From a035261089403de259e74ce4dd196e2715138ed2 Mon Sep 17 00:00:00 2001 From: Thomas Martitz Date: Sat, 7 Jan 2012 19:56:09 +0100 Subject: Move optimized memcpy and friends and strlen to firmware/asm, using the new automatic-asm-picking infrastructure. --- firmware/asm/SOURCES | 8 +- firmware/asm/arm/memcpy.S | 176 +++++++++++ firmware/asm/arm/memmove.S | 190 ++++++++++++ firmware/asm/arm/memset.S | 98 +++++++ firmware/asm/arm/memset16.S | 82 ++++++ firmware/asm/m68k/memcpy.S | 682 +++++++++++++++++++++++++++++++++++++++++++ firmware/asm/m68k/memmove.S | 670 ++++++++++++++++++++++++++++++++++++++++++ firmware/asm/m68k/memset.S | 152 ++++++++++ firmware/asm/m68k/memset16.S | 146 +++++++++ firmware/asm/m68k/strlen.S | 71 +++++ firmware/asm/memcpy.c | 117 ++++++++ firmware/asm/memmove.c | 147 ++++++++++ firmware/asm/memset.c | 110 +++++++ firmware/asm/memset16.c | 78 +++++ firmware/asm/mips/memcpy.S | 143 +++++++++ firmware/asm/mips/memset.S | 239 +++++++++++++++ firmware/asm/sh/memcpy.S | 219 ++++++++++++++ firmware/asm/sh/memmove.S | 222 ++++++++++++++ firmware/asm/sh/memset.S | 109 +++++++ firmware/asm/sh/strlen.S | 96 ++++++ firmware/asm/strlen.c | 93 ++++++ 21 files changed, 3847 insertions(+), 1 deletion(-) create mode 100644 firmware/asm/arm/memcpy.S create mode 100644 firmware/asm/arm/memmove.S create mode 100644 firmware/asm/arm/memset.S create mode 100644 firmware/asm/arm/memset16.S create mode 100644 firmware/asm/m68k/memcpy.S create mode 100644 firmware/asm/m68k/memmove.S create mode 100644 firmware/asm/m68k/memset.S create mode 100644 firmware/asm/m68k/memset16.S create mode 100644 firmware/asm/m68k/strlen.S create mode 100644 firmware/asm/memcpy.c create mode 100644 firmware/asm/memmove.c create mode 100644 firmware/asm/memset.c create mode 100644 firmware/asm/memset16.c create mode 100644 firmware/asm/mips/memcpy.S create mode 100644 firmware/asm/mips/memset.S create mode 100644 firmware/asm/sh/memcpy.S create mode 100644 firmware/asm/sh/memmove.S create mode 100644 firmware/asm/sh/memset.S create mode 100644 firmware/asm/sh/strlen.S create mode 100644 firmware/asm/strlen.c (limited to 'firmware/asm') diff --git a/firmware/asm/SOURCES b/firmware/asm/SOURCES index 8b13789179..805727ea93 100644 --- a/firmware/asm/SOURCES +++ b/firmware/asm/SOURCES @@ -1 +1,7 @@ - +memset16.c +#if (CONFIG_PLATFORM & PLATFORM_NATIVE) || defined(HAVE_ROCKBOX_C_LIBRARY) +memcpy.c +memmove.c +memset.c +strlen.c +#endif diff --git a/firmware/asm/arm/memcpy.S b/firmware/asm/arm/memcpy.S new file mode 100644 index 0000000000..2a55fb5656 --- /dev/null +++ b/firmware/asm/arm/memcpy.S @@ -0,0 +1,176 @@ +/*************************************************************************** + * __________ __ ___. + * Open \______ \ ____ ____ | | _\_ |__ _______ ___ + * Source | _// _ \_/ ___\| |/ /| __ \ / _ \ \/ / + * Jukebox | | ( <_> ) \___| < | \_\ ( <_> > < < + * Firmware |____|_ /\____/ \___ >__|_ \|___ /\____/__/\_ \ + * \/ \/ \/ \/ \/ + * $Id$ + * + * Copyright (C) 2006 Free Software Foundation, Inc. + * This file was originally part of the GNU C Library + * Contributed to glibc by MontaVista Software, Inc. (written by Nicolas Pitre) + * Adapted for Rockbox by Daniel Ankers + * + * This program is free software; you can redistribute it and/or + * modify it under the terms of the GNU General Public License + * as published by the Free Software Foundation; either version 2 + * of the License, or (at your option) any later version. + * + * This software is distributed on an "AS IS" basis, WITHOUT WARRANTY OF ANY + * KIND, either express or implied. + * + ****************************************************************************/ + +#include "config.h" + +/* + * Endian independent macros for shifting bytes within registers. + */ +#ifndef __ARMEB__ +#define pull lsr +#define push lsl +#else +#define pull lsl +#define push lsr +#endif + +/* Prototype: void *memcpy(void *dest, const void *src, size_t n); */ + + .section .icode,"ax",%progbits + + .align 2 + .global memcpy + .type memcpy,%function + +memcpy: + stmfd sp!, {r0, r4, lr} + + subs r2, r2, #4 + blt 8f + ands ip, r0, #3 + bne 9f + ands ip, r1, #3 + bne 10f + +1: subs r2, r2, #(28) + stmfd sp!, {r5 - r8} + blt 5f + +2: +3: +4: ldmia r1!, {r3, r4, r5, r6, r7, r8, ip, lr} + subs r2, r2, #32 + stmia r0!, {r3, r4, r5, r6, r7, r8, ip, lr} + bge 3b + +5: ands ip, r2, #28 + rsb ip, ip, #32 + addne pc, pc, ip @ C is always clear here + b 7f +6: nop + ldr r3, [r1], #4 + ldr r4, [r1], #4 + ldr r5, [r1], #4 + ldr r6, [r1], #4 + ldr r7, [r1], #4 + ldr r8, [r1], #4 + ldr lr, [r1], #4 + + add pc, pc, ip + nop + nop + str r3, [r0], #4 + str r4, [r0], #4 + str r5, [r0], #4 + str r6, [r0], #4 + str r7, [r0], #4 + str r8, [r0], #4 + str lr, [r0], #4 + +7: ldmfd sp!, {r5 - r8} + +8: movs r2, r2, lsl #31 + ldrneb r3, [r1], #1 + ldrcsb r4, [r1], #1 + ldrcsb ip, [r1] + strneb r3, [r0], #1 + strcsb r4, [r0], #1 + strcsb ip, [r0] + + ldmpc regs="r0, r4" + +9: rsb ip, ip, #4 + cmp ip, #2 + ldrgtb r3, [r1], #1 + ldrgeb r4, [r1], #1 + ldrb lr, [r1], #1 + strgtb r3, [r0], #1 + strgeb r4, [r0], #1 + subs r2, r2, ip + strb lr, [r0], #1 + blt 8b + ands ip, r1, #3 + beq 1b + +10: bic r1, r1, #3 + cmp ip, #2 + ldr lr, [r1], #4 + beq 17f + bgt 18f + + + .macro forward_copy_shift pull push + + subs r2, r2, #28 + blt 14f + +11: stmfd sp!, {r5 - r9} + +12: +13: ldmia r1!, {r4, r5, r6, r7} + mov r3, lr, pull #\pull + subs r2, r2, #32 + ldmia r1!, {r8, r9, ip, lr} + orr r3, r3, r4, push #\push + mov r4, r4, pull #\pull + orr r4, r4, r5, push #\push + mov r5, r5, pull #\pull + orr r5, r5, r6, push #\push + mov r6, r6, pull #\pull + orr r6, r6, r7, push #\push + mov r7, r7, pull #\pull + orr r7, r7, r8, push #\push + mov r8, r8, pull #\pull + orr r8, r8, r9, push #\push + mov r9, r9, pull #\pull + orr r9, r9, ip, push #\push + mov ip, ip, pull #\pull + orr ip, ip, lr, push #\push + stmia r0!, {r3, r4, r5, r6, r7, r8, r9, ip} + bge 12b + + ldmfd sp!, {r5 - r9} + +14: ands ip, r2, #28 + beq 16f + +15: mov r3, lr, pull #\pull + ldr lr, [r1], #4 + subs ip, ip, #4 + orr r3, r3, lr, push #\push + str r3, [r0], #4 + bgt 15b + +16: sub r1, r1, #(\push / 8) + b 8b + + .endm + + + forward_copy_shift pull=8 push=24 + +17: forward_copy_shift pull=16 push=16 + +18: forward_copy_shift pull=24 push=8 + diff --git a/firmware/asm/arm/memmove.S b/firmware/asm/arm/memmove.S new file mode 100644 index 0000000000..d8cab048be --- /dev/null +++ b/firmware/asm/arm/memmove.S @@ -0,0 +1,190 @@ +/*************************************************************************** + * __________ __ ___. + * Open \______ \ ____ ____ | | _\_ |__ _______ ___ + * Source | _// _ \_/ ___\| |/ /| __ \ / _ \ \/ / + * Jukebox | | ( <_> ) \___| < | \_\ ( <_> > < < + * Firmware |____|_ /\____/ \___ >__|_ \|___ /\____/__/\_ \ + * \/ \/ \/ \/ \/ + * $Id$ + * + * Copyright (C) 2006 Free Software Foundation, Inc. + * This file was originally part of the GNU C Library + * Contributed to glibc by MontaVista Software, Inc. (written by Nicolas Pitre) + * Adapted for Rockbox by Daniel Ankers + * + * This program is free software; you can redistribute it and/or + * modify it under the terms of the GNU General Public License + * as published by the Free Software Foundation; either version 2 + * of the License, or (at your option) any later version. + * + * This software is distributed on an "AS IS" basis, WITHOUT WARRANTY OF ANY + * KIND, either express or implied. + * + ****************************************************************************/ + +#include "config.h" + +/* + * Endian independent macros for shifting bytes within registers. + */ +#ifndef __ARMEB__ +#define pull lsr +#define push lsl +#else +#define pull lsl +#define push lsr +#endif + + .text + +/* + * Prototype: void *memmove(void *dest, const void *src, size_t n); + * + * Note: + * + * If the memory regions don't overlap, we simply branch to memcpy which is + * normally a bit faster. Otherwise the copy is done going downwards. + */ + + .section .icode,"ax",%progbits + + .align 2 + .global memmove + .type memmove,%function + +memmove: + + subs ip, r0, r1 + cmphi r2, ip + bls memcpy + + stmfd sp!, {r0, r4, lr} + add r1, r1, r2 + add r0, r0, r2 + subs r2, r2, #4 + blt 8f + ands ip, r0, #3 + bne 9f + ands ip, r1, #3 + bne 10f + +1: subs r2, r2, #(28) + stmfd sp!, {r5 - r8} + blt 5f + +2: +3: +4: ldmdb r1!, {r3, r4, r5, r6, r7, r8, ip, lr} + subs r2, r2, #32 + stmdb r0!, {r3, r4, r5, r6, r7, r8, ip, lr} + bge 3b + +5: ands ip, r2, #28 + rsb ip, ip, #32 + addne pc, pc, ip @ C is always clear here + b 7f +6: nop + ldr r3, [r1, #-4]! + ldr r4, [r1, #-4]! + ldr r5, [r1, #-4]! + ldr r6, [r1, #-4]! + ldr r7, [r1, #-4]! + ldr r8, [r1, #-4]! + ldr lr, [r1, #-4]! + + add pc, pc, ip + nop + nop + str r3, [r0, #-4]! + str r4, [r0, #-4]! + str r5, [r0, #-4]! + str r6, [r0, #-4]! + str r7, [r0, #-4]! + str r8, [r0, #-4]! + str lr, [r0, #-4]! + +7: ldmfd sp!, {r5 - r8} + +8: movs r2, r2, lsl #31 + ldrneb r3, [r1, #-1]! + ldrcsb r4, [r1, #-1]! + ldrcsb ip, [r1, #-1] + strneb r3, [r0, #-1]! + strcsb r4, [r0, #-1]! + strcsb ip, [r0, #-1] + ldmpc regs="r0, r4" + +9: cmp ip, #2 + ldrgtb r3, [r1, #-1]! + ldrgeb r4, [r1, #-1]! + ldrb lr, [r1, #-1]! + strgtb r3, [r0, #-1]! + strgeb r4, [r0, #-1]! + subs r2, r2, ip + strb lr, [r0, #-1]! + blt 8b + ands ip, r1, #3 + beq 1b + +10: bic r1, r1, #3 + cmp ip, #2 + ldr r3, [r1, #0] + beq 17f + blt 18f + + + .macro backward_copy_shift push pull + + subs r2, r2, #28 + blt 14f + +11: stmfd sp!, {r5 - r9} + +12: +13: ldmdb r1!, {r7, r8, r9, ip} + mov lr, r3, push #\push + subs r2, r2, #32 + ldmdb r1!, {r3, r4, r5, r6} + orr lr, lr, ip, pull #\pull + mov ip, ip, push #\push + orr ip, ip, r9, pull #\pull + mov r9, r9, push #\push + orr r9, r9, r8, pull #\pull + mov r8, r8, push #\push + orr r8, r8, r7, pull #\pull + mov r7, r7, push #\push + orr r7, r7, r6, pull #\pull + mov r6, r6, push #\push + orr r6, r6, r5, pull #\pull + mov r5, r5, push #\push + orr r5, r5, r4, pull #\pull + mov r4, r4, push #\push + orr r4, r4, r3, pull #\pull + stmdb r0!, {r4 - r9, ip, lr} + bge 12b + + ldmfd sp!, {r5 - r9} + +14: ands ip, r2, #28 + beq 16f + +15: mov lr, r3, push #\push + ldr r3, [r1, #-4]! + subs ip, ip, #4 + orr lr, lr, r3, pull #\pull + str lr, [r0, #-4]! + bgt 15b + +16: add r1, r1, #(\pull / 8) + b 8b + + .endm + + + backward_copy_shift push=8 pull=24 + +17: backward_copy_shift push=16 pull=16 + +18: backward_copy_shift push=24 pull=8 + + diff --git a/firmware/asm/arm/memset.S b/firmware/asm/arm/memset.S new file mode 100644 index 0000000000..682da874ce --- /dev/null +++ b/firmware/asm/arm/memset.S @@ -0,0 +1,98 @@ +/*************************************************************************** + * __________ __ ___. + * Open \______ \ ____ ____ | | _\_ |__ _______ ___ + * Source | _// _ \_/ ___\| |/ /| __ \ / _ \ \/ / + * Jukebox | | ( <_> ) \___| < | \_\ ( <_> > < < + * Firmware |____|_ /\____/ \___ >__|_ \|___ /\____/__/\_ \ + * \/ \/ \/ \/ \/ + * $Id$ + * + * Copyright (C) 2006 by Thom Johansen + * + * This program is free software; you can redistribute it and/or + * modify it under the terms of the GNU General Public License + * as published by the Free Software Foundation; either version 2 + * of the License, or (at your option) any later version. + * + * 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 found in Linux kernel version 2.6.15.3 + * linux/arch/arm/lib/memset.S + * + * Copyright (C) 1995-2000 Russell King + */ + +/* This code will align a pointer for memset, if needed */ +1: cmp r2, #4 @ 1 do we have enough + blt 5f @ 1 bytes to align with? + cmp r3, #2 @ 1 + strgtb r1, [r0, #-1]! @ 1 + strgeb r1, [r0, #-1]! @ 1 + strb r1, [r0, #-1]! @ 1 + sub r2, r2, r3 @ 1 r2 = r2 - r3 + b 2f + + .global memset + .type memset,%function +memset: + add r0, r0, r2 @ we'll write backwards in memory + ands r3, r0, #3 @ 1 unaligned? + bne 1b @ 1 +2: +/* + * we know that the pointer in r0 is aligned to a word boundary. + */ + orr r1, r1, r1, lsl #8 + orr r1, r1, r1, lsl #16 + mov r3, r1 + cmp r2, #16 + blt 5f +/* + * 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 + +3: subs r2, r2, #64 + stmgedb r0!, {r1, r3, ip, lr} @ 64 bytes at a time. + stmgedb r0!, {r1, r3, ip, lr} + stmgedb r0!, {r1, r3, ip, lr} + stmgedb r0!, {r1, r3, ip, lr} + bgt 3b + ldrpc cond=eq @ Now <64 bytes to go. +/* + * No need to correct the count; we're only testing bits from now on + */ + tst r2, #32 + stmnedb r0!, {r1, r3, ip, lr} + stmnedb r0!, {r1, r3, ip, lr} + tst r2, #16 + stmnedb r0!, {r1, r3, ip, lr} + ldr lr, [sp], #4 + +5: tst r2, #8 + stmnedb r0!, {r1, r3} + tst r2, #4 + strne r1, [r0, #-4]! +/* + * When we get here, we've got less than 4 bytes to zero. We + * may have an unaligned pointer as well. + */ +6: tst r2, #2 + strneb r1, [r0, #-1]! + strneb r1, [r0, #-1]! + tst r2, #1 + strneb r1, [r0, #-1]! + bx lr +.end: + .size memset,.end-memset diff --git a/firmware/asm/arm/memset16.S b/firmware/asm/arm/memset16.S new file mode 100644 index 0000000000..5c787b1bed --- /dev/null +++ b/firmware/asm/arm/memset16.S @@ -0,0 +1,82 @@ +/*************************************************************************** + * __________ __ ___. + * Open \______ \ ____ ____ | | _\_ |__ _______ ___ + * Source | _// _ \_/ ___\| |/ /| __ \ / _ \ \/ / + * Jukebox | | ( <_> ) \___| < | \_\ ( <_> > < < + * Firmware |____|_ /\____/ \___ >__|_ \|___ /\____/__/\_ \ + * \/ \/ \/ \/ \/ + * $Id$ + * + * Copyright (C) 2006 by Thom Johansen + * + * This program is free software; you can redistribute it and/or + * modify it under the terms of the GNU General Public License + * as published by the Free Software Foundation; either version 2 + * of the License, or (at your option) any later version. + * + * 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 + ldrpc cond=eq @ 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 diff --git a/firmware/asm/m68k/memcpy.S b/firmware/asm/m68k/memcpy.S new file mode 100644 index 0000000000..9762e31e02 --- /dev/null +++ b/firmware/asm/m68k/memcpy.S @@ -0,0 +1,682 @@ +/*************************************************************************** + * __________ __ ___. + * Open \______ \ ____ ____ | | _\_ |__ _______ ___ + * Source | _// _ \_/ ___\| |/ /| __ \ / _ \ \/ / + * Jukebox | | ( <_> ) \___| < | \_\ ( <_> > < < + * Firmware |____|_ /\____/ \___ >__|_ \|___ /\____/__/\_ \ + * \/ \/ \/ \/ \/ + * $Id$ + * + * Copyright (C) 2004-2005 by Jens Arnold + * + * This program is free software; you can redistribute it and/or + * modify it under the terms of the GNU General Public License + * as published by the Free Software Foundation; either version 2 + * of the License, or (at your option) any later version. + * + * 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 + +#define FULLSPEED /* use burst writing for word aligned destinations */ + .align 2 + .global memcpy + .global __memcpy_fwd_entry + .type memcpy,@function + +/* Copies bytes of data in memory from to + * This version is optimized for speed + * + * arguments: + * (4,%sp) - destination address + * (8,%sp) - source address + * (12,%sp) - length + * + * return value: + * %d0 - destination address (like ANSI version) + * + * register usage: + * %a0 - current source address + * %a1 - current dest address + * %a2 - source end address (in line-copy loops) + * %d0 - data / scratch + * %d1 - source end address (byte and longword copy) / data / scratch + * %d2 - data / scratch + * %d3..