Re: [f-cpu] vsprintf result for GCC & FCPU

[devik <devik@cdi.cz>]

> > :-) The code should work but you have to define macros for
> > jumps. The code is not still able to handle loadaddr and
> > jumps. Also there is still lack of conditional move - it
> > is one which should reduce code size further.
>
> What kind of macros?

like
.macro jmp_direct_nz r,label
  loadcons label
  jmp r,label,r0
.endm

I still can't decide way how to generate lavel & symbol
references. 64 bit symbol load needs 4 loadcons (16 bytes)
and there is a lot of them.
Local labels can use loadaddri (if they are in 16bit range).
I'd rather see to use 32bit addresses for code and 64 for data.

> When you finally find the time, can you please change the assembler
> syntax a little? First, immediate operands should be prefixed with a `$'
> character (that is, in real instructions, not in .pseudo ops) - that way
> it's possible to make a difference between symbolic constants and register
> names (just consider a symbol named `r0') without looking at the opcode.
> Second, please change the register names to r0...r63. There are no
> designated argument or temporary registers, stack or frame pointers.
> Besides that, it's not wise to reflect the calling conventions in the
> register names, and IMHO it's also harder to read.

ok done, see the example of asm I send yesterday.

> Once you're done with that, my assembler should be able to process the
> assembler source files generated by your gcc port. Since I also have ar,
> mcs, nm, size, strings and strip utilities floating around on my disk,
> we would have an almost complete, working f-cpu toolchain (only ld is
> still missing).

can you send me these tools ? So I'd be able at least validate the
output...
devik

*************************************************************
To unsubscribe, send an e-mail to majordomo@seul.org with
unsubscribe f-cpu       in the body. http://f-cpu.seul.org/