[x265] [PATCH 1/2] weight prediction: use sqrtf

[Rafaël Carré]

On 03/06/14 18:56, Steve Borho wrote:
> On Thu, Mar 6, 2014 at 11:44 AM, Rafaël Carré <funman at videolan.org> wrote:
>> On 03/05/14 18:06, Steve Borho wrote:
>>> On Wed, Mar 5, 2014 at 4:38 AM, Rafaël Carré <funman at videolan.org> wrote:
>>>> ---
>>>>  source/encoder/weightPrediction.cpp | 2 +-
>>>>  1 file changed, 1 insertion(+), 1 deletion(-)
>>>>
>>>> diff --git a/source/encoder/weightPrediction.cpp b/source/encoder/weightPrediction.cpp
>>>> index c04361a..c0710fd 100644
>>>> --- a/source/encoder/weightPrediction.cpp
>>>> +++ b/source/encoder/weightPrediction.cpp
>>>> @@ -203,7 +203,7 @@ bool WeightPrediction::checkDenom(int denom)
>>>>                  uint64_t fencVar = fenc->wp_ssd[yuv] + !ref->wp_ssd[yuv];
>>>>                  uint64_t refVar  = ref->wp_ssd[yuv] + !ref->wp_ssd[yuv];
>>>>                  if (fencVar && refVar)
>>>> -                    guessScale[yuv] = Clip3(-2.f, 1.8f, std::sqrt((float)fencVar / refVar));
>>>> +                    guessScale[yuv] = Clip3(-2.f, 1.8f, sqrtf((float)fencVar / refVar));
>>>
>>> Is something wrong with std::sqrt?
>>
>> I got:
>> x265/source/encoder/weightPrediction.cpp:206:57: error: 'sqrt' is not a
>> member of 'std'
>>
>> With g++ 4.8.2
>>
>> Full command line is :
>>
>> cd /home/fun/x265/source/encoder && /usr/bin/c++   -DHAVE_INT_TYPES_H=1
>> -DHIGH_BIT_DEPTH=0 -DLOG_CU_STATISTICS=0 -DX265_ARCH_X86=1 -DX86_64=1
>> -O2 -g -DNDEBUG -I/home/fun/x265/source/. -I/home/fun/x265/source/Lib
>> -I/home/fun/x265/source/common -I/home/fun/x265/source/encoder
>> -I/home/fun/x265/source    -Wall -Wextra -Wshadow -ffast-math
>> -mstackrealign -fPIC -o CMakeFiles/encoder.dir/weightPrediction.cpp.o -c
>> /home/fun/x265/source/encoder/weightPrediction.cpp
>>
>> Perhaps I am missing a -stdc++=something that was added later?
> 
> No, I don't believe so.  I build this regularly with gcc 4.8.1 on both
> Linux and MinGW.
> 
> http://en.cppreference.com/w/cpp/numeric/math/sqrt
> 
> Does adding #include <cmath> to the file help?

Yes, no problem if <cmath> is included.