Author: Matti Picus <[email protected]>
Branch: unicode-utf8-py3
Changeset: r95931:e08614e7b2b3
Date: 2019-02-09 20:38 +0100
http://bitbucket.org/pypy/pypy/changeset/e08614e7b2b3/
Log: merge unicode-utf8 into branch
diff too long, truncating to 2000 out of 4233 lines
diff --git a/.hgtags b/.hgtags
--- a/.hgtags
+++ b/.hgtags
@@ -61,3 +61,9 @@
9112c8071614108b1042bfef0713915107004d62 release-pypy2.7-v7.0.0
1f86f25937b6ae6c8b25236c35228fac587678bf release-pypy3.5-v7.0.0
dab365a465140aa79a5f3ba4db784c4af4d5c195 release-pypy3.6-v7.0.0
+9112c8071614108b1042bfef0713915107004d62 release-pypy2.7-v7.0.0
+c8805ee6d7846ca2722b106eeaa2f128c699aba3 release-pypy2.7-v7.0.0
+1f86f25937b6ae6c8b25236c35228fac587678bf release-pypy3.5-v7.0.0
+928a4f70d3de7d17449456946154c5da6e600162 release-pypy3.5-v7.0.0
+dab365a465140aa79a5f3ba4db784c4af4d5c195 release-pypy3.6-v7.0.0
+fb40f7a5524c77b80e6c468e087d621610137261 release-pypy3.6-v7.0.0
diff --git a/TODO b/TODO
--- a/TODO
+++ b/TODO
@@ -1,8 +1,6 @@
* find a better way to run "find" without creating the index storage, if one
if one is not already readily available (understand cost now, improve after
merge)
-* write the correct jit_elidable in _get_index_storage (Armin)
* improve performance of splitlines (CF)
-* stop using runicode/unicode and move MAXUNICODE to rutf8 (Matti)
* think about cost of utf8 list strategy (CF)
* revisit why runicode import str_decode_utf_8_impl needed instead of runicode
import str_decode_utf_8
diff --git a/pypy/doc/release-v7.0.0.rst b/pypy/doc/release-v7.0.0.rst
--- a/pypy/doc/release-v7.0.0.rst
+++ b/pypy/doc/release-v7.0.0.rst
@@ -39,7 +39,7 @@
The utf8 branch that changes internal representation of unicode to utf8 did not
make it into the release, so there is still more goodness coming.
-You can download the v6.0 releases here:
+You can download the v7.0 releases here:
http://pypy.org/download.html
diff --git a/pypy/doc/whatsnew-head.rst b/pypy/doc/whatsnew-head.rst
--- a/pypy/doc/whatsnew-head.rst
+++ b/pypy/doc/whatsnew-head.rst
@@ -5,6 +5,11 @@
.. this is a revision shortly after release-pypy-7.0.0
.. startrev: 481c69f7d81f
+.. branch: zlib-copying-third-time-a-charm
+
+Make sure zlib decompressobjs have their streams deallocated immediately
+on flush.
+
.. branch: zlib-copying-redux
Fix calling copy on already-flushed compressobjs.
@@ -15,7 +20,11 @@
as they do on CPython.
-.. math-improvements
+.. branch: math-improvements
Improve performance of long operations where one of the operands fits into
-an int.
\ No newline at end of file
+an int.
+
+.. branch: regalloc-playgrounds
+
+Improve register allocation in the JIT.
diff --git a/pypy/doc/whatsnew-pypy2-5.10.0.rst
b/pypy/doc/whatsnew-pypy2-5.10.0.rst
--- a/pypy/doc/whatsnew-pypy2-5.10.0.rst
+++ b/pypy/doc/whatsnew-pypy2-5.10.0.rst
@@ -1,42 +1,42 @@
-==========================
-What's new in PyPy2.7 5.10
-==========================
-
-.. this is a revision shortly after release-pypy2.7-v5.9.0
-.. startrev:d56dadcef996
-
-
-.. branch: cppyy-packaging
-
-Cleanup and improve cppyy packaging
-
-.. branch: docs-osx-brew-openssl
-
-.. branch: keep-debug-symbols
-
-Add a smartstrip tool, which can optionally keep the debug symbols in a
-separate file, instead of just stripping them away. Use it in packaging
-
-.. branch: bsd-patches
-
-Fix failures on FreeBSD, contributed by David Naylor as patches on the issue
-tracker (issues 2694, 2695, 2696, 2697)
-
-.. branch: run-extra-tests
-
-Run extra_tests/ in buildbot
-
-.. branch: vmprof-0.4.10
-
-Upgrade the _vmprof backend to vmprof 0.4.10
-
-.. branch: fix-vmprof-stacklet-switch
-.. branch: fix-vmprof-stacklet-switch-2
-
-Fix a vmprof+continulets (i.e. greenelts, eventlet, gevent, ...)
-
-.. branch: win32-vcvars
-
-.. branch: rdict-fast-hash
-
-Make it possible to declare that the hash function of an r_dict is fast in
RPython.
+==========================
+What's new in PyPy2.7 5.10
+==========================
+
+.. this is a revision shortly after release-pypy2.7-v5.9.0
+.. startrev:d56dadcef996
+
+
+.. branch: cppyy-packaging
+
+Cleanup and improve cppyy packaging
+
+.. branch: docs-osx-brew-openssl
+
+.. branch: keep-debug-symbols
+
+Add a smartstrip tool, which can optionally keep the debug symbols in a
+separate file, instead of just stripping them away. Use it in packaging
+
+.. branch: bsd-patches
+
+Fix failures on FreeBSD, contributed by David Naylor as patches on the issue
+tracker (issues 2694, 2695, 2696, 2697)
+
+.. branch: run-extra-tests
+
+Run extra_tests/ in buildbot
+
+.. branch: vmprof-0.4.10
+
+Upgrade the _vmprof backend to vmprof 0.4.10
+
+.. branch: fix-vmprof-stacklet-switch
+.. branch: fix-vmprof-stacklet-switch-2
+
+Fix a vmprof+continulets (i.e. greenelts, eventlet, gevent, ...)
+
+.. branch: win32-vcvars
+
+.. branch: rdict-fast-hash
+
+Make it possible to declare that the hash function of an r_dict is fast in
RPython.
diff --git a/pypy/doc/whatsnew-pypy2-6.0.0.rst
b/pypy/doc/whatsnew-pypy2-6.0.0.rst
--- a/pypy/doc/whatsnew-pypy2-6.0.0.rst
+++ b/pypy/doc/whatsnew-pypy2-6.0.0.rst
@@ -1,132 +1,128 @@
-===========================
-What's new in PyPy2.7 5.10+
-===========================
-
-.. this is a revision shortly after release-pypy2.7-v5.10.0
-.. startrev: 6b024edd9d12
-
-.. branch: cpyext-avoid-roundtrip
-
-Big refactoring of some cpyext code, which avoids a lot of nonsense when
-calling C from Python and vice-versa: the result is a big speedup in
-function/method calls, up to 6 times faster.
-
-.. branch: cpyext-datetime2
-
-Support ``tzinfo`` field on C-API datetime objects, fixes latest pandas HEAD
-
-
-.. branch: mapdict-size-limit
-
-Fix a corner case of mapdict: When an instance is used like a dict (using
-``setattr`` and ``getattr``, or ``.__dict__``) and a lot of attributes are
-added, then the performance using mapdict is linear in the number of
-attributes. This is now fixed (by switching to a regular dict after 80
-attributes).
-
-
-.. branch: cpyext-faster-arg-passing
-
-When using cpyext, improve the speed of passing certain objects from PyPy to C
-code, most notably None, True, False, types, all instances of C-defined types.
-Before, a dict lookup was needed every time such an object crossed over, now it
-is just a field read.
-
-
-.. branch: 2634_datetime_timedelta_performance
-
-Improve datetime + timedelta performance.
-
-.. branch: memory-accounting
-
-Improve way to describe memory
-
-.. branch: msvc14
-
-Allow compilaiton with Visual Studio 2017 compiler suite on windows
-
-.. branch: winapi
-
-Update _winapi and internal _winbase_cffi (via _winbase_build) for python 3
-
-.. branch: refactor-slots
-
-Refactor cpyext slots.
-
-
-.. branch: call-loopinvariant-into-bridges
-
-Speed up branchy code that does a lot of function inlining by saving one call
-to read the TLS in most bridges.
-
-.. branch: rpython-sprint
-
-Refactor in rpython signatures
-
-.. branch: cpyext-tls-operror2
-
-Store error state thread-locally in executioncontext, fixes issue #2764
-
-.. branch: cpyext-fast-typecheck
-
-Optimize `Py*_Check` for `Bool`, `Float`, `Set`. Also refactor and simplify
-`W_PyCWrapperObject` which is used to call slots from the C-API, greatly
-improving microbenchmarks in https://github.com/antocuni/cpyext-benchmarks
-
-
-.. branch: fix-sre-problems
-
-Fix two (unrelated) JIT bugs manifesting in the re module:
-
-- green fields are broken and were thus disabled, plus their usage removed from
- the _sre implementation
-
-- in rare "trace is too long" situations, the JIT could break behaviour
- arbitrarily.
-
-.. branch: jit-hooks-can-be-disabled
-
-Be more efficient about JIT hooks. Make it possible for the frontend to declare
-that jit hooks are currently not enabled at all. in that case, the list of ops
-does not have to be created in the case of the on_abort hook (which is
-expensive).
