Commits

Show all
Author Commit Message Labels Comments Date
Armin Rigo
Remove deprecated operations.
Branches
smalllong
Armin Rigo
Kill.
Branches
smalllong
Armin Rigo
Add missing doc.
Branches
smalllong
Armin Rigo
Merge trunk into the 'smalllong' branch.
Branches
smalllong
Armin Rigo
import svn externals into the branch
Branches
smalllong
Armin Rigo
Replace the 'svn' link with 'hg'.
Armin Rigo
Alex Gaynor
Allow writing app-level helper functions on AppTests.
Alex Gaynor
The JIT works on 64 bit now.
Amaury Forgeot d'Arc
Fix a leak in a win32 test
Benjamin Peterson
unneeded word
Armin Rigo
Simplify a bit READ_TIMESTAMP on x86-64.
Hakan Ardo
intbound support for shift operations
Hakan Ardo
generalized floordiv intbound propagation
Hakan Ardo
Some integer optimizations and modificatiosn to the integer division implementation to allow the jit to optimize it better in some cases.
Hakan Ardo
hg backout --merge 40391
Branches
jit-unroll-loops
Hakan Ardo
Backed out changeset f342eb27760a
Branches
jit-unroll-loops
Armin Rigo
Tweak the number passed to -j to make.
Armin Rigo
Merge branch/gc-minimark-largeobj again: adds support for young
Hakan Ardo
Propagate intbound through floordiv in a simple case
Branches
jit-unroll-loops
Hakan Ardo
rpythonized
Branches
jit-unroll-loops
Hakan Ardo
Testing with CALL_LOOPINVARIANT and GUARD_NO_EXCEPTION in short preamble
Branches
jit-unroll-loops
Armin Rigo
Fixing the test requires using an identity_dict instead of
Branches
gc-minimark-largeobj
Armin Rigo
Debugging comment.
Branches
gc-minimark-largeobj
Hakan Ardo
More jit friendly overflowcheck for integer divisions. This should produce one operation less in the common case where nothing is known about arguments but there is no OverflowError, and it should allow the entire overflowcheck to be removed in cases where there are intbounds on the arguments.
Branches
jit-unroll-loops
Armin Rigo
More in-progress. Test failure in test_gc that I still don't understand.
Branches
gc-minimark-largeobj
Hakan Ardo
Allow integer divisions to be optimized somewhat if the arguments are known to be possitive. The last 3 ops after int_floordiv should be removed by the backend since their results are never used, right? Where does the "(i8 & i6) == -1" opps come from?
Branches
jit-unroll-loops
Hakan Ardo
hg merge default
Branches
jit-unroll-loops
Carl Friedrich Bolz
kill old and very obscure function
Hakan Ardo
hg merge
  1. Prev
  2. Next