Repository navigation
Conversation
Member
Author
|
If a list writer is used by multiple threads, |
Member
Author
|
This C API is just an abstraction on top of existing PyList functions. |
Documentation build overview
12 files changed ·
|
Member
Author
|
Another advantage of using this C API instead of |
Member
Author
Oh, this failure seems to be unrelated to my change: |
Member
Author
|
I ran a benchmark to compare
I ran the benchmark on Free Threaded build on Linux ( import pyperf, functools, _testcapi
runner = pyperf.Runner()
sizes = (0, 10, 100, 1_000)
for list_len in sizes:
test_list = list(range(list_len))
func = functools.partial(_testcapi.bench_setitem_macro, test_list)
runner.bench_time_func(f'setitem macro {list_len:,}-list', func)
for list_len in sizes:
test_list = list(range(list_len))
func = functools.partial(_testcapi.bench_setitem_func, test_list)
runner.bench_time_func(f'setitem func {list_len:,}-list', func)
for list_len in sizes:
test_list = list(range(list_len))
func = functools.partial(_testcapi.bench_writer, test_list)
runner.bench_time_func(f'writer {list_len:,}-list', func)Results:
On Free Threading, Patch adding 3 bench functions to _testcapi: Detailsdiff --git a/Modules/_testcapimodule.c b/Modules/_testcapimodule.c
index 033ee3234cf..a3a8d9989c0 100644
--- a/Modules/_testcapimodule.c
+++ b/Modules/_testcapimodule.c
@@ -2889,6 +2889,112 @@ uptime_bsd(PyObject *Py_UNUSED(self), PyObject *Py_UNUSED(args))
#endif
+static PyObject*
+bench_setitem_macro(PyObject *Py_UNUSED(self), PyObject *args)
+{
+ PyObject *list;
+ Py_ssize_t loops;
+ if (!PyArg_ParseTuple(args, "On", &list, &loops)) {
+ return NULL;
+ }
+
+ Py_ssize_t len = PyList_GET_SIZE(list);
+
+ PyTime_t t1, t2;
+ (void)PyTime_PerfCounterRaw(&t1);
+ for (Py_ssize_t i=0; i < loops; i++) {
+ PyObject *result = PyList_New(len);
+ if (result == NULL) {
+ return NULL;
+ }
+
+ for (Py_ssize_t j=0; j < len; j++) {
+ PyObject *item = Py_NewRef(PyList_GET_ITEM(list, j));
+ PyList_SET_ITEM(result, j, item);
+ }
+
+ Py_DECREF(result);
+ }
+ (void)PyTime_PerfCounterRaw(&t2);
+
+ return PyFloat_FromDouble(PyTime_AsSecondsDouble(t2 - t1));
+}
+
+
+static PyObject*
+bench_setitem_func(PyObject *Py_UNUSED(self), PyObject *args)
+{
+ PyObject *list;
+ Py_ssize_t loops;
+ if (!PyArg_ParseTuple(args, "On", &list, &loops)) {
+ return NULL;
+ }
+
+ Py_ssize_t len = PyList_GET_SIZE(list);
+
+ PyTime_t t1, t2;
+ (void)PyTime_PerfCounterRaw(&t1);
+ for (Py_ssize_t i=0; i < loops; i++) {
+ PyObject *result = PyList_New(len);
+ if (result == NULL) {
+ return NULL;
+ }
+
+ for (Py_ssize_t j=0; j < len; j++) {
+ PyObject *item = Py_NewRef(PyList_GET_ITEM(list, j));
+ if (PyList_SetItem(result, j, item) < 0) {
+ Py_DECREF(result);
+ return NULL;
+ }
+ }
+
+ Py_DECREF(result);
+ }
+ (void)PyTime_PerfCounterRaw(&t2);
+
+ return PyFloat_FromDouble(PyTime_AsSecondsDouble(t2 - t1));
+}
+
+
+static PyObject*
+bench_writer(PyObject *Py_UNUSED(self), PyObject *args)
+{
+ PyObject *list;
+ Py_ssize_t loops;
+ if (!PyArg_ParseTuple(args, "On", &list, &loops)) {
+ return NULL;
+ }
+
+ Py_ssize_t len = PyList_GET_SIZE(list);
+
+ PyTime_t t1, t2;
+ (void)PyTime_PerfCounterRaw(&t1);
+ for (Py_ssize_t i=0; i < loops; i++) {
+ PyListWriter *writer = PyListWriter_Create(len);
+ if (writer == NULL) {
+ return NULL;
+ }
+
+ for (Py_ssize_t j=0; j < len; j++) {
+ PyObject *item = PyList_GET_ITEM(list, j);
+ if (PyListWriter_Append(writer, item) < 0) {
+ PyListWriter_Discard(writer);
+ return NULL;
+ }
+ }
+
+ PyObject *result = PyListWriter_Finish(writer);
+ if (result == NULL) {
