| Message ID | 1528992297-23920-1-git-send-email-alexey.kodanev@oracle.com |
|---|---|
| State | Accepted |
| Delegated to: | Petr Vorel |
| Headers | show |
| Series | [1/4] network/busy_poll: use smaller packets and a single thread | expand |
Hi Alexey, > This should make the results more stable for better performance > comparison with/without busy polling. > Signed-off-by: Alexey Kodanev <alexey.kodanev@oracle.com> Acked-by: Petr Vorel <pvorel@suse.cz> > --- Unfortunately I don't have any machine with busy polling, but I suppose you know what you're doing. > Tested on ixgbe NIC Kind regards, Petr
diff --git a/testcases/network/busy_poll/busy_poll01.sh b/testcases/network/busy_poll/busy_poll01.sh index 12d3cb5..a9c8cbd 100755 --- a/testcases/network/busy_poll/busy_poll01.sh +++ b/testcases/network/busy_poll/busy_poll01.sh @@ -58,7 +58,7 @@ trap "tst_brkm TBROK 'test interrupted'" INT for x in 50 0; do tst_resm TINFO "set low latency busy poll to $x" set_busy_poll $x - tst_netload -H $(tst_ipaddr rhost) -d res_$x + tst_netload -H $(tst_ipaddr rhost) -n 10 -N 10 -d res_$x done poll_cmp=$(( 100 - ($(cat res_50) * 100) / $(cat res_0) )) diff --git a/testcases/network/busy_poll/busy_poll02.sh b/testcases/network/busy_poll/busy_poll02.sh index 4873dc3..95eeee3 100755 --- a/testcases/network/busy_poll/busy_poll02.sh +++ b/testcases/network/busy_poll/busy_poll02.sh @@ -49,7 +49,7 @@ trap "tst_brkm TBROK 'test interrupted'" INT for x in 50 0; do tst_resm TINFO "set low latency busy poll to $x per socket" set_busy_poll $x - tst_netload -H $(tst_ipaddr rhost) -d res_$x -b $x + tst_netload -H $(tst_ipaddr rhost) -n 10 -N 10 -d res_$x -b $x done poll_cmp=$(( 100 - ($(cat res_50) * 100) / $(cat res_0) )) diff --git a/testcases/network/busy_poll/busy_poll03.sh b/testcases/network/busy_poll/busy_poll03.sh index 0144465..699bee2 100755 --- a/testcases/network/busy_poll/busy_poll03.sh +++ b/testcases/network/busy_poll/busy_poll03.sh @@ -51,7 +51,8 @@ do_test() for x in 50 0; do tst_resm TINFO "set low latency busy poll to $x per $1 socket" set_busy_poll $x - tst_netload -H $(tst_ipaddr rhost) -d res_$x -b $x -T $1 + tst_netload -H $(tst_ipaddr rhost) -n 10 -N 10 -d res_$x \ + -b $x -T $1 done poll_cmp=$(( 100 - ($(cat res_50) * 100) / $(cat res_0) )) diff --git a/testcases/network/busy_poll/busy_poll_lib.sh b/testcases/network/busy_poll/busy_poll_lib.sh index a00c74d..271c01b 100755 --- a/testcases/network/busy_poll/busy_poll_lib.sh +++ b/testcases/network/busy_poll/busy_poll_lib.sh @@ -14,6 +14,9 @@ # You should have received a copy of the GNU General Public License # along with this program. If not, see <http://www.gnu.org/licenses/>. +# for more stable results set to a single thread +TST_NETLOAD_CLN_NUMBER=1 + if tst_kvcmp -lt "3.11"; then tst_brkm TCONF "test must be run with kernel 3.11 or newer" fi
This should make the results more stable for better performance comparison with/without busy polling. Signed-off-by: Alexey Kodanev <alexey.kodanev@oracle.com> --- Tested on ixgbe NIC testcases/network/busy_poll/busy_poll01.sh | 2 +- testcases/network/busy_poll/busy_poll02.sh | 2 +- testcases/network/busy_poll/busy_poll03.sh | 3 ++- testcases/network/busy_poll/busy_poll_lib.sh | 3 +++ 4 files changed, 7 insertions(+), 3 deletions(-)