work in progress

[nbds] / todo

diff --git a/todo b/todo
index 807a31a35e9d2131743fe711675ed4a6f6bd046a..67b75dc6ac0eebafb42b1c188cd9eec75ca18acf 100644 (file)
--- a/todo
+++ b/todo
@@ -1,26 +1,31 @@
-+ fix makefile to compute dependency info as a side-effect of compilation (-MF)
-+ support integer keys for ht
-+ optimize tracing code, still too much overhead
-+ use NULL instead of a sentinal node in skiplist and list
-+ make the interfaces for all data structures consistent 
-+ make list and skiplist use string keys
-+ optimize integer keys
-+ ht_print()
-+ iterators
-- make rcu yield when its buffer gets full instead of throwing an assert
-- alternate memory reclamation schemes, hazard pointers and/or reference count
-- investigate 16 byte CAS; ht can store GUIDs inline instead of pointers to actual keys 
+memory reclamation
+------------------
+- allow threads to dynamically enter and exit rcu's token passing ring
+- augment rcu with heartbeat manager to kill and recover from stalled threads
+- make rcu try yielding when its buffer gets full
+- use alternate memory reclamation schemes: hazard pointers and/or reference counting
+
+quality
+-------
+- verify the memory management of keys in list, skiplist, and hashtable
+- transaction tests
+- validate function arguments in interface functions
 - document usage
 - document algorithms
-- port tests from lib-high-scale
-- 32 bit version of hashtable
-- verify list and skiplist work on 32 bit platforms
-- transaction tests
-- validate the arguments to interface functions
-- shortcut from write-set to entries/nodes
-- use a shared scan for write-set validation, similar to ht copy logic
-- characterize the performance of hashtable, list and skiplist
+
+optimization
+------------
+- investigate 16 byte CAS; ht can store GUIDs inline instead of pointers to actual keys 
+- txn write after write can just update the old update record instead of pushing a new one
+- use a shared scan for write-set validation in txn, similar to ht copy logic
 - experiment with the performance impact of not passing the hash between functions in ht
-- experiment with embedding the keys in the list/skiplist nodes
-- allow values of 0 to be inserted into maps (change DOES_NOT_EXIST to something else)
-- see if it's possible to rename nbd_malloc to malloc
+- experiment with embedding the nstring keys in the list/skiplist nodes
+- lower skiplist's high_water when the actual number of levels in use drops
+- non-power-of 2 sized hashtables for improved memory usage
+- mem2
+
+features
+--------
+- allow values of 0 to be inserted into maps (change DOES_NOT_EXIST to something other than 0)
+- read-committed type transactions
+- recycle free regions across size-classes and between threads