diff options
author | David Howells <dhowells@redhat.com> | 2008-02-08 04:19:31 -0800 |
---|---|---|
committer | Linus Torvalds <torvalds@woody.linux-foundation.org> | 2008-02-08 09:22:30 -0800 |
commit | b920de1b77b72ca9432ac3f97edb26541e65e5dd (patch) | |
tree | 40fa9be1470e929c47927dea7eddf184c0204229 /arch/mn10300/lib | |
parent | ef3d534754f31fed9c3b976fee1ece1b3bc38282 (diff) |
mn10300: add the MN10300/AM33 architecture to the kernel
Add architecture support for the MN10300/AM33 CPUs produced by MEI to the
kernel.
This patch also adds board support for the ASB2303 with the ASB2308 daughter
board, and the ASB2305. The only processor supported is the MN103E010, which
is an AM33v2 core plus on-chip devices.
[akpm@linux-foundation.org: nuke cvs control strings]
Signed-off-by: Masakazu Urade <urade.masakazu@jp.panasonic.com>
Signed-off-by: Koichi Yasutake <yasutake.koichi@jp.panasonic.com>
Signed-off-by: David Howells <dhowells@redhat.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Diffstat (limited to 'arch/mn10300/lib')
-rw-r--r-- | arch/mn10300/lib/Makefile | 7 | ||||
-rw-r--r-- | arch/mn10300/lib/__ashldi3.S | 51 | ||||
-rw-r--r-- | arch/mn10300/lib/__ashrdi3.S | 52 | ||||
-rw-r--r-- | arch/mn10300/lib/__lshrdi3.S | 52 | ||||
-rw-r--r-- | arch/mn10300/lib/ashrdi3.c | 61 | ||||
-rw-r--r-- | arch/mn10300/lib/bitops.c | 51 | ||||
-rw-r--r-- | arch/mn10300/lib/checksum.c | 99 | ||||
-rw-r--r-- | arch/mn10300/lib/delay.c | 50 | ||||
-rw-r--r-- | arch/mn10300/lib/do_csum.S | 162 | ||||
-rw-r--r-- | arch/mn10300/lib/internal.h | 15 | ||||
-rw-r--r-- | arch/mn10300/lib/lshrdi3.c | 60 | ||||
-rw-r--r-- | arch/mn10300/lib/memcpy.S | 135 | ||||
-rw-r--r-- | arch/mn10300/lib/memmove.S | 160 | ||||
-rw-r--r-- | arch/mn10300/lib/memset.S | 121 | ||||
-rw-r--r-- | arch/mn10300/lib/negdi2.c | 57 | ||||
-rw-r--r-- | arch/mn10300/lib/usercopy.c | 166 |
16 files changed, 1299 insertions, 0 deletions
diff --git a/arch/mn10300/lib/Makefile b/arch/mn10300/lib/Makefile new file mode 100644 index 000000000000..fdfa9ec5b5bb --- /dev/null +++ b/arch/mn10300/lib/Makefile @@ -0,0 +1,7 @@ +# +# Makefile for the MN10300-specific library files.. +# + +lib-y = delay.o usercopy.o checksum.o bitops.o memcpy.o memmove.o memset.o +lib-y += do_csum.o +lib-y += __ashldi3.o __ashrdi3.o __lshrdi3.o negdi2.o diff --git a/arch/mn10300/lib/__ashldi3.S b/arch/mn10300/lib/__ashldi3.S new file mode 100644 index 000000000000..a51a9506f00c --- /dev/null +++ b/arch/mn10300/lib/__ashldi3.S @@ -0,0 +1,51 @@ +/* MN10300 64-bit arithmetic left shift + * + * 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> + + .text + .balign L1_CACHE_BYTES + +############################################################################### +# +# unsigned long long __ashldi3(unsigned long long value [D1:D0], +# unsigned by [(12,SP)]) +# +############################################################################### + .globl __ashldi3 + .type __ashldi3,@function +__ashldi3: + mov (12,sp),a0 + and +63,a0 + beq __ashldi3_zero + + cmp +31,a0 + bhi __ashldi3_32plus + + # the count is in the range 1-31 + asl a0,d1 + + mov +32,a1 + sub a0,a1,a1 # a1 = 32 - count + lsr a1,d0,a1 # get overflow from LSW -> MSW + + or_asl a1,d1,a0,d0 # insert overflow into MSW and + # shift the LSW + rets + + .balign L1_CACHE_BYTES + # the count is in the range 32-63 +__ashldi3_32plus: + asl a0,d0,d1 + clr d0 +__ashldi3_zero: + rets + + .size __ashldi3, .