-
-
-.. branch: pyparser-improvements
-
-Improve speed of Python parser, improve ParseError messages slightly.
-
-.. branch: ioctl-arg-size
-
-Work around possible bugs in upstream ioctl users, like CPython allocate at
-least 1024 bytes for the arg in calls to ``ioctl(fd, request, arg)``. Fixes
-issue #2776
-
-.. branch: cpyext-subclass-setattr
-
-Fix for python-level classes that inherit from C-API types, previously the
-`w_obj` was not necessarily preserved throughout the lifetime of the `pyobj`
-which led to cases where instance attributes were lost. Fixes issue #2793
-
-
-.. branch: pyparser-improvements-2
-
-Improve line offsets that are reported by SyntaxError. Improve error messages
-for a few situations, including mismatched parenthesis.
-
-.. branch: issue2752
-
-Fix a rare GC bug that was introduced more than one year ago, but was
-not diagnosed before issue #2752.
-
-.. branch: gc-hooks
-
-Introduce GC hooks, as documented in doc/gc_info.rst
-
-.. branch: gc-hook-better-timestamp
-
-Improve GC hooks
-
-.. branch: cppyy-packaging
-
-Update backend to 0.6.0 and support exceptions through wrappers
+===========================
+What's new in PyPy2.7 5.10+
+===========================
+
+.. this is a revision shortly after release-pypy2.7-v5.10.0
+.. startrev: 6b024edd9d12
+
+.. branch: cpyext-avoid-roundtrip
+
+Big refactoring of some cpyext code, which avoids a lot of nonsense when
+calling C from Python and vice-versa: the result is a big speedup in
+function/method calls, up to 6 times faster.
+
+.. branch: cpyext-datetime2
+
+Support ``tzinfo`` field on C-API datetime objects, fixes latest pandas HEAD
+
+
+.. branch: mapdict-size-limit
+
+Fix a corner case of mapdict: When an instance is used like a dict (using
+``setattr`` and ``getattr``, or ``.__dict__``) and a lot of attributes are
+added, then the performance using mapdict is linear in the number of
+attributes. This is now fixed (by switching to a regular dict after 80
+attributes).
+
+
+.. branch: cpyext-faster-arg-passing
+
+When using cpyext, improve the speed of passing certain objects from PyPy to C
+code, most notably None, True, False, types, all instances of C-defined types.
+Before, a dict lookup was needed every time such an object crossed over, now it
+is just a field read.
+
+
+.. branch: 2634_datetime_timedelta_performance
+
+Improve datetime + timedelta performance.
+
+.. branch: memory-accounting
+
+Improve way to describe memory
+
+.. branch: msvc14
+
+Allow compilaiton with Visual Studio 2017 compiler suite on windows
+
+.. branch: refactor-slots
+
+Refactor cpyext slots.
+
+
+.. branch: call-loopinvariant-into-bridges
+
+Speed up branchy code that does a lot of function inlining by saving one call
+to read the TLS in most bridges.
+
+.. branch: rpython-sprint
+
+Refactor in rpython signatures
+
+.. branch: cpyext-tls-operror2
+
+Store error state thread-locally in executioncontext, fixes issue #2764
+
+.. branch: cpyext-fast-typecheck
+
+Optimize `Py*_Check` for `Bool`, `Float`, `Set`. Also refactor and simplify
+`W_PyCWrapperObject` which is used to call slots from the C-API, greatly
+improving microbenchmarks in https://github.com/antocuni/cpyext-benchmarks
+
+
+.. branch: fix-sre-problems
+
+Fix two (unrelated) JIT bugs manifesting in the re module:
+
+- green fields are broken and were thus disabled, plus their usage removed from
+ the _sre implementation
+
+- in rare "trace is too long" situations, the JIT could break behaviour
+ arbitrarily.
+
+.. branch: jit-hooks-can-be-disabled
+
+Be more efficient about JIT hooks. Make it possible for the frontend to declare
+that jit hooks are currently not enabled at all. in that case, the list of ops
+does not have to be created in the case of the on_abort hook (which is
+expensive).
+
+
+.. branch: pyparser-improvements
+
+Improve speed of Python parser, improve ParseError messages slightly.
+
+.. branch: ioctl-arg-size
+
+Work around possible bugs in upstream ioctl users, like CPython allocate at
+least 1024 bytes for the arg in calls to ``ioctl(fd, request, arg)``. Fixes
+issue #2776
+
+.. branch: cpyext-subclass-setattr
+
+Fix for python-level classes that inherit from C-API types, previously the
+`w_obj` was not necessarily preserved throughout the lifetime of the `pyobj`
+which led to cases where instance attributes were lost. Fixes issue #2793
+
+
+.. branch: pyparser-improvements-2
+
+Improve line offsets that are reported by SyntaxError. Improve error messages
+for a few situations, including mismatched parenthesis.
+
+.. branch: issue2752
+
+Fix a rare GC bug that was introduced more than one year ago, but was
+not diagnosed before issue #2752.
+
+.. branch: gc-hooks
+
+Introduce GC hooks, as documented in doc/gc_info.rst
+
+.. branch: gc-hook-better-timestamp
+
+Improve GC hooks
+
+.. branch: cppyy-packaging
+
+Update backend to 0.6.0 and support exceptions through wrappers
diff --git a/pypy/doc/whatsnew-pypy2-7.0.0.rst
b/pypy/doc/whatsnew-pypy2-7.0.0.rst
--- a/pypy/doc/whatsnew-pypy2-7.0.0.rst
+++ b/pypy/doc/whatsnew-pypy2-7.0.0.rst
@@ -1,81 +1,69 @@
-==========================
-What's new in PyPy2.7 6.0+
-==========================
-
-.. this is a revision shortly after release-pypy-6.0.0
-.. startrev: e50e11af23f1
-
-.. branch: cppyy-packaging
-
-Main items: vastly better template resolution and improved performance. In
-detail: upgrade to backend 1.4, improved handling of templated methods and
-functions (in particular automatic deduction of types), improved pythonization
-interface, range of compatibility fixes for Python3, free functions now take
-fast libffi path when possible, moves for strings (incl. from Python str),
-easier/faster handling of std::vector by numpy, improved and faster object
-identity preservation
-
-.. branch: socket_default_timeout_blockingness
-
-Make sure 'blocking-ness' of socket is set along with default timeout
-
-.. branch: crypt_h
-
-Include crypt.h for crypt() on Linux
-
-.. branch: gc-more-logging
-
-Log additional gc-minor and gc-collect-step info in the PYPYLOG
-
-.. branch: reverse-debugger
-
-The reverse-debugger branch has been merged. For more information, see
-https://bitbucket.org/pypy/revdb
-
-.. branch: pyparser-improvements-3
-
-Small refactorings in the Python parser.
-
-.. branch: fix-readme-typo
-
-.. branch: avoid_shell_injection_in_shutil
-
-.. branch: unicode-utf8-re
-
-.. branch: utf8-io
-
-Utf8 handling for unicode
-
-.. branch: pyparser-improvements-3
-Small refactorings in the Python parser.
-
-Backport CPython fix for possible shell injection issue in `distutils.spawn`,
-https://bugs.python.org/issue34540
-
-.. branch: cffi_dlopen_unicode
-
-Enable use of unicode file names in `dlopen`
-
-.. branch: rlock-in-rpython
-
-Backport CPython fix for `thread.RLock`
-
-
-.. branch: expose-gc-time
-
-Make GC hooks measure time in seconds (as opposed to an opaque unit).
-
-.. branch: cleanup-test_lib_pypy
-
-Update most test_lib_pypy/ tests and move them to extra_tests/.
-
-.. branch: gc-disable
-
-Make it possible to manually manage the GC by using a combination of
-gc.disable() and gc.collect_step(). Make sure to write a proper release
-announcement in which we explain that existing programs could leak memory if
-they run for too much time between a gc.disable()/gc.enable()
-
-.. branch: unicode-utf8
-
-Use utf8 internally to represent unicode
+==========================
+What's new in PyPy2.7 6.0+
+==========================
+
+.. this is a revision shortly after release-pypy-6.0.0
+.. startrev: e50e11af23f1
+
+.. branch: cppyy-packaging
+
+Main items: vastly better template resolution and improved performance. In
+detail: upgrade to backend 1.4, improved handling of templated methods and
+functions (in particular automatic deduction of types), improved pythonization
+interface, range of compatibility fixes for Python3, free functions now take
+fast libffi path when possible, moves for strings (incl. from Python str),
+easier/faster handling of std::vector by numpy, improved and faster object
+identity preservation
+
+.. branch: socket_default_timeout_blockingness
+
+Make sure 'blocking-ness' of socket is set along with default timeout
+
+.. branch: crypt_h
+
+Include crypt.h for crypt() on Linux
+
+.. branch: gc-more-logging
+
+Log additional gc-minor and gc-collect-step info in the PYPYLOG
+
+.. branch: reverse-debugger
+
+The reverse-debugger branch has been merged. For more information, see
+https://bitbucket.org/pypy/revdb
+
+
+.. branch: pyparser-improvements-3
+
+Small refactorings in the Python parser.
+
+.. branch: fix-readme-typo
+
+.. branch: avoid_shell_injection_in_shutil
+
+Backport CPython fix for possible shell injection issue in `distutils.spawn`,
+https://bugs.python.org/issue34540
+
+.. branch: cffi_dlopen_unicode
+
+Enable use of unicode file names in `dlopen`
+
+.. branch: rlock-in-rpython
+
+Backport CPython fix for `thread.RLock`
+
+
+.. branch: expose-gc-time
+
+Make GC hooks measure time in seconds (as opposed to an opaque unit).
