perf test: Make all metrics test more tolerant
authorIan Rogers <irogers@google.com>
Thu, 7 Jul 2022 15:34:49 +0000 (08:34 -0700)
committerArnaldo Carvalho de Melo <acme@redhat.com>
Tue, 12 Jul 2022 11:54:47 +0000 (08:54 -0300)
Metric names are truncated so don't try to match all of one.

Allow AMX metrics to skip as floating point ones do.

Metrics for optane memory can also skip rather than fail.

Add a system wide check for uncore metrics.

Restructure code to avoid extensive nesting.

Some impetus for this in:

https://lore.kernel.org/lkml/d32376b5-5538-ff00-6620-e74ad4b4abf2@huawei.com/

Suggested-by: John Garry <john.garry@huawei.com>
Signed-off-by: Ian Rogers <irogers@google.com>
Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com>
Cc: Andi Kleen <ak@linux.intel.com>
Cc: Ingo Molnar <mingo@redhat.com>
Cc: Jiri Olsa <jolsa@kernel.org>
Cc: John Garry <john.garry@huawei.com>
Cc: Kan Liang <kan.liang@linux.intel.com>
Cc: Mark Rutland <mark.rutland@arm.com>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Stephane Eranian <eranian@google.com>
Link: https://lore.kernel.org/r/20220707153449.202409-2-irogers@google.com
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
tools/perf/tests/shell/stat_all_metrics.sh

index e7c59e5a7a98da3b3ce8e6640103dabc1a806ff0..6e79349e42beff5d1d80086db40f20c3d24fc1b4 100755 (executable)
@@ -1,26 +1,41 @@
-#!/bin/sh
+#!/bin/bash
 # perf all metrics test
 # SPDX-License-Identifier: GPL-2.0
 
-set -e
-
 err=0
 for m in $(perf list --raw-dump metrics); do
   echo "Testing $m"
   result=$(perf stat -M "$m" true 2>&1)
-  if [[ ! "$result" =~ "$m" ]] && [[ ! "$result" =~ "<not supported>" ]]; then
-    # We failed to see the metric and the events are support. Possibly the
-    # workload was too small so retry with something longer.
-    result=$(perf stat -M "$m" perf bench internals synthesize 2>&1)
-    if [[ ! "$result" =~ "$m" ]]; then
-      echo "Metric '$m' not printed in:"
-      echo "$result"
-      if [[ "$result" =~ "FP_ARITH" && "$err" != "1" ]]; then
-        echo "Skip, not fail, for FP issues"
-        err=2
-      else
-        err=1
-      fi
+  if [[ "$result" =~ "${m:0:50}" ]] || [[ "$result" =~ "<not supported>" ]]
+  then
+    continue
+  fi
+  # Failed so try system wide.
+  result=$(perf stat -M "$m" -a true 2>&1)
+  if [[ "$result" =~ "${m:0:50}" ]]
+  then
+    continue
+  fi
+  # Failed again, possibly the workload was too small so retry with something
+  # longer.
+  result=$(perf stat -M "$m" perf bench internals synthesize 2>&1)
+  if [[ "$result" =~ "${m:0:50}" ]]
+  then
+    continue
+  fi
+  echo "Metric '$m' not printed in:"
+  echo "$result"
+  if [[ "$err" != "1" ]]
+  then
+    err=2
+    if [[ "$result" =~ "FP_ARITH" || "$result" =~ "AMX" ]]
+    then
+      echo "Skip, not fail, for FP issues"
+    elif [[ "$result" =~ "PMM" ]]
+    then
+      echo "Skip, not fail, for Optane memory issues"
+    else
+      err=1
     fi
   fi
 done