[PATCH v4] mmc: zynq: parse dt when probing

[Michal Simek]

Hi,

On 15. 04. 20 14:05, Benedikt Grassl wrote:
> Hi Michal,
> 
> I just tried to check if the read performance acutally increased when
> using an 8-bit data bus. As a quick test, I use the fatload command to
> read a chunk of 1 GByte from an eMMC flash (generated with dd from
> Linux). However, I don't see any difference at all:
> 
> mmc info
> Device: mmc at ff160000
> Manufacturer ID: 9d
> OEM: 101
> Name: IS016 
> Bus Speed: 200000000
> Mode: HS200 (200MHz)
> Rd Block Len: 512
> MMC version 5.0
> High Capacity: Yes
> Capacity: 14.6 GiB
> Bus Width: 4-bit
> (...)
> 
> fatload mmc 0 ${loadimage} test.img
> 1070579712 bytes read in 70458 ms (14.5 MiB/s)
> 
> With an 8-bit data bus:
> 
> mmcinfo 
> Device: mmc at ff160000
> Manufacturer ID: 9d
> OEM: 101
> Name: IS016 
> Bus Speed: 200000000
> Mode: HS200 (200MHz)
> Rd Block Len: 512
> MMC version 5.0
> High Capacity: Yes
> Capacity: 14.6 GiB
> Bus Width: 8-bit
> (...)
> 
> fatload mmc 0 ${loadimage} test.img
> 1070579712 bytes read in 70551 ms (14.5 MiB/s)
> 
> I wonder if the fatload command is the bottleneck here? Unfortunately
> I'm working from home right now and don't have an oscilloscope at hand
> to verify in hardware. Do you have any inputs on that?

you can do raw write to avoid fat FS which is not optimal.

Thanks,
Michal