Back to ASM

I've been quite busy lately trying to get a simple dsp algorithm running on an 8-bit processor.
For the first time I've realized that it's probably more challanging for the compiler to output efficient code for such a small processor, rather than a bigger one, say an ARM.
No wonders why you can't easily find an open source toolchain.
I've been amazed by the unexpected number of assembly instructions a simple instruction can translate to.
And sometimes it seems the case that what wouldn't be the quickest way of doing things in C, translates in effect into a fewer number of assembly instructions.