ARM: tegra: power: Prefer movw/movt for loading addresses
authorScott Williams <scwilliams@nvidia.com>
Wed, 29 Jun 2011 18:45:53 +0000 (11:45 -0700)
committerDan Willemsen <dwillemsen@nvidia.com>
Sat, 14 Sep 2013 07:55:14 +0000 (00:55 -0700)
commit2a2188f817c55a2b10d5d51f6773356e218048dd
tree761193c836cea33cf1b6f77f23cb50e56c571f5d
parentd270070b3eaba358a65dc93beb9960778813c9a9
ARM: tegra: power: Prefer movw/movt for loading addresses

The movw/movt instruction pair (encapsulated by the mov32 macro)
is preferred over literals for loading addresses. The use of literals
for singleton data accesses can cause unnecessary cache misses and
evictions for cache lines that are unlikely to be accessed again in
the near future. Furthermore, certain code sequences must refrain
from using data accesses. Therefore, in general, addresses should
be loaded by mov32.

Change-Id: I9bcc3ee191f882996197ce2edc0eb510d4ff7b4a
Reviewed-on: http://git-master/r/40460
Tested-by: Daniel Willemsen <dwillemsen@nvidia.com>
Reviewed-by: Scott Williams <scwilliams@nvidia.com>
Reviewed-by: Daniel Willemsen <dwillemsen@nvidia.com>

Rebase-Id: R7ddd0d9b1e2fc8ab653b9220388acbecdbf4c57f
Signed-off-by: Dan Willemsen <dwillemsen@nvidia.com>
arch/arm/mach-tegra/delay.S
arch/arm/mach-tegra/headsmp.S
arch/arm/mach-tegra/sleep-t20.S
arch/arm/mach-tegra/sleep.S