Commits

Show all
Author Commit Message Labels Comments Date
Armin Rigo
Tweaks.
Branches
gil-improvement
Armin Rigo
Move around the PulseEvents. According to the test (https://bitbucket.org/arigo/arigo/raw/default/hack/stm/misc) it still seems to be a correct placement, and it delays the PulseEvents as much as possible, minimizing the risks of a missed PulseEvent.
Branches
gil-improvement
Armin Rigo
merge heads
Branches
gil-improvement
Armin Rigo
GIL code for Windows. After trying for a while I failed to find a reasonable fully fool-proof equivalent to the Posix code, so we end up with this hackish solution instead.
Branches
gil-improvement
Armin Rigo
hg merge default
Branches
gil-improvement
Armin Rigo
Tweak the hack: it should be equivalent, but have no remaining effect at run-time.
Armin Rigo
Fix test_optimizeopt.test_setarrayitem_followed_by_arraycopy() by putting manually the correct readonly_descrs_arrays and write_descrs_arrays lists.
Armin Rigo
This simplification seems to fix the test.
Armin Rigo
A failing test that shows a real (but rare) case in translation.
Armin Rigo
Bah.
Branches
gil-improvement
Armin Rigo
Fixes.
Branches
gil-improvement
Armin Rigo
For now, always use "python" to run trackgcroot.py. Running pypy is slower, particularly now that each invocation is typically 0.2-0.5 seconds.
Armin Rigo
Force the GIL to be released in this app-test; time.sleep() doesn't do it when non-translated... It likely worked because sys.checkinterval used to be only 100.
Branches
gil-improvement
Alex Gaynor
unroll int_pow if the pow and mod are constant. (based on observing what GCC does)
Armin Rigo
Port the code. Trying with a random but larger value for sys.checkinterval; will experiment more.
Branches
gil-improvement
Armin Rigo
A branch in which to apply https://bitbucket.org/arigo/arigo/raw/default/hack/stm/misc
Branches
gil-improvement
Alex Gaynor
Fix TypeSubClass(object) to work like type(object)
Alex Gaynor
merged upstream
Alex Gaynor
list.pop should call __int__ on its argument.
Hakan Ardo
a setslice mutates the array
Hakan Ardo
Hack to ensure that ll_arraycopy gets a proper effectinfo.write_descrs_arrays
Hakan Ardo
rpython version of this test
Alex Gaynor
fix ropeunicode.
Alex Gaynor
move the allocation of integers around.
Hakan Ardo
maybe nicer with different indexes
Hakan Ardo
can we realy delay the setarrayitem_gc here?
Carl Friedrich Bolz
a failing test for the interaction of the JIT with tagged pointers
Carl Friedrich Bolz
add unicode project idea
Maciej Fijalkowski
write a test and slightly modify conditions. I think I understand typeorder slightly better now
Carl Friedrich Bolz
rename to is_annotation_constant. add a docstring.
  1. Prev
  2. Next