2

I'm on a 64-bit CentOS 6.4 machine. I have a simple benchmark that reads a JPEG into memory or writes a JPEG. I was previously using libjpeg. Now, I installed TurboJpeg, recompiled, and I see no performance improvement. I re-linked against libturbojpeg, I changed the LD_LIBRARY_PATH to point to where the libturbo libraries are (in /opt/libjpeg-turbo), I did everything I can think of, but the performance is exactly the same.

Why is this?

Could it be tha tthe libjpeg.so files that were already in /usr/lib64 on my machine were already from libjpeg-turbo? That's the only thing I can think of.

However, how can I know for sure that libjpeg-turbo is no faster, and I'm not doing something wrong?

Theolodis
  • 4,977
  • 3
  • 34
  • 53
JB_User
  • 3,117
  • 7
  • 31
  • 51
  • Depending on how much image processing you are doing the disk I/O could be the limiting factor. Consider trying putting the files on a RAMdisk and see if the two libraries exhibit a difference then. – Mark Setchell May 18 '14 at 15:55
  • No difference from RAMdisk or regular disk. – JB_User May 19 '14 at 17:40
  • Is there a way I can tell if the libjpeg I'm linking against is from libturbo, or the regular libjpeg? – JB_User May 19 '14 at 17:40
  • could you leak some of the code and the build command you are using? But in fact, you might have to get a fresh libjpeg to make sure that you do in fact not have a libjpeg-turbo named libjpeg. – Theolodis May 24 '14 at 12:48

0 Answers0