I've just posted Magma for Squeak 3.9 (7057) on the MagmaTester project.
The upgrade to work with 3.9 was relatively uneventful; the biggest change was to include serialization of Pragmas in CompiledMethods. There were also some methods renamed.
In order to have just one code base, I had to put a couple of "is3dot9" checks into the code. And due to the extension methods on Pragma and MethodProperties, Monticello will also report a warnning now when trying to load into a 3.8 image. So, for now, I've just put it into MagmaTester to see what anyone else thinks; deal with a couple (literally, two, I think) of ugly conditionals and a warning when loading for 3.7 and 3.8, or maintain two versions (branches) of the code..
I really don't want to maintain two branches, but yet the load warning might be a real turn off for (new) people unless they know to expect it somehow.. Of course, if 3.9 is the future then it won't be an issue.
Enjoy, and thanks for any feedback..
Chris Muller wrote:
I've just posted Magma for Squeak 3.9 (7057) on the MagmaTester project.
Cool Chris, and just when I was getting used to 3.8 ;-)
The logging is really helpful to learn how magma thinks. I am finding tuning reading easier than stubbing out!
For example on my 700Mhz machine.
Obtaining a MagmaSession using a WAMagmaSession takes almost 10 seconds. Whereas the WASharedMagmaSession does not have that overhead
Reading in the front pier page with some text in it.
minimumDepth: 0 takes 45 readGraph operations in: 3408ms minimumDepth: 1 takes 45 readGraph operations in: 4973ms.
i.e. it appears that few of the extra reads generated by increasing minimum depth are actually being used.
so now to some tuning...
Keith
___________________________________________________________ The all-new Yahoo! Mail goes wherever you go - free your email address from your Internet provider. http://uk.docs.yahoo.com/nowyoucan.html
Hi Keith!
Keith Hodges keith_hodges@yahoo.co.uk wrote:
Chris Muller wrote:
I've just posted Magma for Squeak 3.9 (7057) on the MagmaTester project.
Cool Chris, and just when I was getting used to 3.8 ;-)
The logging is really helpful to learn how magma thinks. I am finding tuning reading easier than stubbing out!
For example on my 700Mhz machine.
Obtaining a MagmaSession using a WAMagmaSession takes almost 10 seconds. Whereas the WASharedMagmaSession does not have that overhead
In Gjallar we have Q2MagmaSessionPool (adapted from Cees code originally) so when there is a bunch of sessions we pick one up from there and do an abort to revive it etc. This should typically be a standard component in Magma of course.
regards, Göran
PS. Gonna take a look at your logging stuff, I am just about to play with strategies myself.
magma@lists.squeakfoundation.org
-
Chris Muller -
goran＠krampe.se -
Hilaire Fernandes -
Keith Hodges