Carl Friedrich Bolz-Tereick pushed to branch branch/jit-smaller-miframes at PyPy / pypy
Commits: 3d4a0029 by Carl Friedrich Bolz-Tereick at 2022-04-25T19:16:54+02:00 a branch to experiment using GCREFs to do type erasure using GCREFs (instead of the base instance struct, which excludes erasing things like strs and lists of lists) --HG-- branch : gcref-based-type-erasure - - - - - 6f164215 by Carl Friedrich Bolz-Tereick at 2022-04-25T19:18:15+02:00 cache on the rtyper, not globally --HG-- branch : gcref-based-type-erasure - - - - - 2a6f7de3 by Carl Friedrich Bolz-Tereick at 2022-04-25T20:30:38+02:00 - make gcref erasure opt in, and use it only for lists for now - fix dict.items problems --HG-- branch : gcref-based-type-erasure - - - - - 474e07d4 by Carl Friedrich Bolz-Tereick at 2022-04-25T20:31:02+02:00 not everything has an ll_str method --HG-- branch : gcref-based-type-erasure - - - - - 16c93bcc by Carl Friedrich Bolz-Tereick at 2022-04-25T21:48:18+02:00 fix str/unicode split, rsplit, splitlines --HG-- branch : gcref-based-type-erasure - - - - - a77edf14 by Carl Friedrich Bolz-Tereick at 2022-04-26T12:35:22+02:00 making the values gcrefs just works --HG-- branch : gcref-based-type-erasure - - - - - 1ddec8ed by Carl Friedrich Bolz-Tereick at 2022-04-26T15:54:53+02:00 support a dummy value for gcref --HG-- branch : gcref-based-type-erasure - - - - - 5a5030f1 by Carl Friedrich Bolz-Tereick at 2022-05-26T15:21:07+02:00 fix test on pypy --HG-- branch : gcref-based-type-erasure - - - - - b9a4fb48 by Carl Friedrich Bolz-Tereick at 2022-05-26T21:38:11+02:00 don't use gcrefs when using refcounting, it can't work --HG-- branch : gcref-based-type-erasure - - - - - 24766744 by Carl Friedrich Bolz-Tereick at 2022-05-26T21:38:27+02:00 merge default --HG-- branch : gcref-based-type-erasure - - - - - 5af4ae79 by Carl Friedrich Bolz-Tereick at 2022-09-12T22:12:41+02:00 #3805 implement a sub-quadratic algorithm for int(<some huge string>), O(len(s) ** 1.58) adapted from pure python code in https://github.com/python/cpython/pull/96673 --HG-- branch : rbigint-fromstr-subquadratic - - - - - 902a801b by Carl Friedrich Bolz-Tereick at 2022-09-13T20:55:03+02:00 fix rounding, thanks bjorn martinsson --HG-- branch : rbigint-fromstr-subquadratic - - - - - e10e338f by Carl Friedrich Bolz-Tereick at 2022-09-13T22:27:38+02:00 test for previous commit --HG-- branch : rbigint-fromstr-subquadratic - - - - - 9dd7b25e by Carl Friedrich Bolz-Tereick at 2022-09-15T12:49:31+02:00 merge default --HG-- branch : rbigint-fromstr-subquadratic - - - - - c431c200 by Carl Friedrich Bolz-Tereick at 2022-09-24T12:32:19+02:00 intermediate checkin: resurrect the lopsided karatsuba multiplication case which was wrongly removed it helps a lot for unbalanced huge multiplications. before this commit, unbalanced multiplications would fall back to using O(n**2) base case multiplication, no matter their size not quite sure yet which algorithm to use, CPython's or my own. both fix the complexity, my own seems unexpectedly faster. --HG-- branch : rbigint-fromstr-subquadratic - - - - - dde396a4 by Carl Friedrich Bolz-Tereick at 2022-09-24T12:35:57+02:00 oops --HG-- branch : rbigint-fromstr-subquadratic - - - - - fc674e42 by Carl Friedrich Bolz-Tereick at 2022-09-24T13:22:17+02:00 save a useless multiplication by 0 in the base case --HG-- branch : rbigint-fromstr-subquadratic - - - - - 50615b01 by Carl Friedrich Bolz-Tereick at 2022-09-24T21:04:27+02:00 