On 01/20/2018 02:09 PM, ron minnich wrote:
On Sat, Jan 20, 2018 at 11:07 AM Taiidan@gmx.com Taiidan@gmx.com wrote:
Have you tried the D8/D16 OpenBMC port yet?
I'll take a look. I suspect I am going to do a u-root based openbmc release for contrast. The openbmc work I've seen to date does not impress. I don't think using C++ and Boost was a good decision.
ron
Can you explain why this is a bad thing? And in terms of quality you mean when it comes to the port?, the facebook OpenBMC? or the IBM OpenBMC? (what is the difference anyway? why did facebook make their own?)
Thanks!
On Sat, Jan 20, 2018 at 11:22 AM Taiidan@gmx.com Taiidan@gmx.com wrote:
Can you explain why this is a bad thing? And in terms of quality you mean when it comes to the port?, the facebook OpenBMC? or the IBM OpenBMC? (what is the difference anyway? why did facebook make their own?)
The main complaint that has been put to me by several vendors is the 1 GiB footprint on the OpenBMC they've looked at. They'd like to see if a smaller memory footprint system if possible. Also, since in at least one case they are using NERF on the main x86 CPU, they'd like to have NERF on both the main CPU and the BMC. It simplifies their life immensely to have a common code base on both CPUs, and NERF makes that easy: I've tested NERF on small ARM systems and it works fine, and it hardly needs 1 GiB to function.
ron