rust-discovery: using from Python code
As previously done in other topics, the Rust version is used if it's been
built.
The version fully in Rust of the partialdiscovery class has the performance
advantage over the Python version (actually using the Rust MissingAncestor) if
the undecided set is big enough. Otherwise no sampling occurs, and the
discovery is reasonably fast anyway.
Note: it's hard to predict the size of the initial undecided set, it can
depend on the kind of topological changes between the local and remote graphs.
The point of the Rust version is to make the bad cases acceptable.
More specifically, the performance advantages are:
- faster sampling, especially takefullsample()
- much faster addmissings() in almost all cases (see commit message in
grandparent of the present changeset)
- no conversion cost of the undecided set at the interface between Rust and
Python
== Measurements with big undecided sets
For an extreme example, discovery between mozilla-try and mozilla-unified
(over one million undecided revisions, same case as in dbd0fcca6dfc), we
get roughly a x2.5/x3 better performance:
Growing sample size (5% starting with 200): time goes down from
210 to 72 seconds.
Constant sample size of 200: time down from 1853 to 659 seconds.
With a sample size computed from number of roots and heads of the
undecided set (`respectsize` is `False`), here are perfdiscovery results:
Before ! wall 9.358729 comb 9.360000 user 9.310000 sys 0.050000 (median of 50)
After ! wall 3.793819 comb 3.790000 user 3.750000 sys 0.040000 (median of 50)
In that later case, the sample sizes are routinely in the hundreds of
thousands of revisions. While still faster, the Rust iteration in
addmissings has less of an advantage than with smaller sample sizes, but
one sees addcommons becoming faster, probably a consequence of not having
to copy big sets back and forth.
This example is not a goal in itself, but it showcases several different
areas in which the process can become slow, due to different factors, and
how this full Rust version can help.
== Measurements with small undecided sets
In cases the undecided set is small enough than no sampling occurs,
the Rust version has a disadvantage at init if `targetheads` is really big
(some time is lost in the translation to Rust data structures),
and that is compensated by the faster `addmissings()`.
On a private repository with over one million commits, we still get a minor
improvement, of 6.8%:
Before ! wall 0.593585 comb 0.590000 user 0.550000 sys 0.040000 (median of 50)
After ! wall 0.553035 comb 0.550000 user 0.520000 sys 0.030000 (median of 50)
What's interesting in that case is the first addinfo() at 180ms for Rust and
233ms for Python+C, mostly due to add_missings and the children cache
computation being done in less than 0.2ms on the Rust side vs over 40ms on the
Python side.
The worst case we have on hand is with mozilla-try, prepared with
discovery-helper.sh for 10 heads and depth 10, time goes up 2.2% on the median.
In this case `targetheads` is really huge with 165842 server heads.
Before ! wall 0.823884 comb 0.810000 user 0.790000 sys 0.020000 (median of 50)
After ! wall 0.842607 comb 0.840000 user 0.800000 sys 0.040000 (median of 50)
If that would be considered a problem, more adjustments can be made, which are
prematurate at this stage: cooking special variants of methods of the inner
MissingAncestors object, retrieving local heads directly from Rust to avoid
the cost of conversion. Effort would probably be better spent at this point
improving the surroundings if needed.
Here's another data point with a smaller repository, pypy, where performance
is almost identical
Before ! wall 0.015121 comb 0.030000 user 0.020000 sys 0.010000 (median of 186)
After ! wall 0.015009 comb 0.010000 user 0.010000 sys 0.000000 (median of 184)
Differential Revision: https://phab.mercurial-scm.org/D6430
/*
exewrapper.c - wrapper for calling a python script on Windows
Copyright 2012 Adrian Buehlmann <adrian@cadifra.com> and others
This software may be used and distributed according to the terms of the
GNU General Public License version 2 or any later version.