+
+.. branch: cleanup-test_lib_pypy
+
+Update most test_lib_pypy/ tests and move them to extra_tests/.
+
+.. branch: gc-disable
+
+Make it possible to manually manage the GC by using a combination of
+gc.disable() and gc.collect_step(). Make sure to write a proper release
+announcement in which we explain that existing programs could leak memory if
+they run for too much time between a gc.disable()/gc.enable()
diff --git a/pypy/doc/whatsnew-pypy3-5.10.0.rst
b/pypy/doc/whatsnew-pypy3-5.10.0.rst
--- a/pypy/doc/whatsnew-pypy3-5.10.0.rst
+++ b/pypy/doc/whatsnew-pypy3-5.10.0.rst
@@ -1,21 +1,7 @@
-=========================
-What's new in PyPy3 5.9+
-=========================
-
-.. this is the revision after release-pypy3.5-5.9
-.. startrev: be41e3ac0a29
-
-.. branch: sched_yield
-Add sched_yield posix attribute
-
-.. branch: py3.5-appexec
-Raise if space.is_true(space.appexec()) used in app level tests, fix tests
-that did this
-
-.. branch: py3.5-mac-embedding
-Download and patch dependencies when building cffi-based stdlib modules
-
-.. branch: os_lockf
-
-.. branch: py3.5-xattr
-Add posix.*attr() functions
+========================
+What's new in PyPy3 7.0+
+========================
+
+.. this is the revision after release-pypy3.5-v7.0
+.. startrev: 9d2fa7c63b7c
+
diff --git a/pypy/doc/whatsnew-pypy3-6.0.0.rst
b/pypy/doc/whatsnew-pypy3-6.0.0.rst
--- a/pypy/doc/whatsnew-pypy3-6.0.0.rst
+++ b/pypy/doc/whatsnew-pypy3-6.0.0.rst
@@ -1,28 +1,7 @@
-=========================
-What's new in PyPy3 5.10+
-=========================
+========================
+What's new in PyPy3 7.0+
+========================
-.. this is the revision after release-pypy3.5-v5.10
-.. startrev: 34c63fba0bba
+.. this is the revision after release-pypy3.5-v7.0
+.. startrev: 9d2fa7c63b7c
-.. branch: hroncok/fix-typeerror-str-does-not-support-the-b-1514414905375
-
-Fix for bytestrings in console repl
-
-.. branch: py3-winreg
-
-Update winreg module to use unicode, wide-strings
-
-.. branch: cpyext-py3-instancemethod-attributes
-
-Add missing ``__doc__``, ``__module__``, ``__name__`` attributes to
-``instancemethod``
-
-.. branch: winapi
-
-Update support for _winapi cffi module for python3
-
-.. branch: py3.5-refactor-slots
-
-Refactor cpyext slots.
-
diff --git a/pypy/interpreter/astcompiler/optimize.py
b/pypy/interpreter/astcompiler/optimize.py
--- a/pypy/interpreter/astcompiler/optimize.py
+++ b/pypy/interpreter/astcompiler/optimize.py
@@ -5,7 +5,7 @@
from pypy.tool import stdlib_opcode as ops
from pypy.interpreter.error import OperationError
from rpython.rlib.unroll import unrolling_iterable
-from rpython.rlib.runicode import MAXUNICODE
+from rpython.rlib.rutf8 import MAXUNICODE
from rpython.rlib.objectmodel import specialize
diff --git a/pypy/interpreter/unicodehelper.py
b/pypy/interpreter/unicodehelper.py
--- a/pypy/interpreter/unicodehelper.py
+++ b/pypy/interpreter/unicodehelper.py
@@ -3,7 +3,7 @@
from pypy.interpreter.error import OperationError, oefmt
from rpython.rlib.objectmodel import specialize
from rpython.rlib.rstring import StringBuilder
-from rpython.rlib import rutf8, runicode
+from rpython.rlib import rutf8
from rpython.rlib.rarithmetic import r_uint, intmask
from rpython.rtyper.lltypesystem import rffi
from pypy.module.unicodedata import unicodedb
@@ -40,23 +40,6 @@
space.newtext(msg)]))
return raise_unicode_exception_encode
[email protected]()
-def encode_unicode_error_handler(space):
- # Fast version of the "strict" errors handler.
- def raise_unicode_exception_encode(errors, encoding, msg, uni,
- startingpos, endingpos):
- assert isinstance(uni, unicode)
- u_len = len(uni)
- utf8 = runicode.unicode_encode_utf8sp(uni, u_len)
- raise OperationError(space.w_UnicodeEncodeError,
- space.newtuple([space.newtext(encoding),
- space.newtext(utf8, u_len),
- space.newint(startingpos),
- space.newint(endingpos),
- space.newtext(msg)]))
- return u'', None, 0
- return raise_unicode_exception_encode
-
def default_error_encode(
errors, encoding, msg, u, startingpos, endingpos):
"""A default handler, for tests"""
@@ -1022,23 +1005,45 @@
return result.build()
[email protected]()
+def _encode_unicode_error_handler(space):
+ # Fast version of the "strict" errors handler.
+ from rpython.rlib import runicode
+ def raise_unicode_exception_encode(errors, encoding, msg, uni,
+ startingpos, endingpos):
+ assert isinstance(uni, unicode)
+ u_len = len(uni)
+ utf8 = runicode.unicode_encode_utf8sp(uni, u_len)
+ raise OperationError(space.w_UnicodeEncodeError,
+ space.newtuple([space.newtext(encoding),
+ space.newtext(utf8, u_len),
+ space.newint(startingpos),
+ space.newint(endingpos),
+ space.newtext(msg)]))
+ return u'', None, 0
+ return raise_unicode_exception_encode
+
+
def encode_utf8(space, uni, allow_surrogates=False):
# Note that Python3 tends to forbid *all* surrogates in utf-8.
# If allow_surrogates=True, then revert to the Python 2 behavior
# which never raises UnicodeEncodeError. Surrogate pairs are then
# allowed, either paired or lone. A paired surrogate is considered
# like the non-BMP character it stands for. See also *_utf8sp().
+ from rpython.rlib import runicode
assert isinstance(uni, unicode)
return runicode.unicode_encode_utf_8(
uni, len(uni), "strict",
- errorhandler=encode_unicode_error_handler(space),
+ errorhandler=_encode_unicode_error_handler(space),
allow_surrogates=allow_surrogates)
def encode_utf8sp(space, uni, allow_surrogates=True):
+ xxx
# Surrogate-preserving utf-8 encoding. Any surrogate character
# turns into its 3-bytes encoding, whether it is paired or not.
# This should always be reversible, and the reverse is
# decode_utf8sp().
+ from rpython.rlib import runicode
return runicode.unicode_encode_utf8sp(uni, len(uni))
def decode_utf8sp(space, string):
diff --git a/pypy/module/_codecs/__init__.py b/pypy/module/_codecs/__init__.py
--- a/pypy/module/_codecs/__init__.py
+++ b/pypy/module/_codecs/__init__.py
@@ -1,5 +1,4 @@
from pypy.interpreter.mixedmodule import MixedModule
-from rpython.rlib import runicode
from rpython.rlib.objectmodel import not_rpython
from pypy.module._codecs import interp_codecs
@@ -89,6 +88,7 @@
@not_rpython
def __init__(self, space, *args):
# mbcs codec is Windows specific, and based on rffi.
+ from rpython.rlib import runicode
if (hasattr(runicode, 'str_decode_mbcs')):
self.interpleveldefs['mbcs_encode'] = 'interp_codecs.mbcs_encode'
self.interpleveldefs['mbcs_decode'] = 'interp_codecs.mbcs_decode'
diff --git a/pypy/module/_codecs/interp_codecs.py
b/pypy/module/_codecs/interp_codecs.py
--- a/pypy/module/_codecs/interp_codecs.py
+++ b/pypy/module/_codecs/interp_codecs.py
@@ -2,7 +2,7 @@
from rpython.rlib import jit, rutf8
from rpython.rlib.objectmodel import we_are_translated, not_rpython
from rpython.rlib.rstring import StringBuilder, UnicodeBuilder
-from rpython.rlib import runicode
+from rpython.rlib.rutf8 import MAXUNICODE
from rpython.rlib.runicode import raw_unicode_escape_helper
from pypy.interpreter.error import OperationError, oefmt
@@ -649,7 +649,6 @@
return _call_codec(space, w_decoder, w_obj, "decoding", encoding, errors)
# ____________________________________________________________
-# delegation to runicode/unicodehelper
def _find_implementation(impl_name):
func = getattr(unicodehelper, impl_name)
@@ -721,7 +720,8 @@
]:
make_decoder_wrapper(decoder)
-if hasattr(unicodehelper, 'str_decode_mbcs'):
+from rpython.rlib import runicode
+if hasattr(runicode, 'str_decode_mbcs'):
make_encoder_wrapper('mbcs_encode')
make_decoder_wrapper('mbcs_decode')
diff --git a/pypy/module/cpyext/unicodeobject.py
b/pypy/module/cpyext/unicodeobject.py
--- a/pypy/module/cpyext/unicodeobject.py
+++ b/pypy/module/cpyext/unicodeobject.py
@@ -1,4 +1,5 @@
from rpython.rtyper.lltypesystem import rffi, lltype
+from rpython.rlib import rstring
from rpython.rlib.rarithmetic import widen
from rpython.rlib import rstring, runicode, rutf8
from rpython.tool.sourcetools import func_renamer
@@ -266,7 +267,8 @@
@cpython_api([], Py_UNICODE, error=CANNOT_FAIL)
def PyUnicode_GetMax(space):
"""Get the maximum ordinal for a Unicode character."""
