[GHC] #5059: Pragma to SPECIALISE on value arguments
GHC
cvs-ghc at haskell.org
Tue Aug 23 12:23:12 CEST 2011
#5059: Pragma to SPECIALISE on value arguments
---------------------------------+------------------------------------------
Reporter: batterseapower | Owner:
Type: feature request | Status: new
Priority: normal | Milestone: 7.4.1
Component: Compiler | Version: 7.0.3
Keywords: | Testcase:
Blockedby: | Difficulty:
Os: Unknown/Multiple | Blocking:
Architecture: Unknown/Multiple | Failure: None/Unknown
---------------------------------+------------------------------------------
Comment(by batterseapower):
Edward: that is known as the static argument transformation - see #888
In my tests I found that (with some heuristics about when to do it) it
sped up nofib by 20-30%! I did commit my code at the time but didn't do
the rest of the work to fix all the cases where SAT was making programs
worse. In particular I remember it was causing ++ to be inlined and
specialised a lot more because the second argument is static. We would
like to avoid this.
This might be a nice project if you are interested in this area? I think
the perf win will be big.
--
Ticket URL: <http://hackage.haskell.org/trac/ghc/ticket/5059#comment:10>
GHC <http://www.haskell.org/ghc/>
The Glasgow Haskell Compiler
More information about the Glasgow-haskell-bugs
mailing list