rockbox/firmware/target/arm/ata-as-arm.S
Rafaël Carré 45c7498f59 FS#11335 by me: make ARM assembly functions thumb-friendly
We can't pop into pc on ARMv4t when using thumb: the T bit won't be
modified if we are returning to a thumb function
Code running on ARMv4t should use the new ldrpc / ldmpc macros instead
of ldr pc, [sp], #4 and ldm(cond) sp!, {regs, pc}
No modification on pure ARM builds and ARMv5+

Note: USE_THUMB is currently never defined, no targets can currently be
built with -mthumb, see FS#6734

git-svn-id: svn://svn.rockbox.org/rockbox/trunk@26756 a1c6a512-1295-4272-9138-f99709370657
2010-06-11 04:41:36 +00:00

356 lines
11 KiB
ArmAsm

/***************************************************************************
* __________ __ ___.
* Open \______ \ ____ ____ | | _\_ |__ _______ ___
* Source | _// _ \_/ ___\| |/ /| __ \ / _ \ \/ /
* Jukebox | | ( <_> ) \___| < | \_\ ( <_> > < <
* Firmware |____|_ /\____/ \___ >__|_ \|___ /\____/__/\_ \
* \/ \/ \/ \/ \/
* $Id$
*
* Copyright (C) 2007 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"
#if CONFIG_CPU == PP5002
/* Causes ATA retries on iPod G3 probably related to improper controller
* setup. Needs investigation. */
.section .icode,"ax",%progbits
.equ .ata_port, 0xc00031e0
#elif defined CPU_PP502x
/* Verified working on (PP5020, PP5022) targets */
.section .icode,"ax",%progbits
.equ .ata_port, 0xc30001e0
#elif CONFIG_CPU == S3C2440
/* Untested */
.text
.equ .ata_port, 0x18000000
#elif defined(CREATIVE_ZVx) /* Zen Vision could have an other address */
.text
.equ .ata_port, 0x50FEE000
#elif defined(MROBE_500)
.text
.equ .ata_port, 0x50400000
#else
/* This isn't vaild, but it does it's job, not sure what this should be */
#error ata_port undefined!
#endif
.align 2
.global copy_read_sectors
.type copy_read_sectors,%function
/* Read a number of words from the ATA data port
*
* Optimised for speed; assumes wordcount >= 10
*
* Arguments:
* r0 - buffer address
* r1 - word count
*
* Register usage:
* r0 - current address
* r1 - word count
* r2 - ata port
* r3..r5, r12, lr - read buffers
*/
copy_read_sectors:
stmfd sp!, {r4, r5, lr}
ldr r2, =.ata_port
tst r0, #1 /* 16 bit aligned? */
beq .r_aligned
/* not 16-bit aligned */
sub r1, r1, #1 /* one halfword is handled unconditionally */
ldrh r3, [r2] /* read first halfword */
strb r3, [r0], #1 /* store low byte */
mov r3, r3, lsr #8
tst r0, #2 /* 32 bit aligned? */
beq .r_noword_u
ldrh r4, [r2] /* read second halfword */
orr r3, r3, r4, lsl #8 /* combine with old byte */
strh r3, [r0], #2 /* store */
mov r3, r4, lsr #8
sub r1, r1, #1 /* another halfword taken */
.r_noword_u:
sub r1, r1, #8 /* adjust for zero-check and doing 8 halfwords/loop */
.r_loop_u:
ldrh r4, [r2] /* Read 8 halfwords and combine them into */
orr r3, r3, r4, lsl #8 /* 4 words so that they're properly aligned */
ldrh r4, [r2] /* in memory. Bottom byte of first word is */
orr r3, r3, r4, lsl #24 /* the top byte from the last round. Write */