-__ashldi3 diff --git a/arch/mn10300/lib/__ashrdi3.S b/arch/mn10300/lib/__ashrdi3.S new file mode 100644 index 000000000000..6f42382728cb --- /dev/null +++ b/arch/mn10300/lib/__ashrdi3.S @@ -0,0 +1,52 @@ +/* MN10300 64-bit arithmetic right shift + * + * 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> + + .text + .balign L1_CACHE_BYTES + +############################################################################### +# +# unsigned long long __ashrdi3(unsigned long long value [D1:D0], +# unsigned by [(12,SP)]) +# +############################################################################### + .globl __ashrdi3 + .type __ashrdi3,@function +__ashrdi3: + mov (12,sp),a0 + and +63,a0 + beq __ashrdi3_zero + + cmp +31,a0 + bhi __ashrdi3_32plus + + # the count is in the range 1-31 + lsr a0,d0 + + mov +32,a1 + sub a0,a1,a1 # a1 = 32 - count + asl a1,d1,a1 # get underflow from MSW -> LSW + + or_asr a1,d0,a0,d1 # insert underflow into LSW and + # shift the MSW + rets + + .balign L1_CACHE_BYTES + # the count is in the range 32-63 +__ashrdi3_32plus: + asr a0,d1,d0 + ext d0 # sign-extend result through MDR + mov mdr,d1 +__ashrdi3_zero: + rets + + .size __ashrdi3, .-__ashrdi3 diff --git a/arch/mn10300/lib/__lshrdi3.S b/arch/mn10300/lib/__lshrdi3.S new file mode 100644 index 000000000000..a686aef31e90 --- /dev/null +++ b/arch/mn10300/lib/__lshrdi3.S @@ -0,0 +1,52 @@ +/* MN10300 64-bit logical right shift + * + * 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> + + .text + .balign L1_CACHE_BYTES + +############################################################################### +# +# unsigned long long __lshrdi3(unsigned long long value [D1:D0], +# unsigned by [(12,SP)]) +# +############################################################################### + .globl __lshrdi3 + .type __lshrdi3,@function +__lshrdi3: + mov (12,sp),a0 + and +63,a0 + beq __lshrdi3_zero + + cmp +31,a0 + bhi __lshrdi3_32plus + + # the count is in the range 1-31 + lsr a0,d0 + + mov +32,a1 + sub a0,a1,a1 # a1 = 32 - count + asl a1,d1,a1 # get underflow from MSW -> LSW + + or_lsr a1,d0,a0,d1 # insert underflow into LSW and + # shift the MSW + rets + + .balign L1_CACHE_BYTES + # the count is in the range 32-63 +__lshrdi3_32plus: + lsr a0,d1,d0 + clr d1 +__lshrdi3_zero: + rets + + .size __lshrdi3, .