Commit Graph

13 Commits

Author SHA1 Message Date
Michael Walker
c4eefd4849 Avoid decisions which will immediately block.
If a decision will immediately block without changing the global state, then
there is no point in making it: no state will become reachable from it which
isn't reachable through some other option we have available.

This has three parts:

 - When the prefix runs out and the scheduler is making decisions, it filters
   out decisions which will immediately block.

 - When a subtree is being added, it records which decisions will immediately
   block.

 - When backtracking points are being added, it filters out ones in this block
   list.

This optimisation is likely to only be useful when threads are communicating a
lot. For instance, a `parMap id` is totally unaffected by this, but the test
cases drop from an average of 64 runs to 42.
2015-07-20 16:16:46 +01:00
Michael Walker
aadb27ea1f Implement sleep sets for BPOR.
This further improves performance in all cases, although it's still far worse
on some of the included tests, and I have yet to figure out why.

See also: "Partial-Order Methods for the Verication of Concurrent Systems"
  [Godefroid 1996]
2015-07-17 15:34:52 +01:00
Michael Walker
c12cbcf707 Implement BPOR for SCT, return to listy predicates.
This performs better with "real" code (the Par monad) but surprisingly does far
worse with the included tests! The next thing to do is implement the orthogonal
sleep sets algorithm to cut down on available choices even further and
hopefully correct this issue.

See also: "Bounded Partial-Order Reduction" [Coons, Musuvathi, McKinley 2013]
2015-07-16 22:32:30 +01:00
Michael Walker
23c350c4b1 Fix/ignore hlint warnings 2015-07-10 18:30:34 +01:00
Michael Walker
1c8720a635 Give schedulers one-step lookahead 2015-07-10 18:27:50 +01:00
Michael Walker
89ec649d92 Don't pre-empt guaranteed blocks.
Pre-empting an action which is guaranteed to block is just the same
as letting it block and then inserting a regular non-pre-emptive
context switch.
2015-07-10 15:49:23 +01:00
Michael Walker
2ca7337b01 Avoid scheduling decisions which immediately block (in PB-bounded runner) 2015-07-08 20:06:15 +01:00
Michael Walker
6e01de2e85 Make available the full state of every CVar to SCT runners 2015-07-08 19:15:18 +01:00
Michael Walker
9ca54fea17 Include one-step lookahead in traces 2015-07-08 18:55:30 +01:00
Michael Walker
8794276a3e Add a Show instance for SCTTrees 2015-07-08 18:17:48 +01:00
Michael Walker
8944ea97a5 Use schedule bounding as the primary SCT approach.
This allows results to be naturally reported as lazy trees, rather
than as lists representing a tree traversal. This in turn means
that the actual bound can be moved outwards to the testing code, and
not used at all in the runner. Trees let us do nice things with
shrinking and short-circuiting, if we make the (fairly reasonable)
assumption that the children of a buggy result will exhibit the same
bug.

Storing results as trees does complicate the predicate helper
functions somewhat, but I think the clarity gained in the actual
SCT code is well worth it.
2015-06-19 16:50:51 +01:00
Michael Walker
00ad122b65 Refactor code and update docs to make more not-me friendly 2015-02-01 01:21:42 +00:00
Michael Walker
fbc262c361 Rename and remodularise 2015-01-31 15:50:54 +00:00