svn commit: r322769 - head/sys/arm64/arm64

Andrew Turner andrew at FreeBSD.org
Mon Aug 21 18:12:33 UTC 2017


Author: andrew
Date: Mon Aug 21 18:12:32 2017
New Revision: 322769
URL: https://svnweb.freebsd.org/changeset/base/322769

Log:
  Improve the performance of the arm64 thread switching code.
  
  The full system memory barrier around a TLB invalidation is stricter than
  required. It needs to wait on accesses to main memory, with just the weaker
  store variant before the invalidate. As such use the dsb istst, tlbi, dlb
  ish sequence already used in pmap.
  
  The tlbi instruction in this sequence is also unnecessarily using a
  broadcast invalidate when it just needs to invalidate the local CPUs TLB.
  Switch to a non-broadcast variant of this instruction.
  
  Sponsored by:	DARPA, AFRL

Modified:
  head/sys/arm64/arm64/swtch.S

Modified: head/sys/arm64/arm64/swtch.S
==============================================================================
--- head/sys/arm64/arm64/swtch.S	Mon Aug 21 18:00:26 2017	(r322768)
+++ head/sys/arm64/arm64/swtch.S	Mon Aug 21 18:12:32 2017	(r322769)
@@ -91,9 +91,9 @@ ENTRY(cpu_throw)
 	isb
 
 	/* Invalidate the TLB */
-	dsb	sy
-	tlbi	vmalle1is
-	dsb	sy
+	dsb	ishst
+	tlbi	vmalle1
+	dsb	ish
 	isb
 
 	/* If we are single stepping, enable it */
@@ -192,9 +192,9 @@ ENTRY(cpu_switch)
 	isb
 
 	/* Invalidate the TLB */
-	dsb	sy
-	tlbi	vmalle1is
-	dsb	sy
+	dsb	ishst
+	tlbi	vmalle1
+	dsb	ish
 	isb
 
 	/*


More information about the svn-src-all mailing list