On 01/17/12 21:48, Ilyes Gouta wrote: > I've got this scenario where I'd like to run a slave DirectFB instance > on the master CPU which will transmit the applications' calls to a > second CPU running the master DirectFB instance. This later would > really execute the calls and render to the screen. These two CPUs are > in a AMP configuration (w/ separate OSes) but could be made to share a > bit a uncached memory for implementing a SHM channel.
Interesting architecture, seems trendy. > Could it be possible to use DirectFB's latest Voodoo code to achieve > that use case? Voodoo is just simple socket based IP, so if you've got an IP connection between the kernels you can use Voodoo, of course. > If yes, is it all about implementing a new VoodooConnectionLink interface? I had a kernel based shared memory implementation of that in mind, but did not get to do it. But Fusion allows a dual CPU implementation with shared memory between the two kernels that is used by all Fusionees on both CPUs running transparently in one FusionWorld shared across CPUs/kernels. There's a UML based "emulator" implementation using a host file for shared memory that I still need to push... -- Best regards, Denis Oliver Kropp .------------------------------------------. | DirectFB - Hardware accelerated graphics | | http://www.directfb.org/ | "------------------------------------------" _______________________________________________ directfb-dev mailing list directfb-dev@directfb.org http://mail.directfb.org/cgi-bin/mailman/listinfo/directfb-dev