introduce accessors for size and sign --HG-- branch : rbigint-fromstr-subquadratic - - - - - 8e82e35e by Carl Friedrich Bolz-Tereick at 2022-10-06T15:28:13+02:00 merge default --HG-- branch : rbigint-fromstr-subquadratic - - - - - 11a53919 by Carl Friedrich Bolz-Tereick at 2022-10-06T15:33:13+02:00 Backed out changeset e4c06197fb2d my ideas in that direction didn't help and it breaks other things --HG-- branch : rbigint-fromstr-subquadratic - - - - - 8731e5e2 by Carl Friedrich Bolz-Tereick at 2022-10-09T17:12:39+02:00 merge default --HG-- branch : rbigint-fromstr-subquadratic - - - - - a0b25a34 by Carl Friedrich Bolz-Tereick at 2022-10-09T21:25:18+02:00 implement lopsided mul in the naive way, in my benchmarks it's consistently faster than CPython's approach --HG-- branch : rbigint-fromstr-subquadratic - - - - - 89fdef2c by Carl Friedrich Bolz-Tereick at 2022-10-09T21:25:50+02:00 some more hypothesis tests for string conversion --HG-- branch : rbigint-fromstr-subquadratic - - - - - 713f9727 by Carl Friedrich Bolz-Tereick at 2022-10-09T21:46:38+02:00 make rbigint.floordiv and rbigint.mod also use rbigint.divmod, to benefit from its better complexity for huge inputs --HG-- branch : rbigint-fromstr-subquadratic - - - - - 2dfe03c5 by Matti Picus at 2023-05-04T00:09:13+03:00 start 7.3.13 release, create 7.3.12 release notes - - - - - 26794f45 by Matti Picus at 2023-05-04T06:50:36+03:00 Add a release note. Hightlights still missing - - - - - 10786622 by Matti Picus at 2023-05-04T07:07:53+03:00 typos - - - - - 609d0f44 by Matti Picus at 2023-05-08T08:31:36+03:00 properly cast first argument of pthread_kill to pthread_t, used in python3 - - - - - b5b5062d by Matti Picus at 2023-05-11T15:18:15+03:00 update contributors (one new contributor since the last release) - - - - - 19284cbe by Matti Picus at 2023-05-11T15:25:29+03:00 update release note - - - - - 7f8e25d6 by Matti Picus at 2023-05-11T15:30:38+03:00 update repackage script - - - - - 47014b90 by Matti Picus at 2023-05-11T16:29:42+03:00 update release note with progress on issue 3280 (grpcio) - - - - - f3083223 by Matti Picus at 2023-05-12T08:32:42+03:00 Added tag release-pypy3.9-v7.3.12rc1 for changeset 256b3867989d - - - - - 8f61b810 by Matti Picus at 2023-05-12T08:33:05+03:00 Added tag release-pypy2.7-v7.3.12rc1 for changeset 087971d00098 - - - - - 3b753fad by Matti Picus at 2023-05-12T08:33:36+03:00 Added tag release-pypy3.10-v7.3.12rc1 for changeset 633f94e00474 - - - - - cc71c180 by Stefano Rivera at 2023-05-13T18:38:54-04:00 Tests: Ignore NaN mismatch on app tests under cPython 2.7 - - - - - 3bbb78a3 by Stefano Rivera at 2023-05-13T19:27:30-04:00 Backed out changeset e169c8518a07 I hadn't tested it. I just assumed it was the issue, it wasn't. - - - - - 749a60e0 by Matti Picus at 2023-05-14T09:21:30+03:00 refactor install page to boost conda - - - - - 79f38d7e by Matti Picus at 2023-05-14T09:25:50+03:00 fix release notes, add to index - - - - - 6533d24a by Matti Picus at 2023-05-14T12:02:47+03:00 rephrase slightly - - - - - 97957124 by Stefano Rivera at 2023-05-14T08:09:24-04:00 Backed out changeset b9b73c670a2e Backed out the wrong changeset. Doh. - - - - - cebc4f06 by Stefano Rivera at 2023-05-14T08:31:16-04:00 Backed out changeset 3f8f71f89399 This was the changeset I *meant* to back out earlier. - - - - - e771915d by Stefano Rivera at 2023-05-14T08:36:31-04:00 Use C99 NAN for Py_NAN This uses the standard `NAN` macro provided