process crash on signal 4

[Yogeshwar Shenoy]

Can someone throw some light on what the different reasons for signal 4 (SIGILL)
being sent to a process are? ('Illegal instruction' does not quite make sense in this case).

We are running a server that uses TCP, on Intel Xeon CPU 2.40GHz (hyperthreading disabled)
running FreeBSD 4.8-RELEASE. This process core dumped on signal 4 when
pushing ~15Mbit/sec over an em (gigabit copper) interface. Using gdb I find that the address in
the EIP register is actually the start of my own function; the stack trace does not show a frame 
for abort() either.  So I am not sure what caused this signal to be sent to the process. 

The exact same binary running on a PIII 1.2GHz (FreeBSD 4.6.2-RELEASE), using
an fxp interface has been running fine (also steadily pushing ~15Mbit/sec) for about 8 months.

The binary was compiled on a PIII machine running FreeBSD 4.4-RELEASE machine using
gcc version 2.95.3 20010315 (release) [FreeBSD], with the -O2 option. 

I'd appreciate any help in getting to the bottom of the SIGILL.

Thanks,
Yogeshwar.