Alas on 7/5/21 7:39 PM, I wrote in part: > Beware that some library functions don't behave as expected. For > example, on my machine, sqrt(2.) does not throw the FE_INEXACT > exception.
That's just completely wrong. Sorry. FWIW: I got fooled by a compiler optimization, but that's no excuse. I should have known better. I've been seeing compiler bugs, but that's not one of them. The fPU traps and flags work just fine. ========================== Here's a program that allows you to explore and demonstrate FPU behavior: https://www.av8n.com/cgit/FPU/tree/fpu_exceptions.c The first few lines of the usage message are: Typical usage: :; ./fpu_exceptions sqrt -1 # invalid (flag, not trap) :; ./fpu_exceptions +div div 1.1 0 # divide by zero (trap) :; ./fpu_exceptions +all div 1.1 0 # inexact trap happens before the divide :; ./fpu_exceptions pow 2 1024 # overflow AND inexact (flags, not traps) * Interesting verb phrases include: help # print this message (and immediately exit) atof 1.1 # FE_INEXACT div 1.1 0 # FE_DIVBYZERO (we clear the inexact flag before the div) div 1 10 # FE_INEXACT div 100 89 # FE_INEXACT div 5 10 # OK sqrt 2 # FE_INEXACT sqrt 4 # OK pow 2 1024 # FE_OVERFLOW | FE_INEXACT pow 3 34 # FE_INEXACT # the power is calculated by iterative multiplication; # it you turn on traps it will tell you how many # iterations succeeded. ====================== You can clone the whole thing via: :; git clone https://www.av8n.com/cgit/FPU Compile it using the makefile provided, or simply :; ./fpu_exceptions.c since it's a self-compiling source file (assuming a reasonably conventional linux environment). _______________________________________________ gnumeric-list mailing list gnumeric-list@gnome.org https://mail.gnome.org/mailman/listinfo/gnumeric-list
