[pypy-svn] r39852 - pypy/extradoc/sprintinfo/trillke-2007
cfbolz at codespeak.net
cfbolz at codespeak.net
Sun Mar 4 11:24:11 CET 2007
Author: cfbolz
Date: Sun Mar 4 11:24:09 2007
New Revision: 39852
Modified:
pypy/extradoc/sprintinfo/trillke-2007/planning.txt
Log:
(all): planning for today
Modified: pypy/extradoc/sprintinfo/trillke-2007/planning.txt
==============================================================================
--- pypy/extradoc/sprintinfo/trillke-2007/planning.txt (original)
+++ pypy/extradoc/sprintinfo/trillke-2007/planning.txt Sun Mar 4 11:24:09 2007
@@ -5,29 +5,41 @@
People present: Michael, Arre, Stephan, Anto, Guido, Armin, Maciek, Bea,
Samuele, Georg, Alexander, Holger, Carl Friedrich, Anders
+post-eu funding planning at 16.30, Sunday
+
Tutorial: Georg, Anders DONE
+ * prepare the post-eu funding planning meeting
+ (Carl Friedrich, Holger, Samuele)
+
* think a bit about what needs to be done for 1.0
- * move some of the external functions to the new interface (Samuele, Anto)
+ * move some of the external functions to the new interface
+ IN-PROGRESS, but as good as it's going to get during the sprint
+
+ * fixing translation of the taint space IN-PROGRESS
* Fix the rdict bug DONE
+
+ * fix the new rdict bug
* run application on top of PyPy and turn them into benchmarks: Guido,
Christian
- templess is now a pypy benchmark (5-6x)
- docutils is now a pypy benchmark (5-6x)
- gadfly is working with old-style classes
+ - some progress on making it work with new-style classes
* object optimizations: ropes, list optimizations, cuckoo hashing,
tree-based dicts: after the tutorial, Holger, Armin, Carl Friedrich, Georg
- - ropes DONE (but slow)
- - chunked list DONE
- - multilist broken?! (Alexander, Carl Friedrich)
+ - ropes BUGGY (but not so slow any more)
+ - chunked list DONE (but slower)
+ - multilist not broken any more DONE
- O(sqrt(n)) DONE but is slow and uses too much memory
- - move string interpolation to rpython (Holger, Stephan)
- - chained hash table (Armin), NEARLY DONE
- - discuss global dict lookup ideas (Armin, Georg)
+ - move string interpolation to rpython IN-PROGRESS (Holger, Stephan)
+ - chained hash table, DONE (two times slower on micro-benchmarks)
+ - discuss global dict lookup ideas PFFFF
+ - improve getattr chains DISCUSSED (Anto, Holger to write it down)
* improve the error messages of the rtyper
@@ -37,6 +49,7 @@
(Michael, Holger)
* hunt the strange speed bottleneck of gencli: DONE
+ - wrote RPython microbenchmarks in the process
* fix application-level stackless code: DONE
@@ -48,17 +61,38 @@
* fix the javascript backend work in internet explorer: IN-PROGRESS
(Guido, Maciek)
- * make the usage of the js backend nicer (Guido, Maciek)
+ * make the usage of the js backend nicer SOME PROGRESS (Guido, Maciek)
+
+ * discuss code sharing between the js backend and gencli: DONE
- * discuss code sharing between the js backend and gencli: MOSTLY DONE
+ * make genllvm less of a FrankenBackend (Richard, half of Michael)
+ LOTS OF PROGRESS
+ - ripped out old code
+ - worked in the direction of supporting rctypes
- * make genllvm less of a FrankenBackend (Richard, half of Michael), IN-PROGRESS
+ * moved remaining useful stuff from translator/asm to jit/codegen/ppc
+ DONE
* use "const" for immutable structures
* have a pypy with a gil and threading that works
- * object-level locking for a gil-less pypy
+ * object-level locking for a gil-less pypy: boehm seems to be a problem
* use no locking, multiple interpreters and shared memory instead of
object-level locks
+
+ * solved some of the issues in the tracker: -m working, <Ctrl>-C works
+ better, preventing changes to builtin types DONE
+
+ * start supporting __index__ IN-PROGRESS
+ (Georg, Alexander, Anders, help from Armin)
+
+ * re-think the interface of composable coroutines
+ (Armin, Christian, Stephan)
+
+ * incorporate Richard's rsocket changes (Arre, Richard)
+
+ * kill strdict, move normal dicts to the attic, keep only multidict
+ (Carl Friedrich, ...)
+
More information about the Pypy-commit
mailing list