mov r4, r4, lsr #8 /* all 4 words at once. */
ldrh r5, [r2]
orr r4, r4, r5, lsl #8
ldrh r5, [r2]
orr r4, r4, r5, lsl #24
mov r5, r5, lsr #8
ldrh r12, [r2]
orr r5, r5, r12, lsl #8
ldrh r12, [r2]
orr r5, r5, r12, lsl #24
mov r12, r12, lsr #8
ldrh lr, [r2]
orr r12, r12, lr, lsl #8
ldrh lr, [r2]
orr r12, r12, lr, lsl #24
stmia r0!, {r3, r4, r5, r12}
mov r3, lr, lsr #8
subs r1, r1, #8 /* 8 or more halfwords left? */
bge .r_loop_u
/* No need to adjust the count, only checking bits from now on. */
tst r1, #4 /* 4 or more halfwords left? */
beq .r_end4_u
ldrh r4, [r2]
orr r3, r3, r4, lsl #8
ldrh r4, [r2]
orr r3, r3, r4, lsl #24
mov r4, r4, lsr #8
ldrh r5, [r2]
orr r4, r4, r5, lsl #8
ldrh r5, [r2]
orr r4, r4, r5, lsl #24
stmia r0!, {r3, r4}
mov r3, r5, lsr #8
.r_end4_u:
tst r1, #2 /* 2 or more halfwords left? */
beq .r_end2_u
ldrh r4, [r2]
orr r3, r3, r4, lsl #8
ldrh r4, [r2]
orr r3, r3, r4, lsl #24
str r3, [r0], #4
mov r3, r4, lsr #8
.r_end2_u:
tst r1, #1 /* one halfword left? */
ldrneh r4, [r2]
orrne r3, r3, r4, lsl #8
strneh r3, [r0], #2
movne r3, r4, lsr #8
strb r3, [r0], #1 /* store final byte */
ldmpc regs=r4-r5
/* 16-bit aligned */
.r_aligned:
tst r0, #2 /* 32 bit aligned? */
ldrneh r3, [r2] /* no: read first halfword */
strneh r3, [r0], #2 /* store */
subne r1, r1, #1 /* one halfword taken */
sub r1, r1, #8 /* adjust for zero-check and doing 8 halfwords/loop */
.r_loop_a:
ldrh r3, [r2] /* Read 8 halfwords and combine each pair */
ldrh r4, [r2] /* into a word, then store all at once. */
orr r3, r3, r4, lsl #16
ldrh r4, [r2]
ldrh r5, [r2]
orr r4, r4, r5, lsl #16
ldrh r5, [r2]
ldrh r12, [r2]
orr r5, r5, r12, lsl #16
ldrh r12, [r2]
ldrh lr, [r2]
orr r12, r12, lr, lsl #16
stmia r0!, {r3, r4, r5, r12}
subs r1, r1, #8 /* 8 or more halfwords left? */
bge .r_loop_a
/* No need to adjust the count, only checking bits from now on. */
tst r1, #4 /* 4 or more halfwords left? */
beq .r_end4_a
ldrh r3, [r2]
ldrh r4, [r2]
orr r3, r3, r4, lsl #16
ldrh r4, [r2]
ldrh r5, [r2]
orr r4, r4, r5, lsl #16
stmia r0!, {r3, r4}
.r_end4_a:
tst r1, #2 /* 2 or more halfwords left? */
ldrneh r3, [r2]
ldrneh r4, [r2]
orrne r3, r3, r4, lsl #16
strne r3, [r0], #4
tst r1, #1 /* one halfword left? */
ldrneh r3, [r2]
strneh r3, [r0], #2
ldmpc regs=r4-r5
.r_end:
.size copy_read_sectors,.r_end-copy_read_sectors
.align 2
.global copy_write_sectors
.type copy_write_sectors,%function
/* Write a number of words to the ATA data port
*
* Optimised for speed; assumes wordcount >= 10
*
* Arguments:
* r0 - buffer address
* r1 - word count
*
* Register usage:
* r0 - current address
* r1 - word count
* r2 - ata port
* r3..r5, r12, lr - read buffers
*/
copy_write_sectors:
stmfd sp!, {r4, r5, lr}
ldr r2, =.ata_port
tst r0, #1 /* 16 bit aligned? */
beq .w_aligned
/* not 16-bit aligned */
sub r1, r1, #1 /* one halfword is done unconditionally */
ldrb r3, [r0], #1 /* load 1st byte, now halfword aligned. */
tst r0, #2 /* 32 bit aligned? */
