This is the mail archive of the
mailing list for the GCC project.
Re: Cleanups for the m68k backend
On Tuesday 08 July 2003 02:09, Richard Henderson wrote:
> The best option, IMO, is to stop using asm_fprintf so much,
> and instead use output_asm_insn. Then, both cases would be
> xops = gen_rtx_REG (Pmode, i);
> output_asm_insn ("move%.l %0,%-", xops);
> Even better would be to update the m68k port to emit prologues
> as rtl instead of as text. Yes, I know scheduling is more or
> less pointless on these machines, but we'd like to eventually
> *elimitate* prologue-as-text for maintainence, validation, and
> other assorted reasons.
It's much more work, but I like this idea much more than other
alternatives. It always striked me that the prologue and epilogue
were emitting instructions directly without using RTL.
On second thought, this won't fix the MOTOROLA/MIT duality
everywhere: there's some conditional code even in m68k.md.
I think we need to merge the conditional code first, then
do the asm_fprintf() -> RTL conversion where needed.
// Bernardo Innocenti - Develer S.r.l., R&D dept.
Please don't send Word attachments - http://www.gnu.org/philosophy/no-word-attachments.html