From: Noah Goldstein <goldstein.w.n@gmail.com>
To: libc-alpha@sourceware.org
Cc: goldstein.w.n@gmail.com, hjl.tools@gmail.com, carlos@systemhalted.org
Subject: [PATCH v2 7/7] Bench: Improve benchtests for memchr, strchr, strnlen, strrchr
Date: Tue, 18 Oct 2022 16:19:38 -0700 [thread overview]
Message-ID: <20221018231938.3621554-7-goldstein.w.n@gmail.com> (raw)
In-Reply-To: <20221018231938.3621554-1-goldstein.w.n@gmail.com>
1. Add more complete coverage in the medium size range.
2. In strnlen remove the `1 << i` which was UB (`i` could go beyond
32/64)
3. Add timer for total benchmark runtime (useful for deciding about
tradeoff between coverage and runtime).
---
benchtests/bench-memchr.c | 77 +++++++++++++++++++++++++-----------
benchtests/bench-rawmemchr.c | 30 ++++++++++++--
benchtests/bench-strchr.c | 35 +++++++++++-----
benchtests/bench-strnlen.c | 12 +++---
benchtests/bench-strrchr.c | 28 ++++++++++++-
5 files changed, 137 insertions(+), 45 deletions(-)
diff --git a/benchtests/bench-memchr.c b/benchtests/bench-memchr.c
index 0facda2fa0..2ec9dd86d0 100644
--- a/benchtests/bench-memchr.c
+++ b/benchtests/bench-memchr.c
@@ -126,7 +126,7 @@ do_test (json_ctx_t *json_ctx, size_t align, size_t pos, size_t len,
int
test_main (void)
{
- size_t i;
+ size_t i, j, al, al_max;
int repeats;
json_ctx_t json_ctx;
test_init ();
@@ -147,35 +147,46 @@ test_main (void)
json_array_begin (&json_ctx, "results");
+ al_max = 0;
+#ifdef USE_AS_MEMRCHR
+ al_max = getpagesize () / 2;
+#endif
+
for (repeats = 0; repeats < 2; ++repeats)
{
- for (i = 1; i < 8; ++i)
+ for (al = 0; al <= al_max; al += getpagesize () / 2)
{
- do_test (&json_ctx, 0, 16 << i, 2048, 23, repeats);
- do_test (&json_ctx, i, 64, 256, 23, repeats);
- do_test (&json_ctx, 0, 16 << i, 2048, 0, repeats);
- do_test (&json_ctx, i, 64, 256, 0, repeats);
-
- do_test (&json_ctx, getpagesize () - 15, 64, 256, 0, repeats);
+ for (i = 1; i < 8; ++i)
+ {
+ do_test (&json_ctx, al, 16 << i, 2048, 23, repeats);
+ do_test (&json_ctx, al + i, 64, 256, 23, repeats);
+ do_test (&json_ctx, al, 16 << i, 2048, 0, repeats);
+ do_test (&json_ctx, al + i, 64, 256, 0, repeats);
+
+ do_test (&json_ctx, al + getpagesize () - 15, 64, 256, 0,
+ repeats);
#ifdef USE_AS_MEMRCHR
- /* Also test the position close to the beginning for memrchr. */
- do_test (&json_ctx, 0, i, 256, 23, repeats);
- do_test (&json_ctx, 0, i, 256, 0, repeats);
- do_test (&json_ctx, i, i, 256, 23, repeats);
- do_test (&json_ctx, i, i, 256, 0, repeats);
+ /* Also test the position close to the beginning for memrchr. */
