Matti Picus pushed to branch branch/hpy-0.0.4 at PyPy / pypy
Commits:
ecc8dd42 by Manuel Jacob at 2022-06-30T05:38:12+02:00
Port 211c8c95037f to reverse_debugging.py.
--HG--
branch : py3.8
- - - - -
47ed3769 by Matti Picus at 2022-06-30T17:21:14+03:00
update openssl versions
- - - - -
fd014843 by Matti Picus at 2022-06-30T17:22:44+03:00
merge default
--HG--
branch : py3.8
- - - - -
bd63c59c by Antonio Cuni at 2022-06-30T16:33:19+01:00
merge heads
- - - - -
5b6751ca by Antonio Cuni at 2022-06-30T16:34:03+01:00
merge heads
--HG--
branch : py3.8
- - - - -
7a22f10c by fijal at 2022-07-01T14:43:06+02:00
make sure we have wrappers around variadic calls in the JIT, fixes #3762
- - - - -
0cc2aa62 by Carl Friedrich Bolz-Tereick at 2022-07-02T13:36:06+02:00
give lifetimes a repr
- - - - -
08469c94 by Carl Friedrich Bolz-Tereick at 2022-07-02T15:18:14+02:00
be a bit more debugger friendly
- - - - -
69be3d8f by Carl Friedrich Bolz-Tereick at 2022-07-02T18:14:29+02:00
use a more subtle condition to check whether aliasing is present when doing
malloc removal
so far, we simply checked whether an outgoing link of a block had the variable
that stores the to-be-malloc-removed value twice. that's not precise enough
and
leads to broken graphs in complicated situations.
The new condition is the following:
- we only care about aliasing if there are two distinct creation points. if
there is only one, it does not matter that we ship the same value around in
different variables.
- if there is more than one creation point, it's not good enough to check
whether a single variable is duplicated. instead, we check whether there are
two variables passed along the link that share the same creation point (the
variables might have been copied via same_as or a cast_pointer).
- - - - -
f804cef0 by Matti Picus at 2022-07-02T23:37:33+03:00
use os.open instead of rposix.open where possible for M1 untranslated tests
- - - - -
859bff32 by Matti Picus at 2022-07-03T14:08:52+03:00
tweak test, macos build documentation
- - - - -
f13c8d4d by Matti Picus at 2022-07-03T14:18:19+03:00
backport changes from py3.8
- - - - -
9153a5ab by Matti Picus at 2022-07-03T15:28:34+03:00
fix reordering of calls in pypyjit test after 0a8acf22983f changed something
- - - - -
e2f12d88 by Matti Picus at 2022-07-03T15:29:42+03:00
rearrange so untranslated tests fail only on darwin. A better fix is needed
- - - - -
7817f258 by Matti Picus at 2022-07-03T16:49:35+03:00
remove non-matchging redundant code to fix translation
- - - - -
52bf4302 by Matti Picus at 2022-07-05T09:35:43+03:00
skip test that hengs untranslated on macos arm64, issue 3774
- - - - -
61dab08e by Carl Friedrich Bolz-Tereick at 2022-07-05T11:22:33+02:00
two elidables, add some hypothesis tests for the shift methods
- - - - -
8df404da by Matti Picus at 2022-07-06T06:35:22+03:00
update embedded openssl to 1.1.1p, 3.0.5
- - - - -
600cb49d by Carl Friedrich Bolz-Tereick at 2022-07-06T11:17:33+02:00
all these optimize_* functions that just call "emit" are pointless,
because the
default behaviour is to just call emit
- - - - -
4859d00f by fijal at 2022-07-06T12:25:39+02:00
add more natural arities
- - - - -
eedd062b by Matti Picus at 2022-07-06T17:51:10+03:00
do more name mangling for cpytext so untranslated tests on a pypy2 host run
- - - - -
d5e73b32 by Matti Picus at 2022-07-06T18:07:33+03:00
back out e64d144190c4, test no longer hangs
- - - - -
5e8d5052 by Carl Friedrich Bolz-Tereick at 2022-07-06T21:10:15+02:00
micro-optimize .next() to not allocate quite so many intermediate lists, it
appears relatively high in the profile of optimizeopt
- - - - -
c9a9f046 by Carl Friedrich Bolz-Tereick at 2022-07-06T21:11:07+02:00
only put OptimizationResults into the list for callbacks if the callback would
actually *do* anything
- - - - -
ab9cd9e7 by Matti Picus at 2022-07-07T08:02:25+03:00
mangle another name in cpyext
- - - - -
6cd20a20 by Matti Picus at 2022-07-07T09:10:08+03:00
patch openssl 1.1.1q for issue openssl/openssl/#18720
- - - - -
073ab68c by Carl Friedrich Bolz-Tereick at 2022-07-07T15:37:14+02:00
stop allocating lists for all the arguments of operations in the frontend
--HG--
branch : tracing-speedup-minor
- - - - -
2fa0186d by Carl Friedrich Bolz-Tereick at 2022-07-07T16:17:43+02:00
specialize these
--HG--
branch : tracing-speedup-minor
- - - - -
00bdcc8c by Carl Friedrich Bolz-Tereick at 2022-07-07T16:17:46+02:00
a comment
--HG--
branch : tracing-speedup-minor
- - - - -
4dd26b6e by Carl Friedrich Bolz-Tereick at 2022-07-07T18:22:57+02:00
rename IntBount.getint to get_constant_int
having it share the metho name with ConstantInt means the method moves up to
AbstractValue. Then IntBount.getint shows up as a case that is inlined all over
the metainterp.
--HG--
branch : tracing-speedup-minor
- - - - -
49cefd7e by Carl Friedrich Bolz-Tereick at 2022-07-07T18:39:39+02:00
for constant pointers that come out of jitcodes, use a subclass of ConstPtr
that h
