help-smalltalk
[Top][All Lists]
Advanced

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [Help-smalltalk] faster startup 2/n


From: Mike Anderson
Subject: Re: [Help-smalltalk] faster startup 2/n
Date: Thu, 14 Dec 2006 22:20:15 +0000
User-agent: Thunderbird 1.5.0.5 (X11/20060719)

Paolo Bonzini wrote:
A slightly trickier one. If we don't save directly from the object area, but allocate temporary memory for that, we can save different data than what is in the object memory. So, we can kill CallinProcesses when saving, and we can force looking up CFunctionDescriptor addresses the first time they are called. Finally, we can map the interpreter's primitives to what is in the image, rather than patch the CompiledMethods so that they match the primitive numbers that were selected at compile time.

It seems to me that this has other benefits than just speed; making it easier to re-load images that link with C libraries.

Is it necessarily correct to kill a call-in process, though? Say the process in question was a GUI event that triggered a long-running calculation. There wouldn't be anywhere to return a result to, but it would be the side-effects (displaying the results in the GUI) that were important.

I'm interested because I'm still interested in the idea of persistent Smalltalk images, although I've changed direction slightly from the gsti idea. It would be useful to have a clear idea of what the contract is for an image restart.

Mike




reply via email to

[Prev in Thread] Current Thread [Next in Thread]