Michael Schnell wrote:
Re-implementing SMP on FPGA-based multiCPU systems is, to me, not
necessariyl the right way to go.
Instead of SMP, all applications I have in mind would rather use
separate systems which communicate via some kind of channel. Does
running multiple Linuxes on a multi-processor FPGA with a single RAM
need some extended support ? You of course would not like to have
multiple copies of the OSes code pages in RAM.
Doing it that way might be tough - the common code segment will refer to
the same addresses for many kernel data objects. Some tricky memory
remapping on the data side of the CPU (assuming Harvard architecture)
could allow a shared code segment, with private data, yet somehow
placing all of this in the same physical external memory.
We've build multiCPU linux systems in MicroBlaze before, but the
approach was just to segment the RAM into regions, each one containing a
complete kernel code and data space. Wasteful of memory, but easy to
implement!
John
_______________________________________________
uClinux-dev mailing list
uClinux-dev@uclinux.org
http://mailman.uclinux.org/mailman/listinfo/uclinux-dev
This message was resent by uclinux-dev@uclinux.org
To unsubscribe see:
http://mailman.uclinux.org/mailman/options/uclinux-dev
