Received-SPF: pass (google.com: domain of linux-kernel+bounces-204148-linux.lists.archive=gmail.com@vger.kernel.org designates 2604:1380:45e3:2400::1 as permitted sender) client-ip=2604:1380:45e3:2400::1;
Message-ID: <bc34d0aa-7152-4f0b-a3c4-11b1a63fba9b@arm.com>
Date: Thu, 6 Jun 2024 11:56:46 +0100
Precedence: bulk
MIME-Version: 1.0
User-Agent: Mozilla Thunderbird
Subject: Re: [PATCH v1 0/3] Fix and improve __maps__fixup_overlap_and_insert
To: Ian Rogers <irogers@google.com>
References: <20240521165109.708593-1-irogers@google.com>
Content-Language: en-US
Cc: "Steinar H . Gunderson" <sesse@google.com>,
 Peter Zijlstra <peterz@infradead.org>, Ingo Molnar <mingo@redhat.com>,
 Arnaldo Carvalho de Melo <acme@kernel.org>,
 Namhyung Kim <namhyung@kernel.org>, Mark Rutland <mark.rutland@arm.com>,
 Alexander Shishkin <alexander.shishkin@linux.intel.com>,
 Jiri Olsa <jolsa@kernel.org>, Adrian Hunter <adrian.hunter@intel.com>,
 Kan Liang <kan.liang@linux.intel.com>, linux-perf-users@vger.kernel.org,
 linux-kernel@vger.kernel.org
From: James Clark <james.clark@arm.com>
In-Reply-To: <20240521165109.708593-1-irogers@google.com>
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 7bit


On 21/05/2024 17:51, Ian Rogers wrote:
> Fix latent unlikely bugs in __maps__fixup_overlap_and_insert.
> 
> Improve __maps__fixup_overlap_and_insert's performance 21x in the case
> of overlapping mmaps. sesse@google.com reported slowness opening
> perf.data files from chromium where the files contained a large number
> of overlapping mappings. Improve this case primarily by avoiding
> unnecessary sorting.
> 
> Unscientific timing data processing a perf.data file with overlapping
> mmap events from chromium:
> 
> Before:
> real    0m9.856s
> user    0m9.637s
> sys     0m0.204s
> 
> After:
> real    0m0.675s
> user    0m0.454s
> sys     0m0.196s
> 
> Tested with address/leak sanitizer, invariant checks and validating
> the before and after output are identical.
> 
> Ian Rogers (3):
>   perf maps: Fix use after free in __maps__fixup_overlap_and_insert
>   perf maps: Reduce sorting for overlapping mappings
>   perf maps: Add/use a sorted insert for fixup overlap and insert
> 
>  tools/perf/util/maps.c | 113 +++++++++++++++++++++++++++++++++--------
>  1 file changed, 92 insertions(+), 21 deletions(-)
> 

Reviewed-by: James Clark <james.clark@arm.com>

I'm wondering if there is any point in the non sorted insert any more?

Maps could be always either sorted by name or sorted by address and
insert() is always a sorted/fixup-overlaps insert depending on the sort
style of the list.