Re: [PATCH 2/2] perf tools/tests: Fix topdown groups test on hybrid platforms

[Falcon, Thomas]

On Mon, 2025-02-24 at 08:33 +0000, Dapeng Mi wrote:
> When running topdown groups test on hybrid platforms like LNL/ARL, we
> see the following 2 commands fail.
> 
> perf stat $cputype -e '{instructions,slots},topdown-retiring' true
> perf stat $cputype -e '{instructions,slots},{topdown-retiring}' true
> 
> Take the 1st command as an example, 5 events are created on hybrid
> platform. They are cpu_atom/instructions/, cpu_core/instructions/,
> cpu_core/slots/, cpu_atom/topdown-retiring/ and
> cpu_core/topdown-retiring/ events. The former 3 events are in a group
> and the latter 2 topdown-retiring events are independent events.
> 
> As the limitation of current implementation, the
> cpu_core/topdown-retiring/ event can't be moved into previous group
> as
> long as there are other events before it. That's the reason why we
> see
> the failure.
> 
> Thus add "--cputype core" option to limit only P-core events are
> tested.
> 
> Signed-off-by: Dapeng Mi <dapeng1.mi@xxxxxxxxxxxxxxx>

Tested-by: Thomas Falcon <thomas.falcon@xxxxxxxxx>

> ---
>  tools/perf/arch/x86/util/evlist.c | 26 +++++++++++++++++++++++---
>  tools/perf/tests/shell/stat.sh    | 20 ++++++++++++++++++--
>  2 files changed, 41 insertions(+), 5 deletions(-)
> 
> diff --git a/tools/perf/arch/x86/util/evlist.c
> b/tools/perf/arch/x86/util/evlist.c
> index 447a734e591c..0a71ba975871 100644
> --- a/tools/perf/arch/x86/util/evlist.c
> +++ b/tools/perf/arch/x86/util/evlist.c
> @@ -9,7 +9,7 @@ int arch_evlist__cmp(const struct evsel *lhs, const
> struct evsel *rhs)
>  {
>  	/*
>  	 * Currently the following topdown events sequence are
> supported to
> -	 * move and regroup correctly.
> +	 * move and regroup correctly on non-hybrid platforms.
>  	 *
>  	 * a. all events in a group
>  	 *    perf stat -e "{instructions,topdown-retiring,slots}" -
> C0 sleep 1
> @@ -44,7 +44,7 @@ int arch_evlist__cmp(const struct evsel *lhs, const
> struct evsel *rhs)
>  	 * topdown metrics events must be first event after the
> slots event group,
>  	 * otherwise topdown metrics events can't be regrouped
> correctly, e.g.
>  	 *
> -	 * a. perf stat -e "{instructions,slots},cycles,topdown-
> retiring" -C0 sleep 1
> +	 * e. perf stat -e "{instructions,slots},cycles,topdown-
> retiring" -C0 sleep 1
>  	 *    WARNING: events were regrouped to match PMUs
>  	 *     Performance counter stats for 'CPU(s) 0':
>  	 *         17,923,134      slots
> @@ -56,11 +56,31 @@ int arch_evlist__cmp(const struct evsel *lhs,
> const struct evsel *rhs)
>  	 * has topdown metrics events must contain only the topdown
> metrics event,
>  	 * otherwise topdown metrics event can't be regrouped
> correctly as well, e.g.
>  	 *
> -	 * a. perf stat -e "{instructions,slots},{topdown-
> retiring,cycles}" -C0 sleep 1
> +	 * f. perf stat -e "{instructions,slots},{topdown-
> retiring,cycles}" -C0 sleep 1
>  	 *    WARNING: events were regrouped to match PMUs
>  	 *    Error:
>  	 *    The sys_perf_event_open() syscall returned with 22
> (Invalid argument) for
>  	 *    event (topdown-retiring)
> +	 *