+ return NULL;
+ }
+ Py_DECREF(result);
+ }
+ (void)PyTime_PerfCounterRaw(&t2);
+
+ return PyFloat_FromDouble(PyTime_AsSecondsDouble(t2 - t1));
+}
+
+
static PyMethodDef TestMethods[] = {
{"set_errno", set_errno, METH_VARARGS},
{"test_config", test_config, METH_NOARGS},
@@ -2988,6 +3094,9 @@ static PyMethodDef TestMethods[] = {
#ifdef HAVE_SYSCTLBYNAME
{"uptime_bsd", uptime_bsd, METH_NOARGS},
#endif
+ {"bench_setitem_macro", bench_setitem_macro, METH_VARARGS},
+ {"bench_setitem_func", bench_setitem_func, METH_VARARGS},
+ {"bench_writer", bench_writer, METH_VARARGS},
{NULL, NULL} /* sentinel */
};
|
Member
Author
|
I also ran benchmarks on
Well, it's not surprising that Benchmark script: Detailsimport pyperf, functools, _testcapi
runner = pyperf.Runner()
sizes = (0, 10, 100, 1_000)
for list_len in sizes:
test_list = list(range(list_len))
func = functools.partial(_testcapi.bench_setitem_macro, test_list)
runner.bench_time_func(f'setitem macro {list_len:,}-list', func)
for list_len in sizes:
test_list = list(range(list_len))
func = functools.partial(_testcapi.bench_writer_append, test_list)
runner.bench_time_func(f'writer append {list_len:,}-list', func)
for list_len in sizes:
test_list = list(range(list_len))
func = functools.partial(_testcapi.bench_writer_extend, test_list)
runner.bench_time_func(f'writer extend {list_len:,}-list', func)
for list_len in sizes:
test_list = list(range(list_len))
func = functools.partial(_testcapi.bench_writer_extend_array, test_list)
runner.bench_time_func(f'writer extend array {list_len:,}-list', func)Patch adding C benchmarks: Detailsdiff --git a/Modules/_testcapimodule.c b/Modules/_testcapimodule.c
index 033ee3234cf..c191d12c862 100644
--- a/Modules/_testcapimodule.c
+++ b/Modules/_testcapimodule.c
@@ -2889,6 +2889,185 @@ uptime_bsd(PyObject *Py_UNUSED(self), PyObject *Py_UNUSED(args))
#endif
+static PyObject*
+bench_setitem_macro(PyObject *Py_UNUSED(self), PyObject *args)
+{
+ PyObject *list;
+ Py_ssize_t loops;
+ if (!PyArg_ParseTuple(args, "On", &list, &loops)) {
+ return NULL;
+ }
+
+ Py_ssize_t len = PyList_GET_SIZE(list);
+
+ PyTime_t t1, t2;
+ (void)PyTime_PerfCounterRaw(&t1);
+ for (Py_ssize_t i=0; i < loops; i++) {
+ PyObject *result = PyList_New(len);
+ if (result == NULL) {
+ return NULL;
+ }
+
+ for (Py_ssize_t j=0; j < len; j++) {
+ PyObject *item = Py_NewRef(PyList_GET_ITEM(list, j));
+ PyList_SET_ITEM(result, j, item);
+ }
+
+ Py_DECREF(result);
+ }
+ (void)PyTime_PerfCounterRaw(&t2);
+
+ return PyFloat_FromDouble(PyTime_AsSecondsDouble(t2 - t1));
+}
+
+
+static PyObject*
+bench_setitem_func(PyObject *Py_UNUSED(self), PyObject *args)
+{
+ PyObject *list;
+ Py_ssize_t loops;
+ if (!PyArg_ParseTuple(args, "On", &list, &loops)) {
+ return NULL;
+ }
+
+ Py_ssize_t len = PyList_GET_SIZE(list);
+
+ PyTime_t t1, t2;
+ (void)PyTime_PerfCounterRaw(&t1);
+ for (Py_ssize_t i=0; i < loops; i++) {
+ PyObject *result = PyList_New(len);
+ if (result == NULL) {
+ return NULL;
+ }
+
+ for (Py_ssize_t j=0; j < len; j++) {
+ PyObject *item = Py_NewRef(PyList_GET_ITEM(list, j));
+ if (PyList_SetItem(result, j, item) < 0) {
+ Py_DECREF(result);
+ return NULL;
+ }
+ }
+
+ Py_DECREF(result);
+ }
+ (void)PyTime_PerfCounterRaw(&t2);
+
+ return PyFloat_FromDouble(PyTime_AsSecondsDouble(t2 - t1));
+}
+
+
+static PyObject*
+bench_writer_append(PyObject *Py_UNUSED(self), PyObject *args)
+{
+ PyObject *list;
+ Py_ssize_t loops;
+ if (!PyArg_ParseTuple(args, "On", &list, &loops)) {
+ return NULL;
+ }
+
+ Py_ssize_t len = PyList_GET_SIZE(list);
+
+ PyTime_t t1, t2;