-__lshrdi3 diff --git a/arch/mn10300/lib/ashrdi3.c b/arch/mn10300/lib/ashrdi3.c new file mode 100644 index 000000000000..c54f61ddf0b5 --- /dev/null +++ b/arch/mn10300/lib/ashrdi3.c @@ -0,0 +1,61 @@ +/* ashrdi3.c extracted from gcc-2.7.2/libgcc2.c which is: */ +/* Copyright (C) 1989, 1992, 1993, 1994, 1995 Free Software Foundation, Inc. + +This file is part of GNU CC. + +GNU CC 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, or (at your option) +any later version. + +GNU CC is distributed in the hope that it will be useful, +but WITHOUT ANY WARRANTY; without even the implied warranty of +MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the +GNU General Public Licence for more details. + +You should have received a copy of the GNU General Public Licence +along with GNU CC; see the file COPYING. If not, write to +the Free Software Foundation, 59 Temple Place - Suite 330, +Boston, MA 02111-1307, USA. */ + +#define BITS_PER_UNIT 8 + +typedef int SItype __attribute__((mode(SI))); +typedef unsigned int USItype __attribute__((mode(SI))); +typedef int DItype __attribute__((mode(DI))); +typedef int word_type __attribute__((mode(__word__))); + +struct DIstruct { + SItype low; + SItype high; +}; + +union DIunion { + struct DIstruct s; + DItype ll; +}; + +DItype __ashrdi3(DItype u, word_type b) +{ + union DIunion w; + union DIunion uu; + word_type bm; + + if (b == 0) + return u; + + uu.ll = u; + + bm = (sizeof(SItype) * BITS_PER_UNIT) - b; + if (bm <= 0) { + /* w.s.high = 1..1 or 0..0 */ + w.s.high = uu.s.high >> (sizeof(SItype) * BITS_PER_UNIT - 1); + w.s.low = uu.s.high >> -bm; + } else { + USItype carries = (USItype)uu.s.high << bm; + w.s.high = uu.s.high >> b; + w.s.low = ((USItype)uu.s.low >> b) | carries; + } + + return w.ll; +} diff --git a/arch/mn10300/lib/bitops.c b/arch/mn10300/lib/bitops.c new file mode 100644 index 000000000000..440a7dcbf87b --- /dev/null +++ b/arch/mn10300/lib/bitops.c @@ -0,0 +1,51 @@ +/* MN10300 Non-trivial bit operations + * + * 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 <linux/module.h> +#include <asm/bitops.h> +#include <asm/system.h> + +/* + * try flipping a bit using BSET and BCLR + */ +void change_bit(int nr, volatile void *addr) +{ + if (test_bit(nr, addr)) + goto try_clear_bit; + +try_set_bit: + if (!test_and_set_bit(nr, addr)) + return; + +try_clear_bit: + if (test_and_clear_bit(nr, addr)) + return; + + goto try_set_bit; +} + +/* + * try flipping a bit using BSET and BCLR and returning the old value + */ +int test_and_change_bit(int nr, volatile void *addr) +{ + if (test_bit(nr, addr)) + goto try_clear_bit; + +try_set_bit: + if (!test_and_set_bit(nr, addr)) + return 0; + +try_clear_bit: + if (test_and_clear_bit(nr, addr)) + return 1; + + goto try_set_bit; +} diff --git a/arch/mn10300/lib/checksum.c b/arch/mn10300/lib/checksum.c new file mode 100644 index 000000000000..274f29ec33c1 --- /dev/null +++ b/arch/mn10300/lib/checksum.c @@ -0,0 +1,99 @@ +/* MN10300 Optimised checksumming wrappers + * + * Copyright (C) 2007 Matsushita Electric Industrial Co., Ltd. + * 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 <linux/module.h> +#include <linux/errno.h> +#include <asm/byteorder.h> +#include <asm/uaccess.h> +#include <asm/checksum.h> +#include "internal.h" + +static inline unsigned short from32to16(__wsum sum) +{ + asm(" add %1,%0 \n" + " addc 0xffff,%0 \n" + : "=r" (sum) + : "r" (sum << 16), "0" (sum & 0xffff0000) + ); + return sum >> 16; +} + +__sum16 