+ do_test (&json_ctx, al, i, 256, 23, repeats);
+ do_test (&json_ctx, al, i, 256, 0, repeats);
+ do_test (&json_ctx, al + i, i, 256, 23, repeats);
+ do_test (&json_ctx, al + i, i, 256, 0, repeats);
#endif
+ }
+ for (i = 1; i < 8; ++i)
+ {
+ do_test (&json_ctx, al + i, i << 5, 192, 23, repeats);
+ do_test (&json_ctx, al + i, i << 5, 192, 0, repeats);
+ do_test (&json_ctx, al + i, i << 5, 256, 23, repeats);
+ do_test (&json_ctx, al + i, i << 5, 256, 0, repeats);
+ do_test (&json_ctx, al + i, i << 5, 512, 23, repeats);
+ do_test (&json_ctx, al + i, i << 5, 512, 0, repeats);
+
+ do_test (&json_ctx, al + getpagesize () - 15, i << 5, 256, 23,
+ repeats);
+ }
}
- for (i = 1; i < 8; ++i)
- {
- do_test (&json_ctx, i, i << 5, 192, 23, repeats);
- do_test (&json_ctx, i, i << 5, 192, 0, repeats);
- do_test (&json_ctx, i, i << 5, 256, 23, repeats);
- do_test (&json_ctx, i, i << 5, 256, 0, repeats);
- do_test (&json_ctx, i, i << 5, 512, 23, repeats);
- do_test (&json_ctx, i, i << 5, 512, 0, repeats);
-
- do_test (&json_ctx, getpagesize () - 15, i << 5, 256, 23, repeats);
- }
+
for (i = 1; i < 32; ++i)
{
do_test (&json_ctx, 0, i, i + 1, 23, repeats);
@@ -207,6 +218,24 @@ test_main (void)
do_test (&json_ctx, 0, 2, i + 1, 0, repeats);
#endif
}
+ for (al = 0; al <= al_max; al += getpagesize () / 2)
+ {
+ for (i = (16 / sizeof (CHAR)); i <= (8192 / sizeof (CHAR)); i += i)
+ {
+ for (j = 0; j <= (384 / sizeof (CHAR));
+ j += (32 / sizeof (CHAR)))
+ {
+ do_test (&json_ctx, al, i + j, i, 23, repeats);
+ do_test (&json_ctx, al, i, i + j, 23, repeats);
+ if (j < i)
+ {
+ do_test (&json_ctx, al, i - j, i, 23, repeats);
+ do_test (&json_ctx, al, i, i - j, 23, repeats);
+ }
+ }
+ }
+ }
+
#ifndef USE_AS_MEMRCHR
break;
#endif
diff --git a/benchtests/bench-rawmemchr.c b/benchtests/bench-rawmemchr.c
index b1803afc14..dab77f3858 100644
--- a/benchtests/bench-rawmemchr.c
+++ b/benchtests/bench-rawmemchr.c
@@ -70,7 +70,7 @@ do_test (json_ctx_t *json_ctx, size_t align, size_t pos, size_t len, int seek_ch
size_t i;
char *result;
- align &= 7;
+ align &= getpagesize () - 1;
if (align + len >= page_size)
return;
@@ -106,7 +106,6 @@ test_main (void)
{
json_ctx_t json_ctx;
size_t i;
-
test_init ();
json_init (&json_ctx, 0, stdout);
@@ -120,7 +119,7 @@ test_main (void)
json_array_begin (&json_ctx, "ifuncs");
FOR_EACH_IMPL (impl, 0)
- json_element_string (&json_ctx, impl->name);
+ json_element_string (&json_ctx, impl->name);
json_array_end (&json_ctx);
json_array_begin (&json_ctx, "results");
@@ -137,6 +136,31 @@ test_main (void)
do_test (&json_ctx, 0, i, i + 1, 23);
