The current default memcpy first copies single bytes to align the
amount, then copies the rest as full words. In practice, the start of a
buffer is much more likely to be word-aligned then the end, and aligned
word access are usually more efficient. This patch reorders those
accesses to first copy as many full words as possible and then finish
the rest with byte accesses to optimize this common case.
This fixes a data abort when using USB on ARM without CONFIG_GPL. Due to
some limitations of how DMA memory is set up in coreboot on ARM, it
currently does not support unaligned accesses. (This could be fixed with
a more complicated patch, but it's usually not an issue... unless, of
course, your memcpy happens to be braindead).
Also add word-aligned accesses to memset and memcmp while I'm at it, and
make memcmp's return value standard's compliant.
BUG=chrome-os-partner:24957
TEST=Manual
Original-Change-Id: I2a7bcb35626a05a9a43fcfd99eb958b485d7622a
Original-Signed-off-by: Julius Werner <jwerner@chromium.org>
Original-Reviewed-on: https://chromium-review.googlesource.com/203547
Original-Reviewed-by: Stefan Reinauer <reinauer@chromium.org>
Original-Reviewed-by: David Hendricks <dhendrix@chromium.org>
(cherry picked from commit 05a64d2e107e1675cc3442e6dabe14a341e55673)
Signed-off-by: Marc Jones <marc.jones@se-eng.com>
Change-Id: I0030ca8a203c97587b0da31a0a5e9e11b0be050f
Reviewed-on: http://review.coreboot.org/8126
Tested-by: build bot (Jenkins)
Reviewed-by: Stefan Reinauer <stefan.reinauer@coreboot.org>
The default_ functions in memory.c are only used to initialize a weak
variable. They should not be used outside memory.c. Make them
invisible.
Remove the declaration from libpayload.h. For real this time.
Change-Id: Id54c1fd172c78748f01a958ce4065dd0eb53bbc3
Signed-off-by: Ronald G. Minnich <rminnich@gmail.com>
Reviewed-on: http://review.coreboot.org/2394
Tested-by: build bot (Jenkins)
Reviewed-by: Gabe Black <gabeblack@chromium.org>
The implementations for various stdlib functions in libc/memory.c are very
generic and should work under just about any circumstances. They are
unfortunately also very slow. This change makes them weak symbols so that
faster versions can be defined on a per architecture basis which will
automatically take the place of the slow versions.
Change-Id: Ia1ac90d9dcd45962b2a15f61ecc74b0a4676048d
Signed-off-by: Gabe Black <gabeblack@google.com>
Reviewed-on: http://review.coreboot.org/1725
Reviewed-by: Patrick Georgi <patrick@georgi-clan.de>
Tested-by: build bot (Jenkins)
while others dislike them being extra commits, let's clean them up once and
for all for the existing code. If it's ugly, let it only be ugly once :-)
Signed-off-by: Stefan Reinauer <stepan@coresystems.de>
Acked-by: Stefan Reinauer <stepan@coresystems.de>
git-svn-id: svn://svn.coreboot.org/coreboot/trunk@5507 2b7e53f0-3cfb-0310-b3e9-8179ed1497e1
Thanks to Ulf Jordan for figuring this out!
Signed-off-by: Stefan Reinauer <stepan@coresystems.de>
Acked-by: Stefan Reinauer <stepan@coresystems.de>
git-svn-id: svn://svn.coreboot.org/coreboot/trunk@3527 2b7e53f0-3cfb-0310-b3e9-8179ed1497e1
There was a bit of confusion in the memcpy functions - we could simplify
things slightly without having to revert to 8 bit copies on a 32 bit
system.
Signed-off-by: Jordan Crouse <jordan.crouse@amd.com>
Acked-by: Stefan Reinauer <stepan@coresystems.de>
git-svn-id: svn://svn.coreboot.org/coreboot/trunk@3519 2b7e53f0-3cfb-0310-b3e9-8179ed1497e1
Julius Werner