gh-128213: fast path for bytes creation from list and tuple

Lib/test/test_free_threading/test_bytes_object.py

@@ -0,0 +1,37 @@
+import unittest
+import sys
+from threading import Thread, Barrier
+from test.support import threading_helper
+
+threading_helper.requires_working_threading(module=True)
+
+class BytesThreading(unittest.TestCase):
+
+    @threading_helper.reap_threads
+    def test_conversion_from_list(self):
+        number_of_threads = 10
+        number_of_iterations = 10
+        barrier = Barrier(number_of_threads)
+
+        x = [1, 2, 3, 4, 5]
+        e = [ (ii,)*(2+4*ii) for ii in range(number_of_threads)] # range of sizes to extend
+        def work(ii):
+            barrier.wait()
+            for _ in range(1000):
+                bytes(x)
+                x.extend(e[ii])
+                if len(x) > 10:
+                    x[:] = [0]
+
+        for it in range(number_of_iterations):
+            worker_threads = []
+            for ii in range(number_of_threads):
+                worker_threads.append(
+                    Thread(target=work, args=[ii]))
+            with threading_helper.start_threads(worker_threads):
+                pass
+
+            barrier.reset()
+
+if __name__ == "__main__":
+    unittest.main()

Misc/NEWS.d/next/Core_and_Builtins/2024-12-24-08-44-49.gh-issue-128213.Y71jDi.rst

@@ -0,0 +1,3 @@
+Speed up :class:`bytes` creation from :class:`list` and :class:`tuple` of integers. Benchmarks show that from a list with 1000000 random numbers the time to create a bytes object is reduced by around 31%, or 30% with 10000 numbers, or 27% with 100 numbers.
+
+Patch by Ben Hsing

Objects/bytesobject.c

@@ -6,6 +6,7 @@
 #include "pycore_bytesobject.h"   // _PyBytes_Find(), _PyBytes_Repeat()
 #include "pycore_call.h"          // _PyObject_CallNoArgs()
 #include "pycore_ceval.h"         // _PyEval_GetBuiltin()
+#include "pycore_critical_section.h" // Py_BEGIN_CRITICAL_SECTION_SEQUENCE_FAST()
 #include "pycore_format.h"        // F_LJUST
 #include "pycore_global_objects.h"// _Py_GET_GLOBAL_OBJECT()
 #include "pycore_initconfig.h"    // _PyStatus_OK()
@@ -2854,82 +2855,46 @@ _PyBytes_FromBuffer(PyObject *x)
 }
 
 static PyObject*
-_PyBytes_FromList(PyObject *x)
+_PyBytes_FromSequence(PyObject *x)
 {
-    Py_ssize_t i, size = PyList_GET_SIZE(x);
-    Py_ssize_t value;
-    char *str;
-    PyObject *item;
-    _PyBytesWriter writer;
-
-    _PyBytesWriter_Init(&writer);
-    str = _PyBytesWriter_Alloc(&writer, size);
-    if (str == NULL)
+    Py_ssize_t size = PySequence_Fast_GET_SIZE(x);
+    PyObject *bytes = _PyBytes_FromSize(size, 0);
+    if (bytes == NULL) {
         return NULL;
-    writer.overallocate = 1;
-    size = writer.allocated;
-
-    for (i = 0; i < PyList_GET_SIZE(x); i++) {
-        item = PyList_GET_ITEM(x, i);
-        Py_INCREF(item);
-        value = PyNumber_AsSsize_t(item, NULL);
-        Py_DECREF(item);
-        if (value == -1 && PyErr_Occurred())
-            goto error;
-
-        if (value < 0 || value >= 256) {
-            PyErr_SetString(PyExc_ValueError,
-                            "bytes must be in range(0, 256)");
-            goto error;
-        }
-
-        if (i >= size) {
-            str = _PyBytesWriter_Resize(&writer, str, size+1);
-            if (str == NULL)
-                return NULL;
-            size = writer.allocated;
-        }
-        *str++ = (char) value;
     }
-    return _PyBytesWriter_Finish(&writer, str);
-
-  error:
-    _PyBytesWriter_Dealloc(&writer);
-    return NULL;
-}
-
-static PyObject*
-_PyBytes_FromTuple(PyObject *x)
-{
-    PyObject *bytes;
-    Py_ssize_t i, size = PyTuple_GET_SIZE(x);
-    Py_ssize_t value;
-    char *str;
-    PyObject *item;
-
-    bytes = PyBytes_FromStringAndSize(NULL, size);
-    if (bytes == NULL)
-        return NULL;
-    str = ((PyBytesObject *)bytes)->ob_sval;
-
-    for (i = 0; i < size; i++) {
-        item = PyTuple_GET_ITEM(x, i);
-        value = PyNumber_AsSsize_t(item, NULL);
-        if (value == -1 && PyErr_Occurred())
+    char *str = PyBytes_AS_STRING(bytes);
+    PyObject *const *items = PySequence_Fast_ITEMS(x);
+    Py_BEGIN_CRITICAL_SECTION_SEQUENCE_FAST(x);
+    for (Py_ssize_t i = 0; i < size; i++) {
+        if (!PyLong_Check(items[i])) {
+            Py_DECREF(bytes);
+            /* Py_None as a fallback sentinel to the slow path */
+            bytes = Py_None;
+            goto done;
+        }
+        Py_ssize_t value = PyNumber_AsSsize_t(items[i], NULL);
+        if (value == -1 && PyErr_Occurred()) {
             goto error;
-
+        }
         if (value < 0 || value >= 256) {
             PyErr_SetString(PyExc_ValueError,
                             "bytes must be in range(0, 256)");
             goto error;
         }
         *str++ = (char) value;
     }
-    return bytes;
-
+    goto done;
   error:
     Py_DECREF(bytes);
-    return NULL;
+    bytes = NULL;
+  done:
+    /* some C parsers require a label not to be at the end of a compound
+       statement, which the ending macro of a critical section introduces, so
+       we need an empty statement here to satisfy that syntax rule */
+    ;
+    /* both success and failure need to end the critical section */
+    Py_END_CRITICAL_SECTION_SEQUENCE_FAST();
+    return bytes;
 }
 
 static PyObject *
@@ -3012,11 +2977,13 @@ PyBytes_FromObject(PyObject *x)
     if (PyObject_CheckBuffer(x))
         return _PyBytes_FromBuffer(x);
 
-    if (PyList_CheckExact(x))
-        return _PyBytes_FromList(x);
-
-    if (PyTuple_CheckExact(x))
-        return _PyBytes_FromTuple(x);
+    if (PyList_CheckExact(x) || PyTuple_CheckExact(x)) {
+        PyObject *bytes = _PyBytes_FromSequence(x);
+        /* Py_None as a fallback sentinel to the slow path */
+        if (bytes != Py_None) {
+            return bytes;
+        }
+    }
 
     if (!PyUnicode_Check(x)) {
         it = PyObject_GetIter(x);