| /* $Id: memmove.S,v 1.2 2001/07/27 11:51:09 gniibe Exp $ |
| * |
| * "memmove" implementation of SuperH |
| * |
| * Copyright (C) 1999 Niibe Yutaka |
| * |
| */ |
| |
| /* |
| * void *memmove(void *dst, const void *src, size_t n); |
| * The memory areas may overlap. |
| */ |
| |
| #include <linux/linkage.h> |
| ENTRY(memmove) |
| ! if dest > src, call memcpy (it copies in decreasing order) |
| cmp/hi r5,r4 |
| bf 1f |
| mov.l 2f,r0 |
| jmp @r0 |
| nop |
| .balign 4 |
| 2: .long memcpy |
| 1: |
| sub r5,r4 ! From here, r4 has the distance to r0 |
| tst r6,r6 |
| bt/s 9f ! if n=0, do nothing |
| mov r5,r0 |
| add r6,r5 |
| mov #12,r1 |
| cmp/gt r6,r1 |
| bt/s 8f ! if it's too small, copy a byte at once |
| add #-1,r4 |
| add #1,r4 |
| ! |
| ! [ ... ] DST [ ... ] SRC |
| ! [ ... ] [ ... ] |
| ! : : |
| ! r0+r4--> [ ... ] r0 --> [ ... ] |
| ! : : |
| ! [ ... ] [ ... ] |
| ! r5 --> |
| ! |
| mov r4,r1 |
| mov #3,r2 |
| and r2,r1 |
| shll2 r1 |
| mov r0,r3 ! Save the value on R0 to R3 |
| mova jmptable,r0 |
| add r1,r0 |
| mov.l @r0,r1 |
| jmp @r1 |
| mov r3,r0 ! and back to R0 |
| .balign 4 |
| jmptable: |
| .long case0 |
| .long case1 |
| .long case2 |
| .long case3 |
| |
| ! copy a byte at once |
| 8: mov.b @r0+,r1 |
| cmp/hs r5,r0 |
| bf/s 8b ! while (r0<r5) |
| mov.b r1,@(r0,r4) |
| add #1,r4 |
| 9: |
| add r4,r0 |
| rts |
| sub r6,r0 |
| |
| case_none: |
| bra 8b |
| add #-1,r4 |
| |
| case0: |
| ! |
| ! GHIJ KLMN OPQR --> GHIJ KLMN OPQR |
| ! |
| ! First, align to long word boundary |
| mov r0,r3 |
| and r2,r3 |
| tst r3,r3 |
| bt/s 2f |
| add #-1,r4 |
| mov #4,r2 |
| sub r3,r2 |
| 1: dt r2 |
| mov.b @r0+,r1 |
| bf/s 1b |
| mov.b r1,@(r0,r4) |
| ! |
| 2: ! Second, copy a long word at once |
| add #-3,r4 |
| add #-3,r5 |
| 3: mov.l @r0+,r1 |
| cmp/hs r5,r0 |
| bf/s 3b |
| mov.l r1,@(r0,r4) |
| add #3,r5 |
| ! |
| ! Third, copy a byte at once, if necessary |
| cmp/eq r5,r0 |
| bt/s 9b |
| add #4,r4 |
| bra 8b |
| add #-1,r4 |
| |
| case3: |
| ! |
| ! GHIJ KLMN OPQR --> ...G HIJK LMNO PQR. |
| ! |
| ! First, align to long word boundary |
| mov r0,r3 |
| and r2,r3 |
| tst r3,r3 |
| bt/s 2f |
| add #-1,r4 |
| mov #4,r2 |
| sub r3,r2 |
| 1: dt r2 |
| mov.b @r0+,r1 |
| bf/s 1b |
| mov.b r1,@(r0,r4) |
| ! |
| 2: ! Second, read a long word and write a long word at once |
