notes cleanup

Benchmark.hs

@@ -13,6 +13,21 @@ import Control.Concurrent.Actors
 import Control.Concurrent.Chan.Split
 import TreeExample
 
+
+-- THIS IS A WORK-IN-PROGRESS. PLEASE IGNORE. --
+
+
+{- NOTES & IDEAS
+
+look again at simple case expr for nil
+look at eventlog with equal number inserts and queries (are queries very slow?)
+look at bottom of heap profile, halfway through. Get that flat.
+what about using forkOnIO to keep child actors on the same OS thread?
+better way to handle recursion in Behavior?
+unpacking MVars?
+
+
+ -}
 -- put benchmarking & optimizing on hold until we can figure out how to get
 -- consistent results....
 
@@ -31,7 +46,6 @@ main0 = defaultMain [
   --       ]
     ]
 
--- 484:
 main :: IO ()
 main = testActors (2^10 - 1, 1000) >>= print
 

Control/Concurrent/Actors.lhs

@@ -181,24 +181,26 @@ This module exports a simple, idiomatic implementation of the Actor Model.
 TODO
 -----
 
-0.4
+0.4.1
     - performance tuning / benchmarking:
-        + look at interface file: ghc -ddump-hi Control/Concurrent/Actors.hs -O -c
-        + remove current PRAGMA
-        - close browser and everything, do a fake quick benchmark to get clock info
-        - be more controlled about the source lists (do once before defaultMain), use 'evaluate'
-        - run with +RTS -s and make sure everything is 0
-        - see if case-based nil is better
-        - try storing the same chan (observable sharing) in each node, and use for streaming 
-           send an MVar with messages for the query operation
-        - get accurate baseline comparison between actors and set
-        - use INLINABLE
-        - test again with SPECIALIZE instead
-        - try adding INLINE to all with higher-order args (or higher-order newtype wrappers)
-           and make sure our LHS looks good for inlining
-        - specialize `Action i (Behavior i)` or allow lots of unfolding... ? Optimize those loops, somehow. Rewrite rules?
+        - first optimize TreeExample, by way of Benchmark.hs
+        - criterion and profiling w/r/t lib.:
+            - play with underlying Behavior Monad stack?
+            - be more controlled about the source lists (do once before defaultMain), use 'evaluate'
+            - run with +RTS -s and make sure everything is 0
+            - see if case-based nil is better
+            - try storing the same chan (observable sharing) in each node, and use for streaming 
+               send an MVar with messages for the query operation
+            - get accurate baseline comparison between actors and set
+            - use INLINABLE
+            - test again with SPECIALIZE instead
+            - try adding INLINE to all with higher-order args (or higher-order newtype wrappers)
+               and make sure our LHS looks good for inlining
+            - specialize `Action i (Behavior i)` or allow lots of unfolding... ? Optimize those loops, somehow. Rewrite rules?
+            - look at "let floating" and INLINEABLE to get functions with "fully-applied (syntactically) LHS"
+        - split-chan ChItem in heap profile -hy
         - take a look at threadscope for random tree test
-        - look at "let floating" and INLINEABLE to get functions with "fully-applied (syntactically) LHS"
+        - forkOnIO to keep communicating actors on same HEC?
         - compare with previous version (cp to /tmp to use previous version)
 
 

simple-actors.cabal

@@ -38,8 +38,6 @@ Description:         simple-actors is an EDSL-style library for writing
                      .
                      - old @spawnReading@ removed
                      .
-                     - performance improvements and benchmarks
-                     .
 
 
 -- The license under which the package is released.