do_test (&json_ctx, 0, i, i + 1, 0);
}
+ for (; i < 256; i += 32)
+ {
+ do_test (&json_ctx, 0, i, i + 1, 23);
+ do_test (&json_ctx, 0, i - 1, i, 23);
+ }
+ for (; i < 512; i += 64)
+ {
+ do_test (&json_ctx, 0, i, i + 1, 23);
+ do_test (&json_ctx, 0, i - 1, i, 23);
+ }
+ for (; i < 1024; i += 128)
+ {
+ do_test (&json_ctx, 0, i, i + 1, 23);
+ do_test (&json_ctx, 0, i - 1, i, 23);
+ }
+ for (; i < 2048; i += 256)
+ {
+ do_test (&json_ctx, 0, i, i + 1, 23);
+ do_test (&json_ctx, 0, i - 1, i, 23);
+ }
+ for (; i < 4096; i += 512)
+ {
+ do_test (&json_ctx, 0, i, i + 1, 23);
+ do_test (&json_ctx, 0, i - 1, i, 23);
+ }
json_array_end (&json_ctx);
json_attr_object_end (&json_ctx);
diff --git a/benchtests/bench-strchr.c b/benchtests/bench-strchr.c
index 54640bde7e..aeb882d442 100644
--- a/benchtests/bench-strchr.c
+++ b/benchtests/bench-strchr.c
@@ -287,8 +287,8 @@ int
test_main (void)
{
json_ctx_t json_ctx;
- size_t i;
+ size_t i, j;
test_init ();
json_init (&json_ctx, 0, stdout);
@@ -367,15 +367,30 @@ test_main (void)
do_test (&json_ctx, 0, i, i + 1, 0, BIG_CHAR);
}
- DO_RAND_TEST(&json_ctx, 0, 15, 16, 0.0);
- DO_RAND_TEST(&json_ctx, 0, 15, 16, 0.1);
- DO_RAND_TEST(&json_ctx, 0, 15, 16, 0.25);
- DO_RAND_TEST(&json_ctx, 0, 15, 16, 0.33);
- DO_RAND_TEST(&json_ctx, 0, 15, 16, 0.5);
- DO_RAND_TEST(&json_ctx, 0, 15, 16, 0.66);
- DO_RAND_TEST(&json_ctx, 0, 15, 16, 0.75);
- DO_RAND_TEST(&json_ctx, 0, 15, 16, 0.9);
- DO_RAND_TEST(&json_ctx, 0, 15, 16, 1.0);
+ for (i = 16 / sizeof (CHAR); i <= 8192 / sizeof (CHAR); i += i)
+ {
+ for (j = 32 / sizeof (CHAR); j <= 320 / sizeof (CHAR);
+ j += 32 / sizeof (CHAR))
+ {
+ do_test (&json_ctx, 0, i, i + j, 0, MIDDLE_CHAR);
+ do_test (&json_ctx, 0, i + j, i, 0, MIDDLE_CHAR);
+ if (i > j)
+ {
+ do_test (&json_ctx, 0, i, i - j, 0, MIDDLE_CHAR);
+ do_test (&json_ctx, 0, i - j, i, 0, MIDDLE_CHAR);
+ }
+ }
+ }
+
+ DO_RAND_TEST (&json_ctx, 0, 15, 16, 0.0);
+ DO_RAND_TEST (&json_ctx, 0, 15, 16, 0.1);
+ DO_RAND_TEST (&json_ctx, 0, 15, 16, 0.25);
+ DO_RAND_TEST (&json_ctx, 0, 15, 16, 0.33);
+ DO_RAND_TEST (&json_ctx, 0, 15, 16, 0.5);
+ DO_RAND_TEST (&json_ctx, 0, 15, 16, 0.66);
+ DO_RAND_TEST (&json_ctx, 0, 15, 16, 0.75);
+ DO_RAND_TEST (&json_ctx, 0, 15, 16, 0.9);
+ DO_RAND_TEST (&json_ctx, 0, 15, 16, 1.0);
json_array_end (&json_ctx);
json_attr_object_end (&json_ctx);
diff --git a/benchtests/bench-strnlen.c b/benchtests/bench-strnlen.c
index 13b46b3f57..82c02eb6ed 100644
--- a/benchtests/bench-strnlen.c
+++ b/benchtests/bench-strnlen.c
@@ -195,19 +195,19 @@ test_main (void)
{
for (j = 0; j <= (704 / sizeof (CHAR)); j += (32 / sizeof (CHAR)))
