Hi Heiko,
On 19/04/18 13:10, Heiko Stuebner wrote:
> Hi Enric,
>
> Am Donnerstag, 19. April 2018, 12:40:15 CEST schrieb Enric Balletbo i Serra:
>> DDR3 SDRAM Standard (JESD79-3F) defines some standard speed bins for
>> DDR3 memories. The devfreq/rk3399_dmc.txt binding refers to this file
>> which does not exist, so add a ddr.h file with the standard speed bins
>> for DDR3.
>>
>> Fixes: c1ceb8f7c167 (Documentation: bindings: add dt documentation for
>> rk3399 dmc)
>> Signed-off-by: Enric Balletbo i Serra <[email protected]>
>> ---
>>
>> include/dt-bindings/clock/ddr.h | 34 +++++++++++++++++++++++++++++++++
>> 1 file changed, 34 insertions(+)
>> create mode 100644 include/dt-bindings/clock/ddr.h
>>
>> diff --git a/include/dt-bindings/clock/ddr.h
>> b/include/dt-bindings/clock/ddr.h
>> new file mode 100644
>> index 000000000000..506aef7e609e
>> --- /dev/null
>> +++ b/include/dt-bindings/clock/ddr.h
>> @@ -0,0 +1,34 @@
>> +/* SPDX-License-Identifier: (GPL-2.0+ OR MIT) */
>> +
>> +#ifndef DT_BINDINGS_DDR_H
>> +#define DT_BINDINGS_DDR_H
>> +
>> +/* DDR3-800 Standard Speed Bins */
>> +#define DDR3_800D 15
>> +#define DDR3_800E 18
>> +/* DDR3-1066 Standard Speed Bins */
>> +#define DDR3_1066E 18
>> +#define DDR3_1066F 21
>> +#define DDR3_1066G 24
>
> looking at the mentioned jedec standard, I don't see where these numerical
> values are defined in the standard itself. [I may be blind though]
>
Damm it. No, you're not blind. I did an horrible mistake adding an old version
of this file. It's wrong, that is supposed to be there is DDR3 table available
in the ATF [1] not this.
/* 5-5-5 */
DDR3_800D = 0,
/* 6-6-6 */
DDR3_800E = 1,
/* 6-6-6 */
DDR3_1066E = 2,
/* 7-7-7 */
DDR3_1066F = 3,
...
The value is passed directly to the ATF so the above values are the correct.
One question that I have on my mind is if this file should be called as is or if
will be better name as atf-ddr.h or something like that.
ddr.h -> atf-ddr.h
[1]
https://github.com/ARM-software/arm-trusted-firmware/blob/master/plat/rockchip/rk3399/drivers/dram/dram_spec_timing.h
Sorry about that. Guess it's needed a v2 but let's wait a little bit for if more
discussion arises.
Best regards,
Enric
> Could you explain a bit more where these numerical values are coming from?
>
>
> Thanks
> Heiko
>
