If you could show the generated code (after gcc's register allocation) *and* point out precisely where things go wrong, that would be helpful.
Regards, /Niels The generated code is on the compiler explorer link I provided: http://franke.ms/cex/z/oG53bK There you see the original code on the left and the generated assembly in the middle. You can add the & to "=d" (__umul_tmp1) in the left window and you see immediately the change in the comiled output. I am not the expert here and I do not want to complain or bother you at all. It was just my observation of symptoms I wanted to report. The compiler maintainer says "gcc is ok, the problem is in longlong.h" You say "longlong.h looks ok, we wouldn't hide compiler bugs." For me the workaround (add &) works and all tests pass with it, so can live with the workaround. But somewhere there must be a bug/glitch. If you could prove gcc to be wrong that would also be a really great result! Thanks for all your effort Best regards Alexander _______________________________________________ gmp-bugs mailing list [email protected] https://gmplib.org/mailman/listinfo/gmp-bugs