{
- do_test (&json_ctx, 0, 1 << i, (i + j), BIG_CHAR);
do_test (&json_ctx, 0, i + j, i, BIG_CHAR);
-
- do_test (&json_ctx, 64, 1 << i, (i + j), BIG_CHAR);
do_test (&json_ctx, 64, i + j, i, BIG_CHAR);
+ do_test (&json_ctx, 0, i, i + j, BIG_CHAR);
+ do_test (&json_ctx, 64, i, i + j, BIG_CHAR);
+
if (j < i)
{
- do_test (&json_ctx, 0, 1 << i, i - j, BIG_CHAR);
do_test (&json_ctx, 0, i - j, i, BIG_CHAR);
-
- do_test (&json_ctx, 64, 1 << i, i - j, BIG_CHAR);
do_test (&json_ctx, 64, i - j, i, BIG_CHAR);
+
+ do_test (&json_ctx, 0, i, i - j, BIG_CHAR);
+ do_test (&json_ctx, 64, i, i - j, BIG_CHAR);
}
}
}
diff --git a/benchtests/bench-strrchr.c b/benchtests/bench-strrchr.c
index 7cd2a15484..3fcf3f281d 100644
--- a/benchtests/bench-strrchr.c
+++ b/benchtests/bench-strrchr.c
@@ -151,7 +151,7 @@ int
test_main (void)
{
json_ctx_t json_ctx;
- size_t i, j;
+ size_t i, j, k;
int seek;
test_init ();
@@ -173,7 +173,7 @@ test_main (void)
for (seek = 0; seek <= 23; seek += 23)
{
- for (j = 1; j < 32; j += j)
+ for (j = 1; j <= 256; j = (j * 4))
{
for (i = 1; i < 9; ++i)
{
@@ -197,6 +197,30 @@ test_main (void)
do_test (&json_ctx, getpagesize () - i / 2 - 1, i, i + 1, seek,
SMALL_CHAR, j);
}
+
+ for (i = (16 / sizeof (CHAR)); i <= (288 / sizeof (CHAR)); i += 32)
+ {
+ do_test (&json_ctx, 0, i - 16, i, seek, SMALL_CHAR, j);
+ do_test (&json_ctx, 0, i, i + 16, seek, SMALL_CHAR, j);
+ }
+
+ for (i = (16 / sizeof (CHAR)); i <= (2048 / sizeof (CHAR)); i += i)
+ {
+ for (k = 0; k <= (288 / sizeof (CHAR));
+ k += (48 / sizeof (CHAR)))
+ {
+ do_test (&json_ctx, 0, k, i, seek, SMALL_CHAR, j);
+ do_test (&json_ctx, 0, i, i + k, seek, SMALL_CHAR, j);
+
+ if (k < i)
+ {
+ do_test (&json_ctx, 0, i - k, i, seek, SMALL_CHAR, j);
+ do_test (&json_ctx, 0, k, i - k, seek, SMALL_CHAR, j);
+ do_test (&json_ctx, 0, i, i - k, seek, SMALL_CHAR, j);
+ }
+ }
+ }
+
if (seek == 0)
{
break;
--
2.34.1
next prev parent reply other threads:[~2022-10-18 23:19 UTC|newest]
Thread overview: 41+ messages / expand[flat|nested] mbox.gz Atom feed top
2022-10-18 2:48 [PATCH v1 1/7] x86: Optimize memchr-evex.S and implement with VMM headers Noah Goldstein
2022-10-18 2:48 ` [PATCH v1 2/7] x86: Shrink / minorly optimize strchr-evex " Noah Goldstein
2022-10-18 2:51 ` Noah Goldstein
2022-10-18 2:48 ` [PATCH v1 3/7] x86: Optimize strnlen-evex.S " Noah Goldstein
2022-10-18 2:51 ` Noah Goldstein
2022-10-18 2:48 ` [PATCH v1 4/7] x86: Optimize memrchr-evex.S Noah Goldstein
2022-10-18 2:51 ` Noah Goldstein
2022-10-18 2:48 ` [PATCH v1 5/7] x86: Optimize strrchr-evex.S and implement with VMM headers Noah Goldstein
