"Shawn P. Garbett"
I redid both programs to make them equivalent in action. The haskell program was not actually summing the results, the C program was not checking array bounds (Haskell does), the C program did not initialize the array and the C program didn't print the result.
Times on my laptop (Crusoe 933):
62.061s : Haskell (Array -O2) Note: lot's 'o disk grinding! 18.231s : Haskell (UArray -O2) 18.108s : Haskell (UArray -O2 -fvia-c) 17.443s : Haskell (UArray -O2 -funfolding-update-in-place) 0.807s : C (-O3 without bound check) 1.127s : C (-O3 with bound check)
At best case for Haskell, 15.5 times slower. The thing about bounds checking, in Haskell it's always there. In C, you might have it, you might not there is no certainty by the language, only by design and implementation. So with C, one is free to live dangerously.
The first two sections of http://www.cse.unsw.edu.au/~chak/papers/CK03.html also contain an analysis of some of the overheads in Haskell array codes compiled with GHC. Cheers, Manuel