Author: Armin Rigo <ar...@tunes.org> Branch: stmgc-c8 Changeset: r78270:8d3e51e892af Date: 2015-06-23 17:25 +0200 http://bitbucket.org/pypy/pypy/changeset/8d3e51e892af/
Log: updates diff --git a/TODO b/TODO --- a/TODO +++ b/TODO @@ -6,6 +6,9 @@ ------------------------------------------------------------ fuse the two 32bit setfield_gc for stmflags & tid in the jit +(note: unclear if it makes a major difference: a single +64bit setfield of an immediate that doesn't fit 32bit +requires two instructions too) ------------------------------------------------------------ @@ -13,6 +16,10 @@ rarely break if there are not threads running in parallel. But we need to break sometimes in order to run finalizers... +IMPROVED, but we should check if we break often enough to run +finaliers from time to time, or if we really make infinite +transactions + ------------------------------------------------------------ maybe statically optimize away some stm_become_inevitable(), there @@ -98,7 +105,8 @@ ------------------------------------------------------------ -change the limit of 1 GB +change the memory limit of a few GBs, try to replace it with something +larger or even dynamic ------------------------------------------------------------ @@ -139,15 +147,6 @@ ------------------------------------------------------------ -dicts: have an implementation that follows the principles in -stmgc/hashtable/design.txt - ------------------------------------------------------------- - -replace "atomic transactions" with better management of thread.locks. - ------------------------------------------------------------- - stm_read(p125) cond_call_gc_wb_array(p125...) # don't need the stm_read maybe? _______________________________________________ pypy-commit mailing list pypy-commit@python.org https://mail.python.org/mailman/listinfo/pypy-commit
