Macro expansion problem, RARS limitation or just me

[jknGH]

Hi there - thanks for RARS, I'm finding it useful getting going with RISC-V. I am having trouble converting the macro syntax of a source file I want to assemble, not sure if it's me or if the macro language needs a bit of work.

A minimal case with a contrived macro:

.macro mactest %out
    sw      t0,  0(%out)
    sw      t0,  4(%out)
    sw      t0,  8(%out)
    sw      t0,  12(%out)
.end_macro

Then this assembles:

	mactest a1

but neither of these do

	mactest 0(a1)
	mactest 4(a1)
        # etc.

(I get an error "forward reference or invalid parameters for macro "mactest"")
Is there a workaround I could use?

Thanks a lot, J^n

[TheThirdOne]

It makes sense to me that your first example compiles whereas your second does not. I can't think of any macro system that would allow the second examples. I don't think the expansion sw t0, 8(4(a1)) is meaningful.

More concretely in MARS (and consequently in RARS) all arguments to a macro must be single tokens so 4(a1) could not be a single argument. MARS macro help: https://courses.missouristate.edu/KenVollmar/MARS/Help/MacrosHelp.html

Is there a workaround I could use?

I don't quite know what you want your actual macro to do. If your goal is to offset lw and/or sw with multiple constants, I don't think that's possible in RARS.

A macro like the following is possible though.

.macro mactest %out %n
    sw      t0,  %n(%out)
.end_macro

[jknGH]

Hi, and thanks fro the swift response.

I should say that the macro was not originally in the form I showed; I've had to transform the code a little in order to accommodate other aspects of the RARS assembler syntax (I am not sure which assembler the original code used)

I think the intention is to cause the two numbers to be added, so not sw t0, 8(4(a1)) but sw t0, 12(a1). The macros originally had addressing in a form similar to sw t0, 4+%out but RARS doesn't seem to like that.

I agree that the way things were originally written here implies a more intelligent macro processor than just simple substitution. Since this particular bit of code only occurs at the end of the macro, I think I have a workaround (untested). I will split that part off and write a set of four 'postamble' macros:

.macro postamble0 %out
    sw      t0,  0(%out)
    sw      t0,  4(%out)
    sw      t0,  8(%out)
    sw      t0,  12(%out)
.end_macro

.macro postamble4 %out
    sw      t0,  4(%out)
    sw      t0,  8(%out)
    sw      t0,  12(%out)
    sw      t0,  16(%out)
.end_macro

...

And then put the 'correct' one of those in place after the full body of the larger macro is invoked.

Thanks, J^n

[TheThirdOne]

I think the intention is to cause the two numbers to be added, so not sw t0, 8(4(a1)) but sw t0, 12(a1). The macros originally had addressing in a form similar to sw t0, 4+%out but RARS doesn't seem to like that.

RARS doesn't support any form of arithmetic as part of arguments. That is a feature I would like to introduce some time in the future though.

I agree that the way things were originally written here implies a more intelligent macro processor than just simple substitution. Since this particular bit of code only occurs at the end of the macro, I think I have a workaround (untested). I will split that part off and write a set of four 'postamble' macros:
...
And then put the 'correct' one of those in place after the full body of the larger macro is invoked.

That seems like probably the best workaround that exists currently.

[jknGH]

Hi there - FYI I got things going with a variant of the 'multiple macros' approach I suggest above. Thanks!

One followup - question really, not an issue, sorry; I am using RARS in this way to try to get an idea of time taken for a critical function. I have no hardware as yet and I don't think QEMU can be used to get meaningful results for this. The 'cycle' CSR shown in RARS is giving me good answers(!). Can I trust this to be cycle-accurate?

Thanks once again for RARS, it is proving useful to me.

[TheThirdOne]

I would be cautious of reading too much into the cycles CSR. In RARS, cycles is equivalent to instret, so effectively RARS is a single cycle cpu with a fixed frequency. However if you would be using a single cycle cpu, RARS should be accurate.

I am happy RARS has been useful to you.