- return runicode.UNICHR(runicode.MAXUNICODE)
+ from rpython.rlib import runicode, rutf8
+ return runicode.UNICHR(rutf8.MAXUNICODE)
@cts.decl("int _PyUnicode_Ready(PyObject *unicode)", error=-1)
def _PyUnicode_Ready(space, w_obj):
diff --git a/pypy/module/sys/vm.py b/pypy/module/sys/vm.py
--- a/pypy/module/sys/vm.py
+++ b/pypy/module/sys/vm.py
@@ -3,7 +3,7 @@
"""
from rpython.rlib import jit
-from rpython.rlib.runicode import MAXUNICODE
+from rpython.rlib.rutf8 import MAXUNICODE
from pypy.interpreter import gateway
from pypy.interpreter.error import oefmt
diff --git a/pypy/module/zlib/interp_zlib.py b/pypy/module/zlib/interp_zlib.py
--- a/pypy/module/zlib/interp_zlib.py
+++ b/pypy/module/zlib/interp_zlib.py
@@ -351,6 +351,9 @@
else:
string, finished, unused_len = result
self._save_unconsumed_input(data, finished, unused_len)
+ if finished:
+ rzlib.inflateEnd(self.stream)
+ self.stream = rzlib.null_stream
return space.newbytes(string)
diff --git a/pypy/objspace/std/bytesobject.py b/pypy/objspace/std/bytesobject.py
--- a/pypy/objspace/std/bytesobject.py
+++ b/pypy/objspace/std/bytesobject.py
@@ -438,10 +438,6 @@
space.check_buf_flags(flags, True)
return SimpleView(StringBuffer(self._value))
- def descr_encode(self, space, w_encoding=None, w_errors=None):
- w_uni = self.descr_decode(space, space.newtext('ascii'),
space.newtext('strict'))
- return space.call_method(w_uni, 'encode', w_encoding, w_errors)
-
def descr_getbuffer(self, space, w_flags):
#from pypy.objspace.std.bufferobject import W_Buffer
#return W_Buffer(StringBuffer(self._value))
diff --git a/pypy/objspace/std/unicodeobject.py
b/pypy/objspace/std/unicodeobject.py
--- a/pypy/objspace/std/unicodeobject.py
+++ b/pypy/objspace/std/unicodeobject.py
@@ -988,13 +988,12 @@
descr_rmul = descr_mul
def _get_index_storage(self):
- # XXX write the correct jit.elidable
- if self._index_storage == rutf8.null_storage():
- storage = rutf8.create_utf8_index_storage(self._utf8, self._length)
- else:
- storage = self._index_storage
- if not jit.isconstant(self):
- self._index_storage = storage
+ return jit.conditional_call_elidable(self._index_storage,
+ W_UnicodeObject._compute_index_storage, self)
+
+ def _compute_index_storage(self):
+ storage = rutf8.create_utf8_index_storage(self._utf8, self._length)
+ self._index_storage = storage
return storage
def _getitem_result(self, space, index):
diff --git a/rpython/annotator/binaryop.py b/rpython/annotator/binaryop.py
--- a/rpython/annotator/binaryop.py
+++ b/rpython/annotator/binaryop.py
@@ -727,6 +727,10 @@
return [get_setitem,
op.simple_call(get_setitem.result, v_idx, v_value)]
[email protected]_transform(SomeInstance)
+def contains_SomeInstance(annotator, v_ins, v_idx):
+ get_contains = op.getattr(v_ins, const('__contains__'))
+ return [get_contains, op.simple_call(get_contains.result, v_idx)]
class __extend__(pairtype(SomeIterator, SomeIterator)):
diff --git a/rpython/annotator/test/test_annrpython.py
b/rpython/annotator/test/test_annrpython.py
--- a/rpython/annotator/test/test_annrpython.py
+++ b/rpython/annotator/test/test_annrpython.py
@@ -4065,6 +4065,20 @@
assert len(a.translator.graphs) == 2 # fn, __setitem__
assert isinstance(s, annmodel.SomeInteger)
+ def test_instance_contains(self):
+ class A(object):
+ def __contains__(self, i):
+ return i & 1 == 0
+
+ def fn(i):
+ a = A()
+ return 0 in a and 1 not in a
+
+ a = self.RPythonAnnotator()
+ s = a.build_types(fn, [int])
+ assert len(a.translator.graphs) == 2 # fn, __contains__
+ assert isinstance(s, annmodel.SomeBool)
+
def test_instance_getslice(self):
class A(object):
def __getslice__(self, stop, start):
diff --git a/rpython/jit/backend/llsupport/regalloc.py
b/rpython/jit/backend/llsupport/regalloc.py
--- a/rpython/jit/backend/llsupport/regalloc.py
+++ b/rpython/jit/backend/llsupport/regalloc.py
@@ -1,4 +1,4 @@
-import os
+import sys
from rpython.jit.metainterp.history import Const, REF, JitCellToken
from rpython.rlib.objectmodel import we_are_translated, specialize
from rpython.jit.metainterp.resoperation import rop, AbstractValue
@@ -67,6 +67,7 @@
@specialize.arg(1)
def foreach(self, function, arg):
+ # XXX unused?
node = self.master_node
while node is not None:
function(arg, node.val)
@@ -297,12 +298,12 @@
def is_still_alive(self, v):
# Check if 'v' is alive at the current position.
# Return False if the last usage is strictly before.
- return self.longevity[v][1] >= self.position
+ return self.longevity[v].last_usage >= self.position
def stays_alive(self, v):
# Check if 'v' stays alive after the current position.
# Return False if the last usage is before or at position.
- return self.longevity[v][1] > self.position
+ return self.longevity[v].last_usage > self.position
def next_instruction(self, incr=1):
self.position += incr
@@ -319,7 +320,7 @@
self._check_type(v)
if isinstance(v, Const):
return
- if v not in self.longevity or self.longevity[v][1] <= self.position:
+ if v not in self.longevity or self.longevity[v].last_usage <=
self.position:
if v in self.reg_bindings:
self.free_regs.append(self.reg_bindings[v])
del self.reg_bindings[v]
@@ -355,7 +356,7 @@
for v in self.reg_bindings:
if v not in self.longevity:
llop.debug_print(lltype.Void, "variable %s not in
longevity\n" % v.repr({}))
- assert self.longevity[v][1] > self.position
+ assert self.longevity[v].last_usage > self.position
def try_allocate_reg(self, v, selected_reg=None, need_lower_byte=False):
""" Try to allocate a register, if we have one free.
@@ -366,6 +367,9 @@
returns allocated register or None, if not possible.
"""
self._check_type(v)
+ if isinstance(v, TempVar):
+ self.longevity[v] = Lifetime(self.position, self.position)
+ # YYY all subtly similar code
assert not isinstance(v, Const)
if selected_reg is not None:
res = self.reg_bindings.get(v, None)
@@ -385,42 +389,55 @@
loc = self.reg_bindings.get(v, None)
if loc is not None and loc not in self.no_lower_byte_regs:
return loc
- for i in range(len(self.free_regs) - 1, -1, -1):
- reg = self.free_regs[i]
- if reg not in self.no_lower_byte_regs:
- if loc is not None:
- self.free_regs[i] = loc
- else:
- del self.free_regs[i]
- self.reg_bindings[v] = reg
- return reg
- return None
+ free_regs = [reg for reg in self.free_regs
+ if reg not in self.no_lower_byte_regs]
+ newloc = self.longevity.try_pick_free_reg(
+ self.position, v, free_regs)
+ if newloc is None:
+ return None
+ self.free_regs.remove(newloc)
+ if loc is not None:
+ self.free_regs.append(loc)
+ self.reg_bindings[v] = newloc
+ return newloc
try:
return self.reg_bindings[v]
except KeyError:
- if self.free_regs:
- loc = self.free_regs.pop()
- self.reg_bindings[v] = loc
- return loc
+ loc = self.longevity.try_pick_free_reg(
+ self.position, v, self.free_regs)
+ if loc is None:
+ return None
+ self.reg_bindings[v] = loc
+ self.free_regs.remove(loc)
+ return loc
- def _spill_var(self, v, forbidden_vars, selected_reg,
+ def _spill_var(self, forbidden_vars, selected_reg,
need_lower_byte=False):
- v_to_spill = self._pick_variable_to_spill(v, forbidden_vars,
+ v_to_spill = self._pick_variable_to_spill(forbidden_vars,
selected_reg, need_lower_byte=need_lower_byte)
loc = self.reg_bindings[v_to_spill]
+ self._sync_var_to_stack(v_to_spill)
del self.reg_bindings[v_to_spill]
- if self.frame_manager.get(v_to_spill) is None:
- newloc = self.frame_manager.loc(v_to_spill)
- self.assembler.regalloc_mov(loc, newloc)
return loc
- def _pick_variable_to_spill(self, v, forbidden_vars, selected_reg=None,
- need_lower_byte=False):
- """ Slightly less silly algorithm.
