Source

pypy / pypy / translator / backendopt / malloc.py

Author Commit Message Labels Comments Date
Alex Gaynor
fix a mis-indented block
Armin Rigo
Tentative: kill various pieces of code left and right handling keepalives. Now the inliner no longer inserts tons of keepalives for you (which are almost always unneeded).
Carl Friedrich Bolz
kill the rest of the cases that used traverse
Armin Rigo
Finish the merge of trunk into pyjitpl5.
Alexander Schremmer
Refactor UnionFind API. It calls absorb instead of update now.
Samuele Pedroni
kill flavor_malloc_*, zero_malloc_* now only malloc_* operations should be present in graphs until gc transform malloc and malloc_varsize now always carry a dictionary of flags as second operand, the 'flavor' key must always be set in this dict the flavored_free operation is now just free(p, flavor), flavor is just the flavor string the base gc transform class has now more responsabilities, subclasses need to care only about 'gc'…
Alexander Schremmer
Make backendopts less chatty if verbose is false.
Samuele Pedroni
intermediate: try to regularize clever malloc removal interface, share some common code between it and all
Antonio Cuni
bugfix, and the corresponding test.
Antonio Cuni
Use the right TYPE for the key in newvarsmap; don't remove malloc of BuiltinType
Antonio Cuni
Re-revert to revision 36067, and fix what caused test_all to fail. Things learned: never guess whether type_system should contain 'lltype' or 'lltypesystem'... :-(
Samuele Pedroni
umph, reverting the ootype malloc removal refactoring 36037 (and consequently 36040), it needs more careful review. It broke in a shallow way backendopt/test/test_all.py but once the shallow breakage is repaired we see that a bunch of tests there are failing because malloc removal is not happening anymore it seems. This needs to be fixed in such way that test_all and translator/test/test_simplify pass again and then it can be checked in again. test_all is quite an important test for *all* backendopts!
Antonio Cuni
Malloc removal works with ootype now!
Antonio Cuni
Other steps in porting malloc removal to ootype. Now BaseMallocRemover should be free from any explicit reference to lltype.
Antonio Cuni
First steps in making malloc removal ootype-compatible.
Armin Rigo
Stop being "clever" about which keepalives must be kept during malloc removal, and which ones can be removed.
Armin Rigo
Support for ctypes.Union in genc. (XXX missing llinterp support)
Armin Rigo
Fix and test for another malloc removal bug.
Armin Rigo
(pedronis, arigo) Support for mixing None and py_object() -- not other ctypes objects, only py_object, because the rules for what 'None' means when in ctypes are, let's say, interesting.
Armin Rigo
direct_fieldptr support in malloc removal, for rctypes. The code is getting rather involved, but still has some kind of convoluted elegance to it.
Armin Rigo
A case where malloc removal would hang, trying repeatedly to "remove" the same GcStruct without progress.
Armin Rigo
Minimal support for FixedSizeArray in malloc removal.
Armin Rigo
Fix rather broken code, as better testing shows.
Armin Rigo
Malloc removal in the presence of random 'getsubstruct' operations. This is needed to remove the overhead of rctypes, but it's probably not buying us anything for the rest of PyPy.
Armin Rigo
A test that shows a situation where malloc removal needs an extra 'keepalive' to be valid. Fix.
Samuele Pedroni
count malloc removals
Carl Friedrich Bolz
(johahn, cfbolz): remove debug statement
Carl Friedrich Bolz
(johahn, cfbolz): prevent the malloc removal to remove instances of classes that have a __del__ method (or other structs that have a destructor registered).
Michael Hudson-Doyle
importfun vs translator/ (apart from js) haven't run all the tests yet, will do so on snake pronto.
Michael Hudson-Doyle
remove rpython/lltype.py (easy!) and fix all the references to it to be to rpython/lltypesystem/lltype.py (tedious!). haven't run ALL the tests yet, will do that on snake.
  1. Prev
  2. 1
  3. 2
  4. Next