ip_fast_csum(const void *iph, unsigned int ihl) +{ + return ~do_csum(iph, ihl * 4); +} +EXPORT_SYMBOL(ip_fast_csum); + +__wsum csum_partial(const void *buff, int len, __wsum sum) +{ + __wsum result; + + result = do_csum(buff, len); + result += sum; + if (sum > result) + result++; + return result; +} +EXPORT_SYMBOL(csum_partial); + +__sum16 ip_compute_csum(const void *buff, int len) +{ + return ~from32to16(do_csum(buff, len)); +} +EXPORT_SYMBOL(ip_compute_csum); + +__wsum csum_partial_copy(const void *src, void *dst, int len, __wsum sum) +{ + copy_from_user(dst, src, len); + return csum_partial(dst, len, sum); +} +EXPORT_SYMBOL(csum_partial_copy); + +__wsum csum_partial_copy_nocheck(const void *src, void *dst, + int len, __wsum sum) +{ + sum = csum_partial(src, len, sum); + memcpy(dst, src, len); + return sum; +} +EXPORT_SYMBOL(csum_partial_copy_nocheck); + +__wsum csum_partial_copy_from_user(const void *src, void *dst, + int len, __wsum sum, + int *err_ptr) +{ + int missing; + + missing = copy_from_user(dst, src, len); + if (missing) { + memset(dst + len - missing, 0, missing); + *err_ptr = -EFAULT; + } + + return csum_partial(dst, len, sum); +} +EXPORT_SYMBOL(csum_partial_copy_from_user); + +__wsum csum_and_copy_to_user(const void *src, void *dst, + int len, __wsum sum, + int *err_ptr) +{ + int missing; + + missing = copy_to_user(dst, src, len); + if (missing) { + memset(dst + len - missing, 0, missing); + *err_ptr = -EFAULT; + } + + return csum_partial(src, len, sum); +} +EXPORT_SYMBOL(csum_and_copy_to_user); diff --git a/arch/mn10300/lib/delay.c b/arch/mn10300/lib/delay.c new file mode 100644 index 000000000000..cce66bc0822d --- /dev/null +++ b/arch/mn10300/lib/delay.c @@ -0,0 +1,50 @@ +/* MN10300 Short delay interpolation routines + * + * 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 <linux/module.h> +#include <linux/sched.h> +#include <linux/delay.h> +#include <asm/div64.h> + +/* + * basic delay loop + */ +void __delay(unsigned long loops) +{ + int d0; + + asm volatile( + " bra 1f \n" + " .align 4 \n" + "1: bra 2f \n" + " .align 4 \n" + "2: add -1,%0 \n" + " bne 2b \n" + : "=&d" (d0) + : "0" (loops)); +} +EXPORT_SYMBOL(__delay); + +/* + * handle a delay specified in terms of microseconds + */ +void __udelay(unsigned long usecs) +{ + signed long ioclk, stop; + + /* usecs * CLK / 1E6 */ + stop = __muldiv64u(usecs, MN10300_TSCCLK, 1000000); + stop = TMTSCBC - stop; + + do { + ioclk = TMTSCBC; + } while (stop < ioclk); +} +EXPORT_SYMBOL(__udelay); diff --git a/arch/mn10300/lib/do_csum.S b/arch/mn10300/lib/do_csum.S new file mode 100644 index 000000000000..e138994e1667 --- /dev/null +++ b/arch/mn10300/lib/do_csum.S @@ -0,0 +1,162 @@ +/* Optimised simple memory checksum + * + * 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 + +############################################################################### +# +# unsigned int do_csum(const unsigned char *buff, size_t len) +# +############################################################################### + .globl do_csum + .type do_csum,@function +do_csum: + movm [d2,d3],(sp) + mov d0,(12,sp) + mov d1,(16,sp) + mov d1,d2 # count + mov d0,a0 # buff + clr d1 # accumulator + + cmp +0,d2 + beq do_csum_done # return if zero-length buffer + + # 4-byte align the buffer pointer + btst +3,a0 + beq do_csum_now_4b_aligned + + btst +1,a0 + beq do_csum_addr_not_odd + movbu (a0),d0 + inc a0 + asl +8,d0 + add d0,d1 + addc +0,d1 + add -1,d2 +do_csum_addr_not_odd: + + cmp +2,d2 + bcs do_csum_fewer_than_4 + btst +2,a0 + beq do_csum_now_4b_aligned + movhu (a0+),d0 + add d0,d1 + addc +0,d1 + add -2,d2 + cmp +4,d2 + bcs do_csum_fewer_than_4 + +do_csum_now_4b_aligned: + # we want to checksum as much as we can in chunks of 32 bytes + cmp +31,d2 + bls do_csum_remainder # 4-byte aligned remainder + + add -32,d2 + mov +32,d3 + +do_csum_loop: + mov (a0+),d0 + add d0,d1 + mov (a0+),e0 + addc e0,d1 + mov (a0+),e1 + addc e1,d1 + mov (a0+),e3 + addc e3,d1 + mov (a0+),d0 + addc d0,d1 + mov (a0+),e0 + addc e0,d1 + mov (a0+),e1 + addc e1,d1 + mov (a0+),e3 + addc e3,d1 + addc +0,d1 + + sub d3,d2 + bcc do_csum_loop + + add d3,d2 + beq do_csum_done + +do_csum_remainder: + # cut 16-31 bytes down to 0-15 + cmp +16,d2 + bcs do_csum_fewer_than_16 + mov (a0+),d0 + add d0,d1 + mov (a0+),e0 + addc e0,d1 + mov (a0+),e1 + addc e1,d1 + mov (a0+),e3 + addc e3,d1 + addc +0,d1 + add -16,d2 + beq do_csum_done + +do_csum_fewer_than_16: + # copy the remaining whole words + cmp +4,d2 + bcs do_csum_fewer_than_4 + cmp +8,d2 + bcs do_csum_one_word + cmp +12,d2 + bcs do_csum_two_words + mov (a0+),d0 + add d0,d1 + addc +0,d1 +do_csum_two_words: + mov (a0+),d0 + add d0,d1 + addc +0,d1 +do_csum_one_word: + mov (a0+),d0 + add d0,d1 + addc +0,d1 + +do_csum_fewer_than_4: + and +3,d2 + beq do_csum_done + xor_cmp d0,d0,+2,d2 + bcs do_csum_fewer_than_2 + movhu (a0+),d0 +do_csum_fewer_than_2: + and +1,d2 + beq do_csum_add_last_bit + movbu (a0),d3 + add d3,d0 +do_csum_add_last_bit: + add d0,d1 + addc +0,d1 + +do_csum_done: + # compress the checksum down to 16 bits + mov +0xffff0000,d2 + and d1,d2 + asl +16,d1 + add d2,d1,d0 + addc +0xffff,d0 + lsr +16,d0 + + # flip the halves of the word result if the buffer was oddly aligned + mov (12,sp),d1 + and +1,d1 + beq do_csum_not_oddly_aligned + swaph d0,d0 # exchange bits 15:8 with 7:0 + +do_csum_not_oddly_aligned: + ret [d2,d3],8 + +do_csum_end: + .size do_csum, do_csum_end-do_csum diff --git a/arch/mn10300/lib/internal.h b/arch/mn10300/lib/internal.h new file mode 100644 index 000000000000..0014eee5f04f --- /dev/null +++ b/arch/mn10300/lib/internal.h @@ -0,0 +1,15 @@ +/* Internal definitions for the arch part of the kernel library + * + * 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. + */ + +/* + * do_csum.S + */ +extern unsigned int do_csum(const unsigned char *, size_t); diff --git a/arch/mn10300/lib/lshrdi3.c b/arch/mn10300/lib/lshrdi3.c new file mode 100644 index 000000000000..e05e64e9ce96 --- /dev/null +++ b/arch/mn10300/lib/lshrdi3.c @@ -0,0 +1,60 @@ +/* lshrdi3.c extracted from gcc-2.7.2/libgcc2.c which is: */ +/* Copyright (C) 1989, 1992, 1993, 1994, 1995 Free Software