- """
- cur_max_age = -1
+ def _pick_variable_to_spill(self, forbidden_vars, selected_reg=None,
+ need_lower_byte=False, regs=None):
+
+ # try to spill a variable that has no further real usages, ie that only
+ # appears in failargs or in a jump
+ # if that doesn't exist, spill the variable that has a real_usage that
+ # is the furthest away from the current position
+
+ # YYY check for fixed variable usages
+ if regs is None:
+ regs = self.reg_bindings.keys()
+
+ cur_max_use_distance = -1
+ position = self.position
candidate = None
- for next in self.reg_bindings:
+ cur_max_age_failargs = -1
+ candidate_from_failargs = None
+ for next in regs:
reg = self.reg_bindings[next]
if next in forbidden_vars:
continue
@@ -431,13 +448,25 @@
continue
if need_lower_byte and reg in self.no_lower_byte_regs:
continue
- max_age = self.longevity[next][1]
- if cur_max_age < max_age:
- cur_max_age = max_age
- candidate = next
- if candidate is None:
- raise NoVariableToSpill
- return candidate
+ lifetime = self.longevity[next]
+ if lifetime.is_last_real_use_before(position):
+ # this variable has no "real" use as an argument to an op left
+ # it is only used in failargs, and maybe in a jump. spilling is
+ # fine
+ max_age = lifetime.last_usage
+ if cur_max_age_failargs < max_age:
+ cur_max_age_failargs = max_age
+ candidate_from_failargs = next
+ else:
+ use_distance = lifetime.next_real_usage(position) - position
+ if cur_max_use_distance < use_distance:
+ cur_max_use_distance = use_distance
+ candidate = next
+ if candidate_from_failargs is not None:
+ return candidate_from_failargs
+ if candidate is not None:
+ return candidate
+ raise NoVariableToSpill
def force_allocate_reg(self, v, forbidden_vars=[], selected_reg=None,
need_lower_byte=False):
@@ -450,12 +479,12 @@
"""
self._check_type(v)
if isinstance(v, TempVar):
- self.longevity[v] = (self.position, self.position)
+ self.longevity[v] = Lifetime(self.position, self.position)
loc = self.try_allocate_reg(v, selected_reg,
need_lower_byte=need_lower_byte)
if loc:
return loc
- loc = self._spill_var(v, forbidden_vars, selected_reg,
+ loc = self._spill_var(forbidden_vars, selected_reg,
need_lower_byte=need_lower_byte)
prev_loc = self.reg_bindings.get(v, None)
if prev_loc is not None:
@@ -468,7 +497,7 @@
self.bindings_to_frame_reg[v] = None
def force_spill_var(self, var):
- self._sync_var(var)
+ self._sync_var_to_stack(var)
try:
loc = self.reg_bindings[var]
del self.reg_bindings[var]
@@ -502,7 +531,7 @@
if selected_reg in self.free_regs:
self.assembler.regalloc_mov(immloc, selected_reg)
return selected_reg
- loc = self._spill_var(v, forbidden_vars, selected_reg)
+ loc = self._spill_var(forbidden_vars, selected_reg)
self.free_regs.append(loc)
self.assembler.regalloc_mov(immloc, loc)
return loc
@@ -523,6 +552,7 @@
loc = self.force_allocate_reg(v, forbidden_vars, selected_reg,
need_lower_byte=need_lower_byte)
if prev_loc is not loc:
+ self.assembler.num_reloads += 1
self.assembler.regalloc_mov(prev_loc, loc)
return loc
@@ -530,15 +560,7 @@
reg = self.reg_bindings[from_v]
del self.reg_bindings[from_v]
self.reg_bindings[to_v] = reg
-
- def _move_variable_away(self, v, prev_loc):
- if self.free_regs:
- loc = self.free_regs.pop()
- self.reg_bindings[v] = loc
- self.assembler.regalloc_mov(prev_loc, loc)
- else:
- loc = self.frame_manager.loc(v)
- self.assembler.regalloc_mov(prev_loc, loc)
+ return reg
def force_result_in_reg(self, result_v, v, forbidden_vars=[]):
""" Make sure that result is in the same register as v.
@@ -548,41 +570,39 @@
self._check_type(result_v)
self._check_type(v)
if isinstance(v, Const):
- if self.free_regs:
- loc = self.free_regs.pop()
- else:
- loc = self._spill_var(v, forbidden_vars, None)
- self.assembler.regalloc_mov(self.convert_to_imm(v), loc)
- self.reg_bindings[result_v] = loc
- return loc
- if v not in self.reg_bindings:
- # v not in a register. allocate one for result_v and move v there
- prev_loc = self.frame_manager.loc(v)
- loc = self.force_allocate_reg(result_v, forbidden_vars)
- self.assembler.regalloc_mov(prev_loc, loc)
- return loc
- if self.longevity[v][1] > self.position:
- # we need to find a new place for variable v and
- # store result in the same place
- loc = self.reg_bindings[v]
- del self.reg_bindings[v]
- if self.frame_manager.get(v) is None:
- self._move_variable_away(v, loc)
- self.reg_bindings[result_v] = loc
- else:
- self._reallocate_from_to(v, result_v)
- loc = self.reg_bindings[result_v]
- return loc
+ result_loc = self.force_allocate_reg(result_v, forbidden_vars)
+ self.assembler.regalloc_mov(self.convert_to_imm(v), result_loc)
+ return result_loc
+ v_keeps_living = self.longevity[v].last_usage > self.position
+ # there are two cases where we should allocate a new register for
+ # result:
+ # 1) v is itself not in a register
+ # 2) v keeps on being live. if there is a free register, we need a move
+ # anyway, so we can use force_allocate_reg on result_v to make sure any
+ # fixed registers are used
+ if (v not in self.reg_bindings or (v_keeps_living and self.free_regs)):
+ v_loc = self.loc(v)
+ result_loc = self.force_allocate_reg(result_v, forbidden_vars)
+ self.assembler.regalloc_mov(v_loc, result_loc)
+ return result_loc
+ if v_keeps_living:
+ # since there are no free registers, v needs to go to the stack.
+ # sync it there.
+ self._sync_var_to_stack(v)
+ return self._reallocate_from_to(v, result_v)
- def _sync_var(self, v):
+ def _sync_var_to_stack(self, v):
+ self.assembler.num_spills += 1
if not self.frame_manager.get(v):
reg = self.reg_bindings[v]
to = self.frame_manager.loc(v)
self.assembler.regalloc_mov(reg, to)
+ else:
+ self.assembler.num_spills_to_existing += 1
# otherwise it's clean
def _bc_spill(self, v, new_free_regs):
- self._sync_var(v)
+ self._sync_var_to_stack(v)
new_free_regs.append(self.reg_bindings.pop(v))
def before_call(self, force_store=[], save_all_regs=0):
@@ -650,7 +670,7 @@
move_or_spill = []
for v, reg in self.reg_bindings.items():
- max_age = self.longevity[v][1]
+ max_age = self.longevity[v].last_usage
if v not in force_store and max_age <= self.position:
# variable dies
del self.reg_bindings[v]
@@ -671,45 +691,32 @@
else:
# this is a register like eax/rax, which needs either
# spilling or moving.
- move_or_spill.append((v, max_age))
+ move_or_spill.append(v)
if len(move_or_spill) > 0:
- while len(self.free_regs) > 0:
- new_reg = self.free_regs.pop()
- if new_reg in self.save_around_call_regs:
- new_free_regs.append(new_reg) # not this register...
- continue
- # This 'new_reg' is suitable for moving a candidate to.
- # Pick the one with the smallest max_age. (This
- # is one step of a naive sorting algo, slow in theory,
- # but the list should always be very small so it
- # doesn't matter.)
- best_i = 0
- smallest_max_age = move_or_spill[0][1]
- for i in range(1, len(move_or_spill)):
- max_age = move_or_spill[i][1]
- if max_age < smallest_max_age:
- best_i = i
- smallest_max_age = max_age
- v, max_age = move_or_spill.pop(best_i)
- # move from 'reg' to 'new_reg'
+ free_regs = [reg for reg in self.free_regs
+ if reg not in self.save_around_call_regs]
+ # chose which to spill using the usual spill heuristics
+ while len(move_or_spill) > len(free_regs):
+ v = self._pick_variable_to_spill([], regs=move_or_spill)
+ self._bc_spill(v, new_free_regs)
+ move_or_spill.remove(v)
+ assert len(move_or_spill) <= len(free_regs)
+ for v in move_or_spill:
+ # search next good reg
+ new_reg = None
+ while True:
+ new_reg = self.free_regs.pop()
+ if new_reg in self.save_around_call_regs:
+ new_free_regs.append(new_reg) # not this register...
