arm: hevc_idct: Tune the add_res_8x8 and add_res_32x32 functions
authorMartin Storsjö <martin@martin.st>
Thu, 27 Apr 2017 21:36:40 +0000 (00:36 +0300)
committerMartin Storsjö <martin@martin.st>
Fri, 28 Apr 2017 09:02:14 +0000 (12:02 +0300)
commite1c2453a4fac1f7116244d0d05310935c20887e6
tree8902336cbce2a3479013161a4bf1ec2474fcd72e
parent0d4d43513786f1df4d561e1fac924fb0722c6700
arm: hevc_idct: Tune the add_res_8x8 and add_res_32x32 functions

Before:              Cortex     A7      A8      A9     A53
hevc_add_res_8x8_8_neon:     116.0    58.7    80.2    90.7
hevc_add_res_32x32_8_neon:  1230.0   737.5  1187.5   974.4
After:
hevc_add_res_8x8_8_neon:      97.7    57.0    73.7    80.0
hevc_add_res_32x32_8_neon:  1216.0   698.7  1127.5   827.1

Signed-off-by: Martin Storsjö <martin@martin.st>
libavcodec/arm/hevc_idct.S