Re: [PATCH v12 1/3] lib: Add strongly typed 64bit int_sqrt
From: Crt Mori
Date: Wed Jan 10 2018 - 03:34:03 EST
On 10 January 2018 at 09:15, Crt Mori <cmo@xxxxxxxxxxx> wrote:
> On 9 January 2018 at 20:23, Joe Perches <joe@xxxxxxxxxxx> wrote:
>> On Tue, 2018-01-09 at 16:18 +0100, Crt Mori wrote:
>>> There is no option to perform 64bit integer sqrt on 32bit platform.
>>> Added stronger typed int_sqrt64 enables the 64bit calculations to
>>> be performed on 32bit platforms. Using same algorithm as int_sqrt()
>>> with strong typing provides enough precision also on 32bit platforms,
>>> but it sacrifices some performance.
>> []
>>> diff --git a/lib/int_sqrt.c b/lib/int_sqrt.c
>> []
>>> @@ -36,3 +37,34 @@ unsigned long int_sqrt(unsigned long x)
>>> return y;
>>> }
>>> EXPORT_SYMBOL(int_sqrt);
>>> +
>>> +#if BITS_PER_LONG < 64
>>> +/**
>>> + * int_sqrt64 - strongly typed int_sqrt function when minimum 64 bit input
>>> + * is expected.
>>> + * @x: 64bit integer of which to calculate the sqrt
>>> + */
>>> +u32 int_sqrt64(u64 x)
>>> +{
>>> + u64 b, m;
>>> + u32 y = 0;
>>> +
>>> + if (x <= 1)
>>> + return x;
>>
>> I think this should instead be:
>>
>> if (x <= INT_MAX)
>> return int_sqrt((int)x);
>>
>> to reduce the loop cost below when the
>> value is small enough.
>>
>
> In existing int_sqrt its only 1 and I assume that is more to protect
> from loop execution with 0 or 1. Since there is no difference (except
> fls64) with int_sqrt I assume there is no need to call it to avoid
> loop?
>
Nevermind, I see what you mean (should have thought longer before I
written). The cost of below loop is because of 64bit calculation is
not native on 32bit and we could just use 32bit calculation in that
loop. Will send v13 with a fix for this.
>>> +
>>> + m = 1ULL << (fls64(x) & ~1ULL);
>>> + while (m != 0) {
>>> + b = y + m;
>>> + y >>= 1;
>>> +
>>> + if (x >= b) {
>>> + x -= b;
>>> + y += m;
>>> + }
>>> + m >>= 2;
>>> + }
>>> +
>>> + return y;
>>> +}
>>> +EXPORT_SYMBOL(int_sqrt64);
>>> +#endif