So I add -O2 to the end?
------------------------
CFLAGS_KERNEL  = -fgcse-lm -fgcse-sm -fsched-spec-load -fforce-addr -ffast-math -fsingle-precision-constant -mtune=cortex-a9 -march=armv7-a -mfpu=neon -mvectorize-with-neon-quad -ftree-vectorize -funroll-loops

===================================================================

So like this?
-----------------------
CFLAGS_KERNEL  = -fgcse-lm -fgcse-sm -fsched-spec-load -fforce-addr -ffast-math -fsingle-precision-constant -mtune=cortex-a9 -march=armv7-a -mfpu=neon -mvectorize-with-neon-quad -ftree-vectorize -funroll-loops -O2