beq .w_noword_u
ldrh r4, [r0], #2 /* load a halfword */
orr r3, r3, r4, lsl #8 /* combine with old byte */
strh r3, [r2] /* write halfword */
mov r3, r4, lsr #8
sub r1, r1, #1 /* another halfword taken */
.w_noword_u:
sub r1, r1, #8 /* adjust for zero-check and doing 8 halfwords/loop */
.w_loop_u:
ldmia r0!, {r4, r5, r12, lr}
orr r3, r3, r4, lsl #8 /* Load 4 words at once and decompose them */
strh r3, [r2] /* into 8 halfwords in a way that the words */
mov r3, r3, lsr #16 /* are shifted by 8 bits, putting the high */
strh r3, [r2] /* byte of one word into the low byte of */
mov r4, r4, lsr #24 /* the next. High byte of last word becomes */
orr r4, r4, r5, lsl #8 /* low byte of next round. */
strh r4, [r2]
mov r4, r4, lsr #16
strh r4, [r2]
mov r5, r5, lsr #24
orr r5, r5, r12, lsl #8
strh r5, [r2]
mov r5, r5, lsr #16
strh r5, [r2]
mov r12, r12, lsr #24
orr r12, r12, lr, lsl #8
strh r12, [r2]
mov r12, r12, lsr #16
strh r12, [r2]
mov r3, lr, lsr #24
subs r1, r1, #8 /* 8 or more halfwords left? */
bge .w_loop_u
/* No need to adjust the count, only checking bits from now on. */
tst r1, #4 /* 4 or more halfwords left? */
beq .w_end4_u
ldmia r0!, {r4, r5}
orr r3, r3, r4, lsl #8
strh r3, [r2]
mov r3, r3, lsr #16
strh r3, [r2]
mov r4, r4, lsr #24
orr r4, r4, r5, lsl #8
strh r4, [r2]
mov r4, r4, lsr #16
strh r4, [r2]
mov r3, r5, lsr #24
.w_end4_u:
tst r1, #2 /* 2 or more halfwords left? */
beq .w_end2_u
ldr r4, [r0], #4
orr r3, r3, r4, lsl #8
strh r3, [r2]
mov r3, r3, lsr #16
strh r3, [r2]
mov r3, r4, lsr #24
.w_end2_u:
tst r1, #1 /* one halfword left? */
ldrneh r4, [r0], #2
orrne r3, r3, r4, lsl #8
strneh r3, [r2]
movne r3, r3, lsr #16
ldrb r4, [r0], #1 /* load final byte */
orr r3, r3, r4, lsl #8
strh r3, [r2] /* write final halfword */
ldmpc regs=r4-r5
/* 16-bit aligned */
.w_aligned:
tst r0, #2 /* 32 bit aligned? */
ldrneh r3, [r0], #2 /* no: load first halfword */
strneh r3, [r2] /* write */
subne r1, r1, #1 /* one halfword taken */
sub r1, r1, #8 /* adjust for zero-check and doing 8 halfwords/loop */
.w_loop_a:
ldmia r0!, {r3, r4, r5, r12}
strh r3, [r2] /* Load 4 words and decompose them into */
mov r3, r3, lsr #16 /* 2 halfwords each, and write those. */
strh r3, [r2]
strh r4, [r2]
mov r4, r4, lsr #16
strh r4, [r2]
strh r5, [r2]
mov r5, r5, lsr #16
strh r5, [r2]
strh r12, [r2]
mov r12, r12, lsr #16
strh r12, [r2]
subs r1, r1, #8 /* 8 or more halfwords left? */
bge .w_loop_a
/* No need to adjust the count, only checking bits from now on. */
tst r1, #4 /* 4 or more halfwords left? */
beq .w_end4_a
ldmia r0!, {r3, r4}
strh r3, [r2]
mov r3, r3, lsr #16
strh r3, [r2]
strh r4, [r2]
mov r4, r4, lsr #16
strh r4, [r2]
.w_end4_a:
tst r1, #2 /* 2 or more halfwords left? */
ldrne r3, [r0], #4
strneh r3, [r2]
movne r3, r3, lsr #16
strneh r3, [r2]
tst r1, #1 /* one halfword left? */
ldrneh r3, [r0], #2
strneh r3, [r2]
ldmpc regs=r4-r5
.w_end:
.size copy_write_sectors,.w_end-copy_write_sectors