Source

pypy / pypy / jit / backend / llgraph / runner.py

The branch 'virtual-raw-mallocs' does not exist.
Author Commit Message Labels Comments Date
Antonio Cuni
attach the ts helper to the cpu instead of to the staticdata. This already simplifies things a bit, and moreover assure that ts will be available wherever is_oo is tested
Maciej Fijalkowski
(arigo, fijal) Merge the pyjitpl5-call-if branch
Maciej Fijalkowski
(pedronis, arigo, fijal) Kill overflow flag interface on the CPU
Maciej Fijalkowski
(pedronis, fijal) Change strange interface to set_overflow_error into get_overflow_error that does not store this on CPU first
Armin Rigo
Merge the branch 'pyjitpl5-guardovf', in which 'int_xxx_ovf' no longer raises an exception but just sets a flag that must be checked immediately by 'guard_overflow' or 'guard_no_overflow'.
Armin Rigo
Improve the code generated for ActionFlag().get() and set(), at the end of every loop in the case of PyPy. Now we generate just a "getfield_raw(Const(...), 'value')" instead of a whole "call".
Armin Rigo
Implement do_new for ootype backends. Not directly tested for now (there are some skipped start-of-tests in backend/test/runner_test).
Armin Rigo
RPythonification. (Not sure how it could have worked before, except as a complicated consequence of how Descr is passed around...)
Armin Rigo
New detail in the backends: arraydescr.is_array_of_pointers().
Armin Rigo
Add rop.SAME_AS. Normally not used, because it is just removed by optimize.py, but it's useful to have it during tracing (see test). Implemented it anyway in the backends, for completeness and for tests running without optimize.py.
Armin Rigo
Merge the pyjitpl5-optimize4 branch: a clean-up and rewrite of optimize.py.
Antonio Cuni
- hack test_subclassof until it really emits a subclassof operation and can be run by zrpy_* tests - move do_subclassof to executor.py, as its implementation is backend indipendent - implement llimpl.op_subclassof for the llgraph backend - implement emit_op_subclassof for the cli backend
Antonio Cuni
implement subclassof for ootype jit
Armin Rigo
Rename the 'type' field, which sometimes conflicts with the 'type' class attribute on Boxes and Consts.
Armin Rigo
* added a __slots__ to AbstractDescr. * mark Float as unsupported, even if its size is not too large (on 64-bit machines).
Maciej Fijalkowski
Copy -experiments to pyjitpl5 at point where it made sense
Maciej Fijalkowski
A pretty experimental checkin. Basically I'm toying with the idea of extending args of virtualizables when we discover new nodes in bridges.
Armin Rigo
More support for arrays of voids.
Maciej Fijalkowski
update backends (my mono does not cooperate so I don't know if cli works)
Antonio Cuni
make sure that you return the always same descr for fields defined in the superclasses
Antonio Cuni
make reprs of oo descrs more readable
Armin Rigo
Start supporting the framework GC in addition to Boehm. Infrastructure work only; future work is described in backend/x86/TODO.
Antonio Cuni
- refactor emit_op_call and emit_op_oosend to share common logic - implement emit_op_getfield_gc, to make test_send_to_single_target_method passing
Antonio Cuni
cool, last test passes
Antonio Cuni
implement instanceof for ootype
Armin Rigo
Handle ZeroDivisionError correctly, fixing an XXX in codewriter.py.
Antonio Cuni
translation fixes for ootype (tests are still not passing but progress has been made: now they crashes in the cli backend)
Armin Rigo
Fail if we attempt to translate a non-ready-for-translation CPU.
Antonio Cuni
a failing test and the corresponding fix for handle_residual_oosend
Antonio Cuni
merge more of the pyjitpl5 branch back to pyjitpl5-simplify (the first part of the branch was merged in r64647) svn merge svn+ssh://codespeak.net/svn/pypy/branch/pyjitpl5/ -r64647:HEAD ------------------------------------------------------------------------ r64649 | fijal | 2009-04-24 18:56:39 +0200 (Fri, 24 Apr 2009) | 2 lines Changed paths: M /pypy/branch/pyjitpl5/pypy/jit/backend/minimal/runner.py improve debugging -----------------------------------------------------------…
  1. Prev
  2. 1
  3. 2
  4. 3
  5. 4
  6. 5
  7. 6
  8. 7
  9. 8
  10. Next