+ continue
+ break
+ assert new_reg is not None # must succeed
reg = self.reg_bindings[v]
- if not we_are_translated():
- if move_or_spill:
- assert max_age <= min([_a for _, _a in move_or_spill])
- assert reg in save_sublist
- assert reg in self.save_around_call_regs
- assert new_reg not in self.save_around_call_regs
+ self.assembler.num_moves_calls += 1
self.assembler.regalloc_mov(reg, new_reg)
self.reg_bindings[v] = new_reg # change the binding
new_free_regs.append(reg)
- #
- if len(move_or_spill) == 0:
- break
- else:
- # no more free registers to move to, spill the rest
- for v, max_age in move_or_spill:
- self._bc_spill(v, new_free_regs)
# re-add registers in 'new_free_regs', but in reverse order,
# so that the last ones (added just above, from
@@ -772,7 +779,7 @@
# of COND_CALL don't accept a cc as input
if next_op.getarg(0) is not op:
return False
- if self.longevity[op][1] > i + 1:
+ if self.longevity[op].last_usage > i + 1:
return False
if opnum != rop.COND_CALL:
if op in operations[i + 1].getfailargs():
@@ -786,61 +793,295 @@
descr = op.getdescr()
assert isinstance(descr, JitCellToken)
if op.numargs() == 2:
- self.rm._sync_var(op.getarg(1))
+ self.rm._sync_var_to_stack(op.getarg(1))
return [self.loc(op.getarg(0)), self.fm.loc(op.getarg(1))]
else:
assert op.numargs() == 1
return [self.loc(op.getarg(0))]
+# ____________________________________________________________
+
+
+
+UNDEF_POS = -42
+
+class Lifetime(object):
+ def __init__(self, definition_pos=UNDEF_POS, last_usage=UNDEF_POS):
+ # all positions are indexes into the operations list
+
+ # the position where the variable is defined
+ self.definition_pos = definition_pos
+ # the position where the variable is last used. this includes failargs
+ # and jumps
+ self.last_usage = last_usage
+
+ # *real* usages, ie as an argument to an operation (as opposed to jump
+ # arguments or in failargs)
+ self.real_usages = None
+
+ # fixed registers are positions where the variable *needs* to be in a
+ # specific register
+ self.fixed_positions = None
+
+ # another Lifetime that lives after the current one that would like to
+ # share a register with this variable
+ self.share_with = None
+
+ # the other lifetime will have this variable set to self.definition_pos
+ self._definition_pos_shared = UNDEF_POS
+
+ def last_usage_including_sharing(self):
+ while self.share_with is not None:
+ self = self.share_with
+ return self.last_usage
+
+ def is_last_real_use_before(self, position):
+ if self.real_usages is None:
+ return True
+ return self.real_usages[-1] <= position
+
+ def next_real_usage(self, position):
+ assert position >= self.definition_pos
+ # binary search
+ l = self.real_usages
+ low = 0
+ high = len(l)
+ if position >= l[-1]:
+ return -1
+ while low < high:
+ mid = low + (high - low) // 2 # no overflow ;-)
+ if position < l[mid]:
+ high = mid
+ else:
+ low = mid + 1
+ return l[low]
+
+ def definition_pos_shared(self):
+ if self._definition_pos_shared != UNDEF_POS:
+ return self._definition_pos_shared
+ else:
+ return self.definition_pos
+
+ def fixed_register(self, position, reg):
+ """ registers a fixed register use for the variable at position in
+ register reg. returns the position from where on the register should be
+ held free. """
+ assert self.definition_pos <= position <= self.last_usage
+ if self.fixed_positions is None:
+ self.fixed_positions = []
+ res = self.definition_pos_shared()
+ else:
+ assert position > self.fixed_positions[-1][0]
+ res = self.fixed_positions[-1][0]
+ self.fixed_positions.append((position, reg))
+ return res
+
+ def find_fixed_register(self, opindex):
+ # XXX could use binary search
+ if self.fixed_positions is not None:
+ for (index, reg) in self.fixed_positions:
+ if opindex <= index:
+ return reg
+ if self.share_with is not None:
+ return self.share_with.find_fixed_register(opindex)
+
+ def _check_invariants(self):
+ assert self.definition_pos <= self.last_usage
+ if self.real_usages is not None:
+ assert sorted(self.real_usages) == self.real_usages
+ assert self.last_usage >= max(self.real_usages)
+ assert self.definition_pos < min(self.real_usages)
+
+ def __repr__(self):
+ if self.fixed_positions:
+ s = " " + ", ".join("@%s in %s" % (index, reg) for (index, reg) in
self.fixed_positions)
+ else:
+ s = ""
+ return "%s:%s(%s)%s" % (self.definition_pos, self.real_usages,
self.last_usage, s)
+
+
+class FixedRegisterPositions(object):
+ def __init__(self, register):
+ self.register = register
+
+ self.index_lifetimes = []
+
+ def fixed_register(self, opindex, definition_pos):
+ if self.index_lifetimes:
+ assert opindex > self.index_lifetimes[-1][0]
+ self.index_lifetimes.append((opindex, definition_pos))
+
+ def free_until_pos(self, opindex):
+ # XXX could use binary search
+ for (index, definition_pos) in self.index_lifetimes:
+ if opindex <= index:
+ if definition_pos >= opindex:
+ return definition_pos
+ else:
+ # the variable doesn't exist or didn't make it into the
+ # register despite being defined already. so we don't care
+ # too much, and can say that the variable is free until
+ # index
+ return index
+ return sys.maxint
+
+ def __repr__(self):
+ return "%s: fixed at %s" % (self.register, self.index_lifetimes)
+
+
+class LifetimeManager(object):
+ def __init__(self, longevity):
+ self.longevity = longevity
+
+ # dictionary maps register to FixedRegisterPositions
+ self.fixed_register_use = {}
+
+ def fixed_register(self, opindex, register, var=None):
+ """ Tell the LifetimeManager that variable var *must* be in register at
+ operation opindex. var can be None, if no variable at all can be in
+ that register at the point."""
+ if var is None:
+ definition_pos = opindex
+ else:
+ varlifetime = self.longevity[var]
+ definition_pos = varlifetime.fixed_register(opindex, register)
+ if register not in self.fixed_register_use:
+ self.fixed_register_use[register] =
FixedRegisterPositions(register)
+ self.fixed_register_use[register].fixed_register(opindex,
definition_pos)
+
+ def try_use_same_register(self, v0, v1):
+ """ Try to arrange things to put v0 and v1 into the same register.
+ v0 must be defined before v1"""
+ # only works in limited situations now
+ longevityvar0 = self[v0]
+ longevityvar1 = self[v1]
+ assert longevityvar0.definition_pos < longevityvar1.definition_pos
+ if longevityvar0.last_usage != longevityvar1.definition_pos:
+ return # not supported for now
+ longevityvar0.share_with = longevityvar1
+ longevityvar1._definition_pos_shared =
longevityvar0.definition_pos_shared()
+
+ def longest_free_reg(self, position, free_regs):
+ """ for every register in free_regs, compute how far into the future
+ that register can remain free, according to the constraints of the
+ fixed registers. Find the register that is free the longest. Return a
+ tuple (reg, free_until_pos). """
+ free_until_pos = {}
+ max_free_pos = position
+ best_reg = None
+ # reverse for compatibility with old code
+ for i in range(len(free_regs) - 1, -1, -1):
+ reg = free_regs[i]
+ fixed_reg_pos = self.fixed_register_use.get(reg, None)
+ if fixed_reg_pos is None:
+ return reg, sys.maxint
+ else:
+ free_until_pos = fixed_reg_pos.free_until_pos(position)
+ if free_until_pos > max_free_pos:
+ best_reg = reg
+ max_free_pos = free_until_pos
+ return best_reg, max_free_pos
+
+ def free_reg_whole_lifetime(self, position, v, free_regs):
+ """ try to find a register from free_regs for v at position that's
+ free for the whole lifetime of v. pick the one that is blocked first
+ *after* the lifetime of v. """
+ longevityvar = self[v]
+ min_fixed_use_after = sys.maxint
+ best_reg = None
+ unfixed_reg = None
+ for reg in free_regs:
+ fixed_reg_pos = self.fixed_register_use.get(reg, None)
+ if fixed_reg_pos is None:
+ unfixed_reg = reg
+ continue
+ use_after = fixed_reg_pos.free_until_pos(position)
+ if use_after <= longevityvar.last_usage_including_sharing():
+ # can't fit
+ continue
+ assert use_after >= longevityvar.last_usage_including_sharing()
+ if use_after < min_fixed_use_after:
+ best_reg = reg
+ min_fixed_use_after = use_after
+ if best_reg is not None:
+ return best_reg
+
+ # no fitting fixed registers. pick a non-fixed one
+ return unfixed_reg
+
+ def try_pick_free_reg(self, position, v, free_regs):
+ if not free_regs:
+ return None
+ longevityvar = self[v]
+ # check whether there is a fixed register and whether it's free
+ reg = longevityvar.find_fixed_register(position)
+ if reg is not None and reg in free_regs:
+ return reg
+
+ # try to find a register that's free for the whole lifetime of v
+ # pick the one that is blocked first *after* the lifetime of v
+ loc = self.free_reg_whole_lifetime(position, v, free_regs)
+ if loc is not None:
+ return loc
+
+ # can't fit v completely, so pick the register that's free the longest
+ loc, free_until = self.longest_free_reg(position, free_regs)
+ if loc is not None:
+ return loc
+ # YYY could check whether it's best to spill v here, but hard
+ # to do in the current system
+ return None
+
+ def __contains__(self, var):
+ return var in self.longevity
+
+ def __getitem__(self, var):
+ return self.longevity[var]
+
+ def __setitem__(self, var, val):
+ self.longevity[var] = val
+
def compute_vars_longevity(inputargs, operations):
- # compute a dictionary that maps variables to index in
- # operations that is a "last-time-seen"
-
- # returns a pair longevity/useful. Non-useful variables are ones that
- # never appear in the assembler or it does not matter if they appear on
- # stack or in registers. Main example is loop arguments that go
- # only to guard operations or to jump or to finish
- last_used = {}
- last_real_usage = {}
+ # compute a dictionary that maps variables to Lifetime information
+ # if a variable is not in the dictionary, it's operation is dead because
+ # it's side-effect-free and the result is unused
+ longevity = {}
for i in range(len(operations)-1, -1, -1):
op = operations[i]
- if op.type != 'v':
- if op not in last_used and rop.has_no_side_effect(op.opnum):
+ opnum = op.getopnum()
+ if op not in longevity:
+ if op.type != 'v' and rop.has_no_side_effect(opnum):
+ # result not used, operation has no side-effect, it can be
+ # removed
continue
- opnum = op.getopnum()
+ longevity[op] = Lifetime(definition_pos=i, last_usage=i)
+ else:
+ longevity[op].definition_pos = i
for j in range(op.numargs()):
arg = op.getarg(j)
if isinstance(arg, Const):
continue
- if arg not in last_used:
- last_used[arg] = i
+ if arg not in longevity:
+ lifetime = longevity[arg] = Lifetime(last_usage=i)
+ else:
+ lifetime = longevity[arg]
if opnum != rop.JUMP and opnum != rop.LABEL:
- if arg not in last_real_usage:
- last_real_usage[arg] = i
+ if lifetime.real_usages is None:
+ lifetime.real_usages = []
+ lifetime.real_usages.append(i)
if rop.is_guard(op.opnum):
for arg in op.getfailargs():
if arg is None: # hole
continue
assert not isinstance(arg, Const)
- if arg not in last_used:
- last_used[arg] = i
+ if arg not in longevity:
+ longevity[arg] = Lifetime(last_usage=i)
#
- longevity = {}
- for i, arg in enumerate(operations):
- if arg.type != 'v' and arg in last_used:
- assert not isinstance(arg, Const)
- assert i < last_used[arg]
- longevity[arg] = (i, last_used[arg])
- del last_used[arg]
for arg in inputargs:
assert not isinstance(arg, Const)
- if arg not in last_used:
- longevity[arg] = (-1, -1)
- else:
- longevity[arg] = (0, last_used[arg])
- del last_used[arg]
- assert len(last_used) == 0
+ if arg not in longevity:
+ longevity[arg] = Lifetime(-1, -1)
if not we_are_translated():
produced = {}
@@ -851,9 +1092,15 @@
if not isinstance(arg, Const):
assert arg in produced
produced[op] = None
-
- return longevity, last_real_usage
+ for lifetime in longevity.itervalues():
+ if lifetime.real_usages is not None:
+ lifetime.real_usages.reverse()
+ if not we_are_translated():
+ lifetime._check_invariants()
+ return LifetimeManager(longevity)
+
+# YYY unused?
