Attempt to make Serial_HelloWorld compatible with SDCC 4.4.0

[jz13138]

This Pull Request refers to
https://github.com/free-pdk/free-pdk-examples/issues/13

[spth]

I didn't try it, but from a quick look at the diff it looks ok to me, and should do what the title claims.

Two questions:

• What is the motivation for the PFS154->PFS173 change?
• Do we really need those asm __set1io and __set0io? Why not just use &= and |= with suitable masks?

[cpldcpu]

Do we really need those asm __set1io and __set0io? Why not just use &= and |= with suitable masks?

I believe older versions of SDCC were not able to infer the set instruction for different memory types. Is that now all fixed?

[spth]

Compiling that Serial_HelloWorld, which has both, it looks like there is no difference:

;	../include/serial.h: 28: __set0io(PA, SERIAL_TX_PIN);                // Send 0 on TX Pin
	set0.io	__pa, #7

vs.

;	../include/serial.h: 39: INTEN &= ~INTEN_TM2;                          // Disable TM2 (setup of 16 bit value txdata is non atomic)
	set0.io	__inten, #6

Philipp

P.S.: AFAIK, SDCC does use set0 and set1 for I/O. But it can't do so for general RAM: set0 and1 are only available for a part of the RAM, and which variable will be where in RAM is typically not known before link time.

[jz13138]

Thanks for your feedback.

* What is the motivation for the PFS154->PFS173 change?

I am just evaluating PFS173 at the moment. No need to change that here. Sorry for that.

* Do we really need those asm `__set1io` and `__set0io`? Why not just use `&=` and `|=` with suitable masks?

I was hoping that it will run faster.

However, in a different test case I experienced that
__set1io(ADCC, ADCC_PROCESS_CONTROL_BIT);
could start the ADC conversion whereas |= didn't.

[cpldcpu]

Shall we merge this? Or what are the proposed changes? Changing to PFS154 as default?

[spth]

I'd rather just make the minimal changes necessary as a first step: Replace the mov by mov.io, etc in the inline asm.
Second step: Eliminate inline asm where it doesn't really provide an advantage. For now, I don't see much point in the rest of changes in the pull request (PFS154 default, etc).

P.S.: I've now implemented what I proposed above.