Re: [dpdk-dev] [PATCH] test: provide performance difference ratio and test time

[Jianbo Liu <jianbo.liu@arm.com>]

Hi Herbert,

The 10/23/2017 10:35, Herbert Guan wrote:
> The printed time values presented in TSC is not straight forward
> showing the performance difference.  And if the high resolution
> counter is not enabled, time value is too small to show the actual
> performance (e.g. "1 - 1" seems the same but in fact the perfomance
> diffs by 50% or more).
> With the performance difference ratio caculated and printed, it'll be
> easier for people to read and compare the performance between
> rte_memcpy() and memcpy().
> Since the TSC values' units may diff a lot on different platforms,
> the total execution time of alligned/unaligned memcpy test are
> provided to allow comparation between platforms.
>
> Signed-off-by: Herbert Guan <herbert.guan@arm.com>
> ---
>  test/test/test_memcpy_perf.c | 34 +++++++++++++++++++++++-----------
>  1 file changed, 23 insertions(+), 11 deletions(-)
>
> diff --git a/test/test/test_memcpy_perf.c b/test/test/test_memcpy_perf.c
> index 91de856..30c29e1 100644
> --- a/test/test/test_memcpy_perf.c
> +++ b/test/test/test_memcpy_perf.c
> @@ -35,6 +35,7 @@
>  #include <stdio.h>
>  #include <string.h>
>  #include <stdlib.h>
> +#include <sys/time.h>
>
>  #include <rte_common.h>
>  #include <rte_cycles.h>
> @@ -241,8 +242,9 @@
>              memcpy(dst+dst_addrs[t], src+src_addrs[t], size);               \
>          total_time2 += rte_rdtsc() - start_time;                            \
>      }                                                                       \
> -    printf("%8.0f -",  (double)total_time /TEST_ITERATIONS);                \
> -    printf("%5.0f",  (double)total_time2 / TEST_ITERATIONS);                \
> +    printf("%3.0f -",  (double)total_time /TEST_ITERATIONS);                \
> +    printf("%3.0f",  (double)total_time2 / TEST_ITERATIONS);                \
> +    printf("(%6.2f%%) ", ((double)total_time - total_time2)*100/total_time2);  \
>  } while (0)
>
>  /* Run aligned memcpy tests for each cached/uncached permutation */
> @@ -324,6 +326,8 @@
>  perf_test(void)
>  {
>       int ret;
> +     struct timeval tv_begin, tv_end;
> +     time_t time_aligned, time_unaligned;
>
>       ret = init_buffers();
>       if (ret != 0)
> @@ -340,26 +344,34 @@
>       do_uncached_write(large_buf_write, 0, small_buf_read, 1, SMALL_BUFFER_SIZE);
>
>       printf("\n** rte_memcpy() - memcpy perf. tests (C = compile-time constant) **\n"
> -                "======= ============== ============== ============== ==============\n"
> -                "   Size Cache to cache   Cache to mem   Mem to cache     Mem to mem\n"
> -                "(bytes)        (ticks)        (ticks)        (ticks)        (ticks)\n"
> -                "------- -------------- -------------- -------------- --------------");
> +                "======= ================= ================= ================= =================\n"
> +                "   Size   Cache to cache     Cache to mem      Mem to cache        Mem to mem\n"
> +                "(bytes)          (ticks)          (ticks)           (ticks)           (ticks)\n"
> +                "------- ----------------- ----------------- ----------------- -----------------");
>
> -     printf("\n========================= %2dB aligned ============================",
> +     printf("\n================================= %2dB aligned =================================",
>               alignment_unit);
>       /* Do aligned tests where size is a variable */
> +     gettimeofday(&tv_begin, NULL);
>       perf_test_variable_aligned();
> -     printf("\n------- -------------- -------------- -------------- --------------");
> +     gettimeofday(&tv_end, NULL);
> +     time_aligned = tv_end.tv_sec - tv_begin.tv_sec;
> +     printf("\n------- ----------------- ----------------- ----------------- -----------------");
>       /* Do aligned tests where size is a compile-time constant */
>       perf_test_constant_aligned();
> -     printf("\n=========================== Unaligned =============================");
> +     printf("\n================================== Unaligned ==================================");
>       /* Do unaligned tests where size is a variable */
> +     gettimeofday(&tv_begin, NULL);
>       perf_test_variable_unaligned();
> -     printf("\n------- -------------- -------------- -------------- --------------");
> +     gettimeofday(&tv_end, NULL);
> +     time_unaligned = tv_end.tv_sec - tv_begin.tv_sec;
> +     printf("\n------- ----------------- ----------------- ----------------- -----------------");
>       /* Do unaligned tests where size is a compile-time constant */
>       perf_test_constant_unaligned();
> -     printf("\n======= ============== ============== ============== ==============\n\n");
> +     printf("\n======= ================= ================= ================= =================\n\n");
>
> +     printf("Test Execution Time (Aligned)   = %ld seconds\n", time_aligned);
> +     printf("Test Execution Time (Unaligned) = %ld seconds\n", time_unaligned);

Is there any reason why you didn't include the time cost in constant aligned/unaligned cases?

Thanks!
Jianbo

>       free_buffers();
>
>       return 0;
> --
> 1.8.3.1
>

--
IMPORTANT NOTICE: The contents of this email and any attachments are confidential and may also be privileged. If you are not the intended recipient, please notify the sender immediately and do not disclose the contents to any other person, use it for any purpose, or store or copy the information in any medium. Thank you.