[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-Id: <20230606230658.c1b478f905c82a9f7005034d@kernel.org>
Date: Tue, 6 Jun 2023 23:06:58 +0900
From: Masami Hiramatsu (Google) <mhiramat@...nel.org>
To: Namhyung Kim <namhyung@...nel.org>
Cc: Arnaldo Carvalho de Melo <acme@...nel.org>,
Jiri Olsa <jolsa@...nel.org>, Ian Rogers <irogers@...gle.com>,
Adrian Hunter <adrian.hunter@...el.com>,
Peter Zijlstra <peterz@...radead.org>,
Ingo Molnar <mingo@...nel.org>,
LKML <linux-kernel@...r.kernel.org>,
linux-perf-users@...r.kernel.org, Andi Kleen <ak@...ux.intel.com>,
Masami Hiramatsu <mhiramat@...nel.org>,
Kan Liang <kan.liang@...ux.intel.com>
Subject: Re: [PATCH v2 1/2] perf annotate: Handle x86 instruction suffix
generally
On Wed, 24 May 2023 13:50:53 -0700
Namhyung Kim <namhyung@...nel.org> wrote:
> In AT&T asm syntax, most of x86 instructions can have size suffix like
> b, w, l or q. Instead of adding all these instructions in the table,
> we can handle them in a general way.
>
> For example, it can try to find an instruction as is. If not found,
> assuming it has a suffix and it'd try again without the suffix if it's
> one of the allowed suffixes. This way, we can reduce the instruction
> table size for duplicated entries of the same instructions with a
> different suffix.
>
> If an instruction xyz and others like xyz<suffix> are completely
> different ones, then they both need to be listed in the table so that
> they can be found before the second attempt (without the suffix).
Looks good to me.
Reviewed-by: Masami Hiramatsu (Google) <mhiramat@...nel.org>
>
> Signed-off-by: Namhyung Kim <namhyung@...nel.org>
> ---
> tools/perf/util/annotate.c | 22 ++++++++++++++++++++++
> 1 file changed, 22 insertions(+)
>
> diff --git a/tools/perf/util/annotate.c b/tools/perf/util/annotate.c
> index b708bbc49c9e..7f05f2a2aa83 100644
> --- a/tools/perf/util/annotate.c
> +++ b/tools/perf/util/annotate.c
> @@ -70,6 +70,7 @@ struct arch {
> struct ins_ops *(*associate_instruction_ops)(struct arch *arch, const char *name);
> bool sorted_instructions;
> bool initialized;
> + const char *insn_suffix;
> void *priv;
> unsigned int model;
> unsigned int family;
> @@ -179,6 +180,7 @@ static struct arch architectures[] = {
> .init = x86__annotate_init,
> .instructions = x86__instructions,
> .nr_instructions = ARRAY_SIZE(x86__instructions),
> + .insn_suffix = "bwlq",
> .objdump = {
> .comment_char = '#',
> },
> @@ -720,6 +722,26 @@ static struct ins_ops *__ins__find(struct arch *arch, const char *name)
> }
>
> ins = bsearch(name, arch->instructions, nmemb, sizeof(struct ins), ins__key_cmp);
> + if (ins)
> + return ins->ops;
> +
> + if (arch->insn_suffix) {
> + char tmp[32];
> + char suffix;
> + size_t len = strlen(name);
> +
> + if (len == 0 || len >= sizeof(tmp))
> + return NULL;
> +
> + suffix = name[len - 1];
> + if (strchr(arch->insn_suffix, suffix) == NULL)
> + return NULL;
> +
> + strcpy(tmp, name);
> + tmp[len - 1] = '\0'; /* remove the suffix and check again */
> +
> + ins = bsearch(tmp, arch->instructions, nmemb, sizeof(struct ins), ins__key_cmp);
> + }
> return ins ? ins->ops : NULL;
> }
>
> --
> 2.41.0.rc0.172.g3f132b7071-goog
>
--
Masami Hiramatsu (Google) <mhiramat@...nel.org>
Powered by blists - more mailing lists