From: andrea@suse.de
Subject: low-latency stuff
Patch-mainline: not yet


My point is that preempt and no-preempt should do the same thing there,
otherwise when you benchmark -preempt, you'll get better latency,
but not because of the preempt feature, but just because of unrelated
latency improvements that have nothing to do with preempt.


---
 mm/memory.c |    8 ++++----
 1 file changed, 4 insertions(+), 4 deletions(-)

--- a/mm/memory.c
+++ b/mm/memory.c
@@ -1060,11 +1060,11 @@ static unsigned long unmap_page_range(st
 	return addr;
 }
 
-#ifdef CONFIG_PREEMPT
-# define ZAP_BLOCK_SIZE	(8 * PAGE_SIZE)
+#ifdef CONFIG_SMP
+/* zap one pte page at a time */
+#define ZAP_BLOCK_SIZE		(FREE_PTE_NR * PAGE_SIZE)
 #else
-/* No preempt: go for improved straight-line efficiency */
-# define ZAP_BLOCK_SIZE	(1024 * PAGE_SIZE)
+#define ZAP_BLOCK_SIZE		(253 * PAGE_SIZE)
 #endif
 
 /**