linux-perf-users.vger.kernel.org archive mirror
 help / color / mirror / Atom feed
From: Ian Rogers <irogers@google.com>
To: Peter Zijlstra <peterz@infradead.org>,
	Ingo Molnar <mingo@redhat.com>,
	Arnaldo Carvalho de Melo <acme@kernel.org>,
	Mark Rutland <mark.rutland@arm.com>,
	Alexander Shishkin <alexander.shishkin@linux.intel.com>,
	Jiri Olsa <jolsa@kernel.org>, Namhyung Kim <namhyung@kernel.org>,
	Kajol Jain <kjain@linux.ibm.com>, Andi Kleen <ak@linux.intel.com>,
	Adrian Hunter <adrian.hunter@intel.com>,
	Anshuman Khandual <anshuman.khandual@arm.com>,
	linux-kernel@vger.kernel.org, linux-perf-users@vger.kernel.org,
	Rob Herring <robh@kernel.org>
Cc: Stephane Eranian <eranian@google.com>, Ian Rogers <irogers@google.com>
Subject: [PATCH 2/4] perf: Align user space counter reading with code
Date: Wed,  8 Jun 2022 15:43:51 -0700	[thread overview]
Message-ID: <20220608224353.1176079-3-irogers@google.com> (raw)
In-Reply-To: <20220608224353.1176079-1-irogers@google.com>

Align the user space counter reading documentation with the code in
perf_mmap__read_self. Previously the documentation was based on the perf
rdpmc test, but now general purpose code is provided by libperf.

Signed-off-by: Ian Rogers <irogers@google.com>
---
 include/uapi/linux/perf_event.h       | 32 ++++++++++++++++-----------
 tools/include/uapi/linux/perf_event.h | 32 ++++++++++++++++-----------
 2 files changed, 38 insertions(+), 26 deletions(-)

