Wednesday, July 14, 2004

Floating Point Problems Fixed Shantonu figured out the problem. He's a very smart guy and just continues to prove that over and over. Tested a new kernel he built and it successfully corrects the issue. Without getting into too much detail, opendarwin was not recognizing the features of the chipset. It wasn't correctly saving & restoring some registers (during a context switch?) which caused it to loop forever when a looping structure used a double/float in its boolean test. This only affected Athlon-based systems, so he added code to recognize the chipset and treat it like an Intel chip. Felix has tagged the fix and will provide probably one more Release Candidate. We're close! 6:31:26 AM    comment []