On Wed, 12 Jul 2006, Thomas E Enebo defenestrated me:
>I have been studying block variables (DynamicVariableSet) and
> local variables (Scope) and I am hitting an enlightened state.
> At least I hope I am. I think I may end up killing a stack over
> this stuff. However, will someone confirm
I have been studying block variables (DynamicVariableSet) and
local variables (Scope) and I am hitting an enlightened state.
At least I hope I am. I think I may end up killing a stack over
this stuff. However, will someone confirm this:
Block contains a reference to it's local block variables
Eliminating this stack will be really, really nice. It's just a waste of cycles right now, and we keep having to switch code to CRef anyway.I think there's gotta be some heuristic for the blasted constant logic. We just need to figure out what that is and wire it together the right way. I know Ruby
On Wed, 12 Jul 2006, Charles O Nutter defenestrated me:
>On 7/12/06, Thomas E Enebo <[EMAIL PROTECTED]> wrote:
>It would be nice if the magic load path element would be visible
> and thus removable too. Then that would give the flexibility to
> remove classpath-based loading
Here's the same test run with some other optimizations I've been working on. They're more targeted at interpretation:C:\JRubyWork\jruby-svn>jruby test_fib_compiler.rbTime for interpreted: 48.149Time for compiled:
29.668The 56 -> 48 improvement (14% or so) is from deferring creation of some stacks
On 7/12/06, Thomas E Enebo <[EMAIL PROTECTED]> wrote:
On Wed, 12 Jul 2006, Charles O Nutter defenestrated me:>On 7/12/06, Charles O Nutter <[1][email protected]> wrote:>>Actually what you're suggesting may be better in some ways since it
>would give us a load path element that means "
On Wed, 12 Jul 2006, Charles O Nutter defenestrated me:
>On 7/12/06, Charles O Nutter <[EMAIL PROTECTED]> wrote:
>
>Actually what you're suggesting may be better in some ways since it
>would give us a load path element that means "now search classloader
>resources". My version would
On 7/12/06, Charles O Nutter <[EMAIL PROTECTED]> wrote:
What you're suggesting is basically what I'm hoping to do, except that I think we could just have the load path explicitly contain all jars on the classpath, so that when you do a load it will search classpath jars as well. That part's not too
On 7/12/06, Nick Sieger <[EMAIL PROTECTED]> wrote:
How do these libraries form the absolute path to the files -- is it based on something from "rbconfig" or RUBY_LIB? A wild idea, but what if, when running JRuby with a command like "java -jar
jruby.jar" or when embedding, that we default jruby.ho
On 7/12/06, Charles O Nutter <[EMAIL PROTECTED]> wrote:
If its not too many libraries and the fix looks easy, it might be worth
filing bugs with Ruby. Finding an include through manual file lookupstuff seems like its pretty bad programming practice anyway.
I totally agree. I think it's tres goofy
Some nights you just can't stop hacking.C:\JRubyWork\jruby-svn>jruby test_fib_compiler.rbTime for interpreted: 56.311Time for compiled: 31.636(these times are significantly slower than the others because I'm on my laptop tonight)
fib is very call-intensive, obviously, and the double recursion does
11 matches
Mail list logo
