summaryrefslogtreecommitdiffstats
path: root/arch/sh/lib64/clear_page.S
diff options
context:
space:
mode:
authorPaul Mundt <lethal@linux-sh.org>2007-11-21 22:46:14 +0900
committerPaul Mundt <lethal@linux-sh.org>2008-01-28 13:18:54 +0900
commit9895f9429cb489ba271c06767531083ae4c4bcbe (patch)
tree5f6c410bf94021e9f1f309ad624f0add23eb92fa /arch/sh/lib64/clear_page.S
parentefddb4c882a64b7ca44b2a089d20020b30a2c6dd (diff)
downloadkernel-crypto-9895f9429cb489ba271c06767531083ae4c4bcbe.tar.gz
kernel-crypto-9895f9429cb489ba271c06767531083ae4c4bcbe.tar.xz
kernel-crypto-9895f9429cb489ba271c06767531083ae4c4bcbe.zip
sh: clear/copy_page renames in lib and lib64.
Signed-off-by: Paul Mundt <lethal@linux-sh.org>
Diffstat (limited to 'arch/sh/lib64/clear_page.S')
-rw-r--r--arch/sh/lib64/clear_page.S54
1 files changed, 54 insertions, 0 deletions
diff --git a/arch/sh/lib64/clear_page.S b/arch/sh/lib64/clear_page.S
new file mode 100644
index 00000000000..007ab48ecc1
--- /dev/null
+++ b/arch/sh/lib64/clear_page.S
@@ -0,0 +1,54 @@
+/*
+ Copyright 2003 Richard Curnow, SuperH (UK) Ltd.
+
+ This file is subject to the terms and conditions of the GNU General Public
+ License. See the file "COPYING" in the main directory of this archive
+ for more details.
+
+ Tight version of memset for the case of just clearing a page. It turns out
+ that having the alloco's spaced out slightly due to the increment/branch
+ pair causes them to contend less for access to the cache. Similarly,
+ keeping the stores apart from the allocos causes less contention. => Do two
+ separate loops. Do multiple stores per loop to amortise the
+ increment/branch cost a little.
+
+ Parameters:
+ r2 : source effective address (start of page)
+
+ Always clears 4096 bytes.
+
+ Note : alloco guarded by synco to avoid TAKum03020 erratum
+
+*/
+
+ .section .text..SHmedia32,"ax"
+ .little
+
+ .balign 8
+ .global clear_page
+clear_page:
+ pta/l 1f, tr1
+ pta/l 2f, tr2
+ ptabs/l r18, tr0
+
+ movi 4096, r7
+ add r2, r7, r7
+ add r2, r63, r6
+1:
+ alloco r6, 0
+ synco ! TAKum03020
+ addi r6, 32, r6
+ bgt/l r7, r6, tr1
+
+ add r2, r63, r6
+2:
+ st.q r6, 0, r63
+ st.q r6, 8, r63
+ st.q r6, 16, r63
+ st.q r6, 24, r63
+ addi r6, 32, r6
+ bgt/l r7, r6, tr2
+
+ blink tr0, r63
+
+