HI again,

with this patch [1] time needed to flash the samr21 gets divided 
by factor 3. (numbers below) For someone doing it very often this
could be very interesting. It is still worked on but I tested it 
without error for now. 
Please provide feedback if you use this, as openocd guys are desperately
looking for EDBG testers.

Best, Thomas

with patch:
wrote 32768 bytes from file hello-world.hex in 5.071708s (6.310 KiB/s)          
                                  │
verified 16600 bytes in 1.374114s (11.797 KiB/s)

without patch
wrote 32768 bytes from file hello-world.hex in 16.532793s (1.936 KiB/s)         
                                  
verified 16600 bytes in 1.432996s (11.313 KiB/s)

[1] http://openocd.zylin.com/#/c/2356/

> On 27 Jan 2015, at 13:48, Thomas Eichinger <[email protected]> 
> wrote:
> 
> Hi again,
> 
> after some research I found this post [1] on the openocd mailing list
> explaining why flashing the samr21-xpro is unbearably slow.
> 
> tl;dr 
> The openocd cmsis-dap driver operates in synchronous operation
> and is not ported to a new asynchronous API yet. 
> Fingers crossed they port it soon.
> 
> Best, Thomas
> 
> [1] http://sourceforge.net/p/openocd/mailman/message/32496519/
> 
>> On 14 Jan 2015, at 15:46, Lucas Jenß <[email protected]> wrote:
>> 
>> Hi again,
>> 
>> so it seems that the slowness was caused by virtualization after all. My 
>> previous VM was an Ubuntu 13.10 running inside an older version of VMware 
>> Fusion, which resulted in the ~0.5KiB/s speed. After installing an Ubuntu 
>> 14.10 that went up to ~1.5KiB/s and running OpenOCD directly on the host 
>> gets me close to 2KiB/s when flashing. 
>> 
>> On OS X 10.9:
>> 
>> wrote 32768 bytes from file hello-world.hex in 16.164614s (1.980 KiB/s)
>> verified 16892 bytes in 1.463347s (11.273 KiB/s)
>> 
>> On Ubuntu 14.10:
>> 
>> wrote 32768 bytes from file 
>> /home/lucas/RIOT/examples/hello-world/bin/samr21-xpro/hello-world.hex in 
>> 22.042933s (1.452 KiB/s)
>> verified 16892 bytes in 1.505869s (10.955 KiB/s)
>> 
>> Cheers,
>> Lucas
>> 
>> On 13 Jan 2015, at 11:18, Martin <[email protected]> wrote:
>> 
>>> Hi,
>>> 
>>> my flashing speed is roughly equal to Thomas' for the Samr21-xpro:
>>> 
>>> ```
>>> wrote 65536 bytes from file RIOT/tests/pnet/bin/samr21-xpro/pnet.hex in 
>>> 32.083557s (1.995 KiB/s)
>>> verified 49688 bytes in 4.114729s (11.793 KiB/s)
>>> ```
>>> 
>>> My openocd version:
>>> `Open On-Chip Debugger 0.9.0-dev-00186-g30203b3 (2014-11-12-11:49)`
>>> 
>>> Best regards,
>>> Martin
>>> On 12.01.2015 21:07, Baptiste Clenet wrote:
>>>> Flashing is slow for us too, how do you get the speed?
>>>> 
>>>> 2015-01-12 11:13 GMT+01:00 Lucas Jenß <[email protected]>:
>>>> Hi Thomas,
>>>> 
>>>> verification was much faster as 0.4KiB/s, I think around 10 or so for me.
>>>> I checked out OpenOCD on the 9th. I’m also running Linux inside VMware
>>>> though, so maybe it’s just caused by the virtualization. I’ll see how fast
>>>> it is on the host.
>>>> 
>>>> Cheers,
>>>> Lucas
>>>> 
>>>> A couple of days ago.
>>>> 
>>>> On 12 Jan 2015, at 11:00, Thomas Eichinger <[email protected]> 
>>>> wrote:
>>>> 
>>>>> Hi Lucas,
>>>>> 
>>>>> I was playing with the openocd configuration a bit, mainly
>>>>> `adapter_speed`, back when support for this was added without
>>>>> any significant outcome.
>>>>> Problem is, the EDBG chip, on the bottom of the board, handling
>>>>> communication with the MCU is specified to run on 1MHz and the
>>>>> openocd docs mention, for CMSIS-DAP, it is not advised to let
>>>>> signal frequency exceed half of the operating frequency.
>>>>> (I’d guess Nyquist-Shannon applies)
>>>>> 
>>>>> That said, 0.481KiB/s still seems slow for this. I’m at least
>>>>> reaching 1.787KiB/s for flashing and 11.190KiB/s for verification.
>>>>> When did you check out the OpenOCD code?
>>>>> 
>>>>> Best, Thomas
>>>>> 
>>>>>> On 10 Jan 2015, at 14:25, Lucas Jenß <[email protected]> wrote:
>>>>>> 
>>>>>> Hey everyone,
>>>>>> 
>>>>>> I’ve been playing around with the Samr21 xpro and flashing
>>>>>> the device is _really_ slow, i.e. 0.481 KiB/s. Is this expected
>>>>>> or is there a way to improve it? I’m using the current OpenOCD
>>>>>> Git HEAD because the 0.8.0 release does not seem to contain the
>>>>>> configs for the board yet. I tried to flash the hello-world
>>>>>> example.
>>>>>> 
>>>>>> Cheers,
>>>>>> Lucas
>>>>>> _______________________________________________
>>>>>> devel mailing list
>>>>>> [email protected]
>>>>>> http://lists.riot-os.org/mailman/listinfo/devel
>>>>> 
>>>>> _______________________________________________
>>>>> devel mailing list
>>>>> [email protected]
>>>>> http://lists.riot-os.org/mailman/listinfo/devel
>>>>> 
>>>> 
>>>> _______________________________________________
>>>> devel mailing list
>>>> [email protected]
>>>> http://lists.riot-os.org/mailman/listinfo/devel
>>>> 
>>>> 
>>>> 
>>>> -- 
>>>> Clenet Baptiste
>>>> FR: +33 6 29 73 05 39
>>>> Élève-Ingénieur ESEO Angers, dernière année, spécialisation: Architecte 
>>>> système temps réél embarqué
>>>> Bidiplôme Master Robotics à l'Université de Plymouth en 2013-2014
>>>> 
>>>> 
>>>> 
>>>> _______________________________________________
>>>> devel mailing list
>>>> 
>>>> [email protected]
>>>> http://lists.riot-os.org/mailman/listinfo/devel
>>> 
>>> _______________________________________________
>>> devel mailing list
>>> [email protected]
>>> http://lists.riot-os.org/mailman/listinfo/devel
>> 
>> _______________________________________________
>> devel mailing list
>> [email protected]
>> http://lists.riot-os.org/mailman/listinfo/devel
> 
> _______________________________________________
> devel mailing list
> [email protected]
> http://lists.riot-os.org/mailman/listinfo/devel

_______________________________________________
devel mailing list
[email protected]
http://lists.riot-os.org/mailman/listinfo/devel

Reply via email to