Commits

Show all
Author Commit Message Labels Comments Date
Armin Rigo
Comments.
Branches
mapdict-interp
Armin Rigo
Add a comment.
Branches
mapdict-interp
Armin Rigo
A test, and fix.
Branches
mapdict-interp
Armin Rigo
Trying to fix the bug.
Branches
mapdict-interp
Armin Rigo
A bug.
Branches
mapdict-interp
Armin Rigo
hg backout 8624cde59095
Branches
mapdict-interp
Armin Rigo
A branch in which to re-introduce the fast paths killed in 8624cde59095, to avoid slowing down interpreted execution too much.
Branches
mapdict-interp
Armin Rigo
(cfbolz, arigo) Decided that this code is too brittle. We found two kinds of bugs in it: bugs that were already there, and bugs that arose because of TypeCells. * for the 1st kind of bug: just because the map of an object doesn't change does not mean that getdictvalue() returns from the same spot (case of DevolvedDictTerminator); and in that case, getdictvalue() can actually have random side-effects like changi…
Armin Rigo
(cfbolz, arigo) Fix.
Armin Rigo
(cfbolz, arigo) Found bug B while looking for bug A in mapdict :-(
Antonio Cuni
fix the test, now that we switched to frozenlist
Maciej Fijalkowski
a "typo"
Maciej Fijalkowski
fix tests
Armin Rigo
Skip a failing test about prof-based-inline, which was broken by 5b0e029514d4. The point is that we never use profile-based inlining, so I doubt that it's worth fixing. To make it clear, remove the cmdline options.
Armin Rigo
Uh.
Antonio Cuni
(qbproger) add proper signatures to all remaining sqlite functions, to silence warnings
Maciej Fijalkowski
increase slightly trace limit. This is due to extra operations being introduced that always gets removed by optimizer. Not ideal, but should work for now :/
Maciej Fijalkowski
Put two quasiimmut fields on executioncontext. Should speed up some things slightly.
Antonio Cuni
merge heads
Antonio Cuni
make get_deep_immutable_oplist returning an actual frozen list instead of a tuple, because a lot of places rely on it to be a list (e.g. to do operations + some_other_list)
Antonio Cuni
make sure that rewrite_assembler does not mutate the operations
Antonio Cuni
sanity check to make sure that the backend never mutate the operation list it gets; fails with the x86 backend so far, because of rewrite_assembler
Armin Rigo
merge heads
Armin Rigo
Ignore these files, produced by test runs.
Carl Friedrich Bolz
merge
Carl Friedrich Bolz
merge
Carl Friedrich Bolz
This dict acts as a globals dict, make it one.
Carl Friedrich Bolz
Remove XXXs that don't really apply any more or never denoted things that should be fixed
Antonio Cuni
write an helper function to make sure that we do not mutate the oplist. Not used anywhere so far
Alex Gaynor
Remove some more instructions that were likely moved to the preamble.
  1. Prev
  2. Next