On Thursday 20 January 2011 15:18:51, S. Doaitse Swierstra wrote:
I have small program UULib which i use for checking some timing information.
When I compile with ghc 7 and profiling information I get the timings which are more or less what I expect. If I however recompile without profiling tome consumed goes up by a factor of 20!
Perhaps profiling prevents the application of an optimisation which is in fact a pessimisation for your programme. Does looking at the generated Core reveal something? (if you make the code available, others could look too) What are the times when you compile with -O[2]?
1) Am I misinterpreting the results? 2) If not, does this look familiar to anyone? 3) If so, whether should I look first to see what is going on here?
Look at the Core, ask on glasgow-haskell-users/open a ticket so that you have a better chance of getting the attention of an expert (the GHC team don't regularly follow the cafe).