Re: [9fans] 5l bug

[erik quanstrom <quanstro@labs.coraid.com>]

On Mon Apr 29 14:55:02 EDT 2013, cinap_lenrek@gmx.de wrote:
> following up, theres my naive fix for this. instead of
> emiting conditional division instruction by 5c... dont and
> keep the branch. does this make any sense?
>
> term% hg diff -r 1777 peep.c
> diff -r 5eeb36d3ddd0 sys/src/cmd/5c/peep.c
> --- a/sys/src/cmd/5c/peep.c	Mon Jul 30 19:11:16 2012 +0200
> +++ b/sys/src/cmd/5c/peep.c	Mon Apr 29 20:51:12 2013 +0200
> @@ -1354,6 +1354,15 @@
>  			j->end = r->s2;
>  			return Branch;
>  		}
> +		switch(r->prog->as){
> +		case ADIV:
> +		case ADIVU:
> +		case AMOD:
> +		case AMODU:
> +			/* emulated by 5l, doesnt handle conditionals */
> +			j->end = r->s1;
> +			return Toolong;
> +		}
>  		if (modifiescpsr(r->prog)) {
>  			j->end = r->s1;
>  			return Setcond;

it's possible to do in the linker, but without trying this,
it looks reasonable to me.  what does the resulting asm
look like?

- erik