%d7 - data + * + * For maximum speed this routine reads and writes whole lines using burst + * move (movem.l) where possible. For byte aligned destinations (long+1 and + * long+3) it writes longwords only. Same goes for word aligned destinations + * if FULLSPEED is undefined. + */ +memcpy: + move.l (4,%sp),%a1 /* Destination */ + move.l (8,%sp),%a0 /* Source */ + move.l (12,%sp),%d1 /* Length */ + +__memcpy_fwd_entry: + add.l %a0,%d1 /* %d1 = source end */ + + move.l %a0,%d0 + addq.l #7,%d0 + and.l #0xFFFFFFFC,%d0 /* %d0 = first source long bound + 4 */ + cmp.l %d0,%d1 /* at least one aligned longword to copy? */ + blo.w .bytes2_start /* no, jump directly to trailing byte loop */ + + subq.l #4,%d0 /* %d0 = first source long bound */ + cmp.l %a0,%d0 /* any bytes to copy? */ + jls .bytes1_end /* no: skip byte loop */ + + /* leading byte loop: copies 0..3 bytes */ +.bytes1_loop: + move.b (%a0)+,(%a1)+ /* copy byte */ + cmp.l %a0,%d0 /* runs %a0 up to first long bound */ + jhi .bytes1_loop + +.bytes1_end: + moveq.l #31,%d0 + add.l %a0,%d0 + and.l #0xFFFFFFF0,%d0 /* %d0 = first source line bound + 16 */ + cmp.l %d0,%d1 /* at least one aligned line to copy? */ + blo.w .long_start /* no: jump to longword copy loop */ + + lea.l (-28,%sp),%sp /* free up some registers */ + movem.l %d2-%d7/%a2,(%sp) + + moveq.l #16,%d2 + sub.l %d2,%d0 /* %d0 = first source line bound */ + move.l %d1,%a2 /* %a2 = end address */ + lea.l (-15,%a2),%a2 /* adjust end address for loops doing 16 bytes/ pass */ + move.l %a1,%d1 + moveq.l #3,%d2 /* mask */ + and.l %d2,%d1 + jmp.l (2,%pc,%d1.l*4) /* switch (dest_addr & 3) */ + bra.w .lines_do0_start + bra.w .lines_do1_start + bra.w .lines_do2_start + /* bra.w .lines_do3_start implicit */ + + /* byte aligned destination (long + 3): use line burst reads in main loop */ +.lines_do3_start: + moveq.l #24,%d1 /* shift count for shifting by 3 bytes */ + cmp.l %a0,%d0 /* any leading longwords? */ + jhi .lines_do3_head_start /* yes: leading longword copy */ + + movem.l (%a0),%d4-%d7 /* load first line */ + lea.l (16,%a0),%a0 + move.l %d4,%d2 + lsr.l %d1,%d2 /* get high byte of first longword */ + move.b %d2,(%a1)+ /* store byte */ + jra .lines_do3_entry /* jump into main loop */ + +.lines_do3_head_start: + move.l (%a0)+,%d7 /* load first longword */ + move.l %d7,%d2 + lsr.l %d1,%d2 /* get high byte */ + move.b %d2,(%a1)+ /* store byte */ + jra .lines_do3_head_entry /* jump into leading longword loop */ + +.lines_do3_head_loop: + move.l %d7,%d6 /* move old longword away */ + move.l (%a0)+,%d7 /* load new longword */ + move.l %d7,%d2 + lsr.l %d1,%d2 /* get high byte */ + or.l %d2,%d6 /* combine with old lower 3 bytes */ + move.l %d6,(%a1)+ /* store longword */ +.lines_do3_head_entry: + lsl.l #8,%d7 /* shift up lower 3 bytes */ + cmp.l %a0,%d0 /* runs %a0 up to first line bound */ + jhi .lines_do3_head_loop + +.lines_do3_loop: + move.l %d7,%d3 /* move last longword of old line away */ + movem.l (%a0),%d4-%d7 /* load new line */ + lea.l (16,%a0),%a0 + move.l %d4,%d2 + lsr.l %d1,%d2 /* get high byte of 1st longword */ + or.l %d2,%d3 /* combine with old lower 3 bytes */ + move.l %d3,(%a1)+ /* store longword */ +.lines_do3_entry: + lsl.l #8,%d4 /* shift up lower 3 bytes */ + move.l %d5,%d2 + lsr.l %d1,%d2 /* get high byte of 2nd longword */ + or.l %d2,%d4 /* combine with 1st lower 3 bytes */ + move.l %d4,(%a1)+ /* store longword */ + lsl.l #8,%d5 /* shift up lower 3 bytes */ + move.l %d6,%d2 + lsr.l %d1,%d2 /* get high byte of 3rd longword */ + or.l %d2,%d5 /* combine with 2nd lower 3 bytes */ + move.l %d5,(%a1)+ /* store longword */ + lsl.l #8,%d6 /* shift up lower 3 bytes */ + move.l %d7,%d2 + lsr.l %d1,%d2 /* get high byte of 4th longword */ + or.l %d2,%d6 /* combine with 3rd lower 3 bytes */ + move.l %d6,(%a1)+ /* store longword */ + lsl.l #8,%d7 /* shift up lower 3 bytes */ + cmp.l %a0,%a2 /* runs %a0 up to last line bound */ + jhi .lines_do3_loop + + lea.l (12,%a2),%a2 /* readjust end address for doing longwords */ + cmp.l %a0,%a2 /* any trailing longwords? */ + jls .lines_do3_tail_end /* no: just store last lower 3 bytes */ + +.lines_do3_tail_loop: + move.l %d7,%d6 /* move old longword away */ + move.l (%a0)+,%d7 /* load new longword */ + move.l %d7,%d2 + lsr.l %d1,%d2 /* get high byte */ + or.l %d2,%d6 /* combine with old lower 3 bytes */ + move.l %d6,(%a1)+ /* store longword */ + lsl.l #8,%d7 /* shift up lower 3 bytes */ + cmp.l %a0,%a2 /* runs %a0 up to last long bound */ + jhi .lines_do3_tail_loop + +.lines_do3_tail_end: + swap %d7 /* get high word */ + move.w %d7,(%a1)+ /* store word */ + lsr.l %d1,%d7 /* get moved-up low byte */ + move.b %d7,(%a1)+ /* store byte */ + jra .lines_end + + /* byte aligned destination (long + 1): use line burst reads in main loop */ +.lines_do1_start: + moveq.l #24,%d1 /* shift count for shifting by 3 bytes */ + cmp.l %a0,%d0 /* any leading longwords? */ + jhi .lines_do1_head_start /* yes: leading longword copy */ + + movem.l (%a0),%d4-%d7 /* load first line */ + lea.l (16,%a0),%a0 + move.l %d4,%d2 /* first longword, bytes 3210 */ + lsr.l #8,%d2 /* first longword, bytes .321 */ + swap %d2 /* first longword, bytes 21.3 */ + move.b %d2,(%a1)+ /* store byte */ + swap %d2 /* first longword, bytes .321 */ + move.w %d2,(%a1)+ /* store word */ + jra .lines_do1_entry + +.lines_do1_head_start: + move.l (%a0)+,%d7 /* load first longword */ + move.l %d7,%d2 /* first longword, bytes 3210 */ + lsr.l #8,%d2 /* first longword, bytes .321 */ + swap %d2 /* first longword, bytes 21.3 */ + move.b %d2,(%a1)+ /* store byte */ + swap %d2 /* first longword, bytes .321 */ + move.w %d2,(%a1)+ /* store word */ + jra .lines_do1_head_entry + +.lines_do1_head_loop: + move.l %d7,%d6 /* move old longword away */ + move.l (%a0)+,%d7 /* load new longword */ + move.l %d7,%d2 + lsr.l #8,%d2 /* get upper 3 bytes */ + or.l %d2,%d6 /* combine with old low byte */ + move.l %d6,(%a1)+ /* store longword */ +.lines_do1_head_entry: + lsl.l %d1,%d7 /* shift up low byte */ + cmp.l %a0,%d0 /* runs %a0 up to first line bound */ + jhi .lines_do1_head_loop + +.lines_do1_loop: + move.l %d7,%d3 /* move last longword of old line away */ + movem.l (%a0),%d4-%d7 /* load new line */ + lea.l (16,%a0),%a0 + move.l %d4,%d2 + lsr.l #8,%d2 /* get upper 3 bytes of 1st longword */ + or.l %d2,%d3 /* combine with low byte of old longword */ + move.l %d3,(%a1)+ /* store longword */ +.lines_do1_entry: + lsl.l %d1,%d4 /* shift up low byte */ + move.l %d5,%d2 + lsr.l #8,%d2 /* get upper 3 bytes of 2nd longword */ + or.l %d2,%d4 /* combine with low byte of 1st longword */ + move.l %d4,(%a1)+ /* store longword */ + lsl.l %d1,%d5 /* shift up low byte */ + move.l %d6,%d2 + lsr.l #8,%d2 /* get upper 3 bytes of 3rd longword */ + or.l %d2,%d5 /* combine with low byte of 2nd longword */ + move.l %d5,(%a1)+ /* store longword */ + lsl.l %d1,%d6 /* shift up low byte */ + move.l %d7,%d2 + lsr.l #8,%d2 /* get upper 3 bytes of 4th longword */ + or.l %d2,%d6 /* combine with low byte of 4th longword */ + move.l %d6,(%a1)+ /* store longword */ + lsl.l %d1,%d7 /* shift up low byte */ + cmp.l %a0,%a2 /* runs %a0 up to last line bound */ + jhi .lines_do1_loop + + lea.l (12,%a2),%a2 /* readjust end address for doing longwords */ + cmp.l %a0,%a2 /* any trailing longwords? */ + jls .lines_do1_tail_end /* no: just store last low byte */ + +.lines_do1_tail_loop: + move.l %d7,%d6 /* move old longword away */ + move.l (%a0)+,%d7 /* load new longword */ + move.l %d7,%d2 + lsr.l #8,%d2 /* get upper 3 bytes */ + or.l %d2,%d6 /* combine with old low byte */ + move.l %d6,(%a1)+ /* store longword */ + lsl.l %d1,%d7 /* shift up low byte */ + cmp.l %a0,%a2 /* runs %a0 up to last long bound */ + jhi .lines_do1_tail_loop + +.lines_do1_tail_end: + lsr.l %d1,%d7 /* get shifted-up low byte */ + move.b %d7,(%a1)+ /* store byte */ + jra .lines_end + + /* long aligned destination (line + 0/4/8/12): head */ +.lines_do0_head_loop: + move.l (%a0)+,(%a1)+ /* copy longword */ +.lines_do0_start: + cmp.l %a0,%d0 /* runs %a0 up to first line bound */ + jhi .lines_do0_head_loop + +.lines_do0_head_end: + move.l %a1,%d1 + lsr.l #2,%d1 + moveq.l #3,%d0 /* mask */ + and.l %d0,%d1 + moveq.l #16,%d0 /* address increment for one main loop pass */ + jmp.l (2,%pc,%d1.l*2) /* switch ((dest_addr >> 2) & 3) */ + bra.b .lines_lo0_start + bra.b .lines_lo4_start + bra.b .lines_lo8_start + /* bra.b .lines_lo12_start implicit */ + + /* long aligned destination (line + 12): use line bursts in the loop */ +.lines_lo12_start: + movem.l (%a0),%d4-%d7 /* load first line */ + add.l %d0,%a0 + move.l %d4,(%a1)+ /* store 1st longword */ + cmp.l %a0,%a2 /* any full lines? */ + jls .lines_lo12_end /* no: skip main loop */ + +.lines_lo12_loop: + move.l %d5,%d1 /* move last 3 longwords of old line away */ + move.l %d6,%d2 + move.l %d7,%d3 + movem.l (%a0),%d4-%d7 /* load new line */ + add.l %d0,%a0 + movem.l %d1-%d4,(%a1) /* store line (3 old + 1 new longwords) */ + add.l %d0,%a1 + cmp.l %a0,%a2 /* runs %a0 up to last line bound */ + jhi .lines_lo12_loop + + /* long aligned destination (line + 0/4/8/12): tail */ +.lines_lo12_end: + move.l %d5,(%a1)+ /* store 3rd last longword */ +.lines_lo8_end: + move.l %d6,(%a1)+ /* store 2nd last longword */ +.lines_lo4_end: + move.l %d7,(%a1)+ /* store last longword */ +.lines_lo0_end: + lea.l (12,%a2),%a2 /* readjust end address for doing longwords */ + cmp.l %a0,%a2 /* any trailing longwords? */ + jls .lines_end /* no: get outta here */ + +.lines_do0_tail_loop: + move.l (%a0)+,(%a1)+ /* copy longword */ + cmp.l %a0,%a2 /* runs %a0 up to last long bound */ + jhi .lines_do0_tail_loop + + jra .lines_end + + /* line aligned destination: use line bursts in the loop */ +.lines_lo0_start: +.lines_lo0_loop: + movem.l (%a0),%d4-%d7 /* load line */ + add.l %d0,%a0 + movem.l %d4-%d7,(%a1) /* store line */ + add.l %d0,%a1 + cmp.l %a0,%a2 /* runs %a0 up to last line bound */ + jhi .lines_lo0_loop + + jra .lines_lo0_end /* handle trailing longwords */ + + /* long aligned destination (line + 4): use line bursts in the loop */ +.lines_lo4_start: + movem.l (%a0),%d4-%d7 /* load first line */ + add.l %d0,%a0 + move.l %d4,(%a1)+ /* store 1st longword */ + move.l %d5,(%a1)+ /* store 2nd longword */ + move.l %d6,(%a1)+ /* store 3rd longword */ + cmp.l %a0,%a2 /* any full lines? */ + jls .lines_lo4_end /* no: skip main loop */ + +.lines_lo4_loop: + move.l %d7,%d3 /* move last longword of old line away */ + movem.l (%a0),%d4-%d7 /* load new line */ + add.l %d0,%a0 + movem.l %d3-%d6,(%a1) /* store line (1 old + 3 new longwords) */ + add.l %d0,%a1 + cmp.l %a0,%a2 /* runs %a0 up to last line bound */ + jhi .lines_lo4_loop + + jra .lines_lo4_end /* handle trailing longwords */ + + /* long aligned destination (line + 8): use line bursts in the loop */ +.lines_lo8_start: + movem.l (%a0),%d4-%d7 /* load first line */ + add.l %d0,%a0 + move.l %d4,(%a1)+ /* store 1st longword */ + move.l %d5,(%a1)+ /* store 2nd longword */ + cmp.l %a0,%a2 + jls .lines_lo8_end + +.lines_lo8_loop: + move.l %d6,%d2 /* move last 2 longwords of old line away */ + move.l %d7,%d3 + movem.l (%a0),%d4-%d7 /* load new line */ + add.l %d0,%a0 + movem.l %d2-%d5,(%a1) /* store line (2 old + 2 new longwords) */ + add.l %d0,%a1 + cmp.l %a0,%a2 /* runs %a0 up to last line bound */ + jhi .lines_lo8_loop + + jra .lines_lo8_end /* handle trailing longwords */ + +#ifdef FULLSPEED + + /* word aligned destination (line + 2/6/10/14): head */ +.lines_do2_start: + cmp.l %a0,%d0 /* any leading longwords? */ + jls .lines_do2_selector /* no: jump to mainloop selector */ + + move.l (%a0)+,%d7 /* load first longword */ + swap %d7 /* swap words */ + move.w %d7,(%a1)+ /* store high word */ + cmp.l %a0,%d0 /* any more longword? */ + jls .lines_do2_head_end /* no: skip head loop */ + +.lines_do2_head_loop: + move.l %d7,%d6 /* move old longword away */ + move.l (%a0)+,%d7 /* load new longword */ + swap %d7 /* swap words */ + move.w %d7,%d6 /* combine high word with old low word */ + move.l %d6,(%a1)+ /* store longword */ + cmp.l %a0,%d0 /* runs %a0 up to first line bound */ + jhi .lines_do2_head_loop + +.lines_do2_head_end: + swap %d7 /* undo swap */ + move.w %d7,(%a1)+ /* store word */ + +.lines_do2_selector: + move.l %a1,%d1 + lsr.l #2,%d1 + moveq.l #3,%d0 /* mask */ + and.l %d0,%d1 + moveq.l #16,%d0 /* address increment for one main loop pass */ + jmp.l (2,%pc,%d1.l*4) /* switch ((dest_addr >> 2) & 3) */ + bra.w .lines_lo2_start + bra.w .lines_lo6_start + bra.w .lines_lo10_start + /* bra.w .lines_lo14_start implicit */ + + /* word aligned destination (line + 14): use line bursts in the loop */ +.lines_lo14_start: + movem.l (%a0),%d4-%d7 /* load first line */ + add.l %d0,%a0 + swap %d4 /* swap words of 1st long */ + move.w %d4,(%a1)+ /* store word */ + jra .lines_lo14_entry /* jump into main loop */ + +.lines_lo14_loop: + move.l %d4,%d0 /* move old line away */ + move.l %d5,%d1 + move.l %d6,%d2 + move.l %d7,%d3 + movem.l (%a0),%d4-%d7 /* load new line */ + lea.l (16,%a0),%a0 + swap %d4 /* swap words of 1st long */ + move.w %d4,%d3 /* combine 1st high word with old low word */ + movem.l %d0-%d3,(%a1) /* store line */ + lea.l (16,%a1),%a1 +.lines_lo14_entry: + swap %d5 /* swap words of 2nd long */ + move.w %d5,%d4 /* combine 2nd high word with 1st low word */ + swap %d6 /* swap words of 3rd long */ + move.w %d6,%d5 /* combine 3rd high word with 2nd low word */ + swap %d7 /* swap words of 4th long */ + move.w %d7,%d6 /* combine 4th high word with 3rd low word */ + cmp.l %a0,%a2 /* runs %a0 up to last line bound */ + jhi .lines_lo14_loop + + /* word aligned