pythongh-129201: Use prefetch in GC mark alive phase. (pythongh-129203)

For the free-threaded version of the cyclic GC, restructure the "mark alive" phase to use software prefetch instructions. This gives a speedup in most cases when the number of objects is large enough. The prefetching is enabled conditionally based on the number of long-lived objects the GC finds.
CraZySacX · Feb 5, 2025 · cdcacec · cdcacec
1 parent 5fb019f
commit cdcacec
Show file tree

Hide file tree

Showing 2 changed files with 435 additions and 42 deletions.
diff --git a/Misc/NEWS.d/next/Core_and_Builtins/2025-01-22-14-22-34.gh-issue-129201.wiZzEb.rst b/Misc/NEWS.d/next/Core_and_Builtins/2025-01-22-14-22-34.gh-issue-129201.wiZzEb.rst
@@ -0,0 +1,5 @@
+The free-threaded version of the cyclic garbage collector has been optimized to
+conditionally use CPU prefetch instructions during the collection.  This can
+reduce collection times by making it more likely that data is in the CPU cache
+when it is needed.  The prefetch instructions are enabled if the number of
+long-lived objects (objects surviving a full collection) exceeds a threshold.