Chris Rodie writes: > I have found a solution to the problem: I added "-ffloat-store" to the > CFLAGS declaration in the Makefile in the monte directory. After > recompiling, the test suite ran with no errors. Now the whole library > appears to test OK on this platform (pc-686-mingw). It appears we're > dealing with a difference in precion of values in fpu registers vs. > values in values (see GNU docs on this gcc switch). The problem appears > only in this monte carlo algorithm, nowhere else in the library.
I don't know what default the Mingw people are using for their runtime but if it's extended I would set the processor floating point mode to double-precision and run the same code to see if it is that is the cause, -ffloat-store will change the way it is optimised so it may just mask the problem. -- Brian Gough Network Theory Ltd, Publishing Free Software Manuals --- http://www.network-theory.co.uk/ _______________________________________________ Bug-gsl mailing list [email protected] http://lists.gnu.org/mailman/listinfo/bug-gsl