Foundation, Inc. + +This file is part of GNU CC. + +GNU CC 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, or (at your option) +any later version. + +GNU CC is distributed in the hope that it will be useful, +but WITHOUT ANY WARRANTY; without even the implied warranty of +MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the +GNU General Public Licence for more details. + +You should have received a copy of the GNU General Public Licence +along with GNU CC; see the file COPYING. If not, write to +the Free Software Foundation, 59 Temple Place - Suite 330, +Boston, MA 02111-1307, USA. */ + +#define BITS_PER_UNIT 8 + +typedef int SItype __attribute__((mode(SI))); +typedef unsigned int USItype __attribute__((mode(SI))); +typedef int DItype __attribute__((mode(DI))); +typedef int word_type __attribute__((mode(__word__))); + +struct DIstruct { + SItype low; + SItype high; +}; + +union DIunion { + struct DIstruct s; + DItype ll; +}; + +DItype __lshrdi3(DItype u, word_type b) +{ + union DIunion w; + word_type bm; + union DIunion uu; + + if (b == 0) + return u; + + uu.ll = u; + + bm = (sizeof(SItype) * BITS_PER_UNIT) - b; + if (bm <= 0) { + w.s.high = 0; + w.s.low = (USItype) uu.s.high >> -bm; + } else { + USItype carries = (USItype) uu.s.high << bm; + w.s.high = (USItype) uu.s.high >> b; + w.s.low = ((USItype) uu.s.low >> b) | carries; + } + + return w.ll; +} diff --git a/arch/mn10300/lib/memcpy.S b/arch/mn10300/lib/memcpy.S new file mode 100644 index 000000000000..25fb9bb2604f --- /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 diff --git a/arch/mn10300/lib/memmove.S b/arch/mn10300/lib/memmove.S new file mode 100644 index 000000000000..20b07b62b77c --- /dev/null +++ b/arch/mn10300/lib/memmove.S @@ -0,0 +1,160 @@ +/* MN10300 Optimised simple memory to memory copy, with support for overlapping + * regions + * + * 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 *memmove(void *dst, const void *src, size_t n) +# +############################################################################### + .globl memmove + .type memmove,@function +memmove: + # fall back to memcpy if dst < src to work bottom up + cmp d1,d0 + bcs memmove_memcpy + + # work top down + movm [d2,d3],(sp) + mov d0,(12,sp) + mov d1,(16,sp) + mov (20,sp),d2 # count + add d0,d2,a0 # dst end + add d1,d2,a1 # src end + mov d0,e3 # the return value + + cmp +0,d2 + beq memmove_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 memmove_1 # jump if not + + # we want to transfer as much as we can in chunks of 32 bytes + add -4,a1 + cmp +31,d2 + bls memmove_4_remainder # 4-byte aligned remainder + + add -32,d2 + mov +32,d3 + +memmove_4_loop: + mov (a1),d0 + sub_sub +4,a1,+4,a0 + mov d0,(a0) + mov (a1),d1 + sub_sub +4,a1,+4,a0 + mov d1,(a0) + + mov (a1),d0 + sub_sub +4,a1,+4,a0 + mov d0,(a0) + mov (a1),d1 + sub_sub +4,a1,+4,a0 + mov d1,(a0) + + mov (a1),d0 + sub_sub +4,a1,+4,a0 + mov d0,(a0) + mov (a1),d1 + sub_sub +4,a1,+4,a0 + mov d1,(a0) + + mov (a1),d0 + sub_sub +4,a1,+4,a0 + mov d0,(a0) + mov (a1),d1 + sub_sub +4,a1,+4,a0 + mov d1,(a0) + + sub