DMA is only used for reading because writing seems to be slower with DMA.
Only requests which are cacheline aligned (16 bytes) will use DMA, so many
requests will still use PIO at this point; a later change will align more
reads.
Part of FS#9708, original DMA code by Boris Gjenero (dreamlayers).
git-svn-id: svn://svn.rockbox.org/rockbox/trunk@24405 a1c6a512-1295-4272-9138-f99709370657
This implements the basic functionality for sending inquiries on OS X.
The current implementation has some limitations:
- it will not respect the selected device but pick the first Ipod found.
- it is inefficient due to the way ipodpatcher expects this which doesn't
really match how it works on OS X.
git-svn-id: svn://svn.rockbox.org/rockbox/trunk@24382 a1c6a512-1295-4272-9138-f99709370657
Tested on Clipv1 (sim/target) and Ipod3g (sim)
Each color would be calculated as 0
Note the iteration over the whole range of colors is quite weird, and I
can not really sense what max_iter represents (except it's a number of
different colors, right?)
It should be backported to branch after tomers check if color range is
correct (FS#10935 mentions much smaller range for color targets as well)
git-svn-id: svn://svn.rockbox.org/rockbox/trunk@24377 a1c6a512-1295-4272-9138-f99709370657
With the separation of the SystemInfo values accessing all the values
will create traces, making them rather noisy due to unnecessary multiple
accesses.
git-svn-id: svn://svn.rockbox.org/rockbox/trunk@24373 a1c6a512-1295-4272-9138-f99709370657
Rbutil's (internal) platform names are now identical to the configure
model names except for the cases where further differentiation is
needed. In those cases the platform name is now the configure name with
an additional postfix, separated by a dot. This makes it possible
getting the configure name from the platform name.
git-svn-id: svn://svn.rockbox.org/rockbox/trunk@24371 a1c6a512-1295-4272-9138-f99709370657
- simplify the setting/skin relationship. settings are used as the fallback if it's not specified in the skin
- backdrop buffers are now in the skin buffer (which has also increased slightly to accomodate 1 backdrop for each skin and 2 full colour screens for bmps (up for 1.5))
- if no %X is specified in a skin then the backdrop setting will be used. use %Xd to explicitly disable a skin from displaying a backdrop
- the base skin can now specify a backdrop.
git-svn-id: svn://svn.rockbox.org/rockbox/trunk@24366 a1c6a512-1295-4272-9138-f99709370657
* Use Newton-Raphson divider on ARMv5e and ARMv6, about 7% speedup on Gigabeat S.
* On ARMv4 targets using IRAM, remove insane filter buffer from IRAM, fill available IRAM with LUT of reciprocals for small divisors - speedup varies according to target and available IRAM, APE normal sample is approx. 109% RT on e200.
* Rename apps/codecs/lib/udiv32_armv4.S to apps/codecs/lib/udiv32_arm.S, which includes dividers for all ARM targets specialized for APE.
git-svn-id: svn://svn.rockbox.org/rockbox/trunk@24354 a1c6a512-1295-4272-9138-f99709370657
All: remove obsolete blackjack entry (the manual for some other games is
still incorrect since a menu was added)
Fuze: add missing keymaps
git-svn-id: svn://svn.rockbox.org/rockbox/trunk@24343 a1c6a512-1295-4272-9138-f99709370657
This makes plugins show the menu backdrop in their backdrop so that they don't look like crap if you have an sbs and look more integrated. I've test about all plugins and all work fine.
git-svn-id: svn://svn.rockbox.org/rockbox/trunk@24335 a1c6a512-1295-4272-9138-f99709370657
The proper fix is obviously to make usb_storage_transfer_complete() not get confused, but that's a lot more work, and I prefer things to be correct to them being a bit faster right now
git-svn-id: svn://svn.rockbox.org/rockbox/trunk@24333 a1c6a512-1295-4272-9138-f99709370657