+ (void)PyTime_PerfCounterRaw(&t1);
+ for (Py_ssize_t i=0; i < loops; i++) {
+ PyListWriter *writer = PyListWriter_Create(len);
+ if (writer == NULL) {
+ return NULL;
+ }
+
+ for (Py_ssize_t j=0; j < len; j++) {
+ PyObject *item = PyList_GET_ITEM(list, j);
+ if (PyListWriter_Append(writer, item) < 0) {
+ PyListWriter_Discard(writer);
+ return NULL;
+ }
+ }
+
+ PyObject *result = PyListWriter_Finish(writer);
+ if (result == NULL) {
+ return NULL;
+ }
+ Py_DECREF(result);
+ }
+ (void)PyTime_PerfCounterRaw(&t2);
+
+ return PyFloat_FromDouble(PyTime_AsSecondsDouble(t2 - t1));
+}
+
+
+static PyObject*
+bench_writer_extend(PyObject *Py_UNUSED(self), PyObject *args)
+{
+ PyObject *list;
+ Py_ssize_t loops;
+ if (!PyArg_ParseTuple(args, "On", &list, &loops)) {
+ return NULL;
+ }
+
+ Py_ssize_t len = PyList_GET_SIZE(list);
+
+ PyTime_t t1, t2;
+ (void)PyTime_PerfCounterRaw(&t1);
+ for (Py_ssize_t i=0; i < loops; i++) {
+ PyListWriter *writer = PyListWriter_Create(len);
+ if (writer == NULL) {
+ return NULL;
+ }
+
+ if (PyListWriter_Extend(writer, list) < 0) {
+ PyListWriter_Discard(writer);
+ return NULL;
+ }
+
+ PyObject *result = PyListWriter_Finish(writer);
+ if (result == NULL) {
+ return NULL;
+ }
+ Py_DECREF(result);
+ }
+ (void)PyTime_PerfCounterRaw(&t2);
+
+ return PyFloat_FromDouble(PyTime_AsSecondsDouble(t2 - t1));
+}
+
+
+static PyObject*
+bench_writer_extend_array(PyObject *Py_UNUSED(self), PyObject *args)
+{
+ PyObject *list;
+ Py_ssize_t loops;
+ if (!PyArg_ParseTuple(args, "On", &list, &loops)) {
+ return NULL;
+ }
+
+ PyObject **list_items = _PyList_CAST(list)->ob_item;
+ Py_ssize_t len = PyList_GET_SIZE(list);
+
+ PyTime_t t1, t2;
+ (void)PyTime_PerfCounterRaw(&t1);
+ for (Py_ssize_t i=0; i < loops; i++) {
+ PyListWriter *writer = PyListWriter_Create(len);
+ if (writer == NULL) {
+ return NULL;
+ }
+
+ if (PyListWriter_ExtendArray(writer, list_items, len) < 0) {
+ PyListWriter_Discard(writer);
+ return NULL;
+ }
+
+ PyObject *result = PyListWriter_Finish(writer);
+ if (result == NULL) {
+ return NULL;
+ }
+ Py_DECREF(result);
+ }
+ (void)PyTime_PerfCounterRaw(&t2);
+
+ return PyFloat_FromDouble(PyTime_AsSecondsDouble(t2 - t1));
+}
+
+
static PyMethodDef TestMethods[] = {
{"set_errno", set_errno, METH_VARARGS},
{"test_config", test_config, METH_NOARGS},
@@ -2988,6 +3167,11 @@ static PyMethodDef TestMethods[] = {
#ifdef HAVE_SYSCTLBYNAME
{"uptime_bsd", uptime_bsd, METH_NOARGS},
#endif
+ {"bench_setitem_macro", bench_setitem_macro, METH_VARARGS},
+ {"bench_setitem_func", bench_setitem_func, METH_VARARGS},
+ {"bench_writer_append", bench_writer_append, METH_VARARGS},
+ {"bench_writer_extend", bench_writer_extend, METH_VARARGS},
+ {"bench_writer_extend_array", bench_writer_extend_array, METH_VARARGS},
{NULL, NULL} /* sentinel */
};
|
pitrou
reviewed
Oct 22, 2026
| Extend the list writer with the contents of *items*. | ||
|
|
||
| On success, return ``0``. | ||
| On error, set an exception and return ``-1``. |
Member
There was a problem hiding this comment.
Should say what happens to the refcounts of items on error (also on success: does it steal references?).
Member
Author
There was a problem hiding this comment.
It behaves as PyTuple_FromArray(): new references are created. I will reuse PyTuple_FromArray() phrasing: "copy references from array to the new tuple".
Member
|
+1 on the principle, this looks like a very useful addition. |
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
2 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
PyList_SetItemis ~2.7x slower per item on the free-threaded build #158660