2016-09-16 0:59 GMT+02:00 Luke Kenneth Casson Leighton <lkcl@lkcl.net>:
---
crowd-funded eco-conscious hardware: https://www.crowdsupply.com/eoma68


On Thu, Sep 15, 2016 at 3:38 PM, pelzflorian (Florian Pelz)
<pelzflorian@pelzflorian.de> wrote:
> On 09/15/2016 04:01 PM, Luke Kenneth Casson Leighton wrote:
>>> I stuck S5P6818 in the search at http://elinux.org/ (nope) and http://rhombus-tech.net/ (yep). Looks like a 64 bit chip with no 64 bit support.
>>
>>  correct.  given that it can only address 2GB of RAM that really doesn't matter.
>>
>> l.
>>
>
> What about virtual addressing / swap space? Then you may want more than
> 2GB address space.

 you do _not_ want to be using swap space on raw nand or even eMMC.

Well, Google has released statistical data on their SSD usage and it seems that there is no correlation between number of writes and failure. It's more dependent on 'age'. 

But SSD are 'less' reliable than HDD because of bad cell and bad writes. So error correction becomes more important.

http://www.zdnet.com/article/ssd-reliability-in-the-real-world-googles-experience/

Raw NAND access means you have to do the ECC. With and SSD that taken care of by.....firmware..... 

Now the 3.4 (Lichee)kernel from AW has, AFAICT, shady NAND support. And mainline is growing, proper, NAND support.

Most bootloaders depend on fixed addresses region without ECC. So if the boot region gets "bad" you're device is toast. A20 still boots from SD though.

Still on low speed machines try to avoid swap to any medium. All will get very slow; I/O contention. Memory usually has it's separate/private bus/tracs/connection. The rest, Network, Sata, USB, GPIO, SPI etc. shares a common bus.

So keep away from high profile desktops/compositors like Gnome and KDE on low memory systems.

IOS and Andriod have very strict policies on apps to get exit the system if not used to keep memory free for the active application.

 
or USB-based external storage media.  in fact, you don't want to be
using swap space at all... with the exception possibly of compswap
(the much better version of zram, which linus torvalds refused to
allow the full set of patches for, into the linux kernel).

l.

_______________________________________________
arm-netbook mailing list arm-netbook@lists.phcomp.co.uk
http://lists.phcomp.co.uk/mailman/listinfo/arm-netbook
Send large attachments to arm-netbook@files.phcomp.co.uk