* [PATCH v1] benchtests: Improve reliability of memcmp benchmarks
@ 2021-09-18 21:39 Noah Goldstein
2021-09-23 15:48 ` Noah Goldstein
2021-10-08 19:40 ` H.J. Lu
0 siblings, 2 replies; 3+ messages in thread
From: Noah Goldstein @ 2021-09-18 21:39 UTC (permalink / raw)
To: libc-alpha
No bug. Remove reallocation of bufs between implementation tests. Move
initialization outside of foreach implementation test loop. Increase
iteration count.
Generally before this commit was seeing a great deal of variability
between runs. The goal of this commit is to make the results more
reliable.
Benchtests build and bench-memcmp succeeding.
---
benchtests/bench-memcmp.c | 21 ++++++++++-----------
1 file changed, 10 insertions(+), 11 deletions(-)
diff --git a/benchtests/bench-memcmp.c b/benchtests/bench-memcmp.c
index 744c7ec5ba..0d6a93bf29 100644
--- a/benchtests/bench-memcmp.c
+++ b/benchtests/bench-memcmp.c
@@ -63,7 +63,7 @@ static void
do_one_test (json_ctx_t *json_ctx, impl_t *impl, const CHAR *s1,
const CHAR *s2, size_t len, int exp_result)
{
- size_t i, iters = INNER_LOOP_ITERS8;
+ size_t i, iters = INNER_LOOP_ITERS_LARGE;
timing_t start, stop, cur;
TIMING_NOW (start);
@@ -103,20 +103,19 @@ do_test (json_ctx_t *json_ctx, size_t align1, size_t align2, size_t len,
json_attr_uint (json_ctx, "result", (double) exp_result);
json_array_begin (json_ctx, "timings");
- FOR_EACH_IMPL (impl, 0)
- {
- s1 = (CHAR *) (buf1 + align1);
- s2 = (CHAR *) (buf2 + align2);
+ s1 = (CHAR *)(buf1 + align1);
+ s2 = (CHAR *)(buf2 + align2);
- for (i = 0; i < len; i++)
- s1[i] = s2[i] = 1 + (23 << ((CHARBYTES - 1) * 8)) * i % MAX_CHAR;
+ for (i = 0; i < len; i++)
+ s1[i] = s2[i] = 1 + (23 << ((CHARBYTES - 1) * 8)) * i % MAX_CHAR;
- s1[len] = align1;
- s2[len] = align2;
- s2[len - 1] -= exp_result;
+ s1[len] = align1;
+ s2[len] = align2;
+ s2[len - 1] -= exp_result;
+ FOR_EACH_IMPL (impl, 0)
+ {
do_one_test (json_ctx, impl, s1, s2, len, exp_result);
- alloc_bufs ();
}
json_array_end (json_ctx);
--
2.25.1
^ permalink raw reply [flat|nested] 3+ messages in thread
* Re: [PATCH v1] benchtests: Improve reliability of memcmp benchmarks
2021-09-18 21:39 [PATCH v1] benchtests: Improve reliability of memcmp benchmarks Noah Goldstein
@ 2021-09-23 15:48 ` Noah Goldstein
2021-10-08 19:40 ` H.J. Lu
1 sibling, 0 replies; 3+ messages in thread
From: Noah Goldstein @ 2021-09-23 15:48 UTC (permalink / raw)
To: GNU C Library
On Sat, Sep 18, 2021 at 4:39 PM Noah Goldstein <goldstein.w.n@gmail.com>
wrote:
> No bug. Remove reallocation of bufs between implementation tests. Move
> initialization outside of foreach implementation test loop. Increase
> iteration count.
>
> Generally before this commit was seeing a great deal of variability
> between runs. The goal of this commit is to make the results more
> reliable.
>
> Benchtests build and bench-memcmp succeeding.
> ---
> benchtests/bench-memcmp.c | 21 ++++++++++-----------
> 1 file changed, 10 insertions(+), 11 deletions(-)
>
> diff --git a/benchtests/bench-memcmp.c b/benchtests/bench-memcmp.c
> index 744c7ec5ba..0d6a93bf29 100644
> --- a/benchtests/bench-memcmp.c
> +++ b/benchtests/bench-memcmp.c
> @@ -63,7 +63,7 @@ static void
> do_one_test (json_ctx_t *json_ctx, impl_t *impl, const CHAR *s1,
> const CHAR *s2, size_t len, int exp_result)
> {
> - size_t i, iters = INNER_LOOP_ITERS8;
> + size_t i, iters = INNER_LOOP_ITERS_LARGE;
> timing_t start, stop, cur;
>
> TIMING_NOW (start);
> @@ -103,20 +103,19 @@ do_test (json_ctx_t *json_ctx, size_t align1, size_t
> align2, size_t len,
> json_attr_uint (json_ctx, "result", (double) exp_result);
> json_array_begin (json_ctx, "timings");
>
> - FOR_EACH_IMPL (impl, 0)
> - {
> - s1 = (CHAR *) (buf1 + align1);
> - s2 = (CHAR *) (buf2 + align2);
> + s1 = (CHAR *)(buf1 + align1);
> + s2 = (CHAR *)(buf2 + align2);
>
> - for (i = 0; i < len; i++)
> - s1[i] = s2[i] = 1 + (23 << ((CHARBYTES - 1) * 8)) * i % MAX_CHAR;
> + for (i = 0; i < len; i++)
> + s1[i] = s2[i] = 1 + (23 << ((CHARBYTES - 1) * 8)) * i % MAX_CHAR;
>
> - s1[len] = align1;
> - s2[len] = align2;
> - s2[len - 1] -= exp_result;
> + s1[len] = align1;
> + s2[len] = align2;
> + s2[len - 1] -= exp_result;
>
> + FOR_EACH_IMPL (impl, 0)
> + {
> do_one_test (json_ctx, impl, s1, s2, len, exp_result);
> - alloc_bufs ();
> }
>
> json_array_end (json_ctx);
> --
> 2.25.1
>
>
Going to check this in unless there are any objections.
