Restructure of libbaalue

At the moment i read a very useful book -> Large Scale C++ … one of the main ideas is that software is also physical … i don’t want to repeat the content, but for me, one action is that i will split libbaalue into more independent library’s … in the past i faced one of the problems Lakos mentioned -> how do i use only a small part of libbaalue … i don’t want to use the whole baalue lib for my games like crasonars, when i only want to use the error handling functions (see https://github.com/tjohann/libbaalue/blob/master/src/error.c ).

So one thing i learned is that i have to split a lager library into more smaller ones. I will start with the error handling parts and went on to the network parts.

For newer topics (like seccomp and namespaces) i will create a repository for learning (like Time Triggert parts -> TTEnv ) and then create a single lib for only such topics … At the moment it`s part of libbaalue.)

Release of A20_SDK_V2.9.2

Hi@all,

i released A20_SDK_V2.9.2 wich support a new device -> https://linux-sunxi.org/Sinovoip_Banana_Pi_M3 . This release also included some bug fixes in the supporter scripts.

Images are uploaded, binpkgs with emacs-gtk* packages are also uploaded.

Normally a new device would lead to a A20_SDK_V2.10.0, but i will also add another device -> https://linux-sunxi.org/Xunlong_Orange_Pi_Zero . The reason for this is the good support of jailhouse hypervisor.

So the update to A20_SDK_V2.10.0 will happen when the Orange Pi Zero is included.

So stay tuned …

Thorsten

Update distcc build time@baalue

I have updated the measurements of the kernel build time on my baalue cluster (see Baalue-Distcc-Repo ).

The linux kernel build config is based on my a20 devices. The only difference between the 2 runs is, that it compiles a kernel on a A20 (Cubietruck) and on a A83T (Cubiertuck-Plus) as the master node. The Cluster (8 A20@Bananapi) is the same, so it`s interesting to see how the master node effects the overall performance.

Cubietruck-Plus:

make CC=distcc -j32 LOADADDR=0x40008000 uImage modules dtbs

real 15m56,545s
user 79m49,928s
sys 20m4,033s

Cubietruck:

make CC=distcc -j32 LOADADDR=0x40008000 uImage modules dtbs

real 68m48,674s
user 81m58,231s
sys 44m3,516s

It`s quite impressing to see the difference. The reason is, that the master node is under full load, because it has to distribute the jobs into the cluster. This takes computation time and therefore the cluster nodes are NOT under full load.

see:

 

Release of A20_SDK_V2.9.0

Hi@all,

i`m happy to announce that a new bigger release of my A20_SDK is out. The git repo is already up to date and the images will follow at least on friday.

Due to some problems with dhcpd i decided to use dhclient and let my dhcpd setup the ip address. If you want to do it with static ip`s then pls see UPGRADE_HINTS for more infos.

Finally the cubietruck-plus works with all cores 🙂

Cheers & have fun

Thorsten

Release of A20_SDK_V2.8.3

Hi@all,

the release of A20_SDK_V2.8.3 happened … Don’t forget to do a make mrproper && make init_sdk … also do a make get_toolchain … it is updated to the latest void-linux versions.

Note: this is an in between release to A20_SDK_V2.9.0 which also provides new device images.

Cheers
Thorsten