From: Leo Yan <leo.yan@linaro.org>
To: Arnaldo Carvalho de Melo <acme@kernel.org>,
Peter Zijlstra <peterz@infradead.org>,
Ingo Molnar <mingo@redhat.com>,
Mark Rutland <mark.rutland@arm.com>,
Alexander Shishkin <alexander.shishkin@linux.intel.com>,
Jiri Olsa <jolsa@redhat.com>, Namhyung Kim <namhyung@kernel.org>,
Kan Liang <kan.liang@linux.intel.com>,
Andi Kleen <ak@linux.intel.com>, Ian Rogers <irogers@google.com>,
Joe Mario <jmario@redhat.com>, David Ahern <dsahern@gmail.com>,
Don Zickus <dzickus@redhat.com>, Al Grant <Al.Grant@arm.com>,
James Clark <james.clark@arm.com>,
linux-kernel@vger.kernel.org
Cc: Leo Yan <leo.yan@linaro.org>
Subject: [PATCH v1 2/8] perf c2c: Add dimensions for total LLC hit
Date: Thu, 15 Oct 2020 15:50:35 +0100 [thread overview]
Message-ID: <20201015145041.10953-3-leo.yan@linaro.org> (raw)
In-Reply-To: <20201015145041.10953-1-leo.yan@linaro.org>
Since Arm64 SPE trace data doesn't support HITM, we still want to
explore "perf c2c" tool to analyze cache false sharing. If without HITM
tag, the tool cannot give out accurate result for cache false sharing,
a candidate solution is to sort the LLC hit and connect with the info of
multiple threads, e.g. if multiple threads hit the LLC on the same
cacheline for many times, this hints that it's likely to cause false
sharing issue. Though this solution is not perfect due to lacking HITM
tag, it's pragmatic for detecting false sharing.
To support the sorting with LLC hit, this patch adds dimensions for
total LLC hit and the associated percentage calculation.
Signed-off-by: Leo Yan <leo.yan@linaro.org>
---
tools/perf/builtin-c2c.c | 76 ++++++++++++++++++++++++++++++++++++++++
1 file changed, 76 insertions(+)
diff --git a/tools/perf/builtin-c2c.c b/tools/perf/builtin-c2c.c
index 9c2183957c50..5fb77fcd3c9c 100644
--- a/tools/perf/builtin-c2c.c
+++ b/tools/perf/builtin-c2c.c
@@ -651,6 +651,7 @@ STAT_FN(ld_l1hit)
STAT_FN(ld_l2hit)
STAT_FN(ld_llchit)
STAT_FN(rmt_hit)
+STAT_FN(tot_llchit)
static uint64_t total_records(struct c2c_stats *stats)
{
@@ -856,6 +857,62 @@ percent_hitm_cmp(struct perf_hpp_fmt *fmt __maybe_unused,
return per_left - per_right;
}
+static double percent_llchit(struct c2c_hist_entry *c2c_he)
+{
+ struct c2c_hists *hists;
+ struct c2c_stats *stats;
+ struct c2c_stats *total;
+ int tot = 0, st = 0;
+
+ hists = container_of(c2c_he->he.hists, struct c2c_hists, hists);
+ stats = &c2c_he->stats;
+ total = &hists->stats;
+
+ st = stats->tot_llchit;
+ tot = total->tot_llchit;
+
+ return tot ? (double) st * 100 / tot : 0;
+}
+
+static int
+percent_llchit_entry(struct perf_hpp_fmt *fmt, struct perf_hpp *hpp,
+ struct hist_entry *he)
+{
+ struct c2c_hist_entry *c2c_he;
+ int width = c2c_width(fmt, hpp, he->hists);
+ char buf[10];
+ double per;
+
+ c2c_he = container_of(he, struct c2c_hist_entry, he);
+ per = percent_llchit(c2c_he);
+ return scnprintf(hpp->buf, hpp->size, "%*s", width, PERC_STR(buf, per));
+}
+
+static int
+percent_llchit_color(struct perf_hpp_fmt *fmt, struct perf_hpp *hpp,
+ struct hist_entry *he)
+{
+ return percent_color(fmt, hpp, he, percent_llchit);
+}
+
+static int64_t
+percent_llchit_cmp(struct perf_hpp_fmt *fmt __maybe_unused,
+ struct hist_entry *left, struct hist_entry *right)
+{
+ struct c2c_hist_entry *c2c_left;
+ struct c2c_hist_entry *c2c_right;
+ double per_left;
+ double per_right;
+
