"Slow and Clunky" - that's a nebulous term at best! Slow is relative to what we want to achieve - if a floating point operation takes 1ms and that is acceptable for the application then it is not 'slow' and as for 'clunky' is software floating point any less clunky on a 32 bit cpu? Algorithmically it is identical. Now if our 32 bit has a FPU then that's another kettle of fish!

Back to the RTOS issue - any RTOS has an overhead for the task switch and the RTOS management itself. Again, in light of the application we have to weigh up whether to use a RTOS or not and if we use a RTOS, can we cope with the task switching overhead. I saw a commercial box of tricks actually do pre-emptive switching on a 12clock 8051 at 11.0592MHz that would move the block of registers to/from xram. I was stunned when I saw the code -but it did in fact work and the box did what it was designed to do. Personally I would've suggested the designer go to a more suitable cpu - but he liked his 8051s and that is what he used. The designer is an accomplished assembler programmer - the quality of his code was very good.

Again, I wouldn't recommend a RTOS on a 8051, but if the app had to have an 8051 and had to have a RTOS and the performance was adequate - then so be it. Does running a RTOS make any more sense on a Philips ARM micro simply because it is 32 bit? It does not have any memory managment or specific hardware to support a RTOS. It does however have more ram, a larger stack pointer and more performance than the average 8051 cpu that would make it more preferable for such a task. In days gone by, many people used RTOS's on z80,hc11,6809 etc and continue to do so on Z180S,HC12,H8. I wouldn't suggest that the people doing this are non-sensical.