tree a768522dcaeabf229387f1f854f510a69b32403b
parent 406ed726a23750ef286bdb559b09992c9fed3eff
author Diksha Singh <diksha.singh@ittiam.com> 1662983256 +0530
committer Yunqing Wang <yunqingwang@google.com> 1663972491 +0000

Add dual and quad intrinsic support for CDEF MSE computation

This CL attempts to evaluate the aom_mse_16_bit for two 8x8
and four 4x4 blocks at a time. Also, avoiding few of the unpacks
and 32-bit register operations which can dealt within 16-bit
register.
The overall encode time reduction for RT preset is listed below

                Encode_time
cpu   Testset   Reduction(%)
 7      rtc       0.765
 8      rtc       0.803

Change-Id: I385c7ba00764575a620d563f0f6fed330ff5096d