public inbox for glibc-cvs@sourceware.org
help / color / mirror / Atom feed
* [glibc] stdlib: Tuned down tst-arc4random-thread internal parameters
@ 2022-07-29 12:24 Adhemerval Zanella
  0 siblings, 0 replies; only message in thread
From: Adhemerval Zanella @ 2022-07-29 12:24 UTC (permalink / raw)
  To: glibc-cvs

https://sourceware.org/git/gitweb.cgi?p=glibc.git;h=35363b53cef00c3b9886b2c1f45612869f304960

commit 35363b53cef00c3b9886b2c1f45612869f304960
Author: Adhemerval Zanella <adhemerval.zanella@linaro.org>
Date:   Wed Jul 27 08:59:15 2022 -0300

    stdlib: Tuned down tst-arc4random-thread internal parameters
    
    With new arc4random implementation, the internal parameters might
    require a lot of runtime and/or trigger some contention on older
    kernels (which might trigger spurious timeout failures).
    
    Also, since we are now testing getrandom entropy instead of an
    userspace RNG, there is no much need to extensive testing.
    
    With this change the tst-arc4random-thread goes from about 1m to
    5s on a Ryzen 9 with 5.15.0-41-generic.
    
    Checked on x86_64-linux-gnu.
    
    Reviewed-by: Szabolcs Nagy <szabolcs.nagy@arm.com>

Diff:
---
 stdlib/tst-arc4random-thread.c | 24 ++++++++++++++++++------
 1 file changed, 18 insertions(+), 6 deletions(-)

diff --git a/stdlib/tst-arc4random-thread.c b/stdlib/tst-arc4random-thread.c
index 3373d4d446..daba828fad 100644
--- a/stdlib/tst-arc4random-thread.c
+++ b/stdlib/tst-arc4random-thread.c
@@ -17,23 +17,23 @@
    <https://www.gnu.org/licenses/>.  */
 
 #include <array_length.h>
+#include <sched.h>
 #include <stdio.h>
-#include <string.h>
 #include <stdlib.h>
+#include <string.h>
 #include <support/check.h>
 #include <support/namespace.h>
 #include <support/support.h>
 #include <support/xthread.h>
 
 /* Number of arc4random_buf calls per thread.  */
-enum { count_per_thread = 5000 };
+enum { count_per_thread = 2048 };
 
 /* Number of threads computing randomness.  */
-enum { inner_threads = 5 };
+enum { inner_threads = 4 };
 
-/* Number of threads launching other threads.  Chosen as to not to
-   overload the system.  */
-enum { outer_threads = 7 };
+/* Number of threads launching other threads.  */
+static int outer_threads = 1;
 
 /* Number of launching rounds performed by the outer threads.  */
 enum { outer_rounds = 10 };
@@ -331,6 +331,18 @@ do_test_func (const char *fname, void (*func)(unsigned char *, size_t))
 static int
 do_test (void)
 {
+  /* Do not run more threads than the maximum of schedulable CPUs.  */
+  cpu_set_t cpuset;
+  if (sched_getaffinity (0, sizeof cpuset, &cpuset) == 0)
+    {
+      unsigned int ncpus = CPU_COUNT (&cpuset);
+      /* Limit the number to not overload the system.  */
+      outer_threads = (ncpus / 2) / inner_threads ?: 1;
+    }
+
+  printf ("info: outer_threads=%d inner_threads=%d\n", outer_threads,
+	  inner_threads);
+
   do_test_func ("arc4random", generate_arc4random);
   do_test_func ("arc4random_buf", generate_arc4random_buf);
   do_test_func ("arc4random_uniform", generate_arc4random_uniform);


^ permalink raw reply	[flat|nested] only message in thread

only message in thread, other threads:[~2022-07-29 12:24 UTC | newest]

Thread overview: (only message) (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2022-07-29 12:24 [glibc] stdlib: Tuned down tst-arc4random-thread internal parameters Adhemerval Zanella

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).