public inbox for gdb-cvs@sourceware.org
help / color / mirror / Atom feed
* [binutils-gdb] gprofng docs: provide help for <rate> == <interval>
@ 2022-06-14 21:26 Vladimir Mezentsev
  0 siblings, 0 replies; only message in thread
From: Vladimir Mezentsev @ 2022-06-14 21:26 UTC (permalink / raw)
  To: bfd-cvs, gdb-cvs

https://sourceware.org/git/gitweb.cgi?p=binutils-gdb.git;h=a675e77e55759bad4501cbaaba36fa18dad2339d

commit a675e77e55759bad4501cbaaba36fa18dad2339d
Author: Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
Date:   Tue Jun 14 14:16:24 2022 -0700

    gprofng docs: provide help for <rate> == <interval>
    
    The help message from 'gprofng collect app -h', in
    the section on <rate> == <interval>, had a dangling
    reference to a non-existent manpage. Provide basic
    info, including reasons for caution.

Diff:
---
 gprofng/common/hwctable.c | 20 ++++++++++++++++++--
 1 file changed, 18 insertions(+), 2 deletions(-)

diff --git a/gprofng/common/hwctable.c b/gprofng/common/hwctable.c
index bc441e133a7..a1422b63c11 100644
--- a/gprofng/common/hwctable.c
+++ b/gprofng/common/hwctable.c
@@ -5091,8 +5091,24 @@ hwc_usage_internal (int forKernel, FILE *f_usage, const char *cmd, const char *d
       fprintf (f_usage, GTXT ("           `lo'     per-thread maximum rate of ~10 samples/second\n"));
       fprintf (f_usage, GTXT ("           `on'     per-thread maximum rate of ~100 samples/second\n"));
       fprintf (f_usage, GTXT ("           `hi'     per-thread maximum rate of ~1000 samples/second\n\n"));
-      fprintf (f_usage, GTXT ("        <rate> == <interval>\n"));
-      fprintf (f_usage, GTXT ("           event interval; see collect (1) for details\n\n"));
+      fprintf (f_usage, GTXT ("        <rate> == <interval>\n"
+			      "           Fixed event interval value to trigger a sample.\n"
+			      "           Smaller intervals imply more frequent samples.\n"
+			      "           Example: when counting cycles on a 2 GHz processor,\n"
+			      "           an interval of 2,000,003 implies ~1000 samples/sec\n"
+			      "\n"
+			      "           Use this feature with caution, because:\n"
+			      "             (1) Frequent sampling increases overhead and may disturb \n"
+			      "                 other applications on your system.\n"
+			      "             (2) Event counts vary dramatically depending on the event \n"
+			      "                 and depending on the application.\n"
+			      "             (3) A fixed event interval disables any other gprofng\n"
+			      "                 internal mechanisms that may limit event rates.\n"
+			      "\n"
+			      "           Guidelines:  Aim at <1000 events per second.  Start by \n"
+			      "           collecting with the 'hi' option; in the experiment overview,\n"
+			      "           notice how many events are recorded per second; divide by\n"
+			      "           1000, and use that as your starting point.\n\n"));
 
       fprintf (f_usage, GTXT ("        A comma ',' followed immediately by white space may be omitted.\n\n"));
     }


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

only message in thread, other threads:[~2022-06-14 21:26 UTC | newest]

Thread overview: (only message) (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2022-06-14 21:26 [binutils-gdb] gprofng docs: provide help for <rate> == <interval> Vladimir Mezentsev

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).