Dear Eric Nelson,
Hi Marek,
On 11/14/2013 02:20 PM, Marek Vasut wrote:
Dear Eric Nelson,
This patch updates the memory configuration files for the Nitrogen6X boards to use macros to define the output for imximage. This allows re-use of the same files of register/value pairs to be included from C as a pre-cursor to run-time detection of the memory size on a board.
Signed-off-by: Eric Nelson eric.nelson@boundarydevices.com
board/boundary/nitrogen6x/1066mhz_4x128mx16.cfg | 72
++++++++++++------------- board/boundary/nitrogen6x/1066mhz_4x256mx16.cfg
| 72 ++++++++++++-------------
board/boundary/nitrogen6x/800mhz_2x128mx16.cfg | 72 ++++++++++++------------- board/boundary/nitrogen6x/800mhz_2x256mx16.cfg
| 72 ++++++++++++-------------
board/boundary/nitrogen6x/800mhz_4x128mx16.cfg | 72 ++++++++++++------------- board/boundary/nitrogen6x/800mhz_4x256mx16.cfg
| 72 ++++++++++++------------- board/boundary/nitrogen6x/nitrogen6dl.cfg | | 1 +
board/boundary/nitrogen6x/nitrogen6dl2g.cfg | 1 + board/boundary/nitrogen6x/nitrogen6q.cfg | 1 + board/boundary/nitrogen6x/nitrogen6q2g.cfg | 1 + board/boundary/nitrogen6x/nitrogen6s.cfg | 1 + board/boundary/nitrogen6x/nitrogen6s1g.cfg | 1 + 12 files changed, 222 insertions(+), 216 deletions(-)
I dont mind either way, but just an idea for discussion:
What do you say to implementing the DRAM controller programming in SPL? That way, we can even do proper DRAM calibration on boot and other such nice things. And we'd also get rid of these static magic numbers all around.
We can (and hopefully will very soon) implement some of the configuration in C under SPL, but we can't (and shouldn't) get rid of all of the magic numbers.
That would be really nice.
Some of these numbers come from testing large numbers of devices under a variety of temperature conditions. The best that the automagic calibration facilities can do is set things up based on nominal current conditions.
This is what I'm concerned about. On the other hand, the DRAM controller in iMXes does run-time recalibration of the DDR, does it not ?
That said, I'm sure we can cut down on the total number.
Absolutely. And if we cannot do runtime calibration, we could at least have support for multiple boards in a single image.
My immediate goal is simply to set the stage for SPL and a single binary.
Full ACK on this course ;-)
Best regards, Marek Vasut