| Message ID | 1471461202-22426-1-git-send-email-mike.holmes@linaro.org |
|---|---|
| State | New |
| Headers | show |
On 08/17/16 22:13, Mike Holmes wrote: > A recent bug showed the limitations of using powers of two, add a prime > number of threads as a new test case. > > Signed-off-by: Mike Holmes <mike.holmes@linaro.org> > --- > test/common_plat/performance/odp_scheduling_run.sh | 1 + > test/linux-generic/performance/odp_scheduling_run_proc.sh | 1 + > 2 files changed, 2 insertions(+) > > diff --git a/test/common_plat/performance/odp_scheduling_run.sh b/test/common_plat/performance/odp_scheduling_run.sh > index 755b0c1..29c1929 100755 > --- a/test/common_plat/performance/odp_scheduling_run.sh > +++ b/test/common_plat/performance/odp_scheduling_run.sh > @@ -20,6 +20,7 @@ run() > } > > run 1 > +run 7 > run 8 > it will use cpu number threads if cpu number is lower that that value. Setting 7 you need at least 7 cpus (or even 7 + 2 (if I remember 2 threads there not doing processing.)). So I think it has to be 3 processing threads. (value 3 or 5 check what log says.). Maxim. > exit $ret > diff --git a/test/linux-generic/performance/odp_scheduling_run_proc.sh b/test/linux-generic/performance/odp_scheduling_run_proc.sh > index b3ef26f..3f578f7 100755 > --- a/test/linux-generic/performance/odp_scheduling_run_proc.sh > +++ b/test/linux-generic/performance/odp_scheduling_run_proc.sh > @@ -21,6 +21,7 @@ run() > } > > run 1 > +run 7 > run 8 > > exit $ret
On 17 August 2016 at 15:20, Maxim Uvarov <maxim.uvarov@linaro.org> wrote: > On 08/17/16 22:13, Mike Holmes wrote: > >> A recent bug showed the limitations of using powers of two, add a prime >> number of threads as a new test case. >> >> Signed-off-by: Mike Holmes <mike.holmes@linaro.org> >> --- >> test/common_plat/performance/odp_scheduling_run.sh | 1 + >> test/linux-generic/performance/odp_scheduling_run_proc.sh | 1 + >> 2 files changed, 2 insertions(+) >> >> diff --git a/test/common_plat/performance/odp_scheduling_run.sh >> b/test/common_plat/performance/odp_scheduling_run.sh >> index 755b0c1..29c1929 100755 >> --- a/test/common_plat/performance/odp_scheduling_run.sh >> +++ b/test/common_plat/performance/odp_scheduling_run.sh >> @@ -20,6 +20,7 @@ run() >> } >> run 1 >> +run 7 >> run 8 >> >> > it will use cpu number threads if cpu number is lower that that value. > Setting 7 you need at least 7 cpus (or even 7 + 2 (if I remember 2 threads > there not doing processing.)). > So I think it has to be 3 processing threads. (value 3 or 5 check what log > says.). Right, currently we try for 8, are you suggesting we also change this value to not use "run 8" so that it runs on smaller machines in general ? On my 4 + 4 hyper threading machine the 7 and the 8 tests both use 6 workers, but I assume we were targeting bigger machines. Perhaps we add 3 and 4 as well ? > > > Maxim. > > > > exit $ret >> diff --git a/test/linux-generic/performance/odp_scheduling_run_proc.sh >> b/test/linux-generic/performance/odp_scheduling_run_proc.sh >> index b3ef26f..3f578f7 100755 >> --- a/test/linux-generic/performance/odp_scheduling_run_proc.sh >> +++ b/test/linux-generic/performance/odp_scheduling_run_proc.sh >> @@ -21,6 +21,7 @@ run() >> } >> run 1 >> +run 7 >> run 8 >> exit $ret >> > > -- Mike Holmes Technical Manager - Linaro Networking Group Linaro.org <http://www.linaro.org/> *│ *Open source software for ARM SoCs "Work should be fun and collaborative, the rest follows"