^ permalink raw reply [flat|nested] 3+ messages in thread
* Re: [PATCH v1] benchtests: Improve reliability of memcmp benchmarks
2021-09-18 21:39 [PATCH v1] benchtests: Improve reliability of memcmp benchmarks Noah Goldstein
2021-09-23 15:48 ` Noah Goldstein
@ 2021-10-08 19:40 ` H.J. Lu
1 sibling, 0 replies; 3+ messages in thread
From: H.J. Lu @ 2021-10-08 19:40 UTC (permalink / raw)
To: Noah Goldstein; +Cc: GNU C Library, Carlos O'Donell
On Sat, Sep 18, 2021 at 2:39 PM Noah Goldstein <goldstein.w.n@gmail.com> wrote:
>
> No bug. Remove reallocation of bufs between implementation tests. Move
> initialization outside of foreach implementation test loop. Increase
> iteration count.
>
> Generally before this commit was seeing a great deal of variability
> between runs. The goal of this commit is to make the results more
> reliable.
>
> Benchtests build and bench-memcmp succeeding.
> ---
> benchtests/bench-memcmp.c | 21 ++++++++++-----------
> 1 file changed, 10 insertions(+), 11 deletions(-)
>
> diff --git a/benchtests/bench-memcmp.c b/benchtests/bench-memcmp.c
> index 744c7ec5ba..0d6a93bf29 100644
> --- a/benchtests/bench-memcmp.c
> +++ b/benchtests/bench-memcmp.c
> @@ -63,7 +63,7 @@ static void
> do_one_test (json_ctx_t *json_ctx, impl_t *impl, const CHAR *s1,
> const CHAR *s2, size_t len, int exp_result)
> {
> - size_t i, iters = INNER_LOOP_ITERS8;
> + size_t i, iters = INNER_LOOP_ITERS_LARGE;
> timing_t start, stop, cur;
>
> TIMING_NOW (start);
> @@ -103,20 +103,19 @@ do_test (json_ctx_t *json_ctx, size_t align1, size_t align2, size_t len,
> json_attr_uint (json_ctx, "result", (double) exp_result);
> json_array_begin (json_ctx, "timings");
>
> - FOR_EACH_IMPL (impl, 0)
> - {
> - s1 = (CHAR *) (buf1 + align1);
> - s2 = (CHAR *) (buf2 + align2);
> + s1 = (CHAR *)(buf1 + align1);
> + s2 = (CHAR *)(buf2 + align2);
>
> - for (i = 0; i < len; i++)
> - s1[i] = s2[i] = 1 + (23 << ((CHARBYTES - 1) * 8)) * i % MAX_CHAR;
> + for (i = 0; i < len; i++)
> + s1[i] = s2[i] = 1 + (23 << ((CHARBYTES - 1) * 8)) * i % MAX_CHAR;
>
> - s1[len] = align1;
> - s2[len] = align2;
> - s2[len - 1] -= exp_result;
> + s1[len] = align1;
> + s2[len] = align2;
> + s2[len - 1] -= exp_result;
>
> + FOR_EACH_IMPL (impl, 0)
> + {
> do_one_test (json_ctx, impl, s1, s2, len, exp_result);
> - alloc_bufs ();
> }
>
> json_array_end (json_ctx);
> --
> 2.25.1
>
LGTM.
Reviewed-by: H.J. Lu <hjl.tools@gmail.com>
Thanks.
--
H.J.
^ permalink raw reply [flat|nested] 3+ messages in thread
end of thread, other threads:[~2021-10-08 19:40 UTC | newest]
Thread overview: 3+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2021-09-18 21:39 [PATCH v1] benchtests: Improve reliability of memcmp benchmarks Noah Goldstein
2021-09-23 15:48 ` Noah Goldstein
2021-10-08 19:40 ` H.J. Lu
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for read-only IMAP folder(s) and NNTP newsgroup(s).