diff --git a/include/uapi/linux/perf_event.h b/include/uapi/linux/perf_event.h
index d37629dbad72..3b84e0ad0723 100644
--- a/include/uapi/linux/perf_event.h
+++ b/include/uapi/linux/perf_event.h
@@ -538,9 +538,13 @@ struct perf_event_mmap_page {
 	 *
 	 *     if (pc->cap_usr_time && enabled != running) {
 	 *       cyc = rdtsc();
-	 *       time_offset = pc->time_offset;
 	 *       time_mult   = pc->time_mult;
 	 *       time_shift  = pc->time_shift;
+	 *       time_offset = pc->time_offset;
+	 *       if (pc->cap_user_time_short) {
+	 *         time_cycles = pc->time_cycles;
+	 *         time_mask = pc->time_mask;
+	 *       }
 	 *     }
 	 *
 	 *     index = pc->index;
@@ -548,6 +552,9 @@ struct perf_event_mmap_page {
 	 *     if (pc->cap_user_rdpmc && index) {
 	 *       width = pc->pmc_width;
 	 *       pmc = rdpmc(index - 1);
+	 *       pmc <<= 64 - width;
+	 *       pmc >>= 64 - width;
+	 *       count += pmc;
 	 *     }
 	 *
 	 *     barrier();
@@ -590,25 +597,24 @@ struct perf_event_mmap_page {
 	 * If cap_usr_time the below fields can be used to compute the time
 	 * delta since time_enabled (in ns) using rdtsc or similar.
 	 *
-	 *   u64 quot, rem;
-	 *   u64 delta;
-	 *
-	 *   quot = (cyc >> time_shift);
-	 *   rem = cyc & (((u64)1 << time_shift) - 1);
-	 *   delta = time_offset + quot * time_mult +
-	 *              ((rem * time_mult) >> time_shift);
+	 *   cyc = time_cycles + ((cyc - time_cycles) & time_mask);
+	 *   delta = time_offset + mul_u64_u32_shr(cyc, time_mult, time_shift);
 	 *
 	 * Where time_offset,time_mult,time_shift and cyc are read in the
 	 * seqcount loop described above. This delta can then be added to
-	 * enabled and possible running (if index), improving the scaling:
+	 * enabled and possible running (if index) to improve the scaling. Due
+	 * to event multiplexing, running maybe zero and so care is needed to
+	 * avoid division by zero.
 	 *
 	 *   enabled += delta;
-	 *   if (index)
+	 *   if (idx)
 	 *     running += delta;
 	 *
-	 *   quot = count / running;
-	 *   rem  = count % running;
-	 *   count = quot * enabled + (rem * enabled) / running;
+	 *   if (running != 0) {
+	 *     quot = count / running;
+	 *     rem  = count % running;
+	 *     count = quot * enabled + (rem * enabled) / running;
+	 *   }
 	 */
 	__u16	time_shift;
 	__u32	time_mult;
diff --git a/tools/include/uapi/linux/perf_event.h b/tools/include/uapi/linux/perf_event.h
index d37629dbad72..3b84e0ad0723 100644
--- a/tools/include/uapi/linux/perf_event.h
+++ b/tools/include/uapi/linux/perf_event.h
@@ -538,9 +538,13 @@ struct perf_event_mmap_page {
 	 *
 	 *     if (pc->cap_usr_time && enabled != running) {
 	 *       cyc = rdtsc();
-	 *       time_offset = pc->time_offset;
 	 *       time_mult   = pc->time_mult;
 	 *       time_shift  = pc->time_shift;
+	 *       time_offset = pc->time_offset;
+	 *       if (pc->cap_user_time_short) {
+	 *         time_cycles = pc->time_cycles;
+	 *         time_mask = pc->time_mask;
+	 *       }
 	 *     }
 	 *
 	 *     index = pc->index;
@@ -548,6 +552,9 @@ struct perf_event_mmap_page {
 	 *     if (pc->cap_user_rdpmc && index) {
 	 *       width = pc->pmc_width;
 	 *       pmc = rdpmc(index - 1);
+	 *       pmc <<= 64 - width;
+	 *       pmc >>= 64 - width;
+	 *       count += pmc;
 	 *     }
 	 *
 	 *     barrier();
@@ -590,25 +597,24 @@ struct perf_event_mmap_page {
 	 * If cap_usr_time the below fields can be used to compute the time
 	 * delta since time_enabled (in ns) using rdtsc or similar.
 	 *
-	 *   u64 quot, rem;
-	 *   u64 delta;
-	 *
-	 *   quot = (cyc >> time_shift);
-	 *   rem = cyc & (((u64)1 << time_shift) - 1);
-	 *   delta = time_offset + quot * time_mult +
-	 *              ((rem * time_mult) >> time_shift);
+	 *   cyc = time_cycles + ((cyc - time_cycles) & time_mask);
+	 *   delta = time_offset + mul_u64_u32_shr(cyc, time_mult, time_shift);
 	 *
 	 * Where time_offset,time_mult,time_shift and cyc are read in the
 	 * seqcount loop described above. This delta can then be added to
-	 * enabled and possible running (if index), improving the scaling:
+	 * enabled and possible running (if index) to improve the scaling. Due
+	 * to event multiplexing, running maybe zero and so care is needed to
+	 * avoid division by zero.
 	 *
 	 *   enabled += delta;
-	 *   if (index)
+	 *   if (idx)
 	 *     running += delta;
 	 *
-	 *   quot = count / running;
-	 *   rem  = count % running;
-	 *   count = quot * enabled + (rem * enabled) / running;
+	 *   if (running != 0) {
+	 *     quot = count / running;
+	 *     rem  = count % running;
+	 *     count = quot * enabled + (rem * enabled) / running;
+	 *   }
 	 */
 	__u16	time_shift;
 	__u32	time_mult;
-- 
2.36.1.255.ge46751e96f-goog


  parent reply	other threads:[~2022-06-08 22:44 UTC|newest]

Thread overview: 10+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2022-06-08 22:43 [PATCH 0/4] Tidy user rdpmc documentation and testing Ian Rogers
2022-06-08 22:43 ` [PATCH 1/4] libperf evsel: Open shouldn't leak fd on failure Ian Rogers
2022-06-09 16:47   ` Arnaldo Carvalho de Melo
2022-06-08 22:43 ` Ian Rogers [this message]
2022-06-08 23:24   ` [PATCH 2/4] perf: Align user space counter reading with code Rob Herring
2022-06-09  0:07     ` Ian Rogers
2022-06-08 22:43 ` [PATCH 3/4] perf test: Remove x86 rdpmc test Ian Rogers
2022-06-08 23:29   ` Rob Herring
2022-06-08 23:30     ` Rob Herring
2022-06-08 22:43 ` [PATCH 4/4] perf test: Add user space counter reading tests Ian Rogers

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=20220608224353.1176079-3-irogers@google.com \
    --to=irogers@google.com \
    --cc=acme@kernel.org \
    --cc=adrian.hunter@intel.com \
    --cc=ak@linux.intel.com \
    --cc=alexander.shishkin@linux.intel.com \
    --cc=anshuman.khandual@arm.com \
    --cc=eranian@google.com \
    --cc=jolsa@kernel.org \
    --cc=kjain@linux.ibm.com \
    --cc=linux-kernel@vger.kernel.org \
    --cc=linux-perf-users@vger.kernel.org \
    --cc=mark.rutland@arm.com \
    --cc=mingo@redhat.com \
    --cc=namhyung@kernel.org \
    --cc=peterz@infradead.org \
    --cc=robh@kernel.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 NNTP newsgroup(s).