Richard Biener <richard.guent...@gmail.com> writes:
> On August 4, 2015 4:28:26 PM GMT+02:00, Richard Sandiford
> <richard.sandif...@arm.com> wrote:
>>Richard Biener <richard.guent...@gmail.com> writes:
>>> On Tue, Aug 4, 2015 at 4:21 PM, Richard Biener
>>> <richard.guent...@gmail.com> wrote:
>>>> On Tue, Aug 4, 2015 at 4:15 PM, Richard Sandiford
>>>> <richard.sandif...@arm.com> wrote:
>>>>> Richard Biener <richard.guent...@gmail.com> writes:
>>>>>> in fact the first if () looks redundant if you simply put an else
>>>>>> return NULL
>>>>>> after a else if (wi::exact_log2 (wi::neg (oprnd1)) != -1)
>>>>>>
>>>>>> Note that the issue with INT_MIN is that wi::neg (INT_MIN) is
>>INT_MIN
>>>>>> again, but it seems that wi::exact_log2 returns -1 in that case so
>>you
>>>>>> are fine (and in fact not handling this case).
>>>>>
>>>>> Are you sure it returns -1 for INT_MIN? It isn't supposed to,
>>assuming
>>>>> INT_MIN is shorthand for "minimum value for a signed type".
>>wide_ints
>>>>> aren't signed, so INT_MIN is indistinguishable from an unsigned
>>>>> 1<<(prec-1).
>>>>
>>>> No, not sure. I spotted
>>>>
>>>> /* Reject cases where there are implicit -1 blocks above HIGH. */
>>>> if (x.len * HOST_BITS_PER_WIDE_INT < x.precision && x.sign_mask ()
>>< 0)
>>>> return -1;
>>>>
>>>> and thought that would catch it. I mean the tree value is negative
>>so
>>>> exact_log2 must see it is a negative value.
>>
>>That's handling the "compressed" format, e.g.:
>>
>> {1 << 63}
>>
>>as a 64-bit short-hand for a 256-bit:
>>
>> {1 << 63,-1,-1,-1}
>>
>>In this case more than one of the low x.precision bits are known to be
>>set.
>
> So you are saying exact_log2 is really exact_log2u?
Not sure what you mean, sorry. All I meant was that a number like:
0xffffffff_ffffffff_ffffffff_80000000
(uing 32-bit rather than 64-bit elements for brevity) is stored as
a single {0x80000000}, with the upper 3 elements being implicit. And:
exact_log2 (0xffffffff_ffffffff_ffffffff_80000000)
is obviously -1. That's the case that the code above is handling.
If there are implicit all-1 blocks, the number is not a power of 2.
Thanks,
Richard
