I have run a number of tests, not terribly scientifically organised! Although I haven't found the point where pi starts to fail, it is worth keeping you informed of developments.
Sorry for the long delay, each run takes a while! Your test program: size state error 1000000000 fail Floating point overflow 640000000 pass 650000000 fail Floating point overflow 646456992 fail Floating point overflow 646456991 fail Floating point overflow 646400000 pass 646450000 pass 646456900 fail Floating point overflow 646456800 fail Floating point overflow 646456000 pass 646456500 pass 646456700 fail Floating point overflow 646456650 fail Floating point overflow 646456610 pass 646456620 fail Floating point overflow 646456630 fail Floating point overflow 646456640 fail Floating point overflow 646456610 pass 646456611 pass 646456612 pass 646456613 pass 646456614 pass 646456615 fail Floating point overflow 646456616 fail Floating point overflow 646456617 fail Floating point overflow 646456618 fail Floating point overflow 646456619 fail Floating point overflow So failure occurs at > 646456614 With pi, which takes much longer to run, size state error pi 646456614 fail Internal error: statement in file float/output/cl_F_dprint.cc, line 359 has been reached!!Please send the authors of the program a description how you produced this error!pi 646456615 fail Floating point overflow pi 646456613 fail Internal error: statement in file float/output/cl_F_dprint.cc, line 359 has been reached!!Please send the authors of the program a description how you produced this error! So, it seems I have provoked another, separate problem. But I can confirm that pi with values > 646456614 causes the floating point overflow I originally reported. I will keep looking for pi's failure value. -- Steve -- To UNSUBSCRIBE, email to [EMAIL PROTECTED] with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]