Re: [PATCH bpf-next v2] bpf: test_bpf: Print total time of test in the summary

[Alexei Starovoitov]

On Mon, Aug 23, 2021 at 8:00 PM Tiezhu Yang <yangtiezhu@xxxxxxxxxxx> wrote:
>
> The total time of test is useful to compare the performance
> when bpf_jit_enable is 0 or 1, so print it in the summary.
>
> Signed-off-by: Tiezhu Yang <yangtiezhu@xxxxxxxxxxx>
> ---
>  lib/test_bpf.c | 24 ++++++++++++++++++------
>  1 file changed, 18 insertions(+), 6 deletions(-)
>
> diff --git a/lib/test_bpf.c b/lib/test_bpf.c
> index 830a18e..37f49b7 100644
> --- a/lib/test_bpf.c
> +++ b/lib/test_bpf.c
> @@ -8627,9 +8627,10 @@ static int __run_one(const struct bpf_prog *fp, const void *data,
>         return ret;
>  }
>
> -static int run_one(const struct bpf_prog *fp, struct bpf_test *test)
> +static int run_one(const struct bpf_prog *fp, struct bpf_test *test, u64 *run_one_time)
>  {
>         int err_cnt = 0, i, runs = MAX_TESTRUNS;
> +       u64 time = 0;
>
>         for (i = 0; i < MAX_SUBTESTS; i++) {
>                 void *data;
> @@ -8663,8 +8664,12 @@ static int run_one(const struct bpf_prog *fp, struct bpf_test *test)
>                                 test->test[i].result);
>                         err_cnt++;
>                 }
> +
> +               time += duration;
>         }
>
> +       *run_one_time = time;
> +
>         return err_cnt;
>  }
>
> @@ -8944,9 +8949,11 @@ static __init int test_bpf(void)
>  {
>         int i, err_cnt = 0, pass_cnt = 0;
>         int jit_cnt = 0, run_cnt = 0;
> +       u64 total_time = 0;
>
>         for (i = 0; i < ARRAY_SIZE(tests); i++) {
>                 struct bpf_prog *fp;
> +               u64 run_one_time;
>                 int err;
>
>                 cond_resched();
> @@ -8971,7 +8978,7 @@ static __init int test_bpf(void)
>                 if (fp->jited)
>                         jit_cnt++;
>
> -               err = run_one(fp, &tests[i]);
> +               err = run_one(fp, &tests[i], &run_one_time);
>                 release_filter(fp, i);
>
>                 if (err) {
> @@ -8981,10 +8988,12 @@ static __init int test_bpf(void)
>                         pr_cont("PASS\n");
>                         pass_cnt++;
>                 }
> +
> +               total_time += run_one_time;
>         }
>
> -       pr_info("Summary: %d PASSED, %d FAILED, [%d/%d JIT'ed]\n",
> -               pass_cnt, err_cnt, jit_cnt, run_cnt);
> +       pr_info("Summary: %d PASSED, %d FAILED, [%d/%d JIT'ed] in %llu nsec\n",
> +               pass_cnt, err_cnt, jit_cnt, run_cnt, total_time);
>
>         return err_cnt ? -EINVAL : 0;
>  }
> @@ -9192,6 +9201,7 @@ static __init int test_tail_calls(struct bpf_array *progs)
>  {
>         int i, err_cnt = 0, pass_cnt = 0;
>         int jit_cnt = 0, run_cnt = 0;
> +       u64 total_time = 0;
>
>         for (i = 0; i < ARRAY_SIZE(tail_call_tests); i++) {
>                 struct tail_call_test *test = &tail_call_tests[i];
> @@ -9220,10 +9230,12 @@ static __init int test_tail_calls(struct bpf_array *progs)
>                         pr_cont("ret %d != %d FAIL", ret, test->result);
>                         err_cnt++;
>                 }
> +
> +               total_time += duration;
>         }
>
> -       pr_info("%s: Summary: %d PASSED, %d FAILED, [%d/%d JIT'ed]\n",
> -               __func__, pass_cnt, err_cnt, jit_cnt, run_cnt);
> +       pr_info("%s: Summary: %d PASSED, %d FAILED, [%d/%d JIT'ed] in %llu nsec\n",
> +               __func__, pass_cnt, err_cnt, jit_cnt, run_cnt, total_time);

I think it only adds noise. Pls use dedicated runners like selftests/bpf/bench
for performance measurements. test_bpf.ko also does some.