Intel’s “Intel® 64 and IA-32 Architectures Software Developer’s Manual” page 654 (3-224  Vol. 2A) says that divq produces “#DE;” when high numerator is not less than denominator.
Mach kernel reports “Floating point exception: 8” or at least the shell command reports that.
The operation is ‘divq’ with RDX:RAX = 101000…:000000… and divisor is 10000… .
The bug could be in Intel documentation, in the chip, in the kernel or in the shell.
(Or perhaps in runtime libs.)

<pre>gcc <a href=bug.c>bug.c</a> <a href=q.s>q.s</a> -Wmost
./a.out</pre>
produces:
<pre> Floating point exception: 8</ptr>