From: Leo Yan <leo.yan@linaro.org>
To: Arnaldo Carvalho de Melo <acme@kernel.org>,
Jiri Olsa <jolsa@redhat.com>,
Peter Zijlstra <peterz@infradead.org>,
Ingo Molnar <mingo@redhat.com>,
Mark Rutland <mark.rutland@arm.com>,
Alexander Shishkin <alexander.shishkin@linux.intel.com>,
Namhyung Kim <namhyung@kernel.org>,
Andi Kleen <ak@linux.intel.com>, Ian Rogers <irogers@google.com>,
Kan Liang <kan.liang@linux.intel.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 v2 01/11] perf c2c: Add dimensions for total load hit
Date: Sun, 13 Dec 2020 13:38:40 +0000 [thread overview]
Message-ID: <20201213133850.10070-2-leo.yan@linaro.org> (raw)
In-Reply-To: <20201213133850.10070-1-leo.yan@linaro.org>
Arm SPE trace data doesn't support HITM, but 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 total load operations and connect with
the threads info, e.g. if multiple threads hit the same cache line for
many times, this can give out the hint that it's likely to cause cache
false sharing issue.
Unlike having HITM tag, the proposed solution is not accurate and might
introduce false positive reporting, but it's a pragmatic approach for
detecting false sharing if memory event doesn't support HITM.
To sort with the cache line hit, this patch adds dimensions for total
load hit and the associated percentage calculation.
Signed-off-by: Leo Yan <leo.yan@linaro.org>
---
tools/perf/builtin-c2c.c | 112 +++++++++++++++++++++++++++++++++++++++
1 file changed, 112 insertions(+)
diff --git a/tools/perf/builtin-c2c.c b/tools/perf/builtin-c2c.c
index c5babeaa3b38..3d5a2dc8b4fd 100644
--- a/tools/perf/builtin-c2c.c
+++ b/tools/perf/builtin-c2c.c
@@ -615,6 +615,47 @@ tot_hitm_cmp(struct perf_hpp_fmt *fmt __maybe_unused,
return tot_hitm_left - tot_hitm_right;
}
+#define TOT_LD_HIT(stats) \
+ ((stats)->ld_fbhit + \
+ (stats)->ld_l1hit + \
+ (stats)->ld_l2hit + \
+ (stats)->ld_llchit + \
+ (stats)->lcl_hitm + \
+ (stats)->rmt_hitm + \
+ (stats)->rmt_hit)
+
+static int tot_ld_hit_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);
+ unsigned int tot_hit;
+
+ c2c_he = container_of(he, struct c2c_hist_entry, he);
+ tot_hit = TOT_LD_HIT(&c2c_he->stats);
+
+ return scnprintf(hpp->buf, hpp->size, "%*u", width, tot_hit);
+}
+
+static int64_t tot_ld_hit_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;
+ uint64_t tot_hit_left;
+ uint64_t tot_hit_right;
+
+ c2c_left = container_of(left, struct c2c_hist_entry, he);
+ c2c_right = container_of(right, struct c2c_hist_entry, he);
+
+ tot_hit_left = TOT_LD_HIT(&c2c_left->stats);
+ tot_hit_right = TOT_LD_HIT(&c2c_right->stats);
+
+ return tot_hit_left - tot_hit_right;
+}
+
#define STAT_FN_ENTRY(__f) \
static int \
__f ## _entry(struct perf_hpp_fmt *fmt, struct perf_hpp *hpp, \
@@ -860,6 +901,58 @@ percent_hitm_cmp(struct perf_hpp_fmt *fmt __maybe_unused,
return per_left - per_right;
}
+static double percent_tot_ld_hit(struct c2c_hist_entry *c2c_he)
+{
+ struct c2c_hists *hists;
+ int tot = 0, st = 0;
+
+ hists = container_of(c2c_he->he.hists, struct c2c_hists, hists);
+
+ st = TOT_LD_HIT(&c2c_he->stats);
+ tot = TOT_LD_HIT(&hists->stats);
+
+ return tot ? (double) st * 100 / tot : 0;
+}
+
+static int
+percent_tot_ld_hit_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_tot_ld_hit(c2c_he);
+ return scnprintf(hpp->buf, hpp->size, "%*s", width, PERC_STR(buf, per));
+}
+
+static int
+percent_tot_ld_hit_color(struct perf_hpp_fmt *fmt, struct perf_hpp *hpp,
+ struct hist_entry *he)
+{
