From 88ac32e3bb96d29553ee04923156d5fa4baa8c92 Mon Sep 17 00:00:00 2001
From: Sebastian Sumpf <sebastian.sumpf@genode-labs.com>
Date: Thu, 2 Aug 2012 16:39:16 +0200
Subject: [PATCH] Base: New 'memcpy' implementation

Added CPU specific memcpy function ('memcpy_cpu'), which is tried first in
default 'memcpy'. Improved default 'memcpy' to copy eight byte chunks.
---
 base/include/arm/cpu/string.h | 56 +++++++++++++++++++++++++++++++++++
 base/include/util/string.h    | 24 ++++++++++++---
 base/include/x86/cpu/string.h | 32 ++++++++++++++++++++
 3 files changed, 108 insertions(+), 4 deletions(-)
 create mode 100644 base/include/arm/cpu/string.h
 create mode 100644 base/include/x86/cpu/string.h

diff --git a/base/include/arm/cpu/string.h b/base/include/arm/cpu/string.h
new file mode 100644
index 0000000000..7b37d8d122
--- /dev/null
+++ b/base/include/arm/cpu/string.h
@@ -0,0 +1,56 @@
+/*
+ * \brief  Cpu specifi memcpy
+ * \author Sebastian Sumpf
+ * \author Stefan Kalkowski
+ * \date   2012-08-02
+ */
+
+/*
+ * Copyright (C) 2012 Genode Labs GmbH
+ *
+ * This file is part of the Genode OS framework, which is distributed
+ * under the terms of the GNU General Public License version 2.
+ */
+
+#ifndef _INCLUDE__ARM__CPU__STRING_H_
+#define _INCLUDE__ARM__CPU__STRING_H_
+
+namespace Genode
+{
+	/**
+	 * Copy memory block
+	 *
+	 * \param dst   destination memory block
+	 * \param src   source memory block
+	 * \param size  number of bytes to copy
+	 *
+	 * \return      Number of bytes not copied
+	 */
+	inline size_t memcpy_cpu(void *dst, const void *src, size_t size)
+	{
+		unsigned char *d = (unsigned char *)dst, *s = (unsigned char *)src;
+
+		/* check 4 byte; alignment */
+		size_t d_align = (size_t)d & 0x3;
+		size_t s_align = (size_t)s & 0x3;
+
+		/* at least 32 bytes, 4 byte aligned, same alignment */
+		if (size < 32 || (d_align | s_align) || (d_align ^ s_align))
+			return size;
+
+		/* copy to 4 byte alignment */
+		for (size_t i = 0; i < s_align; i++, *d++ = *s++, size--);
+
+		/* copy 32 byte chunks */
+		for (; size >= 32; size -= 32) {
+			asm volatile ("ldmia %0!, {r3 - r10} \n\t"
+			              "stmia %1!, {r3 - r10} \n\t"
+			              : "+r" (s), "+r" (d)
+			              :: "r3","r4","r5","r6","r7","r8","r9","r10");
+		}
+	
+		return size;
+	}
+}
+
+#endif /* _INCLUDE__ARM__CPU__STRING_H_ */
diff --git a/base/include/util/string.h b/base/include/util/string.h
index a96a32924c..b43b5f19e3 100644
--- a/base/include/util/string.h
+++ b/base/include/util/string.h
@@ -1,6 +1,7 @@
 /*
  * \brief  String utility functions
  * \author Norman Feske
+ * \author Sebastian Sumpf
  * \date   2006-05-10
  */
 
@@ -16,6 +17,7 @@
 
 #include <base/stdint.h>
 #include <util/misc_math.h>
+#include <cpu/string.h>
 
 namespace Genode {
 
@@ -61,10 +63,24 @@ namespace Genode {
 		char *d = (char *)dst, *s = (char *)src;
 		size_t i;
 
-		if (s > d)
-			for (i = 0; i < size; i++, *d++ = *s++);
-		else
-			for (d += size, s += size, i = size; i-- > 0; *(--d) = *(--s));
+		/* try cpu specific version first */
+		if ((i = size - memcpy_cpu(dst, src, size)) == size)
+			return dst;
+
+		d += i; s += i; size -= i;
+
+		/* copy eight byte chunks */
+		for (i = size >> 3; i > 0; i--, *d++ = *s++,
+		                                *d++ = *s++,
+		                                *d++ = *s++,
+		                                *d++ = *s++,
+		                                *d++ = *s++,
+		                                *d++ = *s++,
+		                                *d++ = *s++,
+		                                *d++ = *s++);
+
+		/* copy left over */
+		for (i = 0; i < (size & 0x7); i++, *d++ = *s++);
 
 		return dst;
 	}
diff --git a/base/include/x86/cpu/string.h b/base/include/x86/cpu/string.h
new file mode 100644
index 0000000000..a25e7b096d
--- /dev/null
+++ b/base/include/x86/cpu/string.h
@@ -0,0 +1,32 @@
+/*
+ * \brief  Cpu specifi memcpy
+ * \author Sebastian Sumpf
+ * \date   2012-08-02
+ */
+
+/*
+ * Copyright (C) 2012 Genode Labs GmbH
+ *
+ * This file is part of the Genode OS framework, which is distributed
+ * under the terms of the GNU General Public License version 2.
+ */
+
+#ifndef _INCLUDE__X86__CPU__STRING_H_
+#define _INCLUDE__X86__CPU__STRING_H_
+
+namespace Genode
+{
+	/**
+	 * Copy memory block
+	 *
+	 * \param dst   destination memory block
+	 * \param src   source memory block
+	 * \param size  number of bytes to copy
+	 *
+	 * \return      Number of bytes not copied
+	 */
+	inline size_t memcpy_cpu(void *dst, const void *src, size_t size) {
+		return size; }
+}
+
+#endif /* _INCLUDE__X86__CPU__STRING_H_ */