Commits

Author Commit Message Labels Comments Date
Amaury Forgeot d'Arc
Add PyClassMethod_New()
Amaury Forgeot d'Arc
Implement PyClass_New()
Amaury Forgeot d'Arc
Implement _PyString_Eq()
Amaury Forgeot d'Arc
Implement the (empty) PyType_Modified() function
Amaury Forgeot d'Arc
Implement PyTuple_GetSlice()
Armin Rigo
Get rid of GcRefList, one of the last remaining causes of leaks, as far as I can tell. Kill it by noting that in the common case, with the minimark GC, all ConstPtrs are likely old anyway by the time they are seen by the backend. So the following hack should be enough (see rgc.py).
Branches
jit-write-barrier-from-array
Armin Rigo
Fix test.
Branches
jit-write-barrier-from-array
Armin Rigo
Fix test.
Branches
jit-write-barrier-from-array
Armin Rigo
Fix the XXX: in case the JIT generates a SETARRAYITEM_GC on a list which it cannot prove is short enough, we should really use write_barrier_from_array instead of the default write_barrier. I think that this is what caused the slow-down of slowspitfire on May 25.
Branches
jit-write-barrier-from-array
Antonio Cuni
hg merge default; I had to manually resolve a lot of conflicts in jit/metainterp/logger.py
Branches
jitypes2
Antonio Cuni
port test_ctypes to test_pypy_c_new; we don't test it in a very precise way, because the trace is still too complicated
Branches
jitypes2
Antonio Cuni
add a test checking that we do not allocate the frame when calling an _ffi function
Branches
jitypes2
Antonio Cuni
mention this difference in cpython-differences.rst
Armin Rigo
merge heads
Armin Rigo
ll_arraycopy: optimize the case where the dest value is not virtual, and only the source value is. If the copy is not too long, it can be done as a sequence of residual SETARRAYITEM_GC.
Armin Rigo
Unify these two dictionaries.
Branches
jit-resizable-list
Armin Rigo
No-ops.
Branches
jit-resizable-list
Maciej Fijalkowski
Add project-ideas (draft)
Armin Rigo
Tests and fixes.
Branches
buffer-readline
Amaury Forgeot d'Arc
(nekto0n) cx_Oracle: add support for "twophase" connections
Amaury Forgeot d'Arc
(nekto0n) add cx_Oracle.paramstyle
Armin Rigo
merge heads
Armin Rigo
Even in non-buffering mode, use a very minimal buffering to make readline() have not-too-horrible performance.
Branches
buffer-readline
Armin Rigo
Fix the docstring: readline() had the docstring of readlines()...
David Schneider
use setup_class here
Branches
arm-backed-float
Armin Rigo
In case the variable is already spilled and not living in any register, do nothing.
Branches
arm-backed-float
David Schneider
correct offset calculations for loading and storing spilled floats to and from core registers
Branches
arm-backed-float
David Schneider
make sure to use imm values for offset calculations when possible
Branches
arm-backed-float
David Schneider
implement force_spill operation for testing
Branches
arm-backed-float
David Schneider
(arigo, bivab) add a resoperation for testing called force_spill, that forces a variable to be spilled. It is used to the generate different call patterns with variables that are currently spilled as arguments
Branches
arm-backed-float
  1. Prev
  2. Next