bpo-46504: faster code for trial quotient in x_divrem() (GH-30856) · python/cpython@7c26472 · GitHub

python / cpython Public

• Notifications You must be signed in to change notification settings
• Fork 35.3k
• Star 75k

• Code
• Issues 5k+
• Pull requests 2.6k
• Actions
• Projects
• Security and quality
• Insights

Additional navigation options

Commit 7c26472

Browse filesBrowse the repository at this point in the historyBrowse files

tim-one

authored

bpo-46504: faster code for trial quotient in x_divrem() (GH-30856)

* bpo-46504: faster code for trial quotient in x_divrem() This brings x_divrem() back into synch with x_divrem1(), which was changed in bpo-46406 to generate faster code to find machine-word division quotients and remainders. Modern processors compute both with a single machine instruction, but convincing C to exploit that requires writing _less_ "clever" C code.

1 parent b18fd54 commit 7c26472
Copy full SHA for 7c26472

1 file changed

+8-1Lines changed: 8 additions & 1 deletion

File tree

Expand file treeCollapse file tree

Open diff view settings

Filter options

• Objects
  • longobject.c

Expand file treeCollapse file tree

Open diff view settings

Collapse file

‎Objects/longobject.c‎

Copy file name to clipboardExpand all lines: Objects/longobject.c

+8-1Lines changed: 8 additions & 1 deletion

Original file line number	Diff line number	Diff line change
@@ -2767,8 +2767,15 @@ x_divrem(PyLongObject *v1, PyLongObject *w1, PyLongObject **prem)
2767	2767	vtop = vk[size_w];
2768	2768	assert(vtop <= wm1);
2769	2769	vv = ((twodigits)vtop << PyLong_SHIFT) | vk[size_w-1];
	2770	+ /* The code used to compute the remainder via
	2771	+ * r = (digit)(vv - (twodigits)wm1 * q);
	2772	+ * and compilers generally generated code to do the * and -.
	2773	+ * But modern processors generally compute q and r with a single
	2774	+ * instruction, and modern optimizing compilers exploit that if we
	2775	+ * _don't_ try to optimize it.
	2776	+ */
2770	2777	q = (digit)(vv / wm1);
2771		- r = (digit)(vv - (twodigits)wm1 * q); /* r = vv % wm1 */
	2778	+ r = (digit)(vv % wm1);
2772	2779	while ((twodigits)wm2 * q > (((twodigits)r << PyLong_SHIFT)
2773	2780	| vk[size_w-2])) {
2774	2781	--q;

0 commit comments

Comments

0 (0)