Permalink
Commits on Sep 20, 2009
  1. Working on an interpreter, plus probably some other stuff. Not workin…

    …g yet, but I want to get this stuff committed so I don't lose it.
    committed Sep 20, 2009
Commits on Jul 21, 2009
Commits on Jul 17, 2009
  1. Added a few more things to the list of methods to statically optimize…

    …. The compiler and GC are invoking much fewer calls to the sendMessage_stub now.
    committed Jul 17, 2009
Commits on Jul 16, 2009
  1. Fixed a compiler bug.

    committed Jul 16, 2009
Commits on Jul 13, 2009
  1. The vector and byteVector At: and At:Put: primitives are translated t…

    …o IR nodes now, so their failblocks can be deferred. Still some weird bug in the VM, though.
    committed Jul 13, 2009
Commits on Jul 12, 2009
Commits on Jul 4, 2009
  1. Still in the middle of working on type prediction and block deferring…

    …, but need to get some fixes over to Alex's machine.
    committed Jul 4, 2009
Commits on Jun 22, 2009
  1. Removed an old experiment.

    committed Jun 22, 2009
  2. Small cleanups.

    committed Jun 22, 2009
  3. Added a couple of statistics-gathering menu items to the debugger mor…

    …ph. Also a couple of small fixes.
    committed Jun 22, 2009
  4. Delay cloning of failblocks for some primitives. (So far only impleme…

    …nted for primitives implemented by translating to IR nodes.)
    committed Jun 22, 2009
Commits on Jun 13, 2009
  1. There's no hotspot-optimization stuff yet, but the first test case in…

    …volving dynamic compilation has passed!
    committed Jun 13, 2009
Commits on Jun 7, 2009
Commits on May 15, 2009
  1. Lots of changes - properly using maps instead of mirrors for receiver…

    … types in the compiler, blocks now contain a constant slot pointing to their home scopeDesc...
    committed May 15, 2009
Commits on May 9, 2009
Commits on May 5, 2009
Commits on May 2, 2009
  1. Can do a little bit more inlining now. But it's still not very smart …

    …about figuring out what type the receiver is.
    committed May 2, 2009
Commits on Apr 28, 2009
  1. Klein can do some inlining now!

    There's still lots of inlining that it can't do - blocks, methods containing blocks, etc. Still, it's something. :) The debugger even shows the stack as if the missing frames were present.
    committed Apr 28, 2009
Commits on Apr 24, 2009
Commits on Apr 19, 2009
  1. Did a major refactoring of the compiler, in preparation for implement…

    …ing inlining.
    
    (There's still more refactoring to do, but I reached a point where most stuff seems to work, so I'm committing.)
    committed Apr 19, 2009
Commits on Apr 6, 2009
  1. Now there's a card table. Also fixed a few other bugs. The selfVM tes…

    …ts run to completion now, invoking the GC a few times along the way.
    committed Apr 6, 2009
Commits on Apr 5, 2009
  1. Switched back to segregated byteVectors, because doing a card table s…

    …ounds a lot more complicated with unsegregated ones.
    
    Also made it easier to switch back and forth between segregated and unsegregated.
    committed Apr 5, 2009
Commits on Apr 3, 2009
  1. GC works a bit better. Still doesn't have a card table, though.

    Also fixed that annoying bug in the remote debugger, where it wouldn't update the object table properly.
    committed Apr 3, 2009
Commits on Mar 29, 2009
  1. Removed the breakpoints in the GC process, and made the lotsa-garbage…

    … test do more objects, just to make sure the GC can run more than once.
    committed Mar 29, 2009
  2. I think the GC just passed its first test case!

    Basically it allocates (and then throws away) a bunch of really big objects, about twice as many as the space can actually hold. And when it runs out of memory partway through, it does a scavenge and then keeps going.
    
    Now, the scavenge takes about ten minutes (on a 67 MB space), but still...
    committed Mar 29, 2009
Commits on Mar 26, 2009
  1. Working on GC.

    committed Mar 26, 2009
Commits on Mar 22, 2009