------- Additional Comments From pinskia at gcc dot gnu dot org 2005-09-20 17:16 ------- The difference now is: <L2>:; sum = MEM[base: (int *) ivtmp.278] * MEM[base: (int *) ivtmp.280] + sum; k = k + 1; ivtmp.278 = ivtmp.278 + 4B; ivtmp.280 = ivtmp.280 + 200B; if (k != 50) goto <L2>; else goto <L4>;
For c_style. And <L2>:; sum = sum + C[(unsigned int) (int) ((unsigned int) k + ivtmp.1004)] * D[(unsigned int) (int) ivtmp.991]; k = k + 1; ivtmp.991 = ivtmp.991 + 50; if (k != 50) goto <L2>; else goto <L4>; For oop_style which means we don't do IV-opts on it for some reason. -- http://gcc.gnu.org/bugzilla/show_bug.cgi?id=13745