by C99, as cPython does since (gh-104263). And drops support for Py_NO_NAN. Resolves a test failure in AppTestFloatMacros.test_Py_NAN for Debian. - - - - - 0b41cadd by Matti Picus at 2023-05-15T08:58:22+03:00 update versions.json for 7.3.12rc1 - - - - - 7948e2df by Matti Picus at 2023-05-15T15:14:02+03:00 update release note - - - - - b96d3c4f by Carl Friedrich Bolz-Tereick at 2023-05-18T14:39:43+02:00 merge default --HG-- branch : rbigint-fromstr-subquadratic - - - - - babe43a0 by Carl Friedrich Bolz-Tereick at 2023-05-18T16:26:56+02:00 fix import - - - - - b88b532d by Carl Friedrich Bolz-Tereick at 2023-05-18T16:29:09+02:00 merge heads --HG-- branch : rbigint-fromstr-subquadratic - - - - - 4f8eb614 by Carl Friedrich Bolz-Tereick at 2023-05-18T17:35:28+02:00 some cleanups in the tests --HG-- branch : rbigint-fromstr-subquadratic - - - - - 193767c6 by Carl Friedrich Bolz-Tereick at 2023-05-18T22:04:12+02:00 some additional hypothesis tests that don't rely on the underlying long implementation --HG-- branch : rbigint-fromstr-subquadratic - - - - - 3e71db3d by Carl Friedrich Bolz-Tereick at 2023-05-19T09:10:42+02:00 fix wrong test - - - - - 46c30169 by Carl Friedrich Bolz-Tereick at 2023-05-19T09:22:10+02:00 - ouch, one of the tests was not actually checking anything - add a test to check that the big divmod path is actually used --HG-- branch : rbigint-fromstr-subquadratic - - - - - cca76a1f by Carl Friedrich Bolz-Tereick at 2023-05-20T12:00:19+02:00 some more tests for fromrarith_int --HG-- branch : rbigint-fromstr-subquadratic - - - - - b6b4fcdc by Carl Friedrich Bolz-Tereick at 2023-05-20T12:00:38+02:00 after some more testing: we can reduce the minimum size for when to use the new algorithm for rbigint.fromstr --HG-- branch : rbigint-fromstr-subquadratic - - - - - 001e4396 by Carl Friedrich Bolz-Tereick at 2023-05-20T12:13:53+02:00 use a string builder --HG-- branch : rbigint-fromstr-subquadratic - - - - - 5e853322 by Carl Friedrich Bolz-Tereick at 2023-05-20T20:18:31+02:00 reduce the runtime of this hypothesis test --HG-- branch : rbigint-fromstr-subquadratic - - - - - 4f7edf0c by Carl Friedrich Bolz-Tereick at 2023-05-23T08:49:18+02:00 merge rbigint-fromstr-subquadratic implement the base 10 string-to-int conversion using a divide an conquer algorithm with complexity O(n**1.58). The algorithm is due to Bjorn Martinsson. In the process, I discovered that the "lopsided" case of karatsuba multiplication was removed for no good reason at some point. I re-measured and reimplemented it. - - - - - 69d07ce7 by Matti Picus at 2023-05-24T13:44:23+03:00 update release note - - - - - 1954fe02 by Matti Picus at 2023-05-24T13:48:59+03:00 make note appear - - - - - e1698c85 by Matti Picus at 2023-05-24T13:58:12+03:00 typo - - - - - dd485269 by Matti Picus at 2023-05-24T21:00:36+03:00 Added tag release-pypy2.7-v7.3.12rc2 for changeset 4ac174a992a3 - - - - - 51a67f76 by Matti Picus at 2023-05-24T21:01:05+03:00 Added tag release-pypy3.9-v7.3.12rc2 for changeset a6c2a04c0d03 - - - - - 1b9d1d15 by Matti Picus at 2023-05-24T21:01:23+03:00 Added tag release-pypy3.10-v7.3.12rc2 for changeset 07561e2940ea - - - - - 009e2e91 by Matti Picus at 2023-05-29T00:05:35+03:00 update versions.json for the rc2 release - - - - - 99a7389e by Carl Friedrich Bolz-Tereick at 2023-05-30T14:17:11+02:00 try to get rid of the weird _cache mechanism in the History class --HG-- branch : jit-history-no-cache - - - - - 155a1b3d