[Top][All Lists]

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [Qemu-devel] [PATCH 08/19] target-alpha: use new float64_unordered()

From: Peter Maydell
Subject: Re: [Qemu-devel] [PATCH 08/19] target-alpha: use new float64_unordered() function
Date: Thu, 14 Apr 2011 10:14:33 +0100

On 13 April 2011 16:38, Richard Henderson <address@hidden> wrote:
> (It looks like some of the Alpha code can be cleaned up a bit.
> I don't recall flush_inputs_to_zero option being there before,
> and we do that by hand in helper_ieee_input*.)

While we're on the subject of Alpha and flush-to-zero modes,
do you know what exception bits should get set when Alpha
flushes a denormal output (not input) to zero?

At the moment softfloat doesn't set any flags when it does
this, which is definitely wrong for ARM. I had a look at
the manuals for the other archs which might set flush_to_zero:

 SH: should set underflow flag
 ARM: should set underflow flag
 PPC: should set underflow (I think)
 MIPS: docs didn't say

I think Alpha should set Inexact and not Underflow, but
I'm not sure -- can you confirm/deny?

(Obviously I would prefer it if every architecture just
needed to set underflow, but somehow I don't think it
will work out that cleanly :-))

-- PMM

reply via email to

[Prev in Thread] Current Thread [Next in Thread]