Use AVX2 to increase frequency for FP16<->FP32 Caffe2 ops (#31203)
Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/31203
For multi-instance environment, AVX2 should help increase the clock frequency.
ghstack-source-id: 95502576
Test Plan: buck test //caffe2/caffe2:caffe2_test_cpu -- "Float16"
Reviewed By: jspark1105
Differential Revision: D18962649
fbshipit-source-id: 6532d929a99f41f2f6ad1a1a1962e38ae3ddaecb