Re: [PATCH] powerpc: Optimise the 64bit optimised __clear_user
flat view
From: Olof Johansson <hidden>
Date: 2012-06-04 13:12:35
Hi, On Mon, Jun 4, 2012 at 12:58 AM, Anton Blanchard [off-list ref] wrote:
I blame Mikey for this. He elevated my slightly dubious testcase: # dd if=3D/dev/zero of=3D/dev/null bs=3D1M count=3D10000 to benchmark status. And naturally we need to be number 1 at creating zeros. So lets improve __clear_user some more. As Paul suggests we can use dcbz for large lengths. This patch gets the destination 128 byte aligned then uses dcbz on whole cachelines. Before: 10485760000 bytes (10 GB) copied, 0.414744 s, 25.3 GB/s After: 10485760000 bytes (10 GB) copied, 0.268597 s, 39.0 GB/s 39 GB/s, a new record. Signed-off-by: Anton Blanchard <redacted> --- Index: linux-build/arch/powerpc/lib/string_64.S =3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=
=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D= =3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D
quoted hunk ↗ jump to hunk
--- linux-build.orig/arch/powerpc/lib/string_64.S =A0 =A0 =A0 2012-06-04 =
16:18:56.351604302 +1000
quoted hunk ↗ jump to hunk
+++ linux-build/arch/powerpc/lib/string_64.S =A0 =A02012-06-04 16:47:10.5=
38500871 +1000
quoted hunk ↗ jump to hunk
@@ -78,7 +78,7 @@ _GLOBAL(__clear_user)
[..]
quoted hunk ↗ jump to hunk
+15: +err2; =A0dcbz =A0 =A0r0,r3 + =A0 =A0 =A0 addi =A0 =A0r3,r3,128 + =A0 =A0 =A0 addi =A0 =A0r4,r4,-128 + =A0 =A0 =A0 bdnz =A0 =A015b
This breaks architecture spec (and at least one implementation); cache
lines are not guaranteed to be 128 bytes.
-Olof