Commits

Author Commit Message Labels Comments Date
Armin Rigo
svn merge svn+ssh://codespeak.net/svn/pypy/trunk -r71801:71802
Branches
1.2.x
Armin Rigo
svn merge svn+ssh://codespeak.net/svn/pypy/trunk -r71789:71790
Branches
1.2.x
Armin Rigo
svn merge svn+ssh://codespeak.net/svn/pypy/trunk -r71767:71768
Branches
1.2.x
Armin Rigo
svn merge svn+ssh://codespeak.net/svn/pypy/trunk -r71702:71707
Branches
1.2.x
Armin Rigo
svn merge svn+ssh://codespeak.net/svn/pypy/trunk -r71631:71659
Branches
1.2.x
Armin Rigo
Merge r71729.
Branches
1.2.x
Armin Rigo
Make a release branch.
Branches
1.2.x
convert-repo
closed branch gcremovetypeptr
Branches
gcremovetypeptr
Armin Rigo
Change this to True.
Branches
gcremovetypeptr
Armin Rigo
Make a branch just for running the performance plots.
Branches
gcremovetypeptr
convert-repo
closed branch jit-sandbox
Branches
jit-sandbox
Armin Rigo
Documentation.
Branches
jit-sandbox
Maciej Fijalkowski
Be extra paranoid on rmmap. This is non-issue, since this stuff is unsupported by sandbox anyway, however if one day we start supporting it, there is a test.
Branches
jit-sandbox
Maciej Fijalkowski
Apparently this test should pass. malloc_varsize_marknsweep is called and returns non-zero
Branches
jit-sandbox
Maciej Fijalkowski
A failing test (finally!)
Branches
jit-sandbox
Maciej Fijalkowski
We don't support python 2.3 anyway and that generates DeprecationWarning from pylib, use subprocess from stdlib
Branches
jit-sandbox
Maciej Fijalkowski
Revert 71608
Branches
jit-sandbox
Maciej Fijalkowski
Merge 71607 from trunk
Branches
jit-sandbox
Armin Rigo
Repeat r71577 and r71578: (fijal) I think this is a correct fix for JIT & sandbox - we assumer mmap and friends are sandbox-safe (they can only work on existing file descs anyway). If we compile mmap module, this will probably violate sandbox, but we compile sandbox with no-allworkingmodules anyway Sprinkle sandboxsafe here and there
Branches
jit-sandbox
Armin Rigo
A branch in which to make the JIT cooperate with sandboxing.
Branches
jit-sandbox
convert-repo
closed branch guard-value-counting-2
Branches
guard-value-counting-2
Armin Rigo
Add tracking of which values are seen by failing GUARD_VALUEs. This logic is activated by calling make_a_counter_per_value() on a ResumeGuardDescr. By default each ResumeGuardDescr uses a single counter, like before; but if make_a_counter_per_value() is called, then it will lazily make an instance of ResumeGuardCounters and use that to track which values are seen. The algo in ResumeGuardCounters needs improvements, but it should work as a first draft.
Branches
guard-value-counting-2
Armin Rigo
Move the call to must_compile_from_failure() into the ResumeGuardDescr. Avoids the hack of storing -2 in ResumeGuardForcedDescr.
Branches
guard-value-counting-2
Armin Rigo
Another branch for the same purpose. Here I will try to implement it minimally, without starting with refactoring.
Branches
guard-value-counting-2
convert-repo
closed branch guard-value-counting
Branches
guard-value-counting
Armin Rigo
Check in my working copy. That branch will probably be abandonned and maybe restarted some day.
Branches
guard-value-counting
Armin Rigo
Start implementing the idea that there are really two kinds of Descrs for guards: one before optimization, and the other after optimization. They need slightly different attributes, and the goal is to let optimization decide exactly which subclass of ResumeGuardDescr is needed for each guard.
Branches
guard-value-counting
Armin Rigo
A branch in which to try the following idea: on failing GUARD_VALUE, put not just one counter but one counter per value. This should help because a GUARD_VALUE that keeps seeing new values is not useful to start tracing from.
Branches
guard-value-counting
convert-repo
closed branch fix-blackhole-virt
Branches
fix-blackhole-virt
Armin Rigo
Write a comment about the risks of putting stuff on MetaInterpGlobalData. Too bad that it's not well-tested :-(
Branches
fix-blackhole-virt
  1. Prev
  2. Next