[PATCH 2 of 2] perf: add an option to profile the benchmark section

[Pierre-Yves David]

On 5/23/19 3:14 PM, Martin von Zweigbergk wrote:
> 
> 
> On Wed, May 22, 2019, 07:49 Pierre-Yves David 
> <pierre-yves.david at ens-lyon.org <mailto:pierre-yves.david at ens-lyon.org>> 
> wrote:
> 
>     # HG changeset patch
>     # User Pierre-Yves David <pierre-yves.david at octobus.net
>     <mailto:pierre-yves.david at octobus.net>>
>     # Date 1558445208 -7200
>     #      Tue May 21 15:26:48 2019 +0200
>     # Node ID b5c316b4248f3add425313ac0f443c8888706b79
>     # Parent  8553423d79608bd5d144456448509bdf98c6df11
>     # EXP-Topic perf-profile
>     # Available At https://bitbucket.org/octobus/mercurial-devel/
>     #              hg pull
>     https://bitbucket.org/octobus/mercurial-devel/ -r b5c316b4248f
>     perf: add an option to profile the benchmark section
> 
>     Running a perf command with --profile gather data for the whole command
>     execution, including setup and cleanup. This can significantly alter
>     the data.
> 
>     To work around this we introduce a new option, it trigger the
>     profiling of only one
>     iteration of the benchmarked section.
> 
>     diff --git a/contrib/perf.py b/contrib/perf.py
>     --- a/contrib/perf.py
>     +++ b/contrib/perf.py
>     @@ -18,6 +18,10 @@ Configurations
>       ``pre-run``
>         number of run to perform before starting measurement.
> 
>     +``profile-benchmark``
>     +  Enable profiling for the benchmarked section.
>     +  (The first iteration is benchmarked)
>     +
>       ``run-limits``
>         Control the number of runs each benchmark will perform. The
>     option value
>         should be a list of `<time>-<numberofrun>` pairs. After each run the
>     @@ -109,6 +113,10 @@ try:
>       except ImportError:
>           pass
> 
>     +try:
>     +    from mercurial import profiling
>     +except ImportError:
>     +    profiling = None
> 
>       def identity(a):
>           return a
>     @@ -246,6 +254,9 @@ try:
>           configitem(b'perf', b'pre-run',
>               default=mercurial.configitems.dynamicdefault,
>           )
>     +    configitem(b'perf', b'profile-benchmark',
>     +        default=mercurial.configitems.dynamicdefault,
>     +    )
>           configitem(b'perf', b'run-limits',
>               default=mercurial.configitems.dynamicdefault,
>           )
>     @@ -257,6 +268,13 @@ def getlen(ui):
>               return lambda x: 1
>           return len
> 
>     +class noop(object):
>     +    """dummy context manager"""
>     +    def __enter__(self):
>     +        pass
>     +    def __exit__(self, *args):
>     +        pass
>     +
> 
> 
> Could this be replaced by util.nullcontextmanager()?

It could, but that would make the commands less compatible with older 
version (who don't have `util.nullcontextmanager`), and we care about 
such compatibility in perf.py

-- 
Pierre-Yves David