Re: Update postreload.c to avoid long compilation time

[Paolo Bonzini <bonzini at gnu dot org>]

On 07/31/2010 02:03 AM, Bernd Schmidt wrote:
> It's a bit unfortunate, and probably avoidable, to create additional
> garbage RTL.  Maybe what's really needed is
>    df_set_flags (DF_DEFER_INSN_RESCAN);
> at the top of postreload?  Can you try that?

That's a possibility, but considering postreload is a heavy user of 
note_uses/note_stores, deferring rescans makes it impossible to switch 
it to use the operand caches someday.

> Or, even simpler, avoid
> the SET_REGNO macro, and check all other occurrences of it.  There seem
> to be a few, e.g. in caller-save.c or ira.c, which also probably
> shouldn't invoke df.

Better, yes.

Paolo