| add #-2,r4 |
| mov.l @(r0,r4),r1 |
| add #-7,r5 |
| add #-4,r4 |
| ! |
| #ifdef __LITTLE_ENDIAN__ |
| shll8 r1 |
| 3: mov r1,r3 ! JIHG |
| shlr8 r3 ! xJIH |
| mov.l @r0+,r1 ! NMLK |
| mov r1,r2 |
| shll16 r2 |
| shll8 r2 ! Kxxx |
| or r2,r3 ! KJIH |
| cmp/hs r5,r0 |
| bf/s 3b |
| mov.l r3,@(r0,r4) |
| #else |
| shlr8 r1 |
| 3: mov r1,r3 ! GHIJ |
| shll8 r3 ! HIJx |
| mov.l @r0+,r1 ! KLMN |
| mov r1,r2 |
| shlr16 r2 |
| shlr8 r2 ! xxxK |
| or r2,r3 ! HIJK |
| cmp/hs r5,r0 |
| bf/s 3b |
| mov.l r3,@(r0,r4) |
| #endif |
| add #7,r5 |
| ! |
| ! Third, copy a byte at once, if necessary |
| cmp/eq r5,r0 |
| bt/s 9b |
| add #7,r4 |
| add #-3,r0 |
| bra 8b |
| add #-1,r4 |
| |
| case2: |
| ! |
| ! GHIJ KLMN OPQR --> ..GH IJKL MNOP QR.. |
| ! |
| ! First, align to word boundary |
| tst #1,r0 |
| bt/s 2f |
| add #-1,r4 |
| mov.b @r0+,r1 |
| mov.b r1,@(r0,r4) |
| ! |
| 2: ! Second, read a word and write a word at once |
| add #-1,r4 |
| add #-1,r5 |
| ! |
| 3: mov.w @r0+,r1 |
| cmp/hs r5,r0 |
| bf/s 3b |
| mov.w r1,@(r0,r4) |
| add #1,r5 |
| ! |
| ! Third, copy a byte at once, if necessary |
| cmp/eq r5,r0 |
| bt/s 9b |
| add #2,r4 |
| mov.b @r0,r1 |
| mov.b r1,@(r0,r4) |
| bra 9b |
| add #1,r0 |
| |
| case1: |
| ! |
| ! GHIJ KLMN OPQR --> .GHI JKLM NOPQ R... |
| ! |
| ! First, align to long word boundary |
| mov r0,r3 |
| and r2,r3 |
| tst r3,r3 |
| bt/s 2f |
| add #-1,r4 |
| mov #4,r2 |
| sub r3,r2 |
| 1: dt r2 |
| mov.b @r0+,r1 |
| bf/s 1b |
| mov.b r1,@(r0,r4) |
| ! |
| 2: ! Second, read a long word and write a long word at once |
| mov.l @(r0,r4),r1 |
| add #-7,r5 |
| add #-4,r4 |
| ! |
| #ifdef __LITTLE_ENDIAN__ |
| shll16 r1 |
| shll8 r1 |
| 3: mov r1,r3 ! JIHG |
| shlr16 r3 |
| shlr8 r3 ! xxxJ |
| mov.l @r0+,r1 ! NMLK |
| mov r1,r2 |
| shll8 r2 ! MLKx |
| or r2,r3 ! MLKJ |
| cmp/hs r5,r0 |
| bf/s 3b |
| mov.l r3,@(r0,r4) |
| #else |
| shlr16 r1 |
| shlr8 r1 |
| 3: mov r1,r3 ! GHIJ |
| shll16 r3 |
| shll8 r3 ! Jxxx |
| mov.l @r0+,r1 ! KLMN |
| mov r1,r2 |
| shlr8 r2 ! xKLM |
| or r2,r3 ! JKLM |
| cmp/hs r5,r0 |
| bf/s 3b ! while(r0<r5) |
| mov.l r3,@(r0,r4) |
| #endif |
| add #7,r5 |
| ! |
| ! Third, copy a byte at once, if necessary |
| cmp/eq r5,r0 |
| bt/s 9b |
| add #5,r4 |
| add #-3,r0 |
| bra 8b |
| add #-1,r4 |