> +	 * For hybrid platforms, the sequences 'c' and 'd' are not
> supported as well
> +	 * besides above sequences 'e' and 'f'.
> +	 *
> +	 *    perf stat -e "{instructions,slots},topdown-retiring" -
> C0 sleep 1
> +	 *    perf stat -e "{instructions,slots},{topdown-retiring}"
> -C0 sleep 1
> +	 *
> +	 * On hybrid platforms each event would create an instance
> on all types of PMU
> +	 * if the event can be supported by the PMU, i.e., the
> "topdown-retiring" event
> +	 * would create two instances on hybrid platforms with P-
> cores and E-cores,
> +	 * "cpu_core/topdown-retiring/" and
> "cpu_atom/topdown_retiring".
> +	 *
> +	 * Take the first command as an example, the events list
> would be converted to
> +	 * below list in fact.
> +	 *
> +	 *
> "{cpu_atom/instructions/,cpu_core/instructions/,cpu_core/slots/},
> +	 *  cpu_atom/topdown-retiring/,cpu_core/topdown-retiring/"
> +	 *
> +	 * This is actually same with event list in case 'e',
> "cpu_core/topdown-retiring/"
> +	 * event can't be moved into previous events group.
>  	 */
>  	if (topdown_sys_has_perf_metrics() &&
>  	    (arch_evsel__must_be_in_group(lhs) ||
> arch_evsel__must_be_in_group(rhs))) {
> diff --git a/tools/perf/tests/shell/stat.sh
> b/tools/perf/tests/shell/stat.sh
> index 68323d636fb7..cdfe27c25528 100755
> --- a/tools/perf/tests/shell/stat.sh
> +++ b/tools/perf/tests/shell/stat.sh
> @@ -5,6 +5,16 @@
>  set -e
>  
>  err=0
> +is_hybrid=false
> +
> +check_hybrid_platform() {
> +  pmus=$(ls /sys/bus/event_source/devices/*/cpus 2>/dev/null | wc -
> l)
> +  if [ "$pmus" -gt 1 ]
> +  then
> +    is_hybrid=true
> +  fi
> +}
> +
>  test_default_stat() {
>    echo "Basic stat command test"
>    if ! perf stat true 2>&1 | grep -E -q "Performance counter stats
> for 'true':"
> @@ -62,6 +72,11 @@ test_topdown_groups() {
>    # Topdown events must be grouped with the slots event first. Test
> that
>    # parse-events reorders this.
>    echo "Topdown event group test"
> +  cputype=""
> +  if $is_hybrid
> +  then
> +    cputype="--cputype core"
> +  fi
>    if ! perf stat -e '{slots,topdown-retiring}' true > /dev/null 2>&1
>    then
>      echo "Topdown event group test [Skipped event parsing failed]"
> @@ -85,13 +100,13 @@ test_topdown_groups() {
>      err=1
>      return
>    fi
> -  if perf stat -e '{instructions,slots},topdown-retiring' true 2>&1
> | grep -E -q "<not supported>"
> +  if perf stat $cputype -e '{instructions,slots},topdown-retiring'
> true 2>&1 | grep -E -q "<not supported>"
>    then
>      echo "Topdown event group test [Failed topdown metrics event not
> move into slots group]"
>      err=1
>      return
>    fi
> -  if perf stat -e '{instructions,slots},{topdown-retiring}' true
> 2>&1 | grep -E -q "<not supported>"
> +  if perf stat $cputype -e '{instructions,slots},{topdown-retiring}'
> true 2>&1 | grep -E -q "<not supported>"
>    then
>      echo "Topdown event group test [Failed topdown metrics group not
> merge into slots group]"
>      err=1
> @@ -200,6 +215,7 @@ test_hybrid() {
>    echo "hybrid test [Success]"
>  }
>  
> +check_hybrid_platform
>  test_default_stat
>  test_stat_record_report
>  test_stat_record_script