*/
#include <Python.h>
#include <stdio.h>
#include <tchar.h>
#include <windows.h>
#include "hgpythonlib.h"
#ifdef __GNUC__
int strcat_s(char *d, size_t n, const char *s)
{
return !strncat(d, s, n);
}
int strcpy_s(char *d, size_t n, const char *s)
{
return !strncpy(d, s, n);
}
#define _tcscpy_s strcpy_s
#define _tcscat_s strcat_s
#define _countof(array) (sizeof(array) / sizeof(array[0]))
#endif
static TCHAR pyscript[MAX_PATH + 10];
static TCHAR pyhome[MAX_PATH + 10];
static TCHAR pydllfile[MAX_PATH + 10];
int _tmain(int argc, TCHAR *argv[])
{
TCHAR *p;
int ret;
int i;
int n;
TCHAR **pyargv;
WIN32_FIND_DATA fdata;
HANDLE hfind;
const char *err;
HMODULE pydll;
void(__cdecl * Py_SetPythonHome)(TCHAR * home);
int(__cdecl * Py_Main)(int argc, TCHAR *argv[]);
#if PY_MAJOR_VERSION >= 3
Py_LegacyWindowsStdioFlag = 1;
#endif
if (GetModuleFileName(NULL, pyscript, _countof(pyscript)) == 0) {
err = "GetModuleFileName failed";
goto bail;
}
p = _tcsrchr(pyscript, '.');
if (p == NULL) {
err = "malformed module filename";
goto bail;
}
*p = 0; /* cut trailing ".exe" */
_tcscpy_s(pyhome, _countof(pyhome), pyscript);
hfind = FindFirstFile(pyscript, &fdata);
if (hfind != INVALID_HANDLE_VALUE) {
/* pyscript exists, close handle */
FindClose(hfind);
} else {
/* file pyscript isn't there, take <pyscript>exe.py */
_tcscat_s(pyscript, _countof(pyscript), _T("exe.py"));
}
pydll = NULL;
p = _tcsrchr(pyhome, _T('\\'));
if (p == NULL) {
err = "can't find backslash in module filename";
goto bail;
}
*p = 0; /* cut at directory */
/* check for private Python of HackableMercurial */
_tcscat_s(pyhome, _countof(pyhome), _T("\\hg-python"));
hfind = FindFirstFile(pyhome, &fdata);
if (hfind != INVALID_HANDLE_VALUE) {
/* Path .\hg-python exists. We are probably in HackableMercurial
scenario, so let's load python dll from this dir. */
FindClose(hfind);
_tcscpy_s(pydllfile, _countof(pydllfile), pyhome);
_tcscat_s(pydllfile, _countof(pydllfile),
_T("\\") _T(HGPYTHONLIB) _T(".dll"));
pydll = LoadLibrary(pydllfile);
if (pydll == NULL) {
err = "failed to load private Python DLL " HGPYTHONLIB
".dll";
goto bail;
}
Py_SetPythonHome =
(void *)GetProcAddress(pydll, "Py_SetPythonHome");
if (Py_SetPythonHome == NULL) {
err = "failed to get Py_SetPythonHome";
goto bail;
}
Py_SetPythonHome(pyhome);
}
if (pydll == NULL) {
pydll = LoadLibrary(_T(HGPYTHONLIB) _T(".dll"));
if (pydll == NULL) {
err = "failed to load Python DLL " HGPYTHONLIB ".dll";
goto bail;
}
}
Py_Main = (void *)GetProcAddress(pydll, "Py_Main");
if (Py_Main == NULL) {
err = "failed to get Py_Main";
goto bail;
}
/*
Only add the pyscript to the args, if it's not already there. It may
already be there, if the script spawned a child process of itself, in
the same way as it got called, that is, with the pyscript already in
place. So we optionally accept the pyscript as the first argument
(argv[1]), letting our exe taking the role of the python interpreter.
*/
if (argc >= 2 && _tcscmp(argv[1], pyscript) == 0) {
/*
pyscript is already in the args, so there is no need to copy
the args and we can directly call the python interpreter with
the original args.
*/
return Py_Main(argc, argv);
}
/*
Start assembling the args for the Python interpreter call. We put the
name of our exe (argv[0]) in the position where the python.exe
canonically is, and insert the pyscript next.
*/
pyargv = malloc((argc + 5) * sizeof(TCHAR *));
if (pyargv == NULL) {
err = "not enough memory";
goto bail;
}
n = 0;
pyargv[n++] = argv[0];
pyargv[n++] = pyscript;
/* copy remaining args from the command line */
for (i = 1; i < argc; i++)
pyargv[n++] = argv[i];
/* argv[argc] is guaranteed to be NULL, so we forward that guarantee */
pyargv[n] = NULL;
ret = Py_Main(n, pyargv); /* The Python interpreter call */
free(pyargv);
return ret;
bail:
fprintf(stderr, "abort: %s\n", err);
return 255;
}