by Carl Friedrich Bolz-Tereick at 2023-05-30T14:37:10+02:00 fix the somewhat fiddly exception stuff --HG-- branch : jit-history-no-cache - - - - - c14ddc55 by Carl Friedrich Bolz-Tereick at 2023-05-30T14:53:51+02:00 fix test --HG-- branch : jit-history-no-cache - - - - - a28ccd5c by Carl Friedrich Bolz-Tereick at 2023-05-30T15:15:39+02:00 remove copy_value_from --HG-- branch : jit-history-no-cache - - - - - bd353a7b by Carl Friedrich Bolz-Tereick at 2023-05-30T15:34:13+02:00 fix --HG-- branch : jit-history-no-cache - - - - - c123e4e6 by Carl Friedrich Bolz-Tereick at 2023-05-30T16:17:34+02:00 add the value to the *FrontendOp constructors --HG-- branch : jit-history-no-cache - - - - - 68a32b71 by Carl Friedrich Bolz-Tereick at 2023-05-30T21:01:11+02:00 fix virtualizables maybe --HG-- branch : jit-history-no-cache - - - - - 83f640bb by Matti Picus at 2023-05-31T14:47:29+03:00 update openssl (1.1.1u, 3.0.9) and lzma (5.2.10) - - - - - b320d3c2 by Carl Friedrich Bolz-Tereick at 2023-05-31T18:01:30+02:00 a few fixes --HG-- branch : jit-history-no-cache - - - - - 5b2e7691 by Carl Friedrich Bolz-Tereick at 2023-05-31T21:17:31+02:00 fixes --HG-- branch : jit-history-no-cache - - - - - 18e44485 by Carl Friedrich Bolz-Tereick at 2023-05-31T21:18:11+02:00 merge jit-history-no-cache: small refactoring in the History class, it can now use the proper opencoder encoding of a trace immediately, not only after the inputargs are known. this is a small simplification only. - - - - - 15b62b11 by Carl Friedrich Bolz-Tereick at 2023-05-31T21:20:31+02:00 merge heads - - - - - 07841690 by Carl Friedrich Bolz-Tereick at 2023-05-31T21:36:43+02:00 d77a1ac14a7e leads to the two extra operations, which are a bit annoying but not that costly - - - - - 0edbc28f by Carl Friedrich Bolz-Tereick at 2023-06-01T22:00:20+02:00 don't give all descrs a ei_index and a descr_index field, almost all don't have that - - - - - a683b5c6 by Carl Friedrich Bolz-Tereick at 2023-06-09T09:53:23+02:00 fix comment - - - - - 0d085188 by Carl Friedrich Bolz-Tereick at 2023-06-07T18:28:17+02:00 inline get_field_updater to allow FieldUpdater to be alloc-removed everywhere - - - - - 7c0b4b58 by Carl Friedrich Bolz-Tereick at 2023-06-13T13:19:44+02:00 improve warmup a little bit: - directly put the two most common opcodes into the loop - add a special case for the int_add variant with a constant argument - - - - - cf744857 by Carl Friedrich Bolz-Tereick at 2023-06-13T14:12:50+02:00 use the rbigint.add_int (etc) shortcuts for the int+int overflowing operations as well - - - - - 2c8582d3 by Carl Friedrich Bolz-Tereick at 2023-06-13T15:21:32+02:00 remove some old remnant - - - - - 1561bf26 by Carl Friedrich Bolz-Tereick at 2023-06-13T18:09:05+02:00 do one less list copy for record_known_result and cond_call - - - - - 60e0c97e by Carl Friedrich Bolz-Tereick at 2023-06-13T22:11:36+02:00 optimize the goto_if_not_*: - don't call replace_box for fresh boxes that can't be stored anywhere yet - make the "same box" shortcut faster - - - - - 519f1f7e by Matti Picus at 2023-06-15T12:26:38+03:00 update release note - - - - - 7b3db6f2 by Matti Picus at 2023-06-15T21:59:46+03:00 Added tag release-pypy2.7-v7.3.12 for changeset 8d509266596a - - - - - 979b661c by Matti Picus at 2023-06-15T22:00:19+03:00 Added tag release-pypy3.9-v7.3.12 for changeset 3f3f2298ddc5 - - - - - 0e07b01f by Matti Picus at 2023-06-15T22:00:47+03:00 Added tag release-pypy3.10-v7.3.12 