I hadn't noticed this before but I guess the AVR32 has an MMU and Linux has been ported to it. Maybe we should add TLB and page table support to the XMM cache drivers for P2 so we can boot Linux too! :-)
I hadn't noticed this before but I guess the AVR32 has an MMU and Linux has been ported to it. Maybe we should add TLB and page table support to the XMM cache drivers for P2 so we can boot Linux too! :-)
Actually, the MMU was in an AVR32 version called the AP7 which I guess has been discontinued. You now have to go to an ARM-based chip to run Linux.
Linux is running on an ATMega1284p (also done on an ATMega644a) running an ARM emulator (the ATMega used does not have an MMU). The circuit adds an 80286 30pin SIMM DRAM board and SD card. It is extremely slow to boot (2hrs to boot a bash prompt), but proven to work. P1 could likely do a better job, so just imagine what P2 could do. http://dmitry.gr/index.php?r=05.Projects&proj=07.%20Linux%20on%208bit
Linux is running on an ATMega1284p (also done on an ATMega644a) running an ARM emulator (the ATMega used does not have an MMU). The circuit adds an 80286 30pin SIMM DRAM board and SD card. It is extremely slow to boot (2hrs to boot a bash prompt), but proven to work. P1 could likely do a better job, so just imagine what P2 could do. http://dmitry.gr/index.php?r=05.Projects&proj=07.%20Linux%20on%208bit
I guess I'm talking about running Linux natively. Running it under an emulation of another processor may be an interesting challenge but it won't produce a useable system.
I agree with David on this. It's more of a programmers joke than anything. I'd personally go with porting a OS that doesn't require a MMU and use that as a starting point.
So, it looks like someone has to write something called for example "MMUCog". Then port a Linux to the Propeller2.
Either that or use a version of Linux that doesn't require an MMU like UCLinux. Of course, there is a lot more to porting Linux to a processor than just providing an MMU. It might also be difficult to port Linux to a process that doesn't support interrupts. They can be simulated of course but that is kind of contrary to the Propeller philosophy.
2.11BSD is the last in the series of Berkeley Unixes for the PDP-11. it's footprint is also much smaller than uclinux(BSD needs 1/10th the amount of RAM), so easier to port to memory constrained systems.
2.11BSD is the last in the series of Berkeley Unixes for the PDP-11. it's footprint is also much smaller than uclinux(BSD needs 1/10th the amount of RAM), so easier to port to memory constrained systems.
BTW, somewhat off-topic, but is AVR32 getting very much attention (in particular, design wins)? I really don't hear much about it; OTOH, PIC32 seems to be at least holding its own against the ARM onslaught.
The PIC32 benchmarks(see coremark.org) are equivalent to a STM32F4 running at 168Mhz. So yes it can hold it's own against other embedded ARM's. Microchip is also planning on expanding the line later in the year.
It will be interesting to see what the Coremark is for the P2 once we get real silicon and a fully functional C compiler.
.
The Coremark benchmarks measure not only performance, but also try to measure how much performance you get for the amount of power the chip consumes.
I'd be curious if anyone has tried to move any of the coremark benchmarks to the prop 1 ? There have been some Dhrystone numbers published, though that is not used much anymore, though there are measures for the ARM.
Comments
http://dmitry.gr/index.php?r=05.Projects&proj=07.%20Linux%20on%208bit
2.11BSD is the last in the series of Berkeley Unixes for the PDP-11. it's footprint is also much smaller than uclinux(BSD needs 1/10th the amount of RAM), so easier to port to memory constrained systems.
BTW, somewhat off-topic, but is AVR32 getting very much attention (in particular, design wins)? I really don't hear much about it; OTOH, PIC32 seems to be at least holding its own against the ARM onslaught.
It will be interesting to see what the Coremark is for the P2 once we get real silicon and a fully functional C compiler.
.
I'd be curious if anyone has tried to move any of the coremark benchmarks to the prop 1 ? There have been some Dhrystone numbers published, though that is not used much anymore, though there are measures for the ARM.