[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <555fbea8-7103-4de1-bcd5-accc9c649e62@linaro.org>
Date: Thu, 10 Oct 2024 14:53:38 +0100
From: James Clark <james.clark@...aro.org>
To: Leo Yan <leo.yan@....com>
Cc: Arnaldo Carvalho de Melo <acme@...nel.org>,
Namhyung Kim <namhyung@...nel.org>, Besar Wicaksono <bwicaksono@...dia.com>,
Mark Rutland <mark.rutland@....com>,
Alexander Shishkin <alexander.shishkin@...ux.intel.com>,
Jiri Olsa <jolsa@...nel.org>, Ian Rogers <irogers@...gle.com>,
Adrian Hunter <adrian.hunter@...el.com>,
"Liang, Kan" <kan.liang@...ux.intel.com>,
John Garry <john.g.garry@...cle.com>, Will Deacon <will@...nel.org>,
Mike Leach <mike.leach@...aro.org>, linux-perf-users@...r.kernel.org,
linux-kernel@...r.kernel.org, linux-arm-kernel@...ts.infradead.org
Subject: Re: [PATCH v3 4/7] perf arm-spe: Use metadata to decide the data
source feature
On 03/10/2024 7:53 pm, Leo Yan wrote:
> Use the info in the metadata to decide if the data source feature is
> supported. The CPU MIDR must be in the CPU list for the common data
> source encoding.
>
> For the metadata version 1, it doesn't include info for MIDR. In this
> case, due to absent info for making decision, print out warning to
> remind users to upgrade tool and returns false.
>
> Signed-off-by: Leo Yan <leo.yan@....com>
> ---
> tools/perf/util/arm-spe.c | 67 +++++++++++++++++++++++++++++++++++++--
> 1 file changed, 64 insertions(+), 3 deletions(-)
>
> diff --git a/tools/perf/util/arm-spe.c b/tools/perf/util/arm-spe.c
> index 587943b6bdb8..9221b2f66bbe 100644
> --- a/tools/perf/util/arm-spe.c
> +++ b/tools/perf/util/arm-spe.c
> @@ -278,6 +278,20 @@ static int arm_spe_set_tid(struct arm_spe_queue *speq, pid_t tid)
> return 0;
> }
>
> +static u64 *arm_spe__get_metadata_by_cpu(struct arm_spe *spe, u64 cpu)
> +{
> + u64 i;
> +
> + if (!spe->metadata)
> + return NULL;
> +
> + for (i = 0; i < spe->metadata_nr_cpu; i++)
> + if (spe->metadata[i][ARM_SPE_CPU] == cpu)
> + return spe->metadata[i];
> +
> + return NULL;
> +}
> +
> static struct simd_flags arm_spe__synth_simd_flags(const struct arm_spe_record *record)
> {
> struct simd_flags simd_flags = {};
> @@ -520,10 +534,57 @@ static void arm_spe__synth_memory_level(const struct arm_spe_record *record,
> data_src->mem_lvl |= PERF_MEM_LVL_REM_CCE1;
> }
>
> -static u64 arm_spe__synth_data_source(const struct arm_spe_record *record, u64 midr)
> +static bool arm_spe__is_common_ds_encoding(struct arm_spe_queue *speq)
> +{
> + struct arm_spe *spe = speq->spe;
> + bool is_in_cpu_list;
> + u64 *metadata = NULL;
> + u64 midr = 0;
> +
> + /*
> + * Metadata version 1 doesn't contain any info for MIDR.
> + * Simply return false in this case.
> + */
> + if (spe->metadata_ver == 1) {
> + pr_warning_once("The data file contains metadata version 1, "
> + "which is absent the info for data source. "
> + "Please upgrade the tool to record data.\n");
> + return false;
> + }
> +
> + /* CPU ID is -1 for per-thread mode */
> + if (speq->cpu < 0) {
> + /*
> + * On the heterogeneous system, due to CPU ID is -1,
> + * cannot confirm the data source packet is supported.
> + */
> + if (!spe->is_homogeneous)
> + return false;
Technically you could look at timestamps and context switches to find
the CPU for non-homogeneous per-thread mode. But it's such an edge case
I'm not sure it's even worth leaving a TODO for.
Powered by blists - more mailing lists