def is_comparison_or_ovf_op(opnum):
return rop.is_comparison(opnum) or rop.is_ovf(opnum)
diff --git a/rpython/jit/backend/llsupport/test/test_regalloc.py
b/rpython/jit/backend/llsupport/test/test_regalloc.py
--- a/rpython/jit/backend/llsupport/test/test_regalloc.py
+++ b/rpython/jit/backend/llsupport/test/test_regalloc.py
@@ -1,9 +1,19 @@
import py
+import sys
from rpython.jit.metainterp.history import ConstInt, INT, FLOAT
-from rpython.jit.backend.llsupport.regalloc import FrameManager, LinkedList
-from rpython.jit.backend.llsupport.regalloc import RegisterManager as
BaseRegMan
+from rpython.jit.metainterp.history import BasicFailDescr, TargetToken
+from rpython.jit.metainterp.resoperation import rop
from rpython.jit.metainterp.resoperation import InputArgInt, InputArgRef,\
InputArgFloat
+from rpython.jit.backend.detect_cpu import getcpuclass
+from rpython.jit.backend.llsupport.regalloc import FrameManager, LinkedList
+from rpython.jit.backend.llsupport.regalloc import RegisterManager as
BaseRegMan,\
+ Lifetime as RealLifetime, UNDEF_POS, BaseRegalloc,
compute_vars_longevity,\
+ LifetimeManager
+from rpython.jit.tool.oparser import parse
+from rpython.jit.codewriter.effectinfo import EffectInfo
+from rpython.rtyper.lltypesystem import lltype
+from rpython.rtyper.annlowlevel import llhelper
def newboxes(*values):
return [InputArgInt(v) for v in values]
@@ -11,33 +21,61 @@
def newrefboxes(count):
return [InputArgRef() for _ in range(count)]
+def Lifetime(definition_pos=UNDEF_POS, last_usage=UNDEF_POS,
+ real_usages=UNDEF_POS):
+ if real_usages == UNDEF_POS:
+ real_usages = last_usage
+ lifetime = RealLifetime(definition_pos, last_usage)
+ if isinstance(real_usages, int):
+ real_usages = [real_usages]
+ lifetime.real_usages = real_usages
+ return lifetime
+
+
def boxes_and_longevity(num):
res = []
longevity = {}
for i in range(num):
box = InputArgInt(0)
res.append(box)
- longevity[box] = (0, 1)
+ longevity[box] = Lifetime(0, 1)
return res, longevity
class FakeReg(object):
def __init__(self, i):
self.n = i
+ def _getregkey(self):
+ return self.n
+ def is_memory_reference(self):
+ return False
def __repr__(self):
return 'r%d' % self.n
r0, r1, r2, r3 = [FakeReg(i) for i in range(4)]
+r4, r5, r6, r7, r8, r9 = [FakeReg(i) for i in range(4, 10)]
+
regs = [r0, r1, r2, r3]
class RegisterManager(BaseRegMan):
all_regs = regs
+
+ def __init__(self, longevity, frame_manager=None, assembler=None):
+ if isinstance(longevity, dict):
+ longevity = LifetimeManager(longevity)
+ BaseRegMan.__init__(self, longevity, frame_manager, assembler)
+
def convert_to_imm(self, v):
return v
class FakeFramePos(object):
def __init__(self, pos, box_type):
self.pos = pos
+ self.value = pos
self.box_type = box_type
+ def _getregkey(self):
+ return ~self.value
+ def is_memory_reference(self):
+ return True
def __repr__(self):
return 'FramePos<%d,%s>' % (self.pos, self.box_type)
def __eq__(self, other):
@@ -66,17 +104,311 @@
assert isinstance(loc, FakeFramePos)
return loc.pos
+class FakeCPU(object):
+ def get_baseofs_of_frame_field(self):
+ return 0
+
class MockAsm(object):
def __init__(self):
self.moves = []
-
+ self.emitted = []
+ self.cpu = FakeCPU()
+
+ # XXX register allocation statistics to be removed later
+ self.num_moves_calls = 0
+ self.num_moves_jump = 0
+ self.num_spills = 0
+ self.num_spills_to_existing = 0
+ self.num_reloads = 0
+
+ self.preamble_num_moves_calls = 0
+ self.preamble_num_moves_jump = 0
+ self.preamble_num_spills = 0
+ self.preamble_num_spills_to_existing = 0
+ self.preamble_num_reloads = 0
+
def regalloc_mov(self, from_loc, to_loc):
self.moves.append((from_loc, to_loc))
+ self.emitted.append(("move", to_loc, from_loc))
+
+
+def test_lifetime_next_real_usage():
+ lt = RealLifetime(0, 1000)
+ lt.real_usages = [0, 1, 5, 10, 24, 35, 55, 56, 57, 90, 92, 100]
+ for i in range(100):
+ next = lt.next_real_usage(i)
+ assert next in lt.real_usages
+ assert next > i
+ assert lt.real_usages[lt.real_usages.index(next) - 1] <= i
+ assert lt.next_real_usage(100) == -1
+ assert lt.next_real_usage(101) == -1
+
+def test_fixed_position():
+ b0, b1, b2 = newboxes(0, 0, 0)
+ l0 = Lifetime(0, 5)
+ l1 = Lifetime(2, 9)
+ l2 = Lifetime(0, 9)
+ longevity = LifetimeManager({b0: l0, b1: l1, b2: l2})
+ longevity.fixed_register(1, r0, b0)
+ longevity.fixed_register(4, r2, b0)
+ longevity.fixed_register(5, r1, b1)
+ longevity.fixed_register(8, r1, b1)
+
+ assert l0.fixed_positions == [(1, r0), (4, r2)]
+ assert l1.fixed_positions == [(5, r1), (8, r1)]
+ assert l2.fixed_positions is None
+
+ fpr0 = longevity.fixed_register_use[r0]
+ fpr1 = longevity.fixed_register_use[r1]
+ fpr2 = longevity.fixed_register_use[r2]
+ assert r3 not in longevity.fixed_register_use
+ assert fpr0.index_lifetimes == [(1, 0)]
+ assert fpr1.index_lifetimes == [(5, 2), (8, 5)]
+ assert fpr2.index_lifetimes == [(4, 1)]
+
+def test_fixed_position_none():
+ b0, b1, b2 = newboxes(0, 0, 0)
+ l0 = Lifetime(0, 5)
+ l1 = Lifetime(2, 9)
+ l2 = Lifetime(0, 9)
+ longevity = LifetimeManager({b0: l0, b1: l1, b2: l2})
+ longevity.fixed_register(1, r0)
+ longevity.fixed_register(4, r2)
+ longevity.fixed_register(5, r1)
+ longevity.fixed_register(8, r1)
+
+ fpr0 = longevity.fixed_register_use[r0]
+ fpr1 = longevity.fixed_register_use[r1]
+ fpr2 = longevity.fixed_register_use[r2]
+ assert r3 not in longevity.fixed_register_use
+ assert fpr0.index_lifetimes == [(1, 1)]
+ assert fpr1.index_lifetimes == [(5, 5), (8, 8)]
+ assert fpr2.index_lifetimes == [(4, 4)]
+
+
+def test_free_until_pos_none():
+ longevity = LifetimeManager({})
+ longevity.fixed_register(5, r1, None)
+ longevity.fixed_register(8, r1, None)
+ longevity.fixed_register(35, r1, None)
+
+ fpr1 = longevity.fixed_register_use[r1]
+
+ assert fpr1.free_until_pos(0) == 5
+ assert fpr1.free_until_pos(1) == 5
+ assert fpr1.free_until_pos(2) == 5
+ assert fpr1.free_until_pos(3) == 5
+ assert fpr1.free_until_pos(4) == 5
+ assert fpr1.free_until_pos(5) == 5
