13

I installed nsight with cuda5.5. Everything works fine but I am trying to run opencv functions on the host to generate images. I include opencv header path and also the libs paths that I need. When I try to compile it gives this error /Applications/Xcode.app/Contents/Developer/Toolchains/XcodeDefault.xctoolchain/usr/bin/../lib/clang/5.0/include/emmintrin.h(274): error: identifier "__builtin_ia32_cmpsd" is undefined Anyone knows what could be the reason? Or any way to have make opencv work with nsight eclipse? Thanks!

Sagar Masuti
  • 1,271
  • 2
  • 11
  • 30
Glove
  • 960
  • 6
  • 17
  • 30
  • 6
    I don't think that your error depends on cuda or opencv. For me it seems to be problem with clang. Propably it's only missing a linked library from there. – hubs Nov 29 '13 at 13:11
  • 2
    I would second @hubs in that this looks like OpenCV issue with Clang. – Eugene Dec 02 '13 at 19:29
  • @hubs How did you solve your problem? – Rose Perrone May 26 '14 at 02:39
  • @RosePerrone I didn't solve this problem. I wanted only to note, that this problem doesn't relate to nsight eclipse, cude or opencv. I think there is missing a link against a clang library. – hubs May 26 '14 at 13:06

1 Answers1

1

I solved my issue by removing the '-D HAVE_SSE2' from the macros defined in config.mk.

If you can't do that:

See this solution. In brief, you need to compile the code using SSE2 using gcc and your CUDA code using nvcc and then link them together afterwards.

Apparently NVidia is aware of the problem and "it should be fixed soon", but I have v6.0 and the bug still exists.

Someone mentioned this problem on the NVIDIA developer forums, but nobody replied.

Rose Perrone
  • 61,572
  • 58
  • 208
  • 243