+ return percent_color(fmt, hpp, he, percent_tot_ld_hit);
+}
+
+static int64_t
+percent_tot_ld_hit_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_tot_ld_hit(c2c_left);
+ per_right = percent_tot_ld_hit(c2c_right);
+
+ return per_left - per_right;
+}
+
static struct c2c_stats *he_stats(struct hist_entry *he)
{
struct c2c_hist_entry *c2c_he;
@@ -1412,6 +1505,14 @@ static struct c2c_dimension dim_ld_rmthit = {
.width = 8,
};
+static struct c2c_dimension dim_tot_ld_hit = {
+ .header = HEADER_BOTH("Load Hit", "Total"),
+ .name = "tot_ld_hit",
+ .cmp = tot_ld_hit_cmp,
+ .entry = tot_ld_hit_entry,
+ .width = 8,
+};
+
static struct c2c_dimension dim_tot_recs = {
.header = HEADER_BOTH("Total", "records"),
.name = "tot_recs",
@@ -1460,6 +1561,15 @@ static struct c2c_dimension dim_percent_lcl_hitm = {
.width = 7,
};
+static struct c2c_dimension dim_percent_tot_ld_hit = {
+ .header = HEADER_BOTH("Load Hit", "Pct"),
+ .name = "percent_tot_ld_hit",
+ .cmp = percent_tot_ld_hit_cmp,
+ .entry = percent_tot_ld_hit_entry,
+ .color = percent_tot_ld_hit_color,
+ .width = 8,
+};
+
static struct c2c_dimension dim_percent_stores_l1hit = {
.header = HEADER_SPAN("-- Store Refs --", "L1 Hit", 1),
.name = "percent_stores_l1hit",
@@ -1615,11 +1725,13 @@ static struct c2c_dimension *dimensions[] = {
&dim_ld_l2hit,
&dim_ld_llchit,
&dim_ld_rmthit,
+ &dim_tot_ld_hit,
&dim_tot_recs,
&dim_tot_loads,
&dim_percent_hitm,
&dim_percent_rmt_hitm,
&dim_percent_lcl_hitm,
+ &dim_percent_tot_ld_hit,
&dim_percent_stores_l1hit,
&dim_percent_stores_l1miss,
&dim_dram_lcl,
--
2.17.1
next prev parent reply other threads:[~2020-12-13 13:41 UTC|newest]
Thread overview: 31+ messages / expand[flat|nested] mbox.gz Atom feed top
2020-12-13 13:38 [PATCH v2 00/11] perf c2c: Sort cacheline with all loads Leo Yan
2020-12-13 13:38 ` Leo Yan [this message]
2021-01-06 7:38 ` [PATCH v2 01/11] perf c2c: Add dimensions for total load hit Namhyung Kim
2021-01-11 7:49 ` Leo Yan
2020-12-13 13:38 ` [PATCH v2 02/11] perf c2c: Add dimensions for " Leo Yan
2021-01-06 7:38 ` Namhyung Kim
2021-01-11 8:22 ` Leo Yan
2020-12-13 13:38 ` [PATCH v2 03/11] perf c2c: Add dimensions for load miss Leo Yan
2021-01-06 7:42 ` Namhyung Kim
2021-01-11 8:41 ` Leo Yan
2020-12-13 13:38 ` [PATCH v2 04/11] perf c2c: Rename for shared cache line stats Leo Yan
2021-01-06 7:44 ` Namhyung Kim
2021-01-11 8:42 ` Leo Yan
2020-12-13 13:38 ` [PATCH v2 05/11] perf c2c: Refactor hist entry validation Leo Yan
2020-12-13 13:38 ` [PATCH v2 06/11] perf c2c: Refactor display filter macro Leo Yan
2021-01-06 7:47 ` Namhyung Kim
2021-01-11 8:43 ` Leo Yan
2020-12-13 13:38 ` [PATCH v2 07/11] perf c2c: Refactor node display macro Leo Yan
2021-01-06 7:47 ` Namhyung Kim
2021-01-11 8:44 ` Leo Yan
2020-12-13 13:38 ` [PATCH v2 08/11] perf c2c: Refactor node header Leo Yan
2020-12-13 13:38 ` [PATCH v2 09/11] perf c2c: Add local variables for output metrics Leo Yan
2020-12-13 13:38 ` [PATCH v2 10/11] perf c2c: Sort on all cache hit for load operations Leo Yan
2021-01-06 7:52 ` Namhyung Kim
2021-01-11 8:47 ` Leo Yan
2020-12-13 13:38 ` [PATCH v2 11/11] perf c2c: Update documentation for display option 'all' Leo Yan
2021-01-03 22:52 ` [PATCH v2 00/11] perf c2c: Sort cacheline with all loads Jiri Olsa
2021-01-04 2:09 ` Leo Yan
2021-01-04 9:35 ` Jiri Olsa
2021-01-15 15:17 ` Arnaldo Carvalho de Melo
2021-01-16 0:45 ` 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=20201213133850.10070-2-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