2022-10-18 2:52 ` Noah Goldstein
2022-10-18 2:49 ` [PATCH v1 6/7] x86: Add support for VEC_SIZE == 64 in strcmp-evex.S impl Noah Goldstein
2022-10-20 2:15 ` [PATCH v4] " Noah Goldstein
2022-10-20 3:46 ` H.J. Lu
2022-10-18 2:49 ` [PATCH v1 7/7] Bench: Improve benchtests for memchr, strchr, strnlen, strrchr Noah Goldstein
2022-10-18 21:00 ` H.J. Lu
2022-10-18 21:05 ` Noah Goldstein
2022-10-18 21:53 ` H.J. Lu
2022-10-18 22:58 ` Noah Goldstein
2022-10-18 2:50 ` [PATCH v1 1/7] x86: Optimize memchr-evex.S and implement with VMM headers Noah Goldstein
2022-10-18 23:19 ` [PATCH v2 " Noah Goldstein
2022-10-18 23:19 ` [PATCH v2 2/7] x86: Shrink / minorly optimize strchr-evex " Noah Goldstein
2022-10-18 23:19 ` [PATCH v2 3/7] x86: Optimize strnlen-evex.S " Noah Goldstein
2022-10-18 23:19 ` [PATCH v2 4/7] x86: Optimize memrchr-evex.S Noah Goldstein
2022-10-18 23:19 ` [PATCH v2 5/7] x86: Optimize strrchr-evex.S and implement with VMM headers Noah Goldstein
2022-10-18 23:19 ` [PATCH v2 6/7] x86: Add support for VEC_SIZE == 64 in strcmp-evex.S impl Noah Goldstein
2022-10-18 23:19 ` Noah Goldstein [this message]
2022-10-19 0:01 ` [PATCH v2 7/7] Bench: Improve benchtests for memchr, strchr, strnlen, strrchr H.J. Lu
2022-10-19 0:44 ` Noah Goldstein
2022-10-19 0:44 ` [PATCH v3 1/7] x86: Optimize memchr-evex.S and implement with VMM headers Noah Goldstein
2022-10-19 0:44 ` [PATCH v3 2/7] x86: Shrink / minorly optimize strchr-evex " Noah Goldstein
2022-10-19 16:53 ` H.J. Lu
2022-10-19 0:44 ` [PATCH v3 3/7] x86: Optimize strnlen-evex.S " Noah Goldstein
2022-10-19 16:57 ` H.J. Lu
2022-10-19 0:44 ` [PATCH v3 4/7] x86: Optimize memrchr-evex.S Noah Goldstein
2022-10-19 16:58 ` H.J. Lu
2022-10-19 0:44 ` [PATCH v3 5/7] x86: Optimize strrchr-evex.S and implement with VMM headers Noah Goldstein
2022-10-19 16:58 ` H.J. Lu
2022-10-19 0:44 ` [PATCH v3 6/7] x86: Add support for VEC_SIZE == 64 in strcmp-evex.S impl Noah Goldstein
2022-10-19 16:59 ` H.J. Lu
2022-10-19 0:44 ` [PATCH v3 7/7] Bench: Improve benchtests for memchr, strchr, strnlen, strrchr Noah Goldstein
2022-10-19 17:00 ` H.J. Lu
2022-10-19 16:52 ` [PATCH v3 1/7] x86: Optimize memchr-evex.S and implement with VMM headers H.J. Lu
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=20221018231938.3621554-7-goldstein.w.n@gmail.com \
--to=goldstein.w.n@gmail.com \
--cc=carlos@systemhalted.org \
--cc=hjl.tools@gmail.com \
--cc=libc-alpha@sourceware.org \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
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).