+ assert fpr1.free_until_pos(10) == 35
+ assert fpr1.free_until_pos(20) == 35
+ assert fpr1.free_until_pos(30) == 35
+ assert fpr1.free_until_pos(36) == sys.maxint
+
+def test_free_until_pos():
+ b0, b1, b2 = newboxes(0, 0, 0)
+ l0 = Lifetime(0, 5)
+ l1 = Lifetime(2, 9)
+ l2 = Lifetime(30, 40)
+ longevity = LifetimeManager({b0: l0, b1: l1, b2: l2})
+ longevity.fixed_register(5, r1, b1)
+ longevity.fixed_register(8, r1, b1)
+ longevity.fixed_register(35, r1, b2)
+
+ fpr1 = longevity.fixed_register_use[r1]
+
+ # simple cases: we are before the beginning of the lifetime of the variable
+ # in the fixed register, then it's free until the definition of the
+ # variable
+ assert fpr1.free_until_pos(0) == 2
+ assert fpr1.free_until_pos(1) == 2
+ assert fpr1.free_until_pos(2) == 2
+ assert fpr1.free_until_pos(10) == 30
+ assert fpr1.free_until_pos(20) == 30
+ assert fpr1.free_until_pos(30) == 30
+
+ # after the fixed use, we are fine anyway
+ assert fpr1.free_until_pos(36) == sys.maxint
+ assert fpr1.free_until_pos(50) == sys.maxint
+
+ # asking for a position *after* the definition of the variable in the fixed
+ # register means the variable didn't make it into the fixed register, but
+ # at the latest by the use point it will have to go there
+ assert fpr1.free_until_pos(3) == 5
+ assert fpr1.free_until_pos(4) == 5
+ assert fpr1.free_until_pos(5) == 5
+ assert fpr1.free_until_pos(6) == 8
+ assert fpr1.free_until_pos(7) == 8
+ assert fpr1.free_until_pos(8) == 8
+ assert fpr1.free_until_pos(31) == 35
+ assert fpr1.free_until_pos(32) == 35
+ assert fpr1.free_until_pos(33) == 35
+ assert fpr1.free_until_pos(34) == 35
+ assert fpr1.free_until_pos(35) == 35
+
+def test_free_until_pos_different_regs():
+ b0, b1, b2 = newboxes(0, 0, 0)
+ l0 = Lifetime(0, 5)
+ l1 = Lifetime(2, 9)
+ l2 = Lifetime(30, 40)
+ longevity = LifetimeManager({b0: l0, b1: l1, b2: l2})
+ longevity.fixed_register(1, r0, b0)
+ longevity.fixed_register(4, r2, b0)
+ fpr2 = longevity.fixed_register_use[r2]
+ # the definition of b0 is before the other fixed register use of r0, so the
+ # earliest b0 can be in r2 is that use point at index 1
+ assert fpr2.free_until_pos(0) == 1
+
+
+def test_longest_free_reg():
+ b0, b1, b2 = newboxes(0, 0, 0)
+ l0 = Lifetime(0, 5)
+ l1 = Lifetime(2, 9)
+ l2 = Lifetime(30, 40)
+ longevity = LifetimeManager({b0: l0, b1: l1, b2: l2})
+ longevity.fixed_register(1, r0, b0)
+ longevity.fixed_register(4, r2, b0)
+ longevity.fixed_register(5, r1, b1)
+ longevity.fixed_register(8, r1, b1)
+ longevity.fixed_register(35, r1, b2)
+
+ assert longevity.longest_free_reg(0, [r0, r1, r2]) == (r1, 2)
+
+def test_try_pick_free_reg():
+ b0, b1, b2, b3, b4 = newboxes(0, 0, 0, 0, 0)
+ l0 = Lifetime(0, 4)
+ l1 = Lifetime(2, 20)
+ l2 = Lifetime(6, 20)
+ l3 = Lifetime(8, 20)
+ l4 = Lifetime(0, 10)
+ longevity = LifetimeManager({b0: l0, b1: l1, b2: l2, b3: l3, b4: l4})
+ longevity.fixed_register(3, r1, b1)
+ longevity.fixed_register(7, r2, b2)
+ longevity.fixed_register(9, r3, b3)
+
+ # a best fit
+ loc = longevity.try_pick_free_reg(0, b0, [r1, r2, r3, r4, r5])
+ assert loc is r2
+
+ # does not fit into any of the fixed regs, use a non-fixed one
+ loc = longevity.try_pick_free_reg(0, b4, [r5, r2, r3, r4, r1])
+ assert loc in [r4, r5]
+
+ # all available are fixed but var doesn't fit completely into any of these.
+ # pick the biggest interval
+ loc = longevity.try_pick_free_reg(0, b4, [r1, r2, r3])
+ assert loc is r3
+
+def test_try_pick_free_reg_bug():
+ b0, b1, b2, b3, b4 = newboxes(0, 0, 0, 0, 0)
+ l0 = Lifetime(10, 30)
+ l1 = Lifetime(0, 15)
+ longevity = LifetimeManager({b0: l0, b1: l1})
+ longevity.fixed_register(20, r0, b0)
+
+ # does not fit into r0, use r1
+ loc = longevity.try_pick_free_reg(0, b1, [r0, r1])
+ assert loc == r1
+
+def test_try_pick_free_reg_bug2():
+ b0, b1, b2, b3, b4 = newboxes(0, 0, 0, 0, 0)
+ l0 = Lifetime(1, 2)
+ l1 = Lifetime(2, 4)
+ longevity = LifetimeManager({b0: l0, b1: l1})
+ longevity.fixed_register(4, r1, b1)
+
+ # does not fit into r0, use r1
+ loc = longevity.try_pick_free_reg(0, b0, [r0, r1])
+ assert loc == r0
+
+def test_simple_coalescing():
+ b0, b1, b2, b3, b4 = newboxes(0, 0, 0, 0, 0)
+ l0 = Lifetime(0, 4)
+ l1 = Lifetime(4, 20)
+ l2 = Lifetime(4, 20)
+ longevity = LifetimeManager({b0: l0, b1: l1, b2: l2})
+ longevity.fixed_register(10, r1, b1)
+ longevity.fixed_register(10, r2, b2)
+ longevity.try_use_same_register(b0, b2)
+
+ loc = longevity.try_pick_free_reg(0, b0, [r0, r1, r2, r3, r4])
+ assert loc is r2
+
+def test_coalescing_blocks_regs_correctly():
+ b0, b1, b2, b3, b4 = newboxes(0, 0, 0, 0, 0)
+ l0 = Lifetime(10, 30)
+ l1 = Lifetime(30, 40)
+ l2 = Lifetime(30, 40)
+ l3 = Lifetime(0, 15)
+ l4 = Lifetime(0, 5)
+ longevity = LifetimeManager({b0: l0, b1: l1, b2: l2, b3: l3, b4: l4})
+ longevity.try_use_same_register(b0, b1)
+ longevity.fixed_register(35, r1, b1)
+ longevity.fixed_register(35, r2, b2)
+
+ loc = longevity.try_pick_free_reg(0, b3, [r1, r2])
+ # r2 is picked, otherwise b0 can't end up in r1
+ assert loc is r2
+
+ loc = longevity.try_pick_free_reg(0, b4, [r1, r2])
+ # r1 is picked, because b4 fits before b0
+ assert loc is r1
+
+def test_coalescing_non_fixed_regs():
+ b0, b1, b2, b3, b4 = newboxes(0, 0, 0, 0, 0)
+ l0 = Lifetime(0, 10)
+ l1 = Lifetime(10, 20)
+ l2 = Lifetime(25, 40)
+ l3 = Lifetime(15, 40)
+ longevity = LifetimeManager({b0: l0, b1: l1, b2: l2, b3: l3})
+ longevity.try_use_same_register(b0, b1)
+ longevity.fixed_register(35, r2, b2)
+ longevity.fixed_register(35, r3, b3)
+
+ loc = longevity.try_pick_free_reg(0, b0, [r1, r2, r3])
+ # r2 is picked, otherwise b1 can't end up in the same reg as b0
+ assert loc is r2
+
+
+def test_chained_coalescing():
+ # 5 + b4
+ # |
+ # 10 + b0 |
+ # | |
+ # | 15 +
+ # |
+ # +
+ # 20
_______________________________________________
pypy-commit mailing list
[email protected]
https://mail.python.org/mailman/listinfo/pypy-commit
