Re: [Qemu-devel] [RFC] alpha qemu arithmetic exceptions

[Al Viro <viro@ZenIV.linux.org.uk>]

On Tue, Jul 01, 2014 at 05:34:45AM +0100, Al Viro wrote:

> VAX operations are serious mess, but I'm not sure if we have them actually
> used anywhere in Linux kernel or userland.  Always possible, of course, but...

Grr...  Truncated mail, sorry.  Missing part:

_If_ we decide that we want CVTGQ working correctly, we'll have the following
pile of fun:
	* it needs non-saturating overflow handling, same as cvttq
	* it needs different rounding for CVTGQ and CVTGQ/C
	* CVTGQ/S needs EXC_M_SWC in the word fed to trap in INV case (i.e.
when we see dirty zero or reserved).  I think the right way to do it is to
have it use float_raise() and finish with something similar to
gen_fp_exc_raise(), except that...
	* VAX insns need a slightly different trap handling - fpcr_exc_mask
is IEEE-only.
	* g_to_float64() isn't quite right here - we want e.g. 2^-1023 to
result in 0 *and* we want inexact raised.  As it is, we'll end up with
exact 0.