for changeset af44d0b8114c - - - - - 0f2487ab by Matti Picus at 2023-06-16T08:24:08+03:00 finalize release - - - - - 16d44faa by Carl Friedrich Bolz-Tereick at 2023-06-17T15:02:49+02:00 small quality-of-life improvement in C-level profilers etc: try to invent a sensible name for the smallfuncset pbc dispatchers - - - - - 9aed601f by Carl Friedrich Bolz-Tereick at 2023-06-17T16:38:18+02:00 merge heads - - - - - f8bda533 by Carl Friedrich Bolz-Tereick at 2023-06-17T17:03:04+02:00 merge default --HG-- branch : gcref-based-type-erasure - - - - - 64577e1c by Carl Friedrich Bolz-Tereick at 2023-06-20T09:02:11+02:00 add and fix some tests --HG-- branch : gcref-based-type-erasure - - - - - 0d23e89e by Carl Friedrich Bolz-Tereick at 2023-06-22T13:53:08+02:00 fix this test, which starts failing because some other tests use the same GcArray type to store pointers as l1, l2, l3 --HG-- branch : gcref-based-type-erasure - - - - - 1e95e48d by Carl Friedrich Bolz-Tereick at 2023-06-22T19:07:16+02:00 make odict.items no longer a oopspec, it creates rtyping trouble and serves no benefit --HG-- branch : gcref-based-type-erasure - - - - - 4bf369c3 by Carl Friedrich Bolz-Tereick at 2023-06-26T17:30:45+02:00 make some tagged int tests work --HG-- branch : gcref-based-type-erasure - - - - - fa28c1d5 by Carl Friedrich Bolz-Tereick at 2023-06-26T19:54:07+02:00 merge gcref-based-type-erasure: do type-erasure for rpython dicts and lists more generally, unifying all gced pointer types into one implementation (as opposed to just unifying all *instances* together, but not eg strings). This makes pypy-c ~700KiB smaller. - - - - - a7738d96 by Carl Friedrich Bolz-Tereick at 2023-06-27T10:26:18+02:00 merge default --HG-- branch : jit-smaller-miframes - - - - - 30 changed files: - .hgtags - LICENSE - lib_pypy/pypy_tools/build_cffi_imports.py - pypy/doc/conf.py - pypy/doc/contributor.rst - pypy/doc/index-of-release-notes.rst - pypy/doc/install.rst - pypy/doc/release-v7.3.11.rst - + pypy/doc/release-v7.3.12.rst - pypy/module/cpyext/include/patchlevel.h - pypy/module/cpyext/include/pymath.h - pypy/module/pypyjit/test_pypy_c/test_misc.py - pypy/module/sys/version.py - pypy/objspace/std/intobject.py - pypy/objspace/std/test/test_intobject.py - pypy/tool/release/check_versions.py - pypy/tool/release/repackage.sh - pypy/tool/release/versions.json - rpython/doc/conf.py - rpython/jit/backend/llgraph/runner.py - rpython/jit/backend/llsupport/descr.py - rpython/jit/backend/x86/runner.py - rpython/jit/codewriter/effectinfo.py - rpython/jit/codewriter/liveness.py - rpython/jit/codewriter/support.py - rpython/jit/codewriter/test/test_effectinfo.py - rpython/jit/metainterp/heapcache.py - rpython/jit/metainterp/history.py - rpython/jit/metainterp/opencoder.py - rpython/jit/metainterp/optimizeopt/bridgeopt.py View it on Heptapod: https://foss.heptapod.net/pypy/pypy/-/compare/2347dc64673df7f48013e21ba567a419a5b25418...a7738d9636076cc865cde7ce215be0ec6fec728a -- View it on Heptapod: https://foss.heptapod.net/pypy/pypy/-/compare/2347dc64673df7f48013e21ba567a419a5b25418...a7738d9636076cc865cde7ce215be0ec6fec728a You're receiving this email because of your account on foss.heptapod.net.
_______________________________________________ pypy-commit mailing list -- pypy-commit@python.org To unsubscribe send an email to pypy-commit-le...@python.org https://mail.python.org/mailman3/lists/pypy-commit.python.org/ Member address: arch...@mail-archive.com