We’re investigating if Npp can give any performance over Ipp on image processing, and part of it is to compare nppiCrossCorrValid_NormLevel_8u32f_c1r and ippiCrossCorrValid_NormLevel_8u32f_C1R. The GPU is a GTX 750 Ti, and the CPU is a i7 3770. Source image is 409600 bytes. Template image is 6464…

I tried cuda 7.0 64 bit. Npp execution time went up to 3,000ms.

If you want to provide a complete test code, I will take a look. It needs to be a code that I can copy, paste, compile, and run, without having to add anything or change anything. Likewise, it should include the complete code for the CPU comparison. If that’s a problem, perhaps someone else will …

Not quite sure how I can pull that off, seeing as I don’t know what OS you’re running, where you keep headers / libraries, etc, but this code should run on Windows or Linux. It needs to be linked against ippi.lib, nppi.lib, and cudart.lib. The GPU takes 1600ms to run (GTX 750 ti). CPU is taking …

Apparently for smaller template sizes, NPP can be faster than IPP for this function. However this isn’t one of those cases. Anyway I’ve filed a performance bug with the team responsible for this library. They are aware of the issue. I don’t have any further details on it at this time. One note ab…

Thanks for you help. Is the “warm-up” call just any nnpi function, or is there a special / recommended function to get the initialization done?

For benchmarking purposes, call the function you intend to benchmark. For other use, be aware that the first time you call a function it may take longer to execute.

Hi, Is there any update? cuda 11.2 + RTX3070 system still slow. image size is 1754x6981 and template size is 1744x6971 result size is 11x11 function name is nppiCrossCorrValid_NormLevel_8u32f_C1R IPP takes 200ms but NPP takes 600ms.

Very poor performance with NPP CrossCorrValid

Accelerated Computing GPU-Accelerated Libraries

Robert_Crovella May 25, 2022, 1:10am 9

The issue reported in this thread was addressed in the CUDA 8.0 timeframe. Note the statement about very large template sizes.

If you wish to report an issue, and have a complete test case, you can file a bug.

Topic		Replies	Views
NPP function nppiCrossCorrFull_NormLevel_8u32f_C1R too slow??? CUDA Programming and Performance	8	1650	March 7, 2015
NPP NPP cross correlation CUDA Programming and Performance	3	5199	April 24, 2014
About inaccuracy in nppiCrossCorrValid_NormLevel() GPU-Accelerated Libraries npp	0	629	February 7, 2022
Are Nppi CrossCorrelation functions working? GPU-Accelerated Libraries	0	496	September 11, 2020
NPP_STEP_ERROR when using Cross correlation function from NPP CUDA Programming and Performance	0	1123	April 24, 2014
How to realise 1D Crosscorrelation between tow signals on GPU GPU-Accelerated Libraries npp	0	834	February 9, 2022
Normalized correlation image v.s. template CUDA Programming and Performance	1	407	January 24, 2022
2D cross correlation CUDA Programming and Performance	11	26303	May 19, 2011
help me help you with modern CMake and CUDA: mwe for NPP GPU-Accelerated Libraries	17	3140	February 11, 2021
Could someone give one code example - how to use nppiSqrDistanceValid_Norm_8u32f_C1R_Ctx? GPU-Accelerated Libraries npp	13	1396	October 21, 2021

Very poor performance with NPP CrossCorrValid

Related topics