destination (line + 2/6/10/14): tail */ +.lines_lo14_end: + move.l %d4,(%a1)+ /* store third last longword */ +.lines_lo10_end: + move.l %d5,(%a1)+ /* store second last longword */ +.lines_lo6_end: + move.l %d6,(%a1)+ /* store last longword */ +.lines_lo2_end: + lea.l (12,%a2),%a2 /* readjust end address for doing longwords */ + cmp.l %a0,%a2 /* any trailing longwords? */ + jls .lines_do2_tail_end /* no: skip tail loop */ + +.lines_do2_tail_loop: + move.l %d7,%d6 /* move old longword away */ + move.l (%a0)+,%d7 /* load new longword */ + swap %d7 /* swap words */ + move.w %d7,%d6 /* combine high word with old low word */ + move.l %d6,(%a1)+ /* store longword */ + cmp.l %a0,%a2 /* runs %a0 up to last long bound */ + jhi .lines_do2_tail_loop + +.lines_do2_tail_end: + swap %d7 /* undo swap */ + move.w %d7,(%a1)+ /* store last word */ + jra .lines_end + + /* word aligned destination (line + 2): use line bursts in the loop */ +.lines_lo2_start: + movem.l (%a0),%d4-%d7 /* load first line */ + add.l %d0,%a0 + swap %d4 /* swap words of 1st long */ + move.w %d4,(%a1)+ /* store high word */ + swap %d5 /* swap words of 2nd long */ + move.w %d5,%d4 /* combine 2nd high word with 1st low word */ + swap %d6 /* swap words of 3rd long */ + move.w %d6,%d5 /* combine 3nd high word with 2nd low word */ + swap %d7 /* swap words of 4th long */ + move.w %d7,%d6 /* combine 4th high word with 3rd low word */ + move.l %d4,(%a1)+ /* store 1st longword */ + move.l %d5,(%a1)+ /* store 2nd longword */ + move.l %d6,(%a1)+ /* store 3rd longword */ + cmp.l %a0,%a2 /* any full lines? */ + jls .lines_lo2_end /* no: skip main loop */ + +.lines_lo2_loop: + move.l %d7,%d3 /* move last longword of old line away */ + movem.l (%a0),%d4-%d7 /* load line */ + add.l %d0,%a0 + swap %d4 /* swap words of 1st long */ + move.w %d4,%d3 /* combine 1st high word with old low word */ + swap %d5 /* swap words of 2nd long */ + move.w %d5,%d4 /* combine 2nd high word with 1st low word */ + swap %d6 /* swap words of 3rd long */ + move.w %d6,%d5 /* combine 3rd high word with 2nd low word */ + swap %d7 /* swap words of 4th long */ + move.w %d7,%d6 /* combine 4th high word with 3rd low word */ + movem.l %d3-%d6,(%a1) /* store line */ + add.l %d0,%a1 + cmp.l %a0,%a2 /* runs %a0 up to last line bound */ + jhi .lines_lo2_loop + + jra .lines_lo2_end /* handle trailing longwords */ + + /* word aligned destination (line + 6): use line bursts in the loop */ +.lines_lo6_start: + movem.l (%a0),%d4-%d7 /* load first line */ + add.l %d0,%a0 + swap %d4 /* swap words of 1st long */ + move.w %d4,(%a1)+ /* store high word */ + swap %d5 /* swap words of 2nd long */ + move.w %d5,%d4 /* combine 2nd high word with 1st low word */ + swap %d6 /* swap words of 3rd long */ + move.w %d6,%d5 /* combine 3rd high word with 2nd low word */ + move.l %d4,(%a1)+ /* store 1st longword */ + move.l %d5,(%a1)+ /* store 2nd longword */ + jra .lines_lo6_entry /* jump into main loop */ + +.lines_lo6_loop: + move.l %d6,%d2 /* move last 2 longwords of old line away */ + move.l %d7,%d3 + movem.l (%a0),%d4-%d7 /* load line */ + add.l %d0,%a0 + swap %d4 /* swap words of 1st long */ + move.w %d4,%d3 /* combine 1st high word with old low word */ + swap %d5 /* swap words of 2nd long */ + move.w %d5,%d4 /* combine 2nd high word with 1st low word */ + swap %d6 /* swap words of 3rd long */ + move.w %d6,%d5 /* combine 3rd high word with 2nd low word */ + movem.l %d2-%d5,(%a1) /* store line */ + add.l %d0,%a1 +.lines_lo6_entry: + swap %d7 /* swap words of 4th long */ + move.w %d7,%d6 /* combine 4th high word with 3rd low word */ + cmp.l %a0,%a2 /* runs %a0 up to last line bound */ + jhi .lines_lo6_loop + + jra .lines_lo6_end /* handle trailing longwords */ + + /* word aligned destination (line + 10): use line bursts in the loop */ +.lines_lo10_start: + movem.l (%a0),%d4-%d7 /* load first line */ + add.l %d0,%a0 + swap %d4 /* swap words of 1st long */ + move.w %d4,(%a1)+ /* store high word */ + swap %d5 /* swap words of 2nd long */ + move.w %d5,%d4 /* combine 2nd high word with 1st low word */ + move.l %d4,(%a1)+ /* store 1st longword */ + jra .lines_lo10_entry /* jump into main loop */ + +.lines_lo10_loop: + move.l %d5,%d1 /* move last 3 longwords of old line away */ + move.l %d6,%d2 + move.l %d7,%d3 + movem.l (%a0),%d4-%d7 /* load line */ + add.l %d0,%a0 + swap %d4 /* swap words of 1st long */ + move.w %d4,%d3 /* combine 1st high word with old low word */ + swap %d5 /* swap words of 2nd long */ + move.w %d5,%d4 /* combine 2nd high word with 1st low word */ + movem.l %d1-%d4,(%a1) /* store line */ + add.l %d0,%a1 +.lines_lo10_entry: + swap %d6 /* swap words of 3rd long */ + move.w %d6,%d5 /* combine 3rd high word with 2nd low word */ + swap %d7 /* swap words of 4th long */ + move.w %d7,%d6 /* combine 4th high word with 3rd low word */ + cmp.l %a0,%a2 /* runs %a0 up to last line bound */ + jhi .lines_lo10_loop + + jra .lines_lo10_end /* handle trailing longwords */ + +#else /* !FULLSPEED */ + + /* word aligned destination (long + 2): use line burst reads in the loop */ +.lines_do2_start: + cmp.l %a0,%d0 /* any leading longwords? */ + jhi .lines_do2_head_start /* yes: leading longword copy */ + + movem.l (%a0),%d4-%d7 /* load first line */ + lea.l (16,%a0),%a0 + swap %d4 /* swap words of 1st long */ + move.w %d4,(%a1)+ /* store high word */ + jra .lines_do2_entry /* jump into main loop */ + +.lines_do2_head_start: + move.l (%a0)+,%d7 /* load first longword */ + swap %d7 /* swap words */ + move.w %d7,(%a1)+ /* store high word */ + cmp.l %a0,%d0 /* any full longword? */ + jls .lines_do2_loop /* no: skip head loop */ + +.lines_do2_head_loop: + move.l %d7,%d6 /* move old longword away */ + move.l (%a0)+,%d7 /* load new longword */ + swap %d7 /* swap words */ + move.w %d7,%d6 /* combine high word with old low word */ + move.l %d6,(%a1)+ /* store longword */ + cmp.l %a0,%d0 /* runs %a0 up to first line bound */ + jhi .lines_do2_head_loop + +.lines_do2_loop: + move.l %d7,%d3 /* move last longword of old line away */ + movem.l (%a0),%d4-%d7 /* load line */ + lea.l (16,%a0),%a0 + swap %d4 /* swap words of 1st long */ + move.w %d4,%d3 /* combine 1st high word with old low word */ + move.l %d3,(%a1)+ /* store 1st longword */ +.lines_do2_entry: + swap %d5 /* swap words of 2nd long */ + move.w %d5,%d4 /* combine 2nd high word with 1st low word */ + move.l %d4,(%a1)+ /* store 2nd longword */ + swap %d6 /* swap words of 3rd long */ + move.w %d6,%d5 /* combine 3rd high word with 2nd low word */ + move.l %d5,(%a1)+ /* store 3rd longword */ + swap %d7 /* swap words of 4th long */ + move.w %d7,%d6 /* combine 4th high word with 3rd low word */ + move.l %d6,(%a1)+ /* store 4th longword */ + cmp.l %a0,%a2 /* runs %a0 up to last line bound */ + jhi .lines_do2_loop + +.lines_do2_end: + lea.l (12,%a2),%a2 /* readjust end address for doing longwords */ + cmp.l %a0,%a2 /* any trailing longwords? */ + jls .lines_do2_tail_end /* no: skip tail loop */ + +.lines_do2_tail_loop: + move.l %d7,%d6 /* move old longword away */ + move.l (%a0)+,%d7 /* load new longword */ + swap %d7 /* swap words */ + move.w %d7,%d6 /* combine high word with old low word */ + move.l %d6,(%a1)+ /* store longword */ + cmp.l %a0,%a2 /* runs %a0 up to last long bound */ + jhi .lines_do2_tail_loop + +.lines_do2_tail_end: + swap %d7 /* undo swap */ + move.w %d7,(%a1)+ /* store last word */ + /* jra .lines_end implicit */ + +#endif /* !FULLSPEED */ + +.lines_end: + addq.l #3,%a2 /* readjust end address */ + move.l %a2,%d1 /* end address in %d1 again */ + movem.l (%sp),%d2-%d7/%a2 /* restore registers */ + lea.l (28,%sp),%sp + jra .bytes2_start /* jump to trailing byte loop */ + +.long_start: + subq.l #3,%d1 /* adjust end address for doing 4 bytes/ pass */ + + /* longword copy loop - no lines */ +.long_loop: + move.l (%a0)+,(%a1)+ /* copy longword (write can be unaligned) */ + cmp.l %a0,%d1 /* runs %a0 up to last long bound */ + jhi .long_loop + + addq.l #3,%d1 /* readjust end address */ + cmp.l %a0,%d1 /* any bytes left? */ + jls .bytes2_end /* no: skip trailing byte loop */ + + /* trailing byte loop */ +.bytes2_loop: + move.b (%a0)+,(%a1)+ /* copy byte */ +.bytes2_start: + cmp.l %a0,%d1 /* runs %a0 up to end address */ + jhi .bytes2_loop + +.bytes2_end: + move.l (4,%sp),%d0 /* return destination */ + rts + +.end: + .size memcpy,.end-memcpy diff --git a/firmware/asm/m68k/memmove.S b/firmware/asm/m68k/memmove.S new file mode 100644 index 0000000000..736cd619e1 --- /dev/null +++ b/firmware/asm/m68k/memmove.S @@ -0,0 +1,670 @@ +/*************************************************************************** + * __________ __ ___. + * Open \______ \ ____ ____ | | _\_ |__ _______ ___ + * Source | _// _ \_/ ___\| |/ /| __ \ / _ \ \/ / + * Jukebox | | ( <_> ) \___| < | \_\ ( <_> > < < + * Firmware |____|_ /\____/ \___ >__|_ \|___ /\____/__/\_ \ + * \/ \/ \/ \/ \/ + * $Id$ + * + * Copyright (C) 2006 by Jens Arnold + * + * This program is free software; you can redistribute it and/or + * modify it under the terms of the GNU General Public License + * as published by the Free Software Foundation; either version 2 + * of the License, or (at your option) any later version. + * + * 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 + +#define FULLSPEED /* use burst writing for word aligned destinations */ + .align 2 + .global memmove + .type memmove,@function + +/* Moves bytes of data in memory from to + * Regions may overlap. + * This version is optimized for speed, and needs the corresponding memcpy + * implementation for the forward copy branch. + * + * arguments: + * (4,%sp) - destination address + * (8,%sp) - source address + * (12,%sp) - length + * + * return value: + * %d0 - destination address (like ANSI version) + * + * register usage: + * %a0 - current source address + * %a1 - current dest address + * %a2 - source start address (in line-copy loops) + * %d0 - source start address (byte and longword copy) / data / scratch + * %d1 - data / scratch + * %d2 - data / scratch + * %d3..%d7 - data + * + * For maximum speed this routine reads and writes whole lines using burst + * move (movem.l) where possible. For byte aligned destinations (long-1 and + * long-3) it writes longwords only. Same goes for word aligned destinations + * if FULLSPEED is undefined. + */ +memmove: + move.l (4,%sp),%a1 /* Destination */ + move.l (8,%sp),%a0 /* Source */ + move.l (12,%sp),%d1 /* Length */ + + cmp.l %a0,%a1 + bhi.b .backward /* dest > src -> backward copy */ + jmp __memcpy_fwd_entry + +.backward: + move.l %a0,%d0 /* %d0 = source start */ + add.l %d1,%a0 /* %a0 = source end */ + add.l %d1,%a1 /* %a1 = destination end */ + + move.l %a0,%d1 + and.l #0xFFFFFFFC,%d1 /* %d1 = last source long bound */ + subq.l #4,%d1 + cmp.l %d0,%d1 /* at least one aligned longword to copy? */ + blo.w .bytes2r_start + + addq.l #4,%d1 /* %d1 = last source long bound */ + cmp.l %d1,%a0 /* any bytes to copy */ + jls .bytes1r_end /* no: skip byte loop */ + + /* leading byte loop: copies 0..3 bytes */ +.bytes1r_loop: + move.b -(%a0),-(%a1) /* copy byte */ + cmp.l %d1,%a0 /* runs %a0 down to last long bound */ + jhi .bytes1r_loop + +.bytes1r_end: + moveq.l #-16,%d1 + add.l %a0,%d1 + and.l #0xFFFFFFF0,%d1 /* %d1 = last source line bound - 16 */ + cmp.l %d0,%d1 /* at least one aligned line to copy? */ + blo.w .longr_start /* no: jump to longword copy loop */ + + lea.l (-28,%sp),%sp /* free up some registers */ + movem.l %d2-%d7/%a2,(%sp) + + moveq.l #16,%d2 + add.l %d2,%d1 /* %d1 = last source line bound */ + move.l %d0,%a2 /* %a2 = start address */ + lea.l (15,%a2),%a2 /* adjust start address for loops doing 16 bytes/pass */ + move.l %a1,%d0 + moveq.l #3,%d2 /* mask */ + and.l %d2,%d0 + jmp.l (2,%pc,%d0.l*4) /* switch (dest_addr & 3) */ + bra.w .lines_do0r_start + bra.w .lines_do1r_start + bra.w .lines_do2r_start + /* bra.w .lines_do3r_start implicit */ + + /* byte aligned destination (long - 1): use line burst reads in main loop */ +.lines_do3r_start: + moveq.l #24,%d0 /* shift count for shifting by 3 bytes */ + cmp.l %d1,%a0 /* any leading longwords? */ + jhi .lines_do3r_head_start /* yes: leading longword copy */ + + lea.l (-16,%a0),%a0 + movem.l (%a0),%d3-%d6 /* load initial line */ + move.l %d6,%d2 /* last longword, bytes 3210 */ + move.b %d2,-(%a1) /* store byte */ + lsr.l #8,%d2 /* last longword, bytes .321 */ + move.w %d2,-(%a1) /* store word */ + jra .lines_do3r_entry + +.lines_do3r_head_start: + move.l -(%a0),%d3 /* load initial longword */ + move.l %d3,%d2 /* bytes 3210 */ + move.b %d2,-(%a1) /* store byte */ + lsr.l #8,%d2 /* bytes .321 */ + move.w %d2,-(%a1) /* store word */ + jra .lines_do3r_head_entry + +.lines_do3r_head_loop: + move.l %d3,%d4 /* move old longword away */ + move.l -(%a0),%d3 /* load new longword */ + move.l %d3,%d2 + lsl.l #8,%d2 /* get bytes 210. */ + or.l %d2,%d4 /* combine with old high byte */ + move.l %d4,-(%a1) /* store longword */ +.lines_do3r_head_entry: + lsr.l %d0,%d3 /* shift down high byte */ + cmp.l %d1,%a0 /* run %a0 down to last line bound */ + jhi .lines_do3r_head_loop + +.lines_do3r_loop: + move.l %d3,%d7 /* move first longword of last line away */ + lea.l (-16,%a0),%a0 + movem.l (%a0),%d3-%d6 /* load new line */ + move.l %d6,%d2 + lsl.l #8,%d2 /* get bytes 210. of 4th longword */ + or.l %d2,%d7 /* combine with high byte of old longword */ + move.l %d7,-(%a1) /* store longword */ +.lines_do3r_entry: + lsr.l %d0,%d6 /* shift down high byte */ + move.l %d5,%d2 + lsl.l #8,%d2 /* get bytes 210. of 3rd longword */ + or.l %d2,%d6 /* combine with high byte of 4th longword */ + move.l %d6,-(%a1) /* store longword */ + lsr.l %d0,%d5 /* shift down high byte */ + move.l %d4,%d2 + lsl.l #8,%d2 /* get bytes 210. of 2nd longword */ + or.l %d2,%d5 /* combine with high byte or 3rd longword */ + move.l %d5,-(%a1) /* store longword */ + lsr.l %d0,%d4 /* shift down high byte */ + move.l %d3,%d2 + lsl.l #8,%d2 /* get bytes 210. of 1st longword */ + or.l %d2,%d4 /* combine with high byte of 2nd longword */ + move.l %d4,-(%a1) /* store longword */ + lsr.l %d0,%d3 /* shift down high byte */ + cmp.l %a2,%a0 /* run %a0 down to first line bound */ + jhi .lines_do3r_loop + + lea.l (-12,%a2),%a2 /* readjust start address for doing longwords */ + cmp.l %a2,%a0 /* any trailing longwords? */ + jls .lines_do3r_tail_end /* no: just store last high byte */ + +.lines_do3r_tail_loop: + move.l %d3,%d4 /* move old longword away */ + move.l -(%a0),%d3 /* load new longword */ + move.l %d3,%d2 + lsl.l #8,%d2 /* get bytes 210. */ + or.l %d2,%d4 /* combine with old high byte */ + move.l %d4,-(%a1) /* store longword */ + lsr.l %d0,%d3 /* shift down high byte */ + cmp.l %a2,%a0 /* run %a0 down to first long bound */ + jhi .lines_do3r_tail_loop + +.lines_do3r_tail_end: + move.b %d3,-(%a1) /* store shifted-down high byte */ + jra .linesr_end + + /* byte aligned destination (long - 3): use line burst reads in main loop */ +.lines_do1r_start: + moveq.l #24,%d0 /* shift count for shifting by 3 bytes */ + cmp.l %d1,%a0 /* any leading longwords? */ + jhi .lines_do1r_head_start /* yes: leading longword copy */ + + lea.l (-16,%a0),%a0 + movem.l (%a0),%d3-%d6 /* load initial line */ + move.b %d6,-(%a1) /* store low byte of last longword */ + jra .lines_do1r_entry + +.lines_do1r_head_start: + move.l -(%a0),%d3 /* load initial longword */ + move.b %d3,-(%a1) /* store low byte */ + jra .lines_do1r_head_entry + +.lines_do1r_head_loop: + move.l %d3,%d4 /* move old longword away */ + move.l -(%a0),%d3 /* load new longword */ + move.l %d3,%d2 + lsl.l %d0,%d2 /* get low byte */ + or.l %d2,%d4 /* combine with old bytes .321 */ + move.l %d4,-(%a1) /* store longword */ +.lines_do1r_head_entry: + lsr.l #8,%d3 /* get bytes .321 */ + cmp.l %d1,%a0 /* run %a0 down to last line bound */ + jhi .lines_do1r_head_loop + +.lines_do1r_loop: + move.l %d3,%d7 /* move first longword of old line away */ + lea.l (-16,%a0),%a0 + movem.l (%a0),%d3-%d6 /* load new line */ + move.l %d6,%d2 + lsl.l %d0,%d2 /* get low byte of 4th longword */ + or.l %d2,%d7 /* combine with bytes .321 of old longword */ + move.l %d7,-(%a1) /* store longword */ +.lines_do1r_entry: + lsr.l #8,%d6 /* get bytes .321 */ + move.l %d5,%d2 + lsl.l %d0,%d2 /* get low byte of 3rd longword */ + or.l %d2,%d6 /* combine with bytes .321 of 4th longword */ + move.l %d6,-(%a1) /* store longword */ + lsr.l #8,%d5 /* get bytes .321 */ + move.l %d4,%d2 + lsl.l %d0,%d2 /* get low byte of 2nd longword */ + or.l %d2,%d5 /* combine with bytes .321 of 3rd longword */ + move.l %d5,-(%a1) /* store longword */ + lsr.l #8,%d4 /* get bytes .321 */ + move.l %d3,%d2 + lsl.l %d0,%d2 /* get low byte of 1st longword */ + or.l %d2,%d4 /* combine with bytes .321 of 2nd longword */ + move.l %d4,-(%a1) /* store longword */ + lsr.l #8,%d3 /* get bytes .321 */ + cmp.l %a2,%a0 /* run %a0 down to first line bound */ + jhi .lines_do1r_loop + + lea.l (-12,%a2),%a2 /* readjust start address for doing longwords */ + cmp.l %a2,%a0 /* any trailing longwords? */ + jls .lines_do1r_tail_end /* no: just store last high byte */ + +.lines_do1r_tail_loop: + move.l %d3,%d4 /* move old longword away */ + move.l -(%a0),%d3 /* load new longword */ + move.l %d3,%d2 + lsl.l %d0,%d2 /* get low byte */ + or.l %d2,%d4 /* combine with old bytes .321 */ + move.l %d4,-(%a1) /* store longword */ + lsr.l #8,%d3 /* get bytes .321 */ + cmp.l %a2,%a0 /* run %a0 down to first long bound */ + jhi .lines_do1r_tail_loop + +.lines_do1r_tail_end: + move.w %d3,-(%a1) /* store word 21 */ + swap %d3 + move.b %d3,-(%a1) /* store byte 3 */ + jra .linesr_end + + /* long aligned destination (line - 0/4/8/12): head */ +.lines_do0r_head_loop: + move.l -(%a0),-(%a1) /* copy longword */ +.lines_do0r_start: + cmp.l %d1,%a0 /* run %a0 down to last line bound */ + jhi .lines_do0r_head_loop + +.lines_do0r_head_end: + move.l %a1,%d1 + lsr.l #2,%d1 + moveq.l #3,%d0 /* mask */ + and.l %d0,%d1 + moveq.l #16,%d0 /* address decrement for one main loop pass */ + jmp.l (2,%pc,%d1.l*2) /* switch ((dest_addr >> 2) & 3) */ + bra.b .lines_lo0r_start + bra.b .lines_lo4r_start + bra.b .lines_lo8r_start + /* bra.b .lines_lo12r_start implicit */ + + /* long aligned destination (line - 4): use line bursts in the loop */ +.lines_lo12r_start: + sub.l %d0,%a0 + movem.l (%a0),%d1-%d4 /* load initial line */ + move.l %d4,-(%a1) /* store 4th longword */ + move.l %d3,-(%a1) /* store 3rd longword */ + move.l %d2,-(%a1) /* store 2nd longword */ + cmp.l %a2,%a0 /* any full lines? */ + jls .lines_lo12r_end /* no: skip main loop */ + +.lines_lo12r_loop: + move.l %d1,%d5 /* move first longword of old line away */ + sub.l %d0,%a0 + movem.l (%a0),%d1-%d4 /* load new line */ + sub.l %d0,%a1 + movem.l %d2-%d5,(%a1) /* store line (1 old + 3 new longwords */ + cmp.l %a2,%a0 /* run %a0 down to first line bound */ + jhi .lines_lo12r_loop + + jra .lines_lo12r_end /* handle trailing longwords */ + + /* line aligned destination: use line bursts in the loop */ +.lines_lo0r_start: +.lines_lo0r_loop: + sub.l %d0,%a0 + movem.l (%a0),%d1-%d4 /* load line */ + sub.l %d0,%a1 + movem.l %d1-%d4,(%a1) /* store line */ + cmp.l %a2,%a0 /* run %a0 down to first line bound */ + jhi .lines_lo0r_loop + + jra .lines_lo0r_end /* handle trailing longwords */ + + /* long aligned destination (line - 8): use line bursts in the loop */ +.lines_lo8r_start: + sub.l %d0,%a0 + movem.l (%a0),%d1-%d4 /* load initial line */ + move.l %d4,-(%a1) /* store 4th longword */ + move.l %d3,-(%a1) /* store 3rd longword */ + cmp.l %a2,%a0 /* any full lines? */ + jls .lines_lo8r_end /* no: skip main loop */ + +.lines_lo8r_loop: + move.l %d2,%d6 /* move first 2 longwords of old line away */ + move.l %d1,%d5 + sub.l %d0,%a0 + movem.l (%a0),%d1-%d4 /* load new line */ + sub.l %d0,%a1 + movem.l %d3-%d6,(%a1) /* store line (2 old + 2 new longwords */ + cmp.l %a2,%a0 /* run %a0 down to first line bound */ + jhi .lines_lo8r_loop + + jra .lines_lo8r_end /* handle trailing longwords */ + + /* long aligned destination (line - 12): use line bursts in the loop */ +.lines_lo4r_start: + sub.l %d0,%a0 + movem.l (%a0),%d1-%d4 /* load initial line */ + move.l %d4,-(%a1) /* store 4th longword */ + cmp.l %a2,%a0 /* any full lines? */ + jls .lines_lo4r_end /* no: skip main loop */ + +.lines_lo4r_loop: + move.l %d3,%d7 /* move first 3 longwords of old line away */ + move.l %d2,%d6 + move.l %d1,%d5 + sub.l %d0,%a0 + movem.l (%a0),%d1-%d4 /* load new line */ + sub.l %d0,%a1 + movem.l %d4-%d7,(%a1) /* store line (3 old + 1 new longwords */ + cmp.l %a2,%a0 /* run %a0 down to first line bound */ + jhi .lines_lo4r_loop + + /* long aligned destination (line - 0/4/8/12): tail */ +.lines_lo4r_end: + move.l %d3,-(%a1) /* store 3rd last longword */ +.lines_lo8r_end: + move.l %d2,-(%a1) /* store 2nd last longword */ +.lines_lo12r_end: + move.l %d1,-(%a1) /* store last longword */ +.lines_lo0r_end: + lea.l (-12,%a2),%a2 /* readjust end address for doing longwords */ + cmp.l %a2,%a0 /* any trailing longwords? */ + jls .linesr_end /* no: get outta here */ + +.lines_do0r_tail_loop: + move.l -(%a0),-(%a1) /* copy longword */ + cmp.l %a2,%a0 /* run %a0 down to first long bound */ + jhi .lines_do0r_tail_loop + + jra .linesr_end + +#ifdef FULLSPEED + /* word aligned destination (line - 2/6/10/14): head */ +.lines_do2r_start: + cmp.l %d1,%a0 /* any leading longwords? */ + jls .lines_do2r_selector /* no: jump to mainloop selector */ + + move.l -(%a0),%d3 /* load initial longword */ + move.w %d3,-(%a1) /* store low word */ + cmp.l %d1,%a0 /* any more longwords? */ + jls .lines_do2r_head_end /* no: skip head loop */ + +.lines_do2r_head_loop: + move.l %d3,%d4 /* move old longword away */ + move.l -(%a0),%d3 /* load new longword */ + move.w %d3,%d4 /* combine low word with old high word */ + swap %d4 /* swap words */ + move.l %d4,-(%a1) /* store longword */ + cmp.l %d1,%a0 /* run %a0 down to last line bound */ + jhi .lines_do2r_head_loop + +.lines_do2r_head_end: + swap %d3 /* get high word */ + move.w %d3,-(%a1) /* and store it */ + +.lines_do2r_selector: + move.l %a1,%d1 + lsr.l #2,%d1 + moveq.l #3,%d0 /* mask */ + and.l %d0,%d1 + moveq.l #16,%d7 /* address decrement for one main loop pass */ + jmp.l (2,%pc,%d1.l*4) /* switch ((dest_addr >> 2) & 3) */ + bra.w .lines_lo2r_start + bra.w .lines_lo6r_start + bra.w .lines_lo10r_start + /* bra.w .lines_lo14r_start implicit */ + + /* word aligned destination (line - 2): use line bursts in the loop */ +.lines_lo14r_start: + sub.l %d7,%a0 + movem.l (%a0),%d0-%d3 /* load initial line */ + move.w %d3,-(%a1) /* store last low word */ + move.w %d2,%d3 /* combine 3rd low word with 4th high word */ + swap %d3 /* swap words of 3rd long */ + move.w %d1,%d2 /* combine 2nd low word with 3rd high word */ + swap %d2 /* swap words of 2nd long */ + move.w %d0,%d1 /* combine 1st low word with 2nd high word */ + swap %d1 /* swap words of 1st long */ + move.l %d3,-(%a1) /* store 3rd longword */ + move.l %d2,-(%a1) /* store 2nd longword */ + move.l %d1,-(%a1) /* store 1st longword */ + cmp.l %a2,%a0 /* any full lines? */ + jls .lines_lo14r_end /* no: skip main loop */ + +.lines_lo14r_loop: + move.l %d0,%d4 /* move first longword of old line away */ + sub.l %d7,%a0 + movem.l (%a0),%d0-%d3 /* load line */ + move.w %d3,%d4 /* combine 4th low word with old high word */ + swap %d4 /* swap words of 4th long */ + move.w %d2,%d3 /* combine 3rd low word with 4th high word */ + swap %d3 /* swap words of 3rd long */ + move.w %d1,%d2 /* combine 2nd low word with 3rd high word */ + swap %d2 /* swap words of 2nd long */ + move.w %d0,%d1 /* combine 1st low word with 2nd high word */ + swap %d1 /* swap words of 1st long */ + sub.l %d7,%a1 + movem.l %d1-%d4,(%a1) /* store line */ + cmp.l %a2,%a0 /* run %a0 down to first line bound */ + jhi .lines_lo14r_loop + + jra .lines_lo14r_end /* handle trailing longwords */ + + /* word aligned destination (line - 6): use line bursts in the loop */ +.lines_lo10r_start: + sub.l %d7,%a0 + movem.l (%a0),%d0-%d3 /* load initial line */ + move.w %d3,-(%a1) /* store last low word */ + move.w %d2,%d3 /* combine 3rd low word with 4th high word */ + swap %d3 /* swap words of 3rd long */ + move.w %d1,%d2 /* combine 2nd low word with 3rd high word */ + swap %d2 /* swap words of 2nd long */ + move.l %d3,-(%a1) /* store 3rd longword */ + move.l %d2,-(%a1) /* store 2nd longword */ + jra .lines_lo10r_entry /* jump into main loop */ + +.lines_lo10r_loop: + move.l %d0,%d4 /* move first 2 longwords of old line away */ + move.l %d1,%d5 + sub.l %d7,%a0 + movem.l (%a0),%d0-%d3 /* load line */ + move.w %d3,%d4 /* combine 4th low word with old high word */ + swap %d4 /* swap words of 4th long */ + move.w %d2,%d3 /* combine 3rd low word with 4th high word */ + swap %d3 /* swap words of 3rd long */ + move.w %d1,%d2 /* combine 2nd low word with 3rd high word */ + swap %d2 /* swap words of 2nd long */ + sub.l %d7,%a1 + movem.l %d2-%d5,(%a1) /* store line */ +.lines_lo10r_entry: + move.w %d0,%d1 /* combine 1st low word with 2nd high word */ + swap %d1 /* swap words of 1st long */ + cmp.l %a2,%a0 /* run %a0 down to first line bound */ + jhi .lines_lo10r_loop + + jra .lines_lo10r_end /* handle trailing longwords */ + + /* word aligned destination (line - 10): use line bursts in the loop */ +.lines_lo6r_start: + sub.l %d7,%a0 + movem.l (%a0),%d0-%d3 /* load initial line */ + move.w %d3,-(%a1) /* store last low word */ + move.w %d2,%d3 /* combine 3rd low word with 4th high word */ + swap %d3 /* swap words of 3rd long */ + move.l %d3,-(%a1) /* store 3rd longword */ + jra .lines_lo6r_entry /* jump into main loop */ + +.lines_lo6r_loop: + move.l %d0,%d4 /* move first 3 longwords of old line away */ + move.l %d1,%d5 + move.l %d2,%d6 + sub.l %d7,%a0 + movem.l (%a0),%d0-%d3 /* load line */ + move.w %d3,%d4 /* combine 4th low word with old high word */ + swap %d4 /* swap words of 4th long */ + move.w %d2,%d3 /* combine 3rd low word with 4th high word */ + swap %d3 /* swap words of 3rd long */ + sub.l %d7,%a1 + movem.l %d3-%d6,(%a1) /* store line */ +.lines_lo6r_entry: + move.w %d1,%d2 /* combine 2nd low word with 3rd high word */ + swap %d2 /* swap words of 2nd long */ + move.w %d0,%d1 /* combine 1st low word with 2nd high word */ + swap %d1 /* swap words of 1st long */ + cmp.l %a2,%a0 /* run %a0 down to first line bound */ + jhi .lines_lo6r_loop + + jra .lines_lo6r_end /* handle trailing longwords */ + + /* word aligned destination (line - 14): use line bursts in the loop */ +.lines_lo2r_start: + sub.l %d7,%a0 + movem.l (%a0),%d0-%d3 /* load initial line */ + move.w %d3,-(%a1) /* store last low word */ + jra .lines_lo2r_entry /* jump into main loop */ + +.lines_lo2r_loop: + move.l %d0,%d4 /* move old line away */ + move.l %d1,%d5 + move.l %d2,%d6 + move.l %d3,%d7 + lea.l (-16,%a0),%a0 + movem.l (%a0),%d0-%d3 /* load line */ + move.w %d3,%d4 /* combine 4th low word with old high word */ + swap %d4 /* swap words of 4th long */ + lea.l (-16,%a1),%a1 + movem.l %d4-%d7,(%a1) /* store line */ +.lines_lo2r_entry: + move.w %d2,%d3 /* combine 3rd low word with 4th high word */ + swap %d3 /* swap words of 3rd long */ + move.w %d1,%d2 /* combine 2nd low word with 3rd high word */ + swap %d2 /* swap words of 2nd long */ + move.w %d0,%d1 /* combine 1st low word with 2nd high word */ + swap %d1 /* swap words of 1st long */ + cmp.l %a2,%a0 /* run %a0 down to first line bound */ + jhi .lines_lo2r_loop + + /* word aligned destination (line - 2/6/10/14): tail */ +.lines_lo2r_end: + move.l %d3,-(%a1) /* store third last longword */ +.lines_lo6r_end: + move.l %d2,-(%a1) /* store second last longword */ +.lines_lo10r_end: + move.l %d1,-(%a1) /* store last longword */ +.lines_lo14r_end: + lea.l (-12,%a2),%a2 /* readjust start address for doing longwords */ + cmp.l %a2,%a0 /* any trailing longwords? */ + jls .lines_do2r_tail_end /* no: skip tail loop */ + +.lines_do2r_tail_loop: + move.l %d0,%d1 /* move old longword away */ + move.l -(%a0),%d0 /* load new longword */ + move.w %d0,%d1 /* combine low word with old high word */ + swap %d1 /* swap words */ + move.l %d1,-(%a1) /* store longword */ + cmp.l %a2,%a0 /* run %a0 down to first long bound */ + jhi .lines_do2r_tail_loop + +.lines_do2r_tail_end: + swap %d0 /* get final high word */ + move.w %d0,-(%a1) /* store it */ + /* jra .linesr_end implicit */ + +#else /* !FULLSPEED */ + + /* word aligned destination (long - 2): use line burst reads in the loop */ +.lines_do2r_start: + cmp.l %d1,%a0 /* any leading longwords? */ + jhi .lines_do2r_head_start /* yes: leading longword copy */ + + lea.l (-16,%a0),%a0 + movem.l (%a0),%d3-%d6 /* load initial line */ + move.w %d6,-(%a1) /* store last low word */ + jra .lines_do2r_entry /* jump into main loop */ + +.lines_do2r_head_start: + move.l -(%a0),%d3 /* load initial longword */ + move.w %d3,-(%a1) /* store low word */ + cmp.l %d1,%a0 /* any full longword? */ + jls .lines_do2r_loop /* no: skip head loop */ + +.lines_do2r_head_loop: + move.l %d3,%d4 /* move old longword away */ + move.l -(%a0),%d3 /* load new longword */ + move.w %d3,%d4 /* combine low word with old