Splitting call_insns

[Richard Henderson]

On Thu, Nov 23, 2000 at 12:15:59PM -0200, Alexandre Oliva wrote:
> We seem to be doing some loop unrolling even when -funroll-loops
> isn't given.

Yes.  We unroll small loops with a low trip count when it appears
that the code after unrolling would be smaller than keeping the loop.

> However, we won't do PC-relative calls if optimization isn't
> enabled, because otherwise the intermediate call pattern would
> be split too late.

I think you'd be better off running split_all_insns at the 
beginning of machine_dependent_reorg than have two methods
of call generation.

> 	* emit-rtl.c (try_split): Propagate CALL_INSN_FUNCTION_USAGE
> 	to CALL_INSNs in the split sequence.

Ok.


r~