On 08/17/16 22:51, Mike Holmes wrote: > > > On 17 August 2016 at 15:20, Maxim Uvarov <maxim.uvarov@linaro.org > <mailto:maxim.uvarov@linaro.org>> wrote: > > On 08/17/16 22:13, Mike Holmes wrote: > > A recent bug showed the limitations of using powers of two, > add a prime > number of threads as a new test case. > > Signed-off-by: Mike Holmes <mike.holmes@linaro.org > <mailto:mike.holmes@linaro.org>> > --- > test/common_plat/performance/odp_scheduling_run.sh | 1 + > test/linux-generic/performance/odp_scheduling_run_proc.sh | 1 + > 2 files changed, 2 insertions(+) > > diff --git > a/test/common_plat/performance/odp_scheduling_run.sh > b/test/common_plat/performance/odp_scheduling_run.sh > index 755b0c1..29c1929 100755 > --- a/test/common_plat/performance/odp_scheduling_run.sh > +++ b/test/common_plat/performance/odp_scheduling_run.sh > @@ -20,6 +20,7 @@ run() > } > run 1 > +run 7 > run 8 > > it will use cpu number threads if cpu number is lower that that > value. Setting 7 you need at least 7 cpus (or even 7 + 2 (if I > remember 2 threads there not doing processing.)). > So I think it has to be 3 processing threads. (value 3 or 5 check > what log says.). > > > Right, currently we try for 8, are you suggesting we also change this > value to not use "run 8" so that it runs on smaller machines in general ? > On my 4 + 4 hyper threading machine the 7 and the 8 tests both use 6 > workers, but I assume we were targeting bigger machines. > > Perhaps we add 3 and 4 as well ? I just looked to tests and my suggestion is following: -c 0 - all available cores -c 1 - one thread -c `nproc --all --ignore=1` for all available even cores -f - test that -f option works. Maxim. > > > > Maxim. > > > > exit $ret > diff --git > a/test/linux-generic/performance/odp_scheduling_run_proc.sh > b/test/linux-generic/performance/odp_scheduling_run_proc.sh > index b3ef26f..3f578f7 100755 > --- a/test/linux-generic/performance/odp_scheduling_run_proc.sh > +++ b/test/linux-generic/performance/odp_scheduling_run_proc.sh > @@ -21,6 +21,7 @@ run() > } > run 1 > +run 7 > run 8 > exit $ret > > > > > > -- > Mike Holmes > Technical Manager - Linaro Networking Group > Linaro.org <http://www.linaro.org/>***│ *Open source software for ARM SoCs > "Work should be fun and collaborative, the rest follows" >
diff --git a/test/common_plat/performance/odp_scheduling_run.sh b/test/common_plat/performance/odp_scheduling_run.sh index 755b0c1..29c1929 100755 --- a/test/common_plat/performance/odp_scheduling_run.sh +++ b/test/common_plat/performance/odp_scheduling_run.sh @@ -20,6 +20,7 @@ run() } run 1 +run 7 run 8 exit $ret diff --git a/test/linux-generic/performance/odp_scheduling_run_proc.sh b/test/linux-generic/performance/odp_scheduling_run_proc.sh index b3ef26f..3f578f7 100755 --- a/test/linux-generic/performance/odp_scheduling_run_proc.sh +++ b/test/linux-generic/performance/odp_scheduling_run_proc.sh @@ -21,6 +21,7 @@ run() } run 1 +run 7 run 8 exit $ret
A recent bug showed the limitations of using powers of two, add a prime number of threads as a new test case. Signed-off-by: Mike Holmes <mike.holmes@linaro.org> --- test/common_plat/performance/odp_scheduling_run.sh | 1 + test/linux-generic/performance/odp_scheduling_run_proc.sh | 1 + 2 files changed, 2 insertions(+) -- 2.7.4