> On May 27, 2019, at 5:15 AM, David Marchand <david.march...@redhat.com> wrote:
>
> Hello Dharmik,
Hi David,
>
> On Wed, May 22, 2019 at 9:38 PM Dharmik Thakkar <dharmik.thak...@arm.com>
> wrote:
> > On May 22, 2019, at 12:17 PM, David Marchand <david.march...@redhat.com>
> > wrote:
> > On Wed, May 22, 2019 at 6:27 PM Dharmik Thakkar <dharmik.thak...@arm.com>
> > wrote:
> > > @@ -738,10 +736,9 @@ struct {
> > > rte_eal_remote_launch(test_rwc_reader,
> > > (void
> > > *)(uintptr_t)read_type,
> > >
> > > enabled_core_ids[i]);
> > > - rte_eal_mp_wait_lcore();
> > >
> > > for (i = 1; i <= rwc_core_cnt[n]; i++)
> > > - if (lcore_config[i].ret < 0)
> > > + if (rte_eal_wait_lcore(i) < 0)
> > if (rte_eal_wait_lcore(enabled_core_ids[i]) < 0)
> >
> > (There are similar changes in other functions too.
> > I realize that this is a separate issue than what the patch is aimed for.
> > If you see fit, please integrate it, else I will put out a patch once your
> > patch has been merged.)
> >
> > Indeed, reproduced.
> >
>
> Did you have a chance to look at this ?
> These loops could be reworked by using RTE_LCORE_FOREACH_SLAVE() (so that it
> automatically skips the master lcore) removing the need for
> enabled_core_ids[].
>
Yes, I have tried to use RTE_LCORE_FOREACH_SLAVE() and eliminate
enabled_core_ids[]. However, the code becomes quite complicated to read with if
statements
within RTE_LCORE_FOREACH_SLAVE(), to break the loop (since not all slave cores
are checked for). This is a snip from test_hash_multi_add_lookup():
/* snip */
RTE_LCORE_FOREACH_SLAVE(core_id) {
if (i < rwc_core_cnt[n])
rte_eal_remote_launch(test_rwc_reader,
(void
*)(uintptr_t)read_type,
core_id);
else if (i >= rwc_core_cnt[n] &&
i < (rwc_core_cnt[n] +
rwc_core_cnt[m])) {
rte_eal_remote_launch(test_rwc_multi_writer,
(void
*)(uintptr_t)pos_core,
core_id);
pos_core++;
} else
break;
i++;
}
>
> --
> David Marchand