high word */ + swap %d4 /* swap words */ + move.l %d4,-(%a1) /* store longword */ + cmp.l %d1,%a0 /* run %a0 down to last line bound */ + jhi .lines_do2r_head_loop + +.lines_do2r_loop: + move.l %d3,%d7 /* move first longword of old line away */ + lea.l (-16,%a0),%a0 + movem.l (%a0),%d3-%d6 /* load line */ + move.w %d6,%d7 /* combine 4th low word with old high word */ + swap %d7 /* swap words of 4th long */ + move.l %d7,-(%a1) /* store 4th longword */ +.lines_do2r_entry: + move.w %d5,%d6 /* combine 3rd low word with 4th high word */ + swap %d6 /* swap words of 3rd long */ + move.l %d6,-(%a1) /* store 3rd longword */ + move.w %d4,%d5 /* combine 2nd low word with 3rd high word */ + swap %d5 /* swap words of 2nd long */ + move.l %d5,-(%a1) /* store 2nd longword */ + move.w %d3,%d4 /* combine 1st low word with 2nd high word */ + swap %d4 /* swap words of 1st long */ + move.l %d4,-(%a1) /* store 1st longword */ + cmp.l %a2,%a0 /* run %a0 down to first line bound */ + jhi .lines_do2r_loop + +.lines_do2r_end: + lea.l (-12,%a2),%a2 /* readjust start address for doing longwords */ + cmp.l %a2,%a0 /* any trailing longwords? */ + jls .lines_do2r_tail_end /* no: skip tail loop */ + +.lines_do2r_tail_loop: + move.l %d3,%d4 /* move old longword away */ + move.l -(%a0),%d3 /* load new longword */ + move.w %d3,%d4 /* combine low word with old high word */ + swap %d4 /* swap words */ + move.l %d4,-(%a1) /* store longword */ + cmp.l %a2,%a0 /* run %a0 down to first long bound */ + jhi .lines_do2r_tail_loop + +.lines_do2r_tail_end: + swap %d3 /* get final high word */ + move.w %d3,-(%a1) /* store it */ + /* jra .linesr_end implicit */ + +#endif /* !FULLSPEED */ + +.linesr_end: + subq.l #3,%a2 /* readjust end address */ + move.l %a2,%d0 /* start address in %d0 again */ + movem.l (%sp),%d2-%d7/%a2 /* restore registers */ + lea.l (28,%sp),%sp + jra .bytes2r_start /* jump to trailing byte loop */ + +.longr_start: + addq.l #3,%d0 /* adjust start address for doing 4 bytes/ pass */ + + /* longword copy loop - no lines */ +.longr_loop: + move.l -(%a0),-(%a1) /* copy longword (write can be unaligned) */ + cmp.l %d0,%a0 /* runs %a0 down to first long bound */ + jhi .longr_loop + + subq.l #3,%d0 /* readjust start address */ + cmp.l %d0,%a0 /* any bytes left? */ + jls .bytes2r_end /* no: skip trailing byte loop */ + + /* trailing byte loop */ +.bytes2r_loop: + move.b -(%a0),-(%a1) /* copy byte */ +.bytes2r_start: + cmp.l %d0,%a0 /* runs %a0 down to start address */ + jhi .bytes2r_loop + +.bytes2r_end: + rts /* returns start address */ + +.end: + .size memmove,.end-memmove diff --git a/firmware/asm/m68k/memset.S b/firmware/asm/m68k/memset.S new file mode 100644 index 0000000000..839b305a05 --- /dev/null +++ b/firmware/asm/m68k/memset.S @@ -0,0 +1,152 @@ +/*************************************************************************** + * __________ __ ___. + * Open \______ \ ____ ____ | | _\_ |__ _______ ___ + * Source | _// _ \_/ ___\| |/ /| __ \ / _ \ \/ / + * Jukebox | | ( <_> ) \___| < | \_\ ( <_> > < < + * Firmware |____|_ /\____/ \___ >__|_ \|___ /\____/__/\_ \ + * \/ \/ \/ \/ \/ + * $Id$ + * + * Copyright (C) 2004 by Jens Arnold + * + * This program is free software; you can redistribute it and/or + * modify it under the terms of the GNU General Public License + * as published by the Free Software Foundation; either version 2 + * of the License, or (at your option) any later version. + * + * 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 + .global memset + .type memset,@function + +/* Fills a memory region with specified byte value + * This version is optimized for speed + * + * arguments: + * (4,%sp) - start address + * (8,%sp) - data + * (12,%sp) - length + * + * return value: + * %d0 - start address (like ANSI version) + * + * register usage: + * %d0 - data (spread to all 4 bytes when using long stores) + * %d1 - temporary / data (for burst transfer) + * %d2 - data (for burst transfer) + * %d3 - data (for burst transfer) + * %a0 - start address + * %a1 - current address (runs down from end to start) + * + * For maximum speed this routine uses both long stores and burst mode, + * storing whole lines with movem.l. The routine fills memory from end + * to start in order to ease returning the start address. + */ +memset: + move.l (4,%sp),%a0 /* start address */ + move.l (8,%sp),%d0 /* data */ + move.l (12,%sp),%a1 /* length */ + add.l %a0,%a1 /* %a1 = end address */ + + move.l %a0,%d1 + addq.l #7,%d1 + and.l #0xFFFFFFFC,%d1 /* %d1 = first long bound + 4 */ + cmp.l %d1,%a1 /* at least one aligned longword to fill? */ + blo.b .no_longs /* no, jump directly to byte loop */ + + and.l #0xFF,%d0 /* start: spread data to all 4 bytes */ + move.l %d0,%d1 + lsl.l #8,%d1 + or.l %d1,%d0 /* data now in 2 lower bytes of %d0 */ + move.l %d0,%d1 + swap %d0 + or.l %d1,%d0 /* data now in all 4 bytes of %d0 */ + + move.l %a1,%d1 + and.l #0xFFFFFFFC,%d1 /* %d1 = last long bound */ + cmp.l %d1,%a1 /* any bytes to set? */ + bls.b .end_b1 /* no: skip byte loop */ + + /* leading byte loop: sets 0..3 bytes */ +.loop_b1: + move.b %d0,-(%a1) /* store byte */ + cmp.l %d1,%a1 /* runs %a1 down to last long bound */ + bhi.b .loop_b1 + +.end_b1: + moveq.l #31,%d1 + add.l %a0,%d1 + and.l #0xFFFFFFF0,%d1 /* %d1 = first line bound + 16 */ + cmp.l %d1,%a1 /* at least one full line to fill? */ + blo.b .no_lines /* no, jump to longword loop */ + + mov.l %a1,%d1 + and.l #0xFFFFFFF0,%d1 /* %d1 = last line bound */ + cmp.l %d1,%a1 /* any longwords to set? */ + bls.b .end_l1 /* no: skip longword loop */ + + /* leading longword loop: sets 0..3 longwords */ +.loop_l1: + move.l %d0,-(%a1) /* store longword */ + cmp.l %d1,%a1 /* runs %a1 down to last line bound */ + bhi.b .loop_l1 + +.end_l1: + move.l %d2,-(%sp) /* free some registers */ + move.l %d3,-(%sp) + + move.l %d0,%d1 /* spread data to 4 data registers */ + move.l %d0,%d2 + move.l %d0,%d3 + lea.l (15,%a0),%a0 /* start address += 15, acct. for trl. data */ + + /* main loop: set whole lines utilising burst mode */ +.loop_line: + lea.l (-16,%a1),%a1 /* pre-decrement */ + movem.l %d0-%d3,(%a1) /* store line */ + cmp.l %a0,%a1 /* runs %a1 down to first line bound */ + bhi.b .loop_line + + lea.l (-15,%a0),%a0 /* correct start address */ + move.l (%sp)+,%d3 /* restore registers */ + move.l (%sp)+,%d2 + + move.l %a0,%d1 /* %d1 = start address ... */ + addq.l #3,%d1 /* ... +3, account for possible trailing bytes */ + cmp.l %d1,%a1 /* any longwords left */ + bhi.b .loop_l2 /* yes: jump to longword loop */ + bra.b .no_longs /* no: skip loop */ + +.no_lines: + move.l %a0,%d1 /* %d1 = start address ... */ + addq.l #3,%d1 /* ... +3, account for possible trailing bytes */ + + /* trailing longword loop */ +.loop_l2: + move.l %d0,-(%a1) /* store longword */ + cmp.l %d1,%a1 /* runs %a1 down to first long bound */ + bhi.b .loop_l2 + +.no_longs: + cmp.l %a0,%a1 /* any bytes left? */ + bls.b .end_b2 /* no: skip loop */ + + /* trailing byte loop */ +.loop_b2: + move.b %d0,-(%a1) /* store byte */ + cmp.l %a0,%a1 /* runs %a1 down to start address */ + bhi.b .loop_b2 + +.end_b2: + move.l %a0,%d0 /* return start address */ + rts + +.end: + .size memset,.end-memset diff --git a/firmware/asm/m68k/memset16.S b/firmware/asm/m68k/memset16.S new file mode 100644 index 0000000000..1673038d03 --- /dev/null +++ b/firmware/asm/m68k/memset16.S @@ -0,0 +1,146 @@ +/*************************************************************************** + * __________ __ ___. + * Open \______ \ ____ ____ | | _\_ |__ _______ ___ + * Source | _// _ \_/ ___\| |/ /| __ \ / _ \ \/ / + * Jukebox | | ( <_> ) \___| < | \_\ ( <_> > < < + * Firmware |____|_ /\____/ \___ >__|_ \|___ /\____/__/\_ \ + * \/ \/ \/ \/ \/ + * $Id$ + * + * Copyright (C) 2006 by Jens Arnold + * + * This program is free software; you can redistribute it and/or + * modify it under the terms of the GNU General Public License + * as published by the Free Software Foundation; either version 2 + * of the License, or (at your option) any later version. + * + * 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 + + .global memset16 + .type memset16,@function + +/* Fills a memory region with specified word value + * Start address must be word aligned, length is in words + * This version is optimized for speed + * + * arguments: + * (4,%sp) - start address + * (8,%sp) - data + * (12,%sp) - length + * + * return value: + * %d0 - start address + * + * register usage: + * %d0 - data (spread to both words when using long stores) + * %d1 - temporary / data (for burst transfer) + * %d2 - data (for burst transfer) + * %d3 - data (for burst transfer) + * %a0 - start address + * %a1 - current address (runs down from end to start) + * + * For maximum speed this routine uses both long stores and burst mode, + * storing whole lines with movem.l. The routine fills memory from end + * to start in order to ease returning the start address. + */ +memset16: + move.l (4,%sp),%a0 /* start address */ + move.l (8,%sp),%d0 /* data */ + move.l (12,%sp),%a1 /* length */ + add.l %a1,%a1 + add.l %a0,%a1 /* %a1 = end address */ + + move.l %a0,%d1 + addq.l #6,%d1 + and.l #0xFFFFFFFC,%d1 /* %d1 = first long bound + 4 */ + cmp.l %d1,%a1 /* at least one aligned longword to fill? */ + blo.b .no_longs /* no, jump directly to word loop */ + + and.l #0xFFFF,%d0 /* start: spread data to both words */ + move.l %d0,%d1 + swap %d1 + or.l %d1,%d0 /* data now in both words */ + + move.l %a1,%d1 + and.l #0xFFFFFFFC,%d1 /* %d1 = last long bound */ + cmp.l %d1,%a1 /* one extra word? */ + bls.b .end_w1 /* no: skip */ + + move.w %d0,-(%a1) /* set leading word */ + +.end_w1: + moveq.l #30,%d1 + add.l %a0,%d1 + and.l #0xFFFFFFF0,%d1 /* %d1 = first line bound + 16 */ + cmp.l %d1,%a1 /* at least one full line to fill? */ + blo.b .no_lines /* no, jump to longword loop */ + + mov.l %a1,%d1 + and.l #0xFFFFFFF0,%d1 /* %d1 = last line bound */ + cmp.l %d1,%a1 /* any longwords to set? */ + bls.b .end_l1 /* no: skip longword loop */ + + /* leading longword loop: sets 0..3 longwords */ +.loop_l1: + move.l %d0,-(%a1) /* store longword */ + cmp.l %d1,%a1 /* runs %a1 down to last line bound */ + bhi.b .loop_l1 + +.end_l1: + move.l %d2,-(%sp) /* free some registers */ + move.l %d3,-(%sp) + + move.l %d0,%d1 /* spread data to 4 data registers */ + move.l %d0,%d2 + move.l %d0,%d3 + lea.l (14,%a0),%a0 /* start address += 14, acct. for trl. data */ + + /* main loop: set whole lines utilising burst mode */ +.loop_line: + lea.l (-16,%a1),%a1 /* pre-decrement */ + movem.l %d0-%d3,(%a1) /* store line */ + cmp.l %a0,%a1 /* runs %a1 down to first line bound */ + bhi.b .loop_line + + lea.l (-14,%a0),%a0 /* correct start address */ + move.l (%sp)+,%d3 /* restore registers */ + move.l (%sp)+,%d2 + + move.l %a0,%d1 /* %d1 = start address ... */ + addq.l #2,%d1 /* ... +2, account for possible trailing word */ + cmp.l %d1,%a1 /* any longwords left */ + bhi.b .loop_l2 /* yes: jump to longword loop */ + bra.b .no_longs /* no: skip loop */ + +.no_lines: + move.l %a0,%d1 /* %d1 = start address ... */ + addq.l #2,%d1 /* ... +2, account for possible trailing word */ + + /* trailing longword loop */ +.loop_l2: + move.l %d0,-(%a1) /* store longword */ + cmp.l %d1,%a1 /* runs %a1 down to first long bound */ + bhi.b .loop_l2 + +.no_longs: + cmp.l %a0,%a1 /* any words left? */ + bls.b .end_w2 /* no: skip loop */ + + /* trailing word loop */ +.loop_w2: + move.w %d0,-(%a1) /* store word */ + cmp.l %a0,%a1 /* runs %a1 down to start address */ + bhi.b .loop_w2 + +.end_w2: + move.l %a0,%d0 /* return start address */ + rts + +.end: + .size memset16,.end-memset16 diff --git a/firmware/asm/m68k/strlen.S b/firmware/asm/m68k/strlen.S new file mode 100644 index 0000000000..765969da04 --- /dev/null +++ b/firmware/asm/m68k/strlen.S @@ -0,0 +1,71 @@ +/*************************************************************************** + * __________ __ ___. + * Open \______ \ ____ ____ | | _\_ |__ _______ ___ + * Source | _// _ \_/ ___\| |/ /| __ \ / _ \ \/ / + * Jukebox | | ( <_> ) \___| < | \_\ ( <_> > < < + * Firmware |____|_ /\____/ \___ >__|_ \|___ /\____/__/\_ \ + * \/ \/ \/ \/ \/ + * $Id$ + * + * Copyright (C) 2010 Nils Wallménius + * + * This program is free software; you can redistribute it and/or + * modify it under the terms of the GNU General Public License + * as published by the Free Software Foundation; either version 2 + * of the License, or (at your option) any later version. + * + * This software is distributed on an "AS IS" basis, WITHOUT WARRANTY OF ANY + * KIND, either express or implied. + * + ****************************************************************************/ + +/* size_t strlen(const char *str) */ + + .section .text,"ax",@progbits + .align 2 + .globl strlen + .type strlen, @function + +strlen: + move.l 4(%sp), %a0 /* %a0 = *str */ + move.l %a0, %a1 /* %a1 = start address */ + move.l %a0, %d0 + andi.l #3, %d0 /* %d0 = %a0 & 3 */ + beq.b 1f /* already aligned */ + jmp.l (-2,%pc,%d0.l*4) + tst.b (%a0)+ + beq.b .done + tst.b (%a0)+ + beq.b .done + tst.b (%a0)+ + beq.b .done + + 1: + move.l (%a0)+, %d0 /* load %d0 increment %a0 */ + /* use trick to test the whole word for null bytes */ + move.l %d0, %d1 + subi.l #0x01010101, %d1 + not.l %d0 + and.l %d1, %d0 + andi.l #0x80808080, %d0 + beq.b 1b /* if the test was false repeat */ + + /* ok, so the last word contained a 0 byte, test individual bytes */ + subq.l #4, %a0 + tst.b (%a0)+ + beq.b .done + tst.b (%a0)+ + beq.b .done + tst.b (%a0)+ + beq.b .done + /* last byte must be 0 so we don't need to load it, so we don't increment a0 + so we jump past the subq instr */ + .word 0x51fa /* trapf.w, shadow next instr */ + +.done: + subq.l #1, %a0 /* %a0 is 1 too large due to the last increment */ + sub.l %a1, %a0 /* how many times did we repeat? */ + move.l %a0, %d0 /* return value in %d0 */ + rts + .size strlen, .