open (in category 'instance creation') ----- > open > - "ProcessBrowser open" > - "Create and schedule a ProcessBrowser." > - Smalltalk garbageCollect. > ^ToolBuilder open: self new! > > |
Hi Tobias,
On Tue, Dec 6, 2016 at 1:52 PM, Tobias Pape <[hidden email]> wrote:
The Spur GC is not yet mature. In particular the old space compaction algorithm that I wrote is poor, and we suspect it has very poor performance in rare circumstances. We intend to reimplement asap, certainly before mid 2017. We also plan to implement an incremental global gc which should remove the need to run global GC at all except when one needs to know that everything has been collected now (e.g. the pointer finder).
_,,,^..^,,,_ best, Eliot |
On 07.12.2016, at 20:13, Eliot Miranda <[hidden email]> wrote: > Hi Tobias, > > On Tue, Dec 6, 2016 at 1:52 PM, Tobias Pape <[hidden email]> wrote: > > On 06.12.2016, at 22:50, [hidden email] wrote: > > > Chris Muller uploaded a new version of Tools to project The Inbox: > > http://source.squeak.org/inbox/Tools-cmm.733.mcz > > > > ==================== Summary ==================== > > > > Name: Tools-cmm.733 > > Author: cmm > > Time: 6 December 2016, 3:50:06.783172 pm > > UUID: 46663c9b-de5d-4b65-ae5e-def988146a88 > > Ancestors: Tools-nice.732 > > > > Smalltalk garbageCollect can sometimes take many minutes, > > what? > > The Spur GC is not yet mature. In particular the old space compaction algorithm that I wrote is poor, and we suspect it has very poor performance in rare circumstances. We intend to reimplement asap, certainly before mid 2017. We also plan to implement an incremental global gc which should remove the need to run global GC at all except when one needs to know that everything has been collected now (e.g. the pointer finder). > > Ah. Didn't mean to be harsh, I was just surprised :) > > and there's no need to do it when opening a ProcessBrowser. > > > > =============== Diff against Tools-nice.732 =============== > > > > Item was changed: > > ----- Method: ProcessBrowser class>>open (in category 'instance creation') ----- > > open > > - "ProcessBrowser open" > > - "Create and schedule a ProcessBrowser." > > - Smalltalk garbageCollect. > > ^ToolBuilder open: self new! > > > > > > > > > > -- > _,,,^..^,,,_ > best, Eliot |
Hi Tobias,
On Wed, Dec 7, 2016 at 12:26 PM, Tobias Pape <[hidden email]> wrote:
Don't worry; I was not offended. I try not to take these things personally. But let me take this opportunity to say we're very open to volunteers taking a look at this, or any other, part of the VM :-) Help is most welcome.
_,,,^..^,,,_ best, Eliot |
On Wed, 7 Dec 2016, Eliot Miranda wrote:
> Don't worry; I was not offended. I try not to take these things personally. But let me take this opportunity to say we're very open to volunteers taking a look at this, or any other, part of the VM :-) > Help is most welcome. How long do you think someone could get to the point to understand how the current GC works assuming no former VM developement experience? How long do you think such person could reimplement a new GC? How much effort would it take to decrease the barrier? (I'm asking this, because I think it's way to hard for someone inexperienced to get into the developement without nagging someone experienced with questions for weeks.) Levente |
Free forum by Nabble | Edit this page |