Assembler optimised memset16() for ARM, by Thom Johansen. Should speed up LCD clearing and solid rectangle drawing on colour iPods somewhat.

git-svn-id: svn://svn.rockbox.org/rockbox/trunk@10900 a1c6a512-1295-4272-9138-f99709370657

4 files changed, 83 insertions, 7 deletions

diff --git a/firmware/SOURCES b/firmware/SOURCES
index 9001535150..545227fbdf 100644
--- a/firmware/SOURCES
+++ b/firmware/SOURCES
@@ -53,7 +53,7 @@ target/sh/memset-sh.S
 common/memcpy.c
 common/memmove.c
 target/arm/memset-arm.S
-common/memset16.c
+target/arm/memset16-arm.S
 #else
 common/memcpy.c
 common/memmove.c
diff --git a/firmware/common/memset16.c b/firmware/common/memset16.c
index bc187a5d25..5f0fc3fe08 100755
--- a/firmware/common/memset16.c
+++ b/firmware/common/memset16.c
@@ -22,15 +22,13 @@
 #define UNALIGNED(X)   ((long)X & (sizeof(long) - 1))
 #define TOO_SMALL(LEN) ((LEN) < LBLOCKSIZE)
 
-void *memset16(void *dst, int val, size_t len)
+void memset16(void *dst, int val, size_t len)
 {
 #if defined(PREFER_SIZE_OVER_SPEED) || defined(__OPTIMIZE_SIZE__)
     unsigned short *p = (unsigned short *)dst;
 
     while (len--)
         *p++ = val;
-
-    return dst;
 #else
     unsigned short *p = (unsigned short *)dst;
     unsigned int i;
@@ -73,7 +71,5 @@ void *memset16(void *dst, int val, size_t len)
 
     while (len--)
         *p++ = val;
-
-    return dst;
 #endif /* not PREFER_SIZE_OVER_SPEED */
 }
diff --git a/firmware/include/memory.h b/firmware/include/memory.h
index 2b2a60c7b4..559c6ed96a 100755
--- a/firmware/include/memory.h
+++ b/firmware/include/memory.h
@@ -22,6 +22,6 @@
 
 #include <sys/types.h>
 
-void *memset16(void *dst, int val, size_t len);
+void memset16(void *dst, int val, size_t len);
 
 #endif /* _MEMORY_H_ */
diff --git a/firmware/target/arm/memset16-arm.S b/firmware/target/arm/memset16-arm.S
new file mode 100755
index 0000000000..13213c57f0
--- /dev/null
+++ b/firmware/target/arm/memset16-arm.S
@@ -0,0 +1,80 @@
+/***************************************************************************
+ *             __________               __   ___.
+ *   Open      \______   \ ____   ____ |  | _\_ |__   _______  ___
+ *   Source     |       _//  _ \_/ ___\|  |/ /| __ \ /  _ \  \/  /
+ *   Jukebox    |    |   (  <_> )  \___|    < | \_\ (  <_> > <  <
+ *   Firmware   |____|_  /\____/ \___  >__|_ \|___  /\____/__/\_ \
+ *                     \/            \/     \/    \/            \/
+ * $Id$
+ *
+ * Copyright (C) 2006 by Thom Johansen
+ *
+ * All files in this archive are subject to the GNU General Public License.
+ * See the file COPYING in the source tree root for full license agreement.
+ *
+ * This software is distributed on an "AS IS" basis, WITHOUT WARRANTY OF ANY
+ * KIND, either express or implied.
+ *
+ ****************************************************************************/
+#include "config.h"
+
+    .section    .icode,"ax",%progbits
+
+    .align      2
+
+/*  The following code is based on code from the Linux kernel version 2.6.15.3,
+ *  linux/arch/arm/lib/memset.S
+ *
+ *  Copyright (C) 1995-2000 Russell King
+ */
+
+        .global     memset16
+        .type       memset16,%function
+memset16:
+        tst     r0, #2                  @ unaligned?
+        cmpne   r2, #0
+        strneh  r1, [r0], #2            @ store one halfword to align
+        subne   r2, r2, #1
+
+/*
+ * we know that the pointer in r0 is aligned to a word boundary.
+ */
+        orr     r1, r1, r1, lsl #16
+        mov     r3, r1
+        cmp     r2, #8
+        blt     4f
+/*
+ * We need an extra register for this loop - save the return address and
+ * use the LR
+ */
+        str     lr, [sp, #-4]!
+        mov     ip, r1
+        mov     lr, r1
+
+2:      subs    r2, r2, #32
+        stmgeia r0!, {r1, r3, ip, lr}   @ 64 bytes at a time.
+        stmgeia r0!, {r1, r3, ip, lr}
+        stmgeia r0!, {r1, r3, ip, lr}
+        stmgeia r0!, {r1, r3, ip, lr}
+        bgt     2b
+        ldmeqfd sp!, {pc}               @ Now <64 bytes to go.
+/*
+ * No need to correct the count; we're only testing bits from now on
+ */
+        tst     r2, #16
+        stmneia r0!, {r1, r3, ip, lr}
+        stmneia r0!, {r1, r3, ip, lr}
+        tst     r2, #8
+        stmneia r0!, {r1, r3, ip, lr}
+        ldr     lr, [sp], #4
+
+4:      tst     r2, #4
+        stmneia r0!, {r1, r3}
+        tst     r2, #2
+        strne   r1, [r0], #4
+
+        tst     r2, #1
+        strneh  r1, [r0], #2
+        bx      lr
+.end:
+        .size   memset16,.end-memset16