Does using binary compiled for advanced cpu instructions avx2 fma improve accuracy

[JRMeyer]

>>> Lightnin
[August 27, 2019, 12:21pm]

Hello good people -

I often see this warning when running DeepSpeech: slash
Your CPU supports instructions that this TensorFlow binary was not
compiled to use: AVX2 FMA

Is installing a binary (or compiling my own) that does support those
instructions worthwhile? I'm just using the CPU on my laptop, no GPU.
But I'm wondering if solving this issue would give me improved accuracy,
or improved performance in terms of speed. (Accuracy is important for
me, speed really is not.)

Many thanks in advance-

[This is an archived TTS discussion thread from discourse.mozilla.org/t/does-using-binary-compiled-for-advanced-cpu-instructions-avx2-fma-improve-accuracy]

[JRMeyer]

>>> kdavis
[August 27, 2019, 12:37pm]

Generally, if you compiled your own, it would give improved speed but
not accuracy.

[Archived Post]