+ c2c_left = container_of(left, struct c2c_hist_entry, he);
+ c2c_right = container_of(right, struct c2c_hist_entry, he);
+
+ per_left = percent_llchit(c2c_left);
+ per_right = percent_llchit(c2c_right);
+
+ return per_left - per_right;
+}
+
static struct c2c_stats *he_stats(struct hist_entry *he)
{
struct c2c_hist_entry *c2c_he;
@@ -1392,6 +1449,14 @@ static struct c2c_dimension dim_ld_l2hit = {
.width = 7,
};
+static struct c2c_dimension dim_tot_llchit = {
+ .header = HEADER_BOTH("LLC Hit", "Total"),
+ .name = "tot_llchit",
+ .cmp = tot_llchit_cmp,
+ .entry = tot_llchit_entry,
+ .width = 8,
+};
+
static struct c2c_dimension dim_ld_llchit = {
.header = HEADER_SPAN("- LLC Load Hit --", "LclHit", 1),
.name = "ld_lclhit",
@@ -1438,6 +1503,15 @@ static struct c2c_dimension dim_percent_hitm = {
.width = 7,
};
+static struct c2c_dimension dim_percent_llchit = {
+ .header = HEADER_BOTH("LLC Hit", "Pct"),
+ .name = "percent_llchit",
+ .cmp = percent_llchit_cmp,
+ .entry = percent_llchit_entry,
+ .color = percent_llchit_color,
+ .width = 7,
+};
+
static struct c2c_dimension dim_percent_rmt_hitm = {
.header = HEADER_SPAN("----- HITM -----", "RmtHitm", 1),
.name = "percent_rmt_hitm",
@@ -1611,9 +1685,11 @@ static struct c2c_dimension *dimensions[] = {
&dim_ld_l2hit,
&dim_ld_llchit,
&dim_ld_rmthit,
+ &dim_tot_llchit,
&dim_tot_recs,
&dim_tot_loads,
&dim_percent_hitm,
+ &dim_percent_llchit,
&dim_percent_rmt_hitm,
&dim_percent_lcl_hitm,
&dim_percent_stores_l1hit,
--
2.17.1
next prev parent reply other threads:[~2020-10-15 14:51 UTC|newest]
Thread overview: 19+ messages / expand[flat|nested] mbox.gz Atom feed top
2020-10-15 14:50 [PATCH v1 0/8] perf c2c: Sort cacheline with LLC load Leo Yan
2020-10-15 14:50 ` [PATCH v1 1/8] perf mem: Add structure field c2c_stats::tot_llchit Leo Yan
2020-10-15 14:50 ` Leo Yan [this message]
2020-10-15 14:50 ` [PATCH v1 3/8] perf c2c: Add dimensions for LLC load hit Leo Yan
2020-10-15 14:50 ` [PATCH v1 4/8] perf c2c: Change to general naming for macros Leo Yan
2020-10-15 14:50 ` [PATCH v1 5/8] perf c2c: Rename for shared cache line stats Leo Yan
2020-10-15 14:50 ` [PATCH v1 6/8] perf c2c: Refactor hist entry validation Leo Yan
2020-10-15 14:50 ` [PATCH v1 7/8] perf c2c: Add option '-d llc' for sorting with LLC load Leo Yan
2020-10-20 7:25 ` Jiri Olsa
2020-10-20 8:08 ` Leo Yan
2020-10-22 8:43 ` Jiri Olsa
2020-10-20 7:26 ` Jiri Olsa
2020-10-20 8:14 ` Leo Yan
2020-10-15 14:50 ` [PATCH v1 8/8] perf c2c: Update documentation for display option 'llc' Leo Yan
2020-10-20 7:26 ` Jiri Olsa
2020-10-15 15:05 ` [PATCH v1 0/8] perf c2c: Sort cacheline with LLC load Arnaldo Carvalho de Melo
2020-10-15 15:14 ` Leo Yan
2020-10-20 8:13 ` Namhyung Kim
2020-10-20 8:18 ` Leo Yan
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=20201015145041.10953-3-leo.yan@linaro.org \
--to=leo.yan@linaro.org \
--cc=Al.Grant@arm.com \
--cc=acme@kernel.org \
--cc=ak@linux.intel.com \
--cc=alexander.shishkin@linux.intel.com \
--cc=dsahern@gmail.com \
--cc=dzickus@redhat.com \
--cc=irogers@google.com \
--cc=james.clark@arm.com \
--cc=jmario@redhat.com \
--cc=jolsa@redhat.com \
--cc=kan.liang@linux.intel.com \
--cc=linux-kernel@vger.kernel.org \
--cc=mark.rutland@arm.com \
--cc=mingo@redhat.com \
--cc=namhyung@kernel.org \
--cc=peterz@infradead.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
all inboxes | Powered by JetHome®