Avoid atomic exchange in rb_free_tmp_buffer

Usually RB_ALLOCV_N uses alloca for small allocations, and in that case
the value is 0, and we should not need to atomic exchange it back to 0.

I'm actually not sure why we need atomic operations here anyways.

1 files changed, 1 insertions, 0 deletions

diff --git a/imemo.c b/imemo.c
index ce27ea5601..fc557a3144 100644
--- a/imemo.c
+++ b/imemo.c
@@ -89,6 +89,7 @@ rb_alloc_tmp_buffer_with_count(volatile VALUE *store, size_t size, size_t cnt)
 void
 rb_free_tmp_buffer(volatile VALUE *store)
 {
+    if (!*store) return;
     rb_imemo_tmpbuf_t *s = (rb_imemo_tmpbuf_t*)ATOMIC_VALUE_EXCHANGE(*store, 0);
     if (s) {
         void *ptr = ATOMIC_PTR_EXCHANGE(s->ptr, 0);