deprecated Functions:ippsTone_Direct_32f and ippsr_Sinc_32f

deprecated Functions:ippsTone_Direct_32f and ippsr_Sinc_32f

Аватар пользователя rohitspandey

Hi,

Alot many functions like ippsTone_Direct_32f/sinc etc are deprecated. Is there any replacement function for them?  We use ippsTone_Direct_xx and ippsr_sinc alot in our application. Is ot possible to include them in sp library and not deprecated.

 

If not, them is the combination of slope/ramp function wit ipps_sin/ipps_cos is faster in execution. We need faster replacement for these API. Kindly suggest if IPP sin/cos  is faster compared to tone. How the ipps_cos/sin implemented.

 

Regards

Rohit

6 сообщений / 0 новое
Последнее сообщение
Пожалуйста, обратитесь к странице Уведомление об оптимизации для более подробной информации относительно производительности и оптимизации в программных продуктах компании Intel.
Аватар пользователя Sergey Kostrov

>>...Kindly suggest if IPP sin/cos is faster compared to tone. How the ipps_cos/sin implemented...

That was already explained many times on different IDZ forums that Intel does not release any details on how some API is implemented or what algorithms are used.

With regard to deprecated functions: Please consider to make a feedback on usage of functions you need at:
.
http://software.intel.com/sites/products/ipp-deprecated-features-feedback.

Thanks in advance.

Аватар пользователя rohitspandey

Hi,

 

ok, but can you shed some input on the speed of execution of   IPP sin/cos  when compared to  direct tone?

 

Regards

Rohit

Аватар пользователя Sergey Kostrov

There are several IPP functions from Vector Math domain:

...
IPPAPI( IppStatus, ippsAcos_32f_A11, (const Ipp32f a[],Ipp32f r[],Ipp32s n))
IPPAPI( IppStatus, ippsAcos_32f_A21, (const Ipp32f a[],Ipp32f r[],Ipp32s n))
IPPAPI( IppStatus, ippsAcos_32f_A24, (const Ipp32f a[],Ipp32f r[],Ipp32s n))
IPPAPI( IppStatus, ippsAcos_64f_A26, (const Ipp64f a[],Ipp64f r[],Ipp32s n))
IPPAPI( IppStatus, ippsAcos_64f_A50, (const Ipp64f a[],Ipp64f r[],Ipp32s n))
IPPAPI( IppStatus, ippsAcos_64f_A53, (const Ipp64f a[],Ipp64f r[],Ipp32s n))

IPPAPI( IppStatus, ippsAsin_32f_A11, (const Ipp32f a[],Ipp32f r[],Ipp32s n))
IPPAPI( IppStatus, ippsAsin_32f_A21, (const Ipp32f a[],Ipp32f r[],Ipp32s n))
IPPAPI( IppStatus, ippsAsin_32f_A24, (const Ipp32f a[],Ipp32f r[],Ipp32s n))
IPPAPI( IppStatus, ippsAsin_64f_A26, (const Ipp64f a[],Ipp64f r[],Ipp32s n))
IPPAPI( IppStatus, ippsAsin_64f_A50, (const Ipp64f a[],Ipp64f r[],Ipp32s n))
IPPAPI( IppStatus, ippsAsin_64f_A53, (const Ipp64f a[],Ipp64f r[],Ipp32s n))
...

and performance is related to a precision ( Axx suffix ). It is Not clear what precision do you need in case of '_32f_' functions?

Аватар пользователя rohitspandey

Hi,

Following are the function we would be using in our application to replace (IppStatus ippsTone_Direct_32fc(Ipp32fc* pDst, int len, float magn, float rFreq, float* pPhase, IppHintAlgorithmhint); and IppStatus ippsTone_Direct_64fc(Ipp64fc* pDst, int len, double magn, double rFreq, double* pPhase, IppHintAlgorithmhint);

IPPAPI( IppStatus, ippsAcos_32f_A24, (const Ipp32f a[],Ipp32f r[],Ipp32s n))

IPPAPI( IppStatus, ippsAcos_64f_A53, (const Ipp64f a[],Ipp64f r[],Ipp32s n))

IPPAPI( IppStatus, ippsAsin_64f_A53, (const Ipp64f a[],Ipp64f r[],Ipp32s n))

IPPAPI( IppStatus, ippsAsin_32f_A24, (const Ipp32f a[],Ipp32f r[],Ipp32s n))

 

Can you now tell me if these functions are slower or faster on a intel i5and i7 platform.We would limit  to SSE2 support only.

 

 

Regards

Rohit

Аватар пользователя Sergey Kostrov

>>IPPAPI( IppStatus, ippsAcos_32f_A24, (const Ipp32f a[],Ipp32f r[],Ipp32s n))
>>
>>IPPAPI( IppStatus, ippsAcos_64f_A53, (const Ipp64f a[],Ipp64f r[],Ipp32s n))
>>
>>IPPAPI( IppStatus, ippsAsin_64f_A53, (const Ipp64f a[],Ipp64f r[],Ipp32s n))
>>
>>IPPAPI( IppStatus, ippsAsin_32f_A24, (const Ipp32f a[],Ipp32f r[],Ipp32s n))

Even if these vectorized IPP functions are faster, compared to ippsTone_Direct_32fc, I don't think that it will be a right comparison. Single-precision functions are always faster then double-precision and these vectorized functions just calculate sines or cosines. ippsTone_Direct_32fc does more and it can not be faster. All these functions are very easy to use and you need to create a simple test-case to get exact performance numbers on computer systems you use.

Зарегистрируйтесь, чтобы оставить комментарий.