Author: Matti Picus <matti.pi...@gmail.com> Branch: extradoc Changeset: r5886:3b5b787ccbcf Date: 2018-04-09 14:25 +0200 http://bitbucket.org/pypy/extradoc/changeset/3b5b787ccbcf/
Log: add historical info from sprint diff --git a/sprintinfo/leysin-winter-2018/planning.txt b/sprintinfo/leysin-winter-2018/planning.txt new file mode 100644 --- /dev/null +++ b/sprintinfo/leysin-winter-2018/planning.txt @@ -0,0 +1,100 @@ +What we were working on - Sunday +====================== + +- finish unicode-utf8 - need to finish merge to default +- document the different rpython decorators like enforceargs, signature, and then interp-level unwrap_spec - arianna +- move more things to extra-tests, especially lib_pypy tests +- py3 test runners are too complicated - flub +- see what's missing to get PyGTK running (see also PyGObject and https://github.com/pygobject/pycairo/issues/90 ) - mjacob PAUSED +- revdb: bring it up to date, improve usability, other improvements (xoraxax) +- more cffi tutorial/overview rewrite - jdb + + +Plans for Monday +============= + +- cffi tutorial/overview rewrite - jdb + - document the different rpython decorators like signature, and then interp-level unwrap_spec - arianna DROPPED +- JIT topics: guard-compatible, and the subsequent research project to save and reuse traces across processes (remi) + -> https://docs.google.com/spreadsheets/d/1bIx8NwHOo5DYTOwnenOncN0YKsV2nknvzaHWJazqPnY/edit?usp=sharing + +- revdb: bring it up to date, improve usability, other improvements (xoraxax, armin around, mjacob) +- finish unicode-utf8 - merge of default. Gave up with merge of py3.5 into a new branch off py3.5 ``unicode-utf8-py3.5`` +- cpyext performance (anto, jdb shadowing) +- cpyext cycle detection discussion (Stefan, Armin, anto) +- py3 test runners are too complicated - (flub, ronan) + +Tues +==== + +- clif and cppyy https://gist.github.com/mattip/76be19a4cffab6991ee957059814d95a NOBODY KNOWS +- sqlite3 failure in pypy3.6 branch: http://buildbot.pypy.org/summary/longrepr?testname=unmodified&builder=pypy-c-jit-linux-x86-64&build=5319&mod=lib-python.3.test.test_sqlite - arianna w/manuel SOME PROGRESS, waiting for PR +- continue revdb merge - hunting ll_assert bug (xoraxax, armin around, mjacob around) +- looking for cpyext improvements (anton, mattip) SPED UP TYPECHECKS, more to do + +- JIT topics: understand the performance characteristics of guard-compatible (remi) + - found cases where guard_compatible produces more bridges than it should + - attempted to use vtune, perf, valgrind to look at assembly, but that seems impossible +- presentation about how cffi works (arigo, jdb, flub, etc.) DONE +- cpyext cycle detection implementation (stefan) +- py3 test runners refactoring (ronan, flub) PROGRESS +- support for positional-only arguments? (mjacob) NO OPINION +- fix py3.6 lib-python tests - start with test_asyncgen (mjacob) IN-PROGRESS +- pygame windows binary wheel, appveyor pypy (Ren?) PROGRESS + +Wed - rest day +=========== + +- skiing (arigo, flub, antonio) +- hiking (remi) +- guard_compatible: found that our fix from yesterday was incomplete (remi) +- being sick, laying around (Ren?). +- cpyext: manage to call tp_traverse from the GC, find alternative approaches to collect cycles (stefan) + - use boehm for cpyext + - trace heaps individually: might need several cycles to collect everything (like now), trace at once: only need one cycle +- make new tests less slow + +Thursday +======== + +- guard_compatible (remi) - VTune integration working; new microbenchmarks; found a segfault +- pygame windows binary wheel, appveyor pypy, report differences to cpython in issues.(Ren?) almost done. - error reported, nightly compiled different to release compiled (release user, can not use nightly generated wheel) +- try gc.collect(1) near display.flip(), benchmark to test dropped frames. jit.incremental? import pypyjit... turn on/off (Ren?) - lots of graphs, found GC deallocating all at once, maybe fixed? +- test runner refactoring (ronan) IN PROGRESS +- cpyext: optimize tp_as_mapping slots (anto, ronan around, armin around even if he doesn't want to) MOSTLY DONE, optimized all the slots; more optimizations possible +- cpyext cycle detection: clean up code, continue with implementation (stefan) +- fix test_async DEBUGGING + +Friday +====== + +- guard_compatible debugging (remi) +- vtune support: finish and merge the branch, missing would be giving some better name to the loops than "rpy_loop1" +- Make minimal 'pygame.examples.framedrop' benchmark. check rect benchmark later on.(Ren?) +- finish windows pygame wheel (Ren?) +- cpyext cycle detection: remove generic_cpy_call and directly call tp_traverse, do further cleanup (stefan) +- try to finish and possibly merge the cpyext branch about slots (anto) +- investigate rawrefcount-free-early (anto, armin) +- test runner refactoring (ronan) + +All tasks +======= + +- make win32 builds green +- make packaging more like cpython/portable builds +- get CI builders for PyPy into mainstream projects (Numpy, Scipy, lxml, uwsgi) +- get more of scientific stack working (tensorflow?) +- cpyext performance improvements +- General 3.5 and 3.6 improvements +- update www.pypy.org, speed.pypy.org (web devs needed) +- go over projects at https://bitbucket.org/pypy, delete or document dead projects +- look at google's clif https://github.com/google/clif and try to make a backend for PyPy, how is it different from cppyy +- have a look at https://github.com/aguinet/dragonffi as well +- use cffi instead of ctypes in rpython +- pygame test failures +- what about vmprof ? +- separate cffi versioning from cpyext versioning? + + + + _______________________________________________ pypy-commit mailing list pypy-commit@python.org https://mail.python.org/mailman/listinfo/pypy-commit
