ippiColorTwist too slow

[Anca]

Hello,

I am using ippiColorTwist (more precisely ippiColorTwist32f_16_C3IR) for a 1024x1024 image with 3 samples per pixel, 2 bytes per sample. The average measured time is about 12ms, almost the same as running my C implementation for the same processing. The C implementation performs the operations described herehttp://software.intel.com/sites/products/documentation/hpc/composerxe/en-us/ippxe/ipp_manual_lnx/index.htmplus a simple rounding to 0 for negative numbers.

Why is ippiColorTwist so slow? Am I missing something?

Here are the details about my configurations:

- Intel Core2 6300@1.86 GHz

- Windows 7 Ultimate, 64-bit

-using static linkage with dispatching for the IPP library

- ippInit() is called before any other ipp function

- ippiGetLibVersion() returns:

major = 7

minor = 0

majorBuild = 205

build = 1020

targetCpu = "u8"

Name = "ippiu8_l.lib"

Version = "7.0 build 205.23"

BuildDate = "Sep 1 2010"

- using Visual Studio 2008

- C implementation is compiled for Release, X64 with Full Optimization (/0x)

Thank you.