On 2015-05-19 07:46, Richard Henderson wrote:
> On 05/19/2015 03:26 AM, Aurelien Jarno wrote:
> > @@ -1522,6 +1522,9 @@ static void tcg_liveness_analysis(TCGContext *s)
> > if (dead_temps[arg]) {
> > dead_args |= (1 << i);
> > }
> > + }
> > + for (i = nb_oargs; i < nb_oargs + nb_iargs; i++) {
> > + arg = args[i];
> > dead_temps[arg] = 0;
> > }
> > s->op_dead_args[oi] = dead_args;
>
> How about another line of commentary for each loop?
>
> Something like
>
> /* Record arguments that die in this opcode. */
>
> for the first and
>
> /* Input arguments are live for preceeding opcodes. */
>
> for the second.
Good point.
> As for the same loop for calls, you're right that it may well cause us to do a
> tiny bit of redundant work, but nothing else bad will happen. We'll enter
> temp_dead more times than necessary. I'm always skeptical about knowingly
> giving a compiler bad information though. You tend to not know how data is
> going to be used in future, and *then* get bad results.
You are correct. Anyway I don't think it'll make a big difference in
performance.
I'll send a new version of the patch.
--
Aurelien Jarno GPG: 4096R/1DDD8C9B
aurel...@aurel32.net http://www.aurel32.net