-strlen + diff --git a/firmware/asm/memcpy.c b/firmware/asm/memcpy.c new file mode 100644 index 0000000000..c5456ab41f --- /dev/null +++ b/firmware/asm/memcpy.c @@ -0,0 +1,117 @@ +/* +FUNCTION + <>---copy memory regions + +ANSI_SYNOPSIS + #include + void* memcpy(void *<[out]>, const void *<[in]>, size_t <[n]>); + +TRAD_SYNOPSIS + void *memcpy(<[out]>, <[in]>, <[n]> + void *<[out]>; + void *<[in]>; + size_t <[n]>; + +DESCRIPTION + This function copies <[n]> bytes from the memory region + pointed to by <[in]> to the memory region pointed to by + <[out]>. + + If the regions overlap, the behavior is undefined. + +RETURNS + <> returns a pointer to the first byte of the <[out]> + region. + +PORTABILITY +<> is ANSI C. + +<> requires no supporting OS subroutines. + +QUICKREF + memcpy ansi pure + */ + +#include "config.h" +#include "_ansi.h" /* for _DEFUN */ +#include + +/* Nonzero if either X or Y is not aligned on a "long" boundary. */ +#define UNALIGNED(X, Y) \ + (((long)X & (sizeof (long) - 1)) | ((long)Y & (sizeof (long) - 1))) + +/* How many bytes are copied each iteration of the 4X unrolled loop. */ +#define BIGBLOCKSIZE (sizeof (long) << 2) + +/* How many bytes are copied each iteration of the word copy loop. */ +#define LITTLEBLOCKSIZE (sizeof (long)) + +/* Threshold for punting to the byte copier. */ +#define TOO_SMALL(LEN) ((LEN) < BIGBLOCKSIZE) + +_PTR +_DEFUN (memcpy, (dst0, src0, len0), + _PTR dst0 _AND + _CONST _PTR src0 _AND + size_t len0) ICODE_ATTR; + +_PTR +_DEFUN (memcpy, (dst0, src0, len0), + _PTR dst0 _AND + _CONST _PTR src0 _AND + size_t len0) +{ +#if defined(PREFER_SIZE_OVER_SPEED) || defined(__OPTIMIZE_SIZE__) + char *dst = (char *) dst0; + char *src = (char *) src0; + + _PTR save = dst0; + + while (len0--) + { + *dst++ = *src++; + } + + return save; +#else + char *dst = dst0; + _CONST char *src = src0; + long *aligned_dst; + _CONST long *aligned_src; + unsigned int len = len0; + + /* If the size is small, or either SRC or DST is unaligned, + then punt into the byte copy loop. This should be rare. */ + if (!TOO_SMALL(len) && !UNALIGNED (src, dst)) + { + aligned_dst = (long*)dst; + aligned_src = (long*)src; + + /* Copy 4X long words at a time if possible. */ + while (len >= BIGBLOCKSIZE) + { + *aligned_dst++ = *aligned_src++; + *aligned_dst++ = *aligned_src++; + *aligned_dst++ = *aligned_src++; + *aligned_dst++ = *aligned_src++; + len -= (unsigned int)BIGBLOCKSIZE; + } + + /* Copy one long word at a time if possible. */ + while (len >= LITTLEBLOCKSIZE) + { + *aligned_dst++ = *aligned_src++; + len -= LITTLEBLOCKSIZE; + } + + /* Pick up any residual with a byte copier. */ + dst = (char*)aligned_dst; + src = (char*)aligned_src; + } + + while (len--) + *dst++ = *src++; + + return dst0; +#endif /* not PREFER_SIZE_OVER_SPEED */ +} diff --git a/firmware/asm/memmove.c b/firmware/asm/memmove.c new file mode 100644 index 0000000000..5f423964bb --- /dev/null +++ b/firmware/asm/memmove.c @@ -0,0 +1,147 @@ +/* +FUNCTION + <>---move possibly overlapping memory + +INDEX + memmove + +ANSI_SYNOPSIS + #include + void *memmove(void *<[dst]>, const void *<[src]>, size_t <[length]>); + +TRAD_SYNOPSIS + #include + void *memmove(<[dst]>, <[src]>, <[length]>) + void *<[dst]>; + void *<[src]>; + size_t <[length]>; + +DESCRIPTION + This function moves <[length]> characters from the block of + memory starting at <<*<[src]>>> to the memory starting at + <<*<[dst]>>>. <> reproduces the characters correctly + at <<*<[dst]>>> even if the two areas overlap. + + +RETURNS + The function returns <[dst]> as passed. + +PORTABILITY +<> is ANSI C. + +<> requires no supporting OS subroutines. + +QUICKREF + memmove ansi pure +*/ + +#include "config.h" +#include <_ansi.h> +#include + +/* Nonzero if either X or Y is not aligned on a "long" boundary. */ +#define UNALIGNED(X, Y) \ + (((long)X & (sizeof (long) - 1)) | ((long)Y & (sizeof (long) - 1))) + +/* How many bytes are copied each iteration of the 4X unrolled loop. */ +#define BIGBLOCKSIZE (sizeof (long) << 2) + +/* How many bytes are copied each iteration of the word copy loop. */ +#define LITTLEBLOCKSIZE (sizeof (long)) + +/* Threshhold for punting to the byte copier. */ +#define TOO_SMALL(LEN) ((LEN) < BIGBLOCKSIZE) + +_PTR +_DEFUN (memmove, (dst_void, src_void, length), + _PTR dst_void _AND + _CONST _PTR src_void _AND + size_t length) ICODE_ATTR; + +_PTR +_DEFUN (memmove, (dst_void, src_void, length), + _PTR dst_void _AND + _CONST _PTR src_void _AND + size_t length) +{ +#if defined(PREFER_SIZE_OVER_SPEED) || defined(__OPTIMIZE_SIZE__) + char *dst = dst_void; + _CONST char *src = src_void; + + if (src < dst && dst < src + length) + { + /* Have to copy backwards */ + src += length; + dst += length; + while (length--) + { + *--dst = *--src; + } + } + else + { + while (length--) + { + *dst++ = *src++; + } + } + + return dst_void; +#else + char *dst = dst_void; + _CONST char *src = src_void; + long *aligned_dst; + _CONST long *aligned_src; + unsigned int len = length; + + if (src < dst && dst < src + len) + { + /* Destructive overlap...have to copy backwards */ + src += len; + dst += len; + while (len--) + { + *--dst = *--src; + } + } + else + { + /* Use optimizing algorithm for a non-destructive copy to closely + match memcpy. If the size is small or either SRC or DST is unaligned, + then punt into the byte copy loop. This should be rare. */ + if (!TOO_SMALL(len) && !UNALIGNED (src, dst)) + { + aligned_dst = (long*)dst; + aligned_src = (long*)src; + + /* Copy 4X long words at a time if possible. */ + while (len >= BIGBLOCKSIZE) + { + *aligned_dst++ = *aligned_src++; + *aligned_dst++ = *aligned_src++; + *aligned_dst++ = *aligned_src++; + *aligned_dst++ = *aligned_src++; + len -= BIGBLOCKSIZE; + } + + /* Copy one long word at a time if possible. */ + while (len >= LITTLEBLOCKSIZE) + { + *aligned_dst++ = *aligned_src++; + len -= LITTLEBLOCKSIZE; + } + + /* Pick up any residual with a byte copier. */ + dst = (char*)aligned_dst; + src = (char*)aligned_src; + } + + while (len--) + { + *dst++ = *src++; + } + } + + return dst_void; +#endif /* not PREFER_SIZE_OVER_SPEED */ +} diff --git a/firmware/asm/memset.c b/firmware/asm/memset.c new file mode 100644 index 0000000000..7b8d2137e8 --- /dev/null +++ b/firmware/asm/memset.c @@ -0,0 +1,110 @@ +/* +FUNCTION + <>---set an area of memory + +INDEX + memset + +ANSI_SYNOPSIS + #include + void *memset(const void *<[dst]>, int <[c]>, size_t <[length]>); + +TRAD_SYNOPSIS + #include + void *memset(<[dst]>, <[c]>, <[length]>) + void *<[dst]>; + int <[c]>; + size_t <[length]>; + +DESCRIPTION + This function converts the argument <[c]> into an unsigned + char and fills the first <[length]> characters of the array + pointed to by <[dst]> to the value. + +RETURNS + <> returns the value of <[m]>. + +PORTABILITY +<> is ANSI C. + + <> requires no supporting OS subroutines. + +QUICKREF + memset ansi pure +*/ + +#include +#include "_ansi.h" + +#define LBLOCKSIZE (sizeof(long)) +#define UNALIGNED(X) ((long)X & (LBLOCKSIZE - 1)) +#define TOO_SMALL(LEN) ((LEN) < LBLOCKSIZE) + +_PTR +_DEFUN (memset, (m, c, n), + _PTR m _AND + int c _AND + size_t n) +{ +#if defined(PREFER_SIZE_OVER_SPEED) || defined(__OPTIMIZE_SIZE__) + char *s = (char *) m; + + while (n-- != 0) + { + *s++ = (char) c; + } + + return m; +#else + char *s = (char *) m; + unsigned int i; + unsigned long buffer; + unsigned long *aligned_addr; + + if (!TOO_SMALL (n) && !UNALIGNED (m)) + { + /* If we get this far, we know that n is large and m is word-aligned. */ + + aligned_addr = (unsigned long*)m; + + /* Store C into each char sized location in BUFFER so that + we can set large blocks quickly. */ + c &= 0xff; + if (LBLOCKSIZE == 4) + { + buffer = (c << 8) | c; + buffer |= (buffer << 16); + } + else + { + buffer = 0; + for (i = 0; i < LBLOCKSIZE; i++) + buffer = (buffer << 8) | c; + } + + while (n >= LBLOCKSIZE*4) + { + *aligned_addr++ = buffer; + *aligned_addr++ = buffer; + *aligned_addr++ = buffer; + *aligned_addr++ = buffer; + n -= 4*LBLOCKSIZE; + } + + while (n >= LBLOCKSIZE) + { + *aligned_addr++ = buffer; + n -= LBLOCKSIZE; + } + /* Pick up the remainder with a bytewise loop. */ + s = (char*)aligned_addr; + } + + while (n--) + { + *s++ = (char)c; + } + + return m; +#endif /* not PREFER_SIZE_OVER_SPEED */ +} diff --git a/firmware/asm/memset16.c b/firmware/asm/memset16.c new file mode 100644 index 0000000000..7e31df0cdd --- /dev/null +++ b/firmware/asm/memset16.c @@ -0,0 +1,78 @@ +/*************************************************************************** + * __________ __ ___. + * Open \______ \ ____ ____ | | _\_ |__ _______ ___ + * Source | _// _ \_/ ___\| |/ /| __ \ / _ \ \/ / + * Jukebox | | ( <_> ) \___| < | \_\ ( <_> > < < + * Firmware |____|_ /\____/ \___ >__|_ \|___ /\____/__/\_ \ + * \/ \/ \/ \/ \/ + * $Id$ + * + * Copyright (C) 2006 by Jens Arnold + * + * This program is free software; you can redistribute it and/or + * modify it under the terms of the GNU General Public License + * as published by the Free Software Foundation; either version 2 + * of the License, or (at your option) any later version. + * + * This software is distributed on an "AS IS" basis, WITHOUT WARRANTY OF ANY + * KIND, either express or implied. + * + ****************************************************************************/ + +#include "string-extra.h" /* memset16() */ + +#define LBLOCKSIZE (sizeof(long)/2) +#define UNALIGNED(X) ((long)X & (sizeof(long) - 1)) +#define TOO_SMALL(LEN) ((LEN) < LBLOCKSIZE) + +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; +#else + unsigned short *p = (unsigned short *)dst; + unsigned int i; + unsigned long buffer; + unsigned long *aligned_addr; + + if (!TOO_SMALL(len) && !UNALIGNED(dst)) + { + aligned_addr = (unsigned long *)dst; + + val &= 0xffff; + if (LBLOCKSIZE == 2) + { + buffer = (val << 16) | val; + } + else + { + buffer = 0; + for (i = 0; i < LBLOCKSIZE; i++) + buffer = (buffer << 16) | val; + } + + while (len >= LBLOCKSIZE*4) + { + *aligned_addr++ = buffer; + *aligned_addr++ = buffer; + *aligned_addr++ = buffer; + *aligned_addr++ = buffer; + len -= 4*LBLOCKSIZE; + } + + while (len >= LBLOCKSIZE) + { + *aligned_addr++ = buffer; + len -= LBLOCKSIZE; + } + + p = (unsigned short *)aligned_addr; + } + + while (len--) + *p++ = val; +#endif /* not PREFER_SIZE_OVER_SPEED */ +} diff --git a/firmware/asm/mips/memcpy.S b/firmware/asm/mips/memcpy.S new file mode 100644 index 0000000000..2e7f245c69 --- /dev/null +++ b/firmware/asm/mips/memcpy.S @@ -0,0 +1,143 @@ +/*************************************************************************** + * __________ __ ___. + * Open \______ \ ____ ____ | | _\_ |__ _______ ___ + * Source | _// _ \_/ ___\| |/ /| __ \ / _ \ \/ / + * Jukebox | | ( <_> ) \___| < | \_\ ( <_> > < < + * Firmware |____|_ /\____/ \___ >__|_ \|___ /\____/__/\_ \ + * \/ \/ \/ \/ \/ + * $Id$ + * + * Copyright (C) 2002, 2003 Free Software Foundation, Inc. + * This file was originally part of the GNU C Library + * Contributed to glibc by Hartvig Ekner , 2002 + * Adapted for Rockbox by Maurus Cuelenaere, 2009 + * + * This program is free software; you can redistribute it and/or + * modify it under the terms of the GNU General Public License + * as published by the Free Software Foundation; either version 2 + * of the License, or (at your option) any later version. + * + * This software is distributed on an "AS IS" basis, WITHOUT WARRANTY OF ANY + * KIND, either express or implied. + * + ****************************************************************************/ + +#include "config.h" +#include "mips.h" + +/* void *memcpy(void *s1, const void *s2, size_t n); */ + +#ifdef ROCKBOX_BIG_ENDIAN +# define LWHI lwl /* high part is left in big-endian */ +# define SWHI swl /* high part is left in big-endian */ +# define LWLO lwr /* low part is right in big-endian */ +# define SWLO swr /* low part is right in big-endian */ +#else +# define LWHI lwr /* high part is right in little-endian */ +# define SWHI swr /* high part is right in little-endian */ +# define LWLO lwl /* low part is left in little-endian */ +# define SWLO swl /* low part is left in little-endian */ +#endif + + .section .icode, "ax", %progbits + + .global memcpy + .type memcpy, %function + + .set noreorder + +memcpy: + slti t0, a2, 8 # Less than 8? + bne t0, zero, last8 + move v0, a0 # Setup exit value before too late + + xor t0, a1, a0 # Find a0/a1 displacement + andi t0, 0x3 + bne t0, zero, shift # Go handle the unaligned case + subu t1, zero, a1 + andi t1, 0x3 # a0/a1 are aligned, but are we + beq t1, zero, chk8w # starting in the middle of a word? + subu a2, t1 + LWHI t0, 0(a1) # Yes we are... take care of that + addu a1, t1 + SWHI t0, 0(a0) + addu a0, t1 + +chk8w: + andi t0, a2, 0x1f # 32 or more bytes left? + beq t0, a2, chk1w + subu a3, a2, t0 # Yes + addu a3, a1 # a3 = end address of loop + move a2, t0 # a2 = what will be left after loop +lop8w: + lw t0, 0(a1) # Loop taking 8 words at a time + lw t1, 4(a1) + lw t2, 8(a1) + lw t3, 12(a1) + lw t4, 16(a1) + lw t5, 20(a1) + lw t6, 24(a1) + lw t7, 28(a1) + addiu a0, 32 + addiu a1, 32 + sw t0, -32(a0) + sw t1, -28(a0) + sw t2, -24(a0) + sw t3, -20(a0) + sw t4, -16(a0) + sw t5, -12(a0) + sw t6, -8(a0) + bne a1, a3, lop8w + sw t7, -4(a0) + +chk1w: + andi t0, a2, 0x3 # 4 or more bytes left? + beq t0, a2, last8 + subu a3, a2, t0 # Yes, handle them one word at a time + addu a3, a1 # a3 again end address + move a2, t0 +lop1w: + lw t0, 0(a1) + addiu a0, 4 + addiu a1, 4 + bne a1, a3, lop1w + sw t0, -4(a0) + +last8: + blez a2, lst8e # Handle last 8 bytes, one at a time + addu a3, a2, a1 +lst8l: + lb t0, 0(a1) + addiu a0, 1 + addiu a1, 1 + bne a1, a3, lst8l + sb t0, -1(a0) +lst8e: + jr ra # Bye, bye + nop + +shift: + subu a3, zero, a0 # Src and Dest unaligned + andi a3, 0x3 # (unoptimized case...) + beq a3, zero, shft1 + subu a2, a3 # a2 = bytes left + LWHI t0, 0(a1) # Take care of first odd part + LWLO t0, 3(a1) + addu a1, a3 + SWHI t0, 0(a0) + addu a0, a3 +shft1: + andi t0, a2, 0x3 + subu a3, a2, t0 + addu a3, a1 +shfth: + LWHI t1, 0(a1) # Limp through, word by word + LWLO t1, 3(a1) + addiu a0, 4 + addiu a1, 4 + bne a1, a3, shfth + sw t1, -4(a0) + b last8 # Handle anything which may be left + move a2, t0 + + .set reorder diff --git a/firmware/asm/mips/memset.S b/firmware/asm/mips/memset.S new file mode 100644 index 0000000000..8db76d9123 --- /dev/null +++ b/firmware/asm/mips/memset.S @@ -0,0 +1,239 @@ +/*************************************************************************** + * __________ __ ___. + * Open \______ \ ____ ____ | | _\_ |__ _______ ___ + * Source | _// _ \_/ ___\| |/ /| __ \ / _ \ \/ / + * Jukebox | | ( <_> ) \___| < | \_\ ( <_> > < < + * Firmware |____|_ /\____/ \___ >__|_ \|___ /\____/__/\_ \ + * \/ \/ \/ \/ \/ + * $Id$ + * + * This file was originally part of the Linux/MIPS GNU C Library + * Copyright (C) 1998 by Ralf Baechle + * Adapted for Rockbox by Maurus Cuelenaere, 2009 + * + * This program is free software; you can redistribute it and/or + * modify it under the terms of the GNU General Public License + * as published by the Free Software Foundation; either version 2 + * of the License, or (at your option) any later