Source

pypy / pypy / translator / backendopt / inline.py

Author Commit Message Labels Comments Date
Alex Gaynor
some random style + deadcode cleanups
Armin Rigo
Change _always_inline_=True to crash if inlining fails. Use _always_inline_='try' to not crash if it fails. Fixes left and right.
Antonio Cuni
completely change the strategy for inlining jit_merge_points in the
Branches
autoreds
Armin Rigo
Get rid of 'isstartblock'.
Armin Rigo
hg backout 94d415eff8b4
Armin Rigo
Backout the merge of stacklet. It seems to give nonsense in non-jitted
Armin Rigo
In-progress: kill the stackless transformer and the llops depending on it.
Branches
stacklet
Carl Friedrich Bolz
kill more remnants of resume points
Armin Rigo
Tentative: kill various pieces of code left and right handling
Carl Friedrich Bolz
get rid of some more traverse calls
Antonio Cuni
in rare cases, it might happend that we want to downcast instead of upcasting. Unfortunately, I spent 1 day searching for a failing test without success :-/
Carl Friedrich Bolz
(mikke, cfbolz): kill jit inlining heuristic, as it was replaced by a new
Mikael Schönenberg
(cbolz, micke): do not inline functions marked _dont_look_inside_ if you use the jit heuristic
Carl Friedrich Bolz
(micke, cfbolz): rename the heuristic to be called jit, because it does more
Carl Friedrich Bolz
make the jit inliner not inline pure functions
Armin Rigo
issue431 testing
Armin Rigo
When doing secondary backendopts, don't inline a graph
Armin Rigo
Rename the 'dont_inline' flag to '_dont_inline_' for better
Alexander Schremmer
(cfbolz, xoraxax): Remove extfunctable and suggested_primitive support. Refactored exttypeinfo-usage
Maciej Fijalkowski
Add a possibility to iterate over all callsites
Maciej Fijalkowski
Speed up inlining (especially important when inlining helpers)
Carl Friedrich Bolz
don't insert keepalives any more when inlining. it's no longer necessary and
Carl Friedrich Bolz
issue313 resolved
Armin Rigo
Finish the merge of the pypy-more-rtti-inprogress branch.
Antonio Cuni
- add a comment for the new check in checkgraph
Alexander Schremmer
Make backendopts less chatty if verbose is false.
Samuele Pedroni
let's specify an heuristic function for the various inlinings, generalize such that it can return a large number until
Samuele Pedroni
oops, this should have been checked in separately
Samuele Pedroni
intermediate: explicit threshold for malloc prediction for now.
Samuele Pedroni
intermediate: try to regularize clever malloc removal interface, share some common code between it and all
  1. Prev
  2. 1
  3. 2
  4. 3
  5. 4
  6. Next