[PATCH]: Whoops, ignore that last patch
Richard Henderson
rth@redhat.com
Sun Aug 19 22:51:00 GMT 2001
On Sun, Aug 19, 2001 at 02:08:59PM -0400, Daniel Berlin wrote:
> * df.h (struct df): Add rts_order variable.
>
> * df.c (df_visit_next_rts): New function.
> (df_visit_next): Renamed to df_visit_next_rc
> (df_analyse_1): Allocate/compute/free rts_order as well.
> (df_rd_global_compute): Use df_visit_next_rc instead of
> df_visit_next.
> (df_ru_global_compute): Use df_visit_next_rts instead of
> df_visit_next.
>
> * flow.c (flow_reverse_top_sort_order_compute): New function.
>
> * basic-block.h: Add prototype.
Ok, but,
> + static int
> + df_visit_next_rts (df, blocks)
> + struct df *df ATTRIBUTE_UNUSED;
df is most definitely used.
> + sbitmap blocks;
> + {
> + int i=0;
> + for (i = 0; i < n_basic_blocks; i++)
double init of i.
> + if (TEST_BIT (blocks, df->rts_order[i]))
> + return df->rts_order[i];
> + return sbitmap_first_set_bit (blocks);
when would the loop ever not cover all blocks, and thus
when would sbitmap_first_set_bit ever find a bit set
that the loop wouldn't?
r~
More information about the Gcc-patches
mailing list