mn10300: add the MN10300/AM33 architecture to the kernel

[pandora-kernel.git] / arch / mn10300 / lib / memcpy.S

diff --git a/arch/mn10300/lib/memcpy.S b/arch/mn10300/lib/memcpy.S
new file mode 100644 (file)
index 0000000..25fb9bb
--- /dev/null
+++ b/arch/mn10300/lib/memcpy.S
@@ -0,0 +1,135 @@
+/* MN10300 Optimised simple memory to memory copy
+ *
+ * Copyright (C) 2007 Red Hat, Inc. All Rights Reserved.
+ * Written by David Howells (dhowells@redhat.com)
+ *
+ * This program is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU General Public Licence
+ * as published by the Free Software Foundation; either version
+ * 2 of the Licence, or (at your option) any later version.
+ */
+#include <asm/cache.h>
+
+        .section .text
+        .balign        L1_CACHE_BYTES
+
+###############################################################################
+#
+# void *memcpy(void *dst, const void *src, size_t n)
+#
+###############################################################################
+       .globl  memcpy
+        .type  memcpy,@function
+memcpy:
+       movm    [d2,d3],(sp)
+       mov     d0,(12,sp)
+       mov     d1,(16,sp)
+       mov     (20,sp),d2                      # count
+       mov     d0,a0                           # dst
+       mov     d1,a1                           # src
+       mov     d0,e3                           # the return value
+
+       cmp     +0,d2
+       beq     memcpy_done                     # return if zero-length copy
+
+       # see if the three parameters are all four-byte aligned
+       or      d0,d1,d3
+       or      d2,d3
+       and     +3,d3
+       bne     memcpy_1                        # jump if not
+
+       # we want to transfer as much as we can in chunks of 32 bytes
+       cmp     +31,d2
+       bls     memcpy_4_remainder              # 4-byte aligned remainder
+
+       movm    [exreg1],(sp)
+       add     -32,d2
+       mov     +32,d3
+
+memcpy_4_loop:
+       mov     (a1+),d0
+       mov     (a1+),d1
+       mov     (a1+),e0
+       mov     (a1+),e1
+       mov     (a1+),e4
+       mov     (a1+),e5
+       mov     (a1+),e6
+       mov     (a1+),e7
+       mov     d0,(a0+)
+       mov     d1,(a0+)
+       mov     e0,(a0+)
+       mov     e1,(a0+)
+       mov     e4,(a0+)
+       mov     e5,(a0+)
+       mov     e6,(a0+)
+       mov     e7,(a0+)
+
+       sub     d3,d2
+       bcc     memcpy_4_loop
+
+       movm    (sp),[exreg1]
+       add     d3,d2
+       beq     memcpy_4_no_remainder
+
+memcpy_4_remainder:
+       # cut 4-7 words down to 0-3
+       cmp     +16,d2
+       bcs     memcpy_4_three_or_fewer_words
+       mov     (a1+),d0
+       mov     (a1+),d1
+       mov     (a1+),e0
+       mov     (a1+),e1
+       mov     d0,(a0+)
+       mov     d1,(a0+)
+       mov     e0,(a0+)
+       mov     e1,(a0+)
+       add     -16,d2
+       beq     memcpy_4_no_remainder
+
+       # copy the remaining 1, 2 or 3 words
+memcpy_4_three_or_fewer_words:
+       cmp     +8,d2
+       bcs     memcpy_4_one_word
+       beq     memcpy_4_two_words
+       mov     (a1+),d0
+       mov     d0,(a0+)
+memcpy_4_two_words:
+       mov     (a1+),d0
+       mov     d0,(a0+)
+memcpy_4_one_word:
+       mov     (a1+),d0
+       mov     d0,(a0+)
+
+memcpy_4_no_remainder:
+       # check we copied the correct amount
+       # TODO: REMOVE CHECK
+       sub     e3,a0,d2
+       mov     (20,sp),d1
+       cmp     d2,d1
+       beq     memcpy_done
+       break
+       break
+       break
+
+memcpy_done:
+       mov     e3,a0
+       ret     [d2,d3],8
+
+       # handle misaligned copying
+memcpy_1:
+       add     -1,d2
+       mov     +1,d3
+       setlb                                   # setlb requires the next insns
+                                               # to occupy exactly 4 bytes
+
+       sub     d3,d2
+       movbu   (a1),d0
+       movbu   d0,(a0)
+       add_add d3,a1,d3,a0
+       lcc
+
+       mov     e3,a0
+       ret     [d2,d3],8
+
+memcpy_end:
+       .size   memcpy, memcpy_end-memcpy