[Pharo-project] Mars, stuck on callbacks (again)... need opinions
Hi guys, I'm doing some "advance callbacks" for Mars, and I'm having several problems with table renderers (they are called A LOT of times and needs to be efficient). The ObjectiveCBridge is not as efficient as that calls needs to be executed, and well... I was planing to spend some time on this, trying to see how can I improve the callback velocity. And then a question came to my mind: do I spend time improving current ObjectiveCBridge implementation... or I better use Alien and migrate everything to it? some context: 1) it was always planned to "sonner or latter" migrate the ObjectiveCBridge stuff to alien (there is another . 2) Alien is not working -yet- for iPhone. It is planned (bah... it is "thought") to run Alien in iPhoneVM too. 3) Alien callback implementation is a lot cleaner than the current ObjectiveCBridge implementation (which relies on both sides semaphores, when alien callbacks relies on "image ownership") So... it will delay a little a release of Mars? sure it will... but in the long way it will be better. I would like to hear your opinions, cheers, Esteban
On 23 December 2010 20:35, Esteban Lorenzano <estebanlm@gmail.com> wrote:
Hi guys, I'm doing some "advance callbacks" for Mars, and I'm having several problems with table renderers (they are called A LOT of times and needs to be efficient). The ObjectiveCBridge is not as efficient as that calls needs to be executed, and well... I was planing to spend some time on this, trying to see how can I improve the callback velocity. And then a question came to my mind:
do I spend time improving current ObjectiveCBridge implementation... or I better use Alien and migrate everything to it?
some context:
1) it was always planned to "sonner or latter" migrate the ObjectiveCBridge stuff to alien (there is another . 2) Alien is not working -yet- for iPhone. It is planned (bah... it is "thought") to run Alien in iPhoneVM too. 3) Alien callback implementation is a lot cleaner than the current ObjectiveCBridge implementation (which relies on both sides semaphores, when alien callbacks relies on "image ownership") .. So... it will delay a little a release of Mars? sure it will... but in the long way it will be better.
i wanted to offer a NativeBoost, but it is not ready for use under MacOS yet :) About iPhoneVM - definitely, you will need to use something else than NB. And Eliot also having somethings to offer :)
I would like to hear your opinions,
cheers, Esteban
-- Best regards, Igor Stasenko AKA sig.
Esteban, FW(very little)IW, my first reaction as I started reading was "has he thought about native boost or alien?" Callbacks have been in a possible-but-not-supported state for a long time. I believe Eliot recently mentioned callbacks in FFI, so it seems to be an idea that is solidifying. At present, I have .so that exports functions in lieu of callbacks. I have to recompile the library to change anything :( but it's number crunching so I get C speed for my trouble :) The functions are used small numbers of times, so Smalltalk speed and flexibility would be preferable for all of my uses to date ~:0 Good hunting! Bill
Hi Bill, Yes, Alien already support callbacks... maybe you can try it instead recompiling every time :) Cheers, Esteban El 23/12/2010, a las 6:22p.m., Schwab,Wilhelm K escribió:
Esteban,
FW(very little)IW, my first reaction as I started reading was "has he thought about native boost or alien?" Callbacks have been in a possible-but-not-supported state for a long time. I believe Eliot recently mentioned callbacks in FFI, so it seems to be an idea that is solidifying.
At present, I have .so that exports functions in lieu of callbacks. I have to recompile the library to change anything :( but it's number crunching so I get C speed for my trouble :) The functions are used small numbers of times, so Smalltalk speed and flexibility would be preferable for all of my uses to date ~:0
Good hunting!
Bill
I'm still watching. Alien didn't really exist on Linux (so it was not an option) when I last looked, and native boost was starting to look pretty attractive. FFI gets some points because I know more about (at least using) it than the others. The recompiling is easy enough for now, and it ain't broke. ________________________________________ From: pharo-project-bounces@lists.gforge.inria.fr [pharo-project-bounces@lists.gforge.inria.fr] On Behalf Of Esteban Lorenzano [estebanlm@gmail.com] Sent: Thursday, December 23, 2010 4:28 PM To: Pharo-project@lists.gforge.inria.fr Subject: Re: [Pharo-project] Mars, stuck on callbacks (again)... need opinions Hi Bill, Yes, Alien already support callbacks... maybe you can try it instead recompiling every time :) Cheers, Esteban El 23/12/2010, a las 6:22p.m., Schwab,Wilhelm K escribió:
Esteban,
FW(very little)IW, my first reaction as I started reading was "has he thought about native boost or alien?" Callbacks have been in a possible-but-not-supported state for a long time. I believe Eliot recently mentioned callbacks in FFI, so it seems to be an idea that is solidifying.
At present, I have .so that exports functions in lieu of callbacks. I have to recompile the library to change anything :( but it's number crunching so I get C speed for my trouble :) The functions are used small numbers of times, so Smalltalk speed and flexibility would be preferable for all of my uses to date ~:0
Good hunting!
Bill
participants (3)
-
Esteban Lorenzano -
Igor Stasenko -
Schwab,Wilhelm K