On Wed, Jan 11, 2017 at 02:09:34PM +0100, Andrew Jones wrote: > On Wed, Jan 11, 2017 at 12:00:23PM +0100, Andrew Jones wrote: > > On Wed, Jan 11, 2017 at 01:29:35PM +0800, Peter Xu wrote: > > > run_task.sh is getting slow. This patch is trying to make it faster by > > > running the tests concurrently. > > > > > > We provide a new parameter "-j" for the run_tests.sh, which can be used > > > to specify how many run queues we want for the tests. Default queue > > > length is 1, which is the old behavior. > > > > > > Quick test on my laptop (4 cores, 2 threads each) shows 3x speed boost: > > > > > > |-----------------+-----------| > > > | command | time used | > > > |-----------------+-----------| > > > | run_test.sh | 75s | > > > | run_test.sh -j8 | 27s | > > > |-----------------+-----------| > > > > > > Signed-off-by: Peter Xu <peterx@xxxxxxxxxx> > > > --- > > > run_tests.sh | 12 ++++++++++-- > > > scripts/common.bash | 16 +++++++++++++++- > > > 2 files changed, 25 insertions(+), 3 deletions(-) > > > > > > diff --git a/run_tests.sh b/run_tests.sh > > > index b6a1059..477d4fb 100755 > > > --- a/run_tests.sh > > > +++ b/run_tests.sh > > > @@ -13,10 +13,11 @@ function usage() > > > { > > > cat <<EOF > > > > > > -Usage: $0 [-g group] [-h] [-v] > > > +Usage: $0 [-g group] [-h] [-v] [-j num_run_queues] > > > > > > -g: Only execute tests in the given group > > > -h: Output this help text > > > + -j: Execute tests in parallel > > > -v: Enables verbose mode > > > > > > Set the environment variable QEMU=/path/to/qemu-system-ARCH to > > > @@ -28,7 +29,7 @@ EOF > > > RUNTIME_arch_run="./$TEST_DIR/run" > > > source scripts/runtime.bash > > > > > > -while getopts "g:hv" opt; do > > > +while getopts "g:hj:v" opt; do > > > case $opt in > > > g) > > > only_group=$OPTARG > > > @@ -37,6 +38,13 @@ while getopts "g:hv" opt; do > > > usage > > > exit > > > ;; > > > + j) > > > + unittest_run_queues=$OPTARG > > > + if (( $unittest_run_queues <= 0 )); then > > > + echo "Invalid -j option: $unittest_run_queues" > > > + exit 1 > > > > We should probably use 'exit 2' here, and below in the *) case. Will fix this one first, with another one line patch for the below one (we can either take that new one, or squash it into this patch). > > > > > + fi > > > + ;; > > > v) > > > verbose="yes" > > > ;; > > > diff --git a/scripts/common.bash b/scripts/common.bash > > > index 2dd7360..ef103ee 100644 > > > --- a/scripts/common.bash > > > +++ b/scripts/common.bash > > > @@ -1,11 +1,19 @@ > > > : ${unittest_log_dir:=logs} > > > +: ${unittest_run_queues:=1} > > > > > > function run_task() > > > { > > > local testname="$2" > > > > > > + while (( $(jobs | wc -l) == $unittest_run_queues )); do > > > + # wait for any background test to finish > > > + wait -n > > > + done > > > + > > > RUNTIME_log_file="${unittest_log_dir}/${testname}.log" > > > - "$@" > > > + > > > + # start the testcase in the background > > > + "$@" & > > > > If you check the logs before and after applying this patch series you'll > > see a bunch of "stty: 'standard input': Inappropriate ioctl for device" > > are now present. These messages come from the stty calls in run_qemu, > > which we need to avoid the loss of terminal echo when QEMU aborts. We > > can get rid of these new "inappropriate ioctl" messages by changing the > > above line to > > > > "$@" <$(tty) & > > Actually I just came up with a better solution for this. Since we don't > need input to our unit tests, i.e. we never attempt to read from the > serial port within them, then we can just use /dev/null for stdin. That > allows us to leave your code above alone, and also remove the stty stuff > in run_qemu. I'll send a patch. Thanks, then I'll keep it as it is. > > drew > > > > > > > > } > > > > > > function for_each_unittest() > > > @@ -22,6 +30,8 @@ function for_each_unittest() > > > local accel > > > local timeout > > > > > > + trap "wait; exit 130" SIGINT > > > + > > > exec {fd}<"$unittests" > > > > > > while read -u $fd line; do > > > @@ -55,5 +65,9 @@ function for_each_unittest() > > > fi > > > done > > > run_task "$cmd" "$testname" "$groups" "$smp" "$kernel" "$opts" "$arch" "$check" "$accel" "$timeout" > > > + > > > + # wait until all task finish > > > > Still not quite right :-) s/task/tasks/ Sorry! Fixing up. -- peterx -- To unsubscribe from this list: send the line "unsubscribe kvm" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html