Changes between Version 4 and Version 5 of PerformanceWarnings


Ignore:
Timestamp:
Oct 5, 2010 5:50:58 PM (4 years ago)
Author:
dons
Comment:

tweaks

Legend:

Unmodified
Added
Removed
Modified
  • PerformanceWarnings

    v4 v5  
    1717'''The compiler should warn about unnecessary uses of bang patterns or seq.''' 
    1818 
     19Simon PJ: ''Warning about unnecessary bangs or seqs is going to be hard, because the program has been shaken around so much between source and the end of the optimisation pipeline. Even deciding what "unnecessary" means is tricky.. is a bang unnecessary if its absence depends on a subtle 
     20strictness property of a function that is called in the body?'' 
     21 
    1922=== Heuristics for "Probably a Performance Bug" === 
    2023 
     
    2730 
    2831We can improve users' code by warning of any boxed types as arguments, or in data structures, for these types. Particularly accumulating parameters, or fields in record types. Additionally, if they are strict, but not unboxed, that should also be a warning. 
     32 
     33Simon PJ: ''Warning about every boxed Int might yield a lot of warnings!'' 
    2934 
    3035=== Other heuristics === 
     
    3944 * Warn about possible loss of sharing due to `-fstate-hack` 
    4045 * SpecConstr warnings 
     46 
     47== Issues == 
     48 
     49 * Warning about (>>=) not inlined will warn about all monad-overloaded functions. 
     50 * Something simple would still give useful information. 
    4151 
    4252== Output format ==