IPP 8.1 - ippsFIRGetStateSize_32f

[Michal_B_Intel]

Hello,

I have a question about FIR filtering functionality. When I ask for the FIR state buffer size using method below:

ippsFIRGetStateSize_32f( int tapsLen, int *pBufferSize )

I get unusually huge values e. g. 17792 for tapsLen=1. Is this expected? From what I understood from documentation the FIR state structure contains taps values and delay buffer. I see no reason why this should be this big.

[Chao_Y_Intel]

Hi Michal,

What is the processor you find the problem?  Is it a 32 bit code, or Intel 64 bit code?

Thanks,
Chao

[Michal_B_Intel]

It's 64-bit.

Michal

[Michal_B_Intel]

Hello,

Is there any chance for an update on this issue?

Michal

[Igor_A_Intel]

HI Michal,

I guess that for 1 tap FIR it's better to use ippsMulC_32f function :)

As the same state structure is used for in-place and out-of-place FIR processing functions, the main part of this state structure is temporal buffer for in-place function - 4K floats = 16K bytes.

regards, Igor

[Michal_B_Intel]

Hello.

The 1-tap filter was an example I plan to use longer ones but the memory usage grows too rapidly with each tap.

Thanks for clarifying this, I guess that I'll have to implement my own FIR.

Michal

[Igor_A_Intel]

Hi Michal,

did you receive my private message for you?

regards, Igor