d3,d2 + bcc memmove_4_loop + + add d3,d2 + beq memmove_4_no_remainder + +memmove_4_remainder: + # cut 4-7 words down to 0-3 + cmp +16,d2 + bcs memmove_4_three_or_fewer_words + mov (a1),d0 + sub_sub +4,a1,+4,a0 + mov d0,(a0) + mov (a1),d1 + sub_sub +4,a1,+4,a0 + mov d1,(a0) + mov (a1),e0 + sub_sub +4,a1,+4,a0 + mov e0,(a0) + mov (a1),e1 + sub_sub +4,a1,+4,a0 + mov e1,(a0) + add -16,d2 + beq memmove_4_no_remainder + + # copy the remaining 1, 2 or 3 words +memmove_4_three_or_fewer_words: + cmp +8,d2 + bcs memmove_4_one_word + beq memmove_4_two_words + mov (a1),d0 + sub_sub +4,a1,+4,a0 + mov d0,(a0) +memmove_4_two_words: + mov (a1),d0 + sub_sub +4,a1,+4,a0 + mov d0,(a0) +memmove_4_one_word: + mov (a1),d0 + sub_sub +4,a1,+4,a0 + mov d0,(a0) + +memmove_4_no_remainder: + # check we copied the correct amount + # TODO: REMOVE CHECK + sub e3,a0,d2 + beq memmove_done + break + break + break + +memmove_done: + mov e3,a0 + ret [d2,d3],8 + + # handle misaligned copying +memmove_1: + add -1,a1 + add -1,d2 + mov +1,d3 + setlb # setlb requires the next insns + # to occupy exactly 4 bytes + + sub d3,d2 + movbu (a1),d0 + sub_sub d3,a1,d3,a0 + movbu d0,(a0) + lcc + + mov e3,a0 + ret [d2,d3],8 + +memmove_memcpy: + jmp memcpy + +memmove_end: + .size memmove, memmove_end-memmove diff --git a/arch/mn10300/lib/memset.S b/arch/mn10300/lib/memset.S new file mode 100644 index 000000000000..bc02e39629b7 --- /dev/null +++ b/arch/mn10300/lib/memset.S @@ -0,0 +1,121 @@ +/* Optimised simple memory fill + * + * 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 *memset(void *dst, int c, size_t n) +# +############################################################################### + .globl memset + .type memset,@function +memset: + movm [d2,d3],(sp) + mov d0,(12,sp) + mov d1,(16,sp) + mov (20,sp),d2 # count + mov d0,a0 # dst + mov d0,e3 # the return value + + cmp +0,d2 + beq memset_done # return if zero-length fill + + # see if the region parameters are four-byte aligned + or d0,d2,d3 + and +3,d3 + bne memset_1 # jump if not + + extbu d1 + mov_asl d1,d3,8,d1 + or_asl d1,d3,8,d1 + or_asl d1,d3,8,d1 + or d3,d1 + + # we want to transfer as much as we can in chunks of 32 bytes + cmp +31,d2 + bls memset_4_remainder # 4-byte aligned remainder + + add -32,d2 + mov +32,d3 + +memset_4_loop: + mov d1,(a0+) + mov d1,(a0+) + mov d1,(a0+) + mov d1,(a0+) + mov d1,(a0+) + mov d1,(a0+) + mov d1,(a0+) + mov d1,(a0+) + + sub d3,d2 + bcc memset_4_loop + + add d3,d2 + beq memset_4_no_remainder + +memset_4_remainder: + # cut 4-7 words down to 0-3 + cmp +16,d2 + bcs memset_4_three_or_fewer_words + mov d1,(a0+) + mov d1,(a0+) + mov d1,(a0+) + mov d1,(a0+) + add -16,d2 + beq memset_4_no_remainder + + # copy the remaining 1, 2 or 3 words +memset_4_three_or_fewer_words: + cmp +8,d2 + bcs memset_4_one_word + beq memset_4_two_words + mov d1,(a0+) +memset_4_two_words: + mov d1,(a0+) +memset_4_one_word: + mov d1,(a0+) + +memset_4_no_remainder: + # check we set the correct amount + # TODO: REMOVE CHECK + sub e3,a0,d2 + mov (20,sp),d1 + cmp d2,d1 + beq memset_done + break |