[PATCH 01 of 11] bdiff: use Python memory allocator in fixws

Jun Wu quark at fb.com
Thu Mar 16 02:16:20 EDT 2017


Excerpts from Jun Wu's message of 2017-03-15 23:15:13 -0700:
> I think it makes sense to migrate Python modules to PyMem_*.
> 
> However, I'm also a bit conservative about the "calloc" perf issue mentioned
> by Yuya. I think it's safe to take Patch 1, 2, 4, 5 where there is no calloc

Sorry, it should be "Patch 1, 2, 4, and 6", not 5. I cannot count.

> change for now. We can figure out what to do with calloc / pure C code
> later.
> 
> Excerpts from Gregory Szorc's message of 2017-03-09 13:59:09 -0800:
> > # HG changeset patch
> > # User Gregory Szorc <gregory.szorc at gmail.com>
> > # Date 1489089265 28800
> > #      Thu Mar 09 11:54:25 2017 -0800
> > # Node ID 7a8ce919a13a9ff6e73f391fe14d228e18387f15
> > # Parent  1871a1ee64ed49172b1568b89cdbab126284b309
> > bdiff: use Python memory allocator in fixws
> > 
> > Python has its own memory allocation APIs. For allocations
> > <= 512 bytes, it allocates memory from arenas. This means that
> > average small allocations don't call the system allocator, which
> > makes them faster. Also, arena allocations cut down on memory
> > fragmentation, which can matter for performance in long-running
> > processes.
> > 
> > Another advantage of using the Python memory allocator is that
> > allocations are tracked by Python. This is a bigger deal in
> > Python 3, as modern versions of Python have some decent built-in
> > tools for examining memory usage, leaks, etc.
> > 
> > This patch converts a trivial malloc() + free() in the bdiff code
> > to use the Python allocator APIs. Since the object being
> > operated on is a line, chances are it will use an arena. So,
> > this could have a net positive impact on performance (although
> > I didn't measure it).
> > 
> > diff --git a/mercurial/bdiff_module.c b/mercurial/bdiff_module.c
> > --- a/mercurial/bdiff_module.c
> > +++ b/mercurial/bdiff_module.c
> > @@ -158,7 +158,7 @@ static PyObject *fixws(PyObject *self, P
> >      r = PyBytes_AsString(s);
> >      rlen = PyBytes_Size(s);
> >  
> > -    w = (char *)malloc(rlen ? rlen : 1);
> > +    w = (char *)PyMem_Malloc(rlen ? rlen : 1);
> >      if (!w)
> >          goto nomem;
> >  
> > @@ -178,7 +178,7 @@ static PyObject *fixws(PyObject *self, P
> >      result = PyBytes_FromStringAndSize(w, wlen);
> >  
> >  nomem:
> > -    free(w);
> > +    PyMem_Free(w);
> >      return result ? result : PyErr_NoMemory();
> >  }
> >  


More information about the Mercurial-devel mailing list