version. + * + * This software is distributed on an "AS IS" basis, WITHOUT WARRANTY OF ANY + * KIND, either express or implied. + * + ****************************************************************************/ + +#include "config.h" +#include "mips.h" + +#define FILL256(dst, offset, val) \ + sw val, (offset + 0x00)(dst); \ + sw val, (offset + 0x04)(dst); \ + sw val, (offset + 0x08)(dst); \ + sw val, (offset + 0x0c)(dst); \ + sw val, (offset + 0x10)(dst); \ + sw val, (offset + 0x14)(dst); \ + sw val, (offset + 0x18)(dst); \ + sw val, (offset + 0x1c)(dst); \ + sw val, (offset + 0x20)(dst); \ + sw val, (offset + 0x24)(dst); \ + sw val, (offset + 0x28)(dst); \ + sw val, (offset + 0x2c)(dst); \ + sw val, (offset + 0x30)(dst); \ + sw val, (offset + 0x34)(dst); \ + sw val, (offset + 0x38)(dst); \ + sw val, (offset + 0x3c)(dst); \ + sw val, (offset + 0x40)(dst); \ + sw val, (offset + 0x44)(dst); \ + sw val, (offset + 0x48)(dst); \ + sw val, (offset + 0x4c)(dst); \ + sw val, (offset + 0x50)(dst); \ + sw val, (offset + 0x54)(dst); \ + sw val, (offset + 0x58)(dst); \ + sw val, (offset + 0x5c)(dst); \ + sw val, (offset + 0x60)(dst); \ + sw val, (offset + 0x64)(dst); \ + sw val, (offset + 0x68)(dst); \ + sw val, (offset + 0x6c)(dst); \ + sw val, (offset + 0x70)(dst); \ + sw val, (offset + 0x74)(dst); \ + sw val, (offset + 0x78)(dst); \ + sw val, (offset + 0x7c)(dst); \ + sw val, (offset + 0x80)(dst); \ + sw val, (offset + 0x84)(dst); \ + sw val, (offset + 0x88)(dst); \ + sw val, (offset + 0x8c)(dst); \ + sw val, (offset + 0x90)(dst); \ + sw val, (offset + 0x94)(dst); \ + sw val, (offset + 0x98)(dst); \ + sw val, (offset + 0x9c)(dst); \ + sw val, (offset + 0xa0)(dst); \ + sw val, (offset + 0xa4)(dst); \ + sw val, (offset + 0xa8)(dst); \ + sw val, (offset + 0xac)(dst); \ + sw val, (offset + 0xb0)(dst); \ + sw val, (offset + 0xb4)(dst); \ + sw val, (offset + 0xb8)(dst); \ + sw val, (offset + 0xbc)(dst); \ + sw val, (offset + 0xc0)(dst); \ + sw val, (offset + 0xc4)(dst); \ + sw val, (offset + 0xc8)(dst); \ + sw val, (offset + 0xcc)(dst); \ + sw val, (offset + 0xd0)(dst); \ + sw val, (offset + 0xd4)(dst); \ + sw val, (offset + 0xd8)(dst); \ + sw val, (offset + 0xdc)(dst); \ + sw val, (offset + 0xe0)(dst); \ + sw val, (offset + 0xe4)(dst); \ + sw val, (offset + 0xe8)(dst); \ + sw val, (offset + 0xec)(dst); \ + sw val, (offset + 0xf0)(dst); \ + sw val, (offset + 0xf4)(dst); \ + sw val, (offset + 0xf8)(dst); \ + sw val, (offset + 0xfc)(dst); + +#define FILL128(dst, offset, val) \ + sw val, (offset + 0x00)(dst); \ + sw val, (offset + 0x04)(dst); \ + sw val, (offset + 0x08)(dst); \ + sw val, (offset + 0x0c)(dst); \ + sw val, (offset + 0x10)(dst); \ + sw val, (offset + 0x14)(dst); \ + sw val, (offset + 0x18)(dst); \ + sw val, (offset + 0x1c)(dst); \ + sw val, (offset + 0x20)(dst); \ + sw val, (offset + 0x24)(dst); \ + sw val, (offset + 0x28)(dst); \ + sw val, (offset + 0x2c)(dst); \ + sw val, (offset + 0x30)(dst); \ + sw val, (offset + 0x34)(dst); \ + sw val, (offset + 0x38)(dst); \ + sw val, (offset + 0x3c)(dst); \ + sw val, (offset + 0x40)(dst); \ + sw val, (offset + 0x44)(dst); \ + sw val, (offset + 0x48)(dst); \ + sw val, (offset + 0x4c)(dst); \ + sw val, (offset + 0x50)(dst); \ + sw val, (offset + 0x54)(dst); \ + sw val, (offset + 0x58)(dst); \ + sw val, (offset + 0x5c)(dst); \ + sw val, (offset + 0x60)(dst); \ + sw val, (offset + 0x64)(dst); \ + sw val, (offset + 0x68)(dst); \ + sw val, (offset + 0x6c)(dst); \ + sw val, (offset + 0x70)(dst); \ + sw val, (offset + 0x74)(dst); \ + sw val, (offset + 0x78)(dst); \ + sw val, (offset + 0x7c)(dst); + +#define FILL64(dst, offset, val) \ + sw val, (offset + 0x00)(dst); \ + sw val, (offset + 0x04)(dst); \ + sw val, (offset + 0x08)(dst); \ + sw val, (offset + 0x0c)(dst); \ + sw val, (offset + 0x10)(dst); \ + sw val, (offset + 0x14)(dst); \ + sw val, (offset + 0x18)(dst); \ + sw val, (offset + 0x1c)(dst); \ + sw val, (offset + 0x20)(dst); \ + sw val, (offset + 0x24)(dst); \ + sw val, (offset + 0x28)(dst); \ + sw val, (offset + 0x2c)(dst); \ + sw val, (offset + 0x30)(dst); \ + sw val, (offset + 0x34)(dst); \ + sw val, (offset + 0x38)(dst); \ + sw val, (offset + 0x3c)(dst); + +#define FILL32(dst, offset, val) \ + sw val, (offset + 0x00)(dst); \ + sw val, (offset + 0x04)(dst); \ + sw val, (offset + 0x08)(dst); \ + sw val, (offset + 0x0c)(dst); \ + sw val, (offset + 0x10)(dst); \ + sw val, (offset + 0x14)(dst); \ + sw val, (offset + 0x18)(dst); \ + sw val, (offset + 0x1c)(dst); + +#define FILL 64 +#define F_FILL FILL64 + + +#ifdef ROCKBOX_BIG_ENDIAN +# define SWHI swl /* high part is left in big-endian */ +#else +# define SWHI swr /* high part is right in little-endian */ +#endif + +/* + * memset(void *s, int c, size_t n) + * + * a0: start of area to clear + * a1: char to fill with + * a2: size of area to clear + */ + .section .icode, "ax", %progbits + + .global memset + .type memset, %function + + .set noreorder + .align 5 +memset: + beqz a1, 1f + move v0, a0 /* result */ + + andi a1, 0xff /* spread fillword */ + sll t1, a1, 8 + or a1, t1 + sll t1, a1, 16 + or a1, t1 +1: + + sltiu t0, a2, 4 /* very small region? */ + bnez t0, small_memset + andi t0, a0, 3 /* aligned? */ + + beqz t0, 1f + subu t0, 4 /* alignment in bytes */ + + SWHI a1, (a0) /* make word aligned */ + subu a0, t0 /* word align ptr */ + addu a2, t0 /* correct size */ + +1: ori t1, a2, (FILL-1) /* # of full blocks */ + xori t1, (FILL-1) + beqz t1, memset_partial /* no block to fill */ + andi t0, a2, (FILL-4) + + addu t1, a0 /* end address */ + .set reorder +1: addiu a0, FILL + F_FILL( a0, -FILL, a1 ) + bne t1, a0, 1b + .set noreorder + +memset_partial: + la t1, 2f /* where to start */ + subu t1, t0 + jr t1 + addu a0, t0 /* dest ptr */ + + F_FILL( a0, -FILL, a1 ) /* ... but first do words ... */ +2: andi a2, 3 /* 0 <= n <= 3 to go */ + + beqz a2, 1f + addu a0, a2 /* What's left */ + SWHI a1, -1(a0) +1: jr ra + move a2, zero + +small_memset: + beqz a2, 2f + addu t1, a0, a2 + +1: addiu a0, 1 /* fill bytewise */ + bne t1, a0, 1b + sb a1, -1(a0) + +2: jr ra /* done */ + move a2, zero + + .set reorder diff --git a/firmware/asm/sh/memcpy.S b/firmware/asm/sh/memcpy.S new file mode 100644 index 0000000000..e23a579b05 --- /dev/null +++ b/firmware/asm/sh/memcpy.S @@ -0,0 +1,219 @@ +/*************************************************************************** + * __________ __ ___. + * Open \______ \ ____ ____ | | _\_ |__ _______ ___ + * Source | _// _ \_/ ___\| |/ /| __ \ / _ \ \/ / + * Jukebox | | ( <_> ) \___| < | \_\ ( <_> > < < + * Firmware |____|_ /\____/ \___ >__|_ \|___ /\____/__/\_ \ + * \/ \/ \/ \/ \/ + * $Id$ + * + * Copyright (C) 2004-2005 by Jens Arnold + * + * This program is free software; you can redistribute it and/or + * modify it under the terms of the GNU General Public License + * as published by the Free Software Foundation; either version 2 + * of the License, or (at your option) any later version. + * + * 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 + .global _memcpy + .global ___memcpy_fwd_entry + .type _memcpy,@function + +/* Copies bytes of data in memory from to + * This version is optimized for speed + * + * arguments: + * r4 - destination address + * r5 - source address + * r6 - length + * + * return value: + * r0 - destination address (like ANSI version) + * + * register usage: + * r0 - data / scratch + * r1 - 2nd data / scratch + * r2 - scratch + * r3 - first long bound / adjusted end address (only if >= 11 bytes) + * r4 - current dest address + * r5 - current source address + * r6 - source end address + * r7 - stored dest start address + * + * The instruction order is devised in a way to utilize the pipelining + * of the SH1 to the max. The routine also tries to utilize fast page mode. + */ + +_memcpy: + mov r4,r7 /* store dest for returning */ +___memcpy_fwd_entry: + add #-8,r4 /* offset for early increment (max. 2 longs) */ + mov #11,r0 + cmp/hs r0,r6 /* at least 11 bytes to copy? (ensures 2 aligned longs) */ + add r5,r6 /* r6 = source_end */ + bf .start_b2 /* no: jump directly to byte loop */ + + mov #3,r0 + neg r5,r3 + and r0,r3 /* r3 = (4 - align_offset) % 4 */ + tst r3,r3 /* already aligned? */ + bt .end_b1 /* yes: skip leading byte loop */ + + add r5,r3 /* r3 = first source long bound */ + + /* leading byte loop: copies 0..3 bytes */ +.loop_b1: + mov.b @r5+,r0 /* load byte & increment source addr */ + add #1,r4 /* increment dest addr */ + mov.b r0,@(7,r4) /* store byte */ + cmp/hi r5,r3 /* runs r5 up to first long bound */ + bt .loop_b1 + /* now r5 is always at a long boundary */ + /* -> memory reading is done in longs for all dest alignments */ + + /* selector for main copy loop */ +.end_b1: + mov #3,r1 + and r4,r1 /* r1 = dest alignment offset */ + mova .jmptab,r0 + mov.b @(r0,r1),r1 /* select appropriate main loop */ + add r0,r1 + mov r6,r3 /* move end address to r3 */ + jmp @r1 /* and jump to it */ + add #-7,r3 /* adjust end addr for main loops doing 2 longs/pass */ + + /** main loops, copying 2 longs per pass to profit from fast page mode **/ + + /* long aligned destination (fastest) */ + .align 2 +.loop_do0: + mov.l @r5+,r1 /* load first long & increment source addr */ + add #16,r4 /* increment dest addr & account for decrementing stores */ + mov.l @r5+,r0 /* load second long & increment source addr */ + cmp/hi r5,r3 /* runs r5 up to last or second last long bound */ + mov.l r0,@-r4 /* store second long */ + mov.l r1,@-r4 /* store first long; NOT ALIGNED - no speed loss here! */ + bt .loop_do0 + + add #4,r3 /* readjust end address */ + cmp/hi r5,r3 /* one long left? */ + bf .start_b2 /* no, jump to trailing byte loop */ + + mov.l @r5+,r0 /* load last long & increment source addr */ + add #4,r4 /* increment dest addr */ + bra .start_b2 /* jump to trailing byte loop */ + mov.l r0,@(4,r4) /* store last long */ + + /* word aligned destination (long + 2) */ + .align 2 +.loop_do2: + mov.l @r5+,r1 /* load first long & increment source addr */ + add #16,r4 /* increment dest addr */ + mov.l @r5+,r0 /* load second long & increment source addr */ + cmp/hi r5,r3 /* runs r5 up to last or second last long bound */ + mov.w r0,@-r4 /* store low word of second long */ + xtrct r1,r0 /* extract low word of first long & high word of second long */ + mov.l r0,@-r4 /* and store as long */ + swap.w r1,r0 /* get high word of first long */ + mov.w r0,@-r4 /* and store it */ + bt .loop_do2 + + add #4,r3 /* readjust end address */ + cmp/hi r5,r3 /* one long left? */ + bf .start_b2 /* no, jump to trailing byte loop */ + + mov.l @r5+,r0 /* load last long & increment source addr */ + add #4,r4 /* increment dest addr */ + mov.w r0,@(6,r4) /* store low word */ + shlr16 r0 /* get high word */ + bra .start_b2 /* jump to trailing byte loop */ + mov.w r0,@(4,r4) /* and store it */ + + /* jumptable for loop selector */ + .align 2 +.jmptab: + .byte .loop_do0 - .jmptab /* placed in the middle because the SH1 */ + .byte .loop_do1 - .jmptab /* loads bytes sign-extended. Otherwise */ + .byte .loop_do2 - .jmptab /* the last loop would be out of reach */ + .byte .loop_do3 - .jmptab /* of the offset range. */ + + /* byte aligned destination (long + 1) */ + .align 2 +.loop_do1: + mov.l @r5+,r1 /* load first long & increment source addr */ + add #16,r4 /* increment dest addr */ + mov.l @r5+,r0 /* load second long & increment source addr */ + cmp/hi r5,r3 /* runs r5 up to last or second last long bound */ + mov.b r0,@-r4 /* store low byte of second long */ + shlr8 r0 /* get upper 3 bytes */ + mov r1,r2 /* copy first long */ + shll16 r2 /* move low byte of first long all the way up, .. */ + shll8 r2 + or r2,r0 /* ..combine with the 3 bytes of second long.. */ + mov.l r0,@-r4 /* ..and store as long */ + shlr8 r1 /* get middle 2 bytes */ + mov.w r1,@-r4 /* store as word */ + shlr16 r1 /* get upper byte */ + mov.b r1,@-r4 /* and store */ + bt .loop_do1 + + add #4,r3 /* readjust end address */ +.last_do13: + cmp/hi r5,r3 /* one long left? */ + bf .start_b2 /* no, jump to trailing byte loop */ + + mov.l @r5+,r0 /* load last long & increment source addr */ + add #12,r4 /* increment dest addr */ + mov.b r0,@-r4 /* store low byte */ + shlr8 r0 /* get middle 2 bytes */ + mov.w r0,@-r4 /* store as word */ + shlr16 r0 /* get upper byte */ + mov.b r0,@-r4 /* and store */ + bra .start_b2 /* jump to trailing byte loop */ + add #-4,r4 /* readjust destination */ + + /* byte aligned destination (long + 3) */ + .align 2 +.loop_do3: + mov.l @r5+,r1 /* load first long & increment source addr */ + add #16,r4 /* increment dest addr */ + mov.l @r5+,r0 /* load second long & increment source addr */ + mov r1,r2 /* copy first long */ + mov.b r0,@-r4 /* store low byte of second long */ + shlr8 r0 /* get middle 2 bytes */ + mov.w r0,@-r4 /* store as word */ + shlr16 r0 /* get upper byte */ + shll8 r2 /* move lower 3 bytes of first long one up.. */ + or r2,r0 /* ..combine with the 1 byte of second long.. */ + mov.l r0,@-r4 /* ..and store as long */ + shlr16 r1 /* get upper byte of first long.. */ + shlr8 r1 + cmp/hi r5,r3 /* runs r5 up to last or second last long bound */ + mov.b r1,@-r4 /* ..and store */ + bt .loop_do3 + + bra .last_do13 /* handle last longword: reuse routine for (long + 1) */ + add #4,r3 /* readjust end address */ + + /* trailing byte loop: copies 0..3 bytes (or all for < 11 in total) */ + .align 2 +.loop_b2: + mov.b @r5+,r0 /* load byte & increment source addr */ + add #1,r4 /* increment dest addr */ + mov.b r0,@(7,r4) /* store byte */ +.start_b2: + cmp/hi r5,r6 /* runs r5 up to end address */ + bt .loop_b2 + + rts + mov r7,r0 /* return dest start address */ +.end: + .size _memcpy,.end-_memcpy diff --git a/firmware/asm/sh/memmove.S b/firmware/asm/sh/memmove.S new file mode 100644 index 0000000000..d5a7160043 --- /dev/null +++ b/firmware/asm/sh/memmove.S @@ -0,0 +1,222 @@ +/*************************************************************************** + * __________ __ ___. + * Open \______ \ ____ ____ | | _\_ |__ _______ ___ + * Source | _// _ \_/ ___\| |/ /| __ \ / _ \ \/ / + * Jukebox | | ( <_> ) \___| < | \_\ ( <_> > < < + * Firmware |____|_ /\____/ \___ >__|_ \|___ /\____/__/\_ \ + * \/ \/ \/ \/ \/ + * $Id$ + * + * Copyright (C) 2006 by Jens Arnold + * + * This program is free software; you can redistribute it and/or + * modify it under the terms of the GNU General Public License + * as published by the Free Software Foundation; either version 2 + * of the License, or (at your option) any later version. + * + * 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 + .global _memmove + .type _memmove,@function + +/* Moves bytes of data in memory from to + * Regions may overlap. + * This version is optimized for speed, and needs the corresponding memcpy + * implementation for the forward copy branch. + * + * arguments: + * r4 - destination address + * r5 - source address + * r6 - length + * + * return value: + * r0 - destination address (like ANSI version) + * + * register usage: + * r0 - data / scratch + * r1 - 2nd data / scratch + * r2 - scratch + * r3 - last long bound / adjusted start address (only if >= 11 bytes) + * r4 - current dest address + * r5 - source start address + * r6 - current source address + * + * The instruction order is devised in a way to utilize the pipelining + * of the SH1 to the max. The routine also tries to utilize fast page mode. + */ + +_memmove: + cmp/hi r4,r5 /* source > destination */ + bf .backward /* no: backward copy */ + mov.l .memcpy_fwd,r0 + jmp @r0 + mov r4,r7 /* store dest for returning */ + + .align 2 +.memcpy_fwd: + .long ___memcpy_fwd_entry + +.backward: + add r6,r4 /* r4 = destination end */ + mov #11,r0 + cmp/hs r0,r6 /* at least 11 bytes to copy? (ensures 2 aligned longs) */ + add #-8,r5 /* adjust for late decrement (max. 2 longs) */ + add r5,r6 /* r6 = source end - 8 */ + bf .start_b2r /* no: jump directly to byte loop */ + + mov #-4,r3 /* r3 = 0xfffffffc */ + and r6,r3 /* r3 = last source long bound */ + cmp/hi r3,r6 /* already aligned? */ + bf .end_b1r /* yes: skip leading byte loop */ + +.loop_b1r: + mov.b @(7,r6),r0 /* load byte */ + add #-1,r6 /* decrement source addr */ + mov.b r0,@-r4 /* store byte */ + cmp/hi r3,r6 /* runs r6 down to last long bound */ + bt .loop_b1r + +.end_b1r: + mov #3,r1 + and r4,r1 /* r1 = dest alignment offset */ + mova .jmptab_r,r0 + mov.b @(r0,r1),r1 /* select appropriate main loop.. */ + add r0,r1 + mov r5,r3 /* copy start adress to r3 */ + jmp @r1 /* ..and jump to it */ + add #7,r3 /* adjust end addr for main loops doing 2 longs/pass */ + + /** main loops, copying 2 longs per pass to profit from fast page mode **/ + + /* long aligned destination (fastest) */ + .align 2 +.loop_do0r: + mov.l @r6,r1 /* load first long */ + add #-8,r6 /* decrement source addr */ + mov.l @(12,r6),r0 /* load second long */ + cmp/hi r3,r6 /* runs r6 down to first or second long bound */ + mov.l r0,@-r4 /* store second long */ + mov.l r1,@-r4 /* store first long; NOT ALIGNED - no speed loss here! */ + bt .loop_do0r + + add #-4,r3 /* readjust end address */ + cmp/hi r3,r6 /* first long left? */ + bf .start_b2r /* no, jump to trailing byte loop */ + + mov.l @(4,r6),r0 /* load first long */ + add #-4,r6 /* decrement source addr */ + bra .start_b2r /* jump to trailing byte loop */ + mov.l r0,@-r4 /* store first long */ + + /* word aligned destination (long + 2) */ + .align 2 +.loop_do2r: + mov.l @r6,r1 /* load first long */ + add #-8,r6 /* decrement source addr */ + mov.l @(12,r6),r0 /* load second long */ + cmp/hi r3,r6 /* runs r6 down to first or second long bound */ + mov.w r0,@-r4 /* store low word of second long */ + xtrct r1,r0 /* extract low word of first long & high word of second long */ + mov.l r0,@-r4 /* and store as long */ + shlr16 r1 /* get high word of first long */ + mov.w r1,@-r4 /* and store it */ + bt .loop_do2r + + add #-4,r3 /* readjust end address */ + cmp/hi r3,r6 /* first long left? */ + bf .start_b2r /* no, jump to trailing byte loop */ + + mov.l @(4,r6),r0 /* load first long & decrement source addr */ + add #-4,r6 /* decrement source addr */ + mov.w r0,@-r4 /* store low word */ + shlr16 r0 /* get high word */ + bra .start_b2r /* jump to trailing byte loop */ + mov.w r0,@-r4 /* and store it */ + + /* jumptable for loop selector */ + .align 2 +.jmptab_r: + .byte .loop_do0r - .jmptab_r /* placed in the middle because the SH1 */ + .byte .loop_do1r - .jmptab_r /* loads bytes sign-extended. Otherwise */ + .byte .loop_do2r - .jmptab_r /* the last loop would be out of reach */ + .byte .loop_do3r - .jmptab_r /* of the offset range. */ + + /* byte aligned destination (long + 1) */ + .align 2 +.loop_do1r: + mov.l @r6,r1 /* load first long */ + add #-8,r6 /* decrement source addr */ + mov.l @(12,r6),r0 /* load second long */ + cmp/hi r3,r6 /* runs r6 down to first or second long bound */ + mov.b r0,@-r4 /* store low byte of second long */ + shlr8 r0 /* get upper 3 bytes */ + mov r1,r2 /* copy first long */ + shll16 r2 /* move low byte of first long all the way up, .. */ + shll8 r2 + or r2,r0 /* ..combine with the 3 bytes of second long.. */ + mov.l r0,@-r4 /* ..and store as long */ + shlr8 r1 /* get middle 2 bytes */ + mov.w r1,@-r4 /* store as word */ + shlr16 r1 /* get upper byte */ + mov.b r1,@-r4 /* and store */ + bt .loop_do1r + + add #-4,r3 /* readjust end address */ +.last_do13r: + cmp/hi r3,r6 /* first long left? */ + bf .start_b2r /* no, jump to trailing byte loop */ + + nop /* alignment */ + mov.l @(4,r6),r0 /* load first long */ + add #-4,r6 /* decrement source addr */ + mov.b r0,@-r4 /* store low byte */ + shlr8 r0 /* get middle 2 bytes */ + mov.w r0,@-r4 /* store as word */ + shlr16 r0 /* get upper byte */ + bra .start_b2r /* jump to trailing byte loop */ + mov.b r0,@-r4 /* and store */ + + /* byte aligned destination (long + 3) */ + .align 2 +.loop_do3r: + mov.l @r6,r1 /* load first long */ + add #-8,r6 /* decrement source addr */ + mov.l @(12,r6),r0 /* load second long */ + mov r1,r2 /* copy first long */ + mov.b r0,@-r4 /* store low byte of second long */ + shlr8 r0 /* get middle 2 bytes */ + mov.w r0,@-r4 /* store as word */ + shlr16 r0 /* get upper byte */ + shll8 r2 /* move lower 3 bytes of first long one up.. */ + or r2,r0 /* ..combine with the 1 byte of second long.. */ + mov.l r0,@-r4 /* ..and store as long */ + shlr16 r1 /* get upper byte of first long */ + shlr8 r1 + cmp/hi r3,r6 /* runs r6 down to first or second long bound */ + mov.b r1,@-r4 /* ..and store */ + bt .loop_do3r + + bra .last_do13r /* handle first longword: reuse routine for (long + 1) */ + add #-4,r3 /* readjust end address */ + + /* trailing byte loop: copies 0..3 bytes (or all for < 11 in total) */ + .align 2 +.loop_b2r: + mov.b @(7,r6),r0 /* load byte */ + add #-1,r6 /* decrement source addr */ + mov.b r0,@-r4 /* store byte */ +.start_b2r: + cmp/hi r5,r6 /* runs r6 down to start address */ + bt .loop_b2r + + rts + mov r4,r0 /* return dest start address */ +.end: + .size _memmove,.end-_memmove diff --git a/firmware/asm/sh/memset.S b/firmware/asm/sh/memset.S new file mode 100644 index 0000000000..8cae1ea112 --- /dev/null +++ b/firmware/asm/sh/memset.S @@ -0,0 +1,109 @@ +/*************************************************************************** + * __________ __ ___. + * Open \______ \ ____ ____ | | _\_ |__ _______ ___ + * Source | _// _ \_/ ___\| |/ /| __ \ / _ \ \/ / + * Jukebox | | ( <_> ) \___| < | \_\ ( <_> > < < + * Firmware |____|_ /\____/ \___ >__|_ \|___ /\____/__/\_ \ + * \/ \/ \/ \/ \/ + * $Id$ + * + * Copyright (C) 2004 by Jens Arnold + * + * This program is free software; you can redistribute it and/or + * modify it under the terms of the GNU General Public License + * as published by the Free Software Foundation; either version 2 + * of the License, or (at your option) any later version. + * + * 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 + .global _memset + .type _memset,@function + +/* Fills a memory region with specified byte value + * This version is optimized for speed + * + * arguments: + * r4 - start address + * r5 - data + * r6 - length + * + * return value: + * r0 - start address (like ANSI version) + * + * register usage: + * r0 - temporary + * r1 - start address +11 for main loop + * r4 - start address + * r5 - data (spread to all 4 bytes when using long stores) + * r6 - current address (runs down from end to start) + * + * The instruction order below is devised in a way to utilize the pipelining + * of the SH1 to the max. The routine fills memory from end to start in + * order to utilize the auto-decrementing store instructions. + */ + +_memset: + neg r4,r0 + and #3,r0 /* r0 = (4 - align_offset) % 4 */ + add #4,r0 + cmp/hs r0,r6 /* at least one aligned longword to fill? */ + add r4,r6 /* r6 = end_address */ + bf .no_longs /* no, jump directly to byte loop */ + + extu.b r5,r5 /* start: spread data to all 4 bytes */ + swap.b r5,r0 + or r0,r5 /* data now in 2 lower bytes of r5 */ + swap.w r5,r0 + or r0,r5 /* data now in all 4 bytes of r5 */ + + mov r6,r0 + tst #3,r0 /* r0 already long aligned? */ + bt .end_b1 /* yes: skip loop */ + + /* leading byte loop: sets 0..3 bytes */ +.loop_b1: + mov.b r5,@-r0 /* store byte */ + tst #3,r0 /* r0 long aligned? */ + bf .loop_b1 /* runs r0 down until long aligned */ + + mov r0,r6 /* r6 = last long bound */ + nop /* keep alignment */ + +.end_b1: + mov r4,r1 /* r1 = start_address... */ + add #11,r1 /* ... + 11, combined for rounding and offset */ + xor r1,r0 + tst #4,r0 /* bit 2 tells whether an even or odd number of */ + bf .loop_odd /* longwords to set */ + + /* main loop: set 2 longs per pass */ +.loop_2l: + mov.l r5,@-r6 /* store first long */ +.loop_odd: + cmp/hi r1,r6 /* runs r6 down to first long bound */ + mov.l r5,@-r6 /* store second long */ + bt .loop_2l + +.no_longs: + cmp/hi r4,r6 /* any bytes left? */ + bf .end_b2 /* no: skip loop */ + + /* trailing byte loop */ +.loop_b2: + mov.b r5,@-r6 /* store byte */ + cmp/hi r4,r6 /* runs r6 down to the start address */ + bt .loop_b2 + +.end_b2: + rts + mov r4,r0 /* return start address */ + +.end: + .size _memset,.end-_memset diff --git a/firmware/asm/sh/strlen.S b/firmware/asm/sh/strlen.S new file mode 100644 index 0000000000..e7169e25db --- /dev/null +++ b/firmware/asm/sh/strlen.S @@ -0,0 +1,96 @@ +/*************************************************************************** + * __________ __ ___. + * Open \______ \ ____ ____ | | _\_ |__ _______ ___ + * Source | _// _ \_/ ___\| |/ /| __ \ / _ \ \/ / + * Jukebox | | ( <_> ) \___| < | \_\ ( <_> > < < + * Firmware |____|_ /\____/ \___ >__|_ \|___ /\____/__/\_ \ + * \/ \/ \/ \/ \/ + * $Id$ + * + * Copyright (C) 2005 by Jens Arnold + * + * This program is free software; you can redistribute it and/or + * modify it under the terms of the GNU General Public License + * as published by the Free Software Foundation; either version 2 + * of the License, or (at your option) any later version. + * + * 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 + .global _strlen + .type _strlen,@function + +/* Works out the length of a string + * This version is optimized for speed + * + * arguments: + * r4 - start address + * + * return value: + * r0 - string length + * + * register usage: + * r0 - current address + * r1 - current value (byte/long) + * r2 - mask for alignment / zero (for cmp/str) + * r4 - start address + * + */ + +_strlen: + mov r4,r0 /* r0 = start address */ + tst #3,r0 /* long aligned? */ + bt .start_l /* yes, jump directly to the longword loop */ + + /* not long aligned: check the first 3 bytes */ + mov.b @r0+,r1 /* fetch first byte */ + tst r1,r1 /* byte == 0 ? */ + bt .hitzero /* yes, string end found */ + mov.b @r0+,r1 /* fetch second byte */ + mov #3,r2 /* prepare mask: r2 = 0..00000011b */ + tst r1,r1 /* byte == 0 ? */ + bt .hitzero /* yes, string end found */ + mov.b @r0+,r1 /* fetch third byte */ + not r2,r2 /* prepare mask: r2 = 1..11111100b */ + tst r1,r1 /* byte == 0 ? */ + bt .hitzero /* yes, string end found */ + + /* not yet found, fall through into longword loop */ + and r2,r0 /* align down to long bound */ + + /* main loop: check longwords */ +.start_l: + mov #0,r2 /* zero longword for cmp/str */ +.loop_l: + mov.l @r0+,r1 /* fetch long word */ + cmp/str r1,r2 /* any zero byte within? */ + bf .loop_l /* no, loop */ + add #-4,r0 /* set address back to start of this longword */ + + /* the last longword contains the string end: figure out the byte */ + mov.b @r0+,r1 /* fetch first byte */ + tst r1,r1 /* byte == 0 ? */ + bt .hitzero /* yes, string end found */ + mov.b @r0+,r1 /* fetch second byte */ + tst r1,r1 /* byte == 0 ? */ + bt .hitzero /* yes, string end found */ + mov.b @r0+,r1 /* fetch third byte */ + tst r1,r1 /* byte == 0 ? */ + bt .hitzero /* yes, string end found */ + rts /* must be the fourth byte */ + sub r4,r0 /* len = string_end - string_start */ + +.hitzero: + add #-1,r0 /* undo address increment */ + rts + sub r4,r0 /* len = string_end - string_start */ + +.end: + .size _strlen,.end-_strlen + diff --git a/firmware/asm/strlen.c b/firmware/asm/strlen.c new file mode 100644 index 0000000000..649df6764b --- /dev/null +++ b/firmware/asm/strlen.c @@ -0,0 +1,93 @@ +/* +FUNCTION + <>---character string length + +INDEX + strlen + +ANSI_SYNOPSIS + #include + size_t strlen(const char *<[str]>); + +TRAD_SYNOPSIS + #include + size_t strlen(<[str]>) + char *<[src]>; + +DESCRIPTION + The <> function works out the length of the string + starting at <<*<[str]>>> by counting chararacters until it + reaches a <> character. + +RETURNS + <> returns the character count. + +PORTABILITY +<> is ANSI C. + +<> requires no supporting OS subroutines. + +QUICKREF + strlen ansi pure +*/ + +#include "config.h" +#include "_ansi.h" +#include +#include + +#define LBLOCKSIZE (sizeof (long)) +#define UNALIGNED(X) ((long)X & (LBLOCKSIZE - 1)) + +#if LONG_MAX == 2147483647L +#define DETECTNULL(X) (((X) - 0x01010101) & ~(X) & 0x80808080) +#else +#if LONG_MAX == 9223372036854775807L +/* Nonzero if X (a long int) contains a NULL byte. */ +#define DETECTNULL(X) (((X) - 0x0101010101010101) & ~(X) & 0x8080808080808080) +#else +#error long int is not a 32bit or 64bit type. +#endif +#endif + +#ifndef DETECTNULL +#error long int is not a 32bit or 64bit byte +#endif + +size_t +_DEFUN (strlen, (str), + _CONST char *str) ICODE_ATTR; + +size_t +_DEFUN (strlen, (str), + _CONST char *str) +{ +#if defined(PREFER_SIZE_OVER_SPEED) || defined(__OPTIMIZE_SIZE__) + _CONST char *start = str; + + while (*str) + str++; + + return str - start; +#else + _CONST char *start = str; + unsigned long *aligned_addr; + + if (!UNALIGNED (str)) + { + /* If the string is word-aligned, we can check for the presence of + a null in each word-sized block. */ + aligned_addr = (unsigned long*)str; + while (!DETECTNULL (*aligned_addr)) + aligned_addr++; + + /* Once a null is detected, we check each byte in that block for a + precise position of the null. */ + str = (char*)aligned_addr; + } + + while (*str) + str++; + return str - start; +#endif /* not PREFER_SIZE_OVER_SPEED */ +} -- cgit v1.2.3