Is there a way to get Tensorflow's C API with AVX support?

Question

I am currently using Tensorflow's C API for a project. However during the first inference run I get this message:

Your CPU supports instructions that this TensorFlow binary was not compiled to use: SSE4.1 SSE4.2 AVX AVX2 FMA

Is there a way to use the C API with AVX or AVX2 support?

You can compile it for your machine: https://stackoverflow.com/questions/41293077/how-to-compile-tensorflow-with-sse4-2-and-avx-instructions — Kampi, Jun 12 '19 at 07:18
this is not an error, tensorflow is just reminding you that it can do better if it was compiled with different settings. @Kampi 's link can guide you if you really want to compile it yourself. — Mete Han Kahraman, Jun 12 '19 at 07:30
Possible duplicate of [How to compile Tensorflow with SSE4.2 and AVX instructions?](https://stackoverflow.com/questions/41293077/how-to-compile-tensorflow-with-sse4-2-and-avx-instructions) — E_net4, Jun 12 '19 at 10:18
Hi, thanks for your replies. However as I never compiled Tensorflow on my own I have a question regarding this. Using the C API seems to be possible without any installation, I just had to link it (see: https://www.tensorflow.org/install/lang_c), so I did not use bazel, conda, pip or whatever. Is there something precompiled or how would the procedure look like for my case? Would the C API even recognize that I compiled Tensorflow with AVX when it does not require an installation at all? — tre95, Jun 12 '19 at 11:00

score 0 · Accepted Answer · answered Jun 28 '19 at 13:29

0

Found it out some time ago:

answered Jun 28 '19 at 13:29

tre95

1 Answers1