floating point problems in current EGCS?

Stephen L Moshier moshier@mediaone.net
Tue Nov 30 23:39:00 GMT 1999


> problems found running "paranoia"

Making -fforce_mem true by default several years ago broke paranoia
pretty solidly on ix86 and it has never recovered.

You can get through it by setting your math coprocessor rounding
precision to be the same as the precision of the arithmetic under
test.



More information about the Gcc-bugs mailing list