Re: [Qemu-devel] [PATCH 4/4] target-m68k: add fscale, fgetman and fgetexp

[Laurent Vivier]

Le 03/07/2017 à 22:31, Richard Henderson a écrit :
> On 07/03/2017 12:50 PM, Laurent Vivier wrote:
>>         fmove.x #0x000000000ABCDEF12345,%fp0
>>         fgetman.x %fp0,%fp6
>> fp0            2.1518178707571747286191852003521627e-4938    (raw
>> 0x0000000000000abcdef12345)
>> fp6            1.342222103012886691431049257516861    (raw
>> 0x3fff0000abcdef1234500000)
> 
> This one shows exactly what I was thinking about.
> 
>> So I guess the mantissa must be shifted to left until we have a 1 in the
>> explicit integer part bit?
> 
> Yes.
> 
> Please try fgetexp on this same input.

                fmove.x #0x000000000ABCDEF12345,%fp0
                fgetexp.x %fp0,%fp6
fp0            (raw 0x0000000000000abcdef12345)
fp6            -16403

> I suspect the answer is -16384 - clz64(val->l.lower).

It looks like -16383 - clz64(val->l.lower)

                fmove.x #0x000056789ABCDEF12345,%fp0
                fgetexp.x %fp0,%fp6
fp0            (raw 0x0000000056789abcdef12345)
fp6            -16384

                fmove.x #0x000086789ABCDEF12345,%fp0
                fgetexp.x %fp0,%fp6
fp0            (raw 0x0000000086789abcdef12345)
fp6            -16383

Thanks,
Laurent