kolibrios-fun/programs/develop/libraries/libmpg123/synth_i586.S

337 lines
5.9 KiB
ArmAsm
Raw Normal View History

/*
decode_i586: asm synth
copyright ?-2006 by the mpg123 project - free software under the terms of the LGPL 2.1
see COPYING and AUTHORS files in distribution or http://mpg123.org
initially written by Stefan Bieschewski
synth_1to1 works the same way as the c version of this
file. only two types of changes have been made:
- reordered floating point instructions to
prevent pipline stalls
- made WRITE_SAMPLE use integer instead of
(slower) floating point
all kinds of x86 processors should benefit from these
modifications.
useful sources of information on optimizing x86 code include:
Intel Architecture Optimization Manual
http://www.intel.com/design/pentium/manuals/242816.htm
Cyrix 6x86 Instruction Set Summary
ftp://ftp.cyrix.com/6x86/6x-dbch6.pdf
AMD-K5 Processor Software Development
http://www.amd.com/products/cpg/techdocs/appnotes/20007e.pdf
Stefan Bieschewski <stb@acm.org>
$Id: decode_i586.s 1 2004-09-18 13:30:08Z thomas $
*/
#include "mangle.h"
.data
#ifndef __APPLE__
.section .rodata
#endif
ALIGN8
.LC0:
.long 0x0,0x40dfffc0
ALIGN8
.LC1:
.long 0x0,0xc0e00000
ALIGN8
.text
/* int synth_1to1_i586_asm(real *bandPtr, int channel, unsigned char *out, unsigned char *buffs, int *bo, real *decwin); */
.globl ASM_NAME(synth_1to1_i586_asm)
ASM_NAME(synth_1to1_i586_asm):
subl $12,%esp
pushl %ebp
pushl %edi
pushl %esi
pushl %ebx
/* stack: 0=ebx, 4=esi, 8=edi, 12=ebp, 16,20,24=local, 28=back, 32=bandPtr, 36=channel, 40=out, 44=buffs, 48=bo, 52=decwin */
movl 32(%esp),%eax /* *bandPtr */
movl 40(%esp),%esi /* *out */
movl 48(%esp),%edi /* *bo */
movl (%edi),%ebp /* store bo value in ebp */
xorl %edi,%edi
cmpl %edi,36(%esp)
jne .L48 /* if(!channel) */
decl %ebp /* bo-- */
andl $15,%ebp /* bo &= 0xf */
movl 48(%esp), %edi /* *bo */
movl %ebp,(%edi) /* write back bo */
xorl %edi,%edi /* restore %edi to 0; it's used later */
movl 44(%esp),%ecx /* use buffs */
jmp .L49
.L48: /* if(channel) use buffs+2176 */
addl $2,%esi
movl 44(%esp),%ecx /* *buffs */
addl $2176,%ecx
.L49:
testl $1,%ebp
je .L50
movl %ecx,%ebx
movl %ebp,16(%esp)
pushl %eax
movl 20(%esp),%edx
leal (%ebx,%edx,4),%eax
pushl %eax
movl 24(%esp),%eax
incl %eax
andl $15,%eax
leal 1088(,%eax,4),%eax
addl %ebx,%eax
jmp .L74
.L50:
leal 1088(%ecx),%ebx
leal 1(%ebp),%edx
movl %edx,16(%esp)
pushl %eax
leal 1092(%ecx,%ebp,4),%eax
pushl %eax
leal (%ecx,%ebp,4),%eax
.L74:
pushl %eax
call ASM_NAME(dct64_i386)
addl $12,%esp
/* stack now back on track */
movl 16(%esp),%edx
leal 0(,%edx,4),%edx
movl 52(%esp),%eax /* decwin */
addl $64,%eax
movl %eax,%ecx
subl %edx,%ecx
movl $16,%ebp
.L55:
flds (%ecx)
fmuls (%ebx)
flds 4(%ecx)
fmuls 4(%ebx)
fxch %st(1)
flds 8(%ecx)
fmuls 8(%ebx)
fxch %st(2)
fsubrp %st,%st(1)
flds 12(%ecx)
fmuls 12(%ebx)
fxch %st(2)
faddp %st,%st(1)
flds 16(%ecx)
fmuls 16(%ebx)
fxch %st(2)
fsubrp %st,%st(1)
flds 20(%ecx)
fmuls 20(%ebx)
fxch %st(2)
faddp %st,%st(1)
flds 24(%ecx)
fmuls 24(%ebx)
fxch %st(2)
fsubrp %st,%st(1)
flds 28(%ecx)
fmuls 28(%ebx)
fxch %st(2)
faddp %st,%st(1)
flds 32(%ecx)
fmuls 32(%ebx)
fxch %st(2)
fsubrp %st,%st(1)
flds 36(%ecx)
fmuls 36(%ebx)
fxch %st(2)
faddp %st,%st(1)
flds 40(%ecx)
fmuls 40(%ebx)
fxch %st(2)
fsubrp %st,%st(1)
flds 44(%ecx)
fmuls 44(%ebx)
fxch %st(2)
faddp %st,%st(1)
flds 48(%ecx)
fmuls 48(%ebx)
fxch %st(2)
fsubrp %st,%st(1)
flds 52(%ecx)
fmuls 52(%ebx)
fxch %st(2)
faddp %st,%st(1)
flds 56(%ecx)
fmuls 56(%ebx)
fxch %st(2)
fsubrp %st,%st(1)
flds 60(%ecx)
fmuls 60(%ebx)
fxch %st(2)
subl $4,%esp
faddp %st,%st(1)
fxch %st(1)
fsubrp %st,%st(1)
fistpl (%esp)
popl %eax
cmpl $32767,%eax
jg 1f
cmpl $-32768,%eax
jl 2f
movw %ax,(%esi)
jmp 4f
1: movw $32767,(%esi)
jmp 3f
2: movw $-32768,(%esi)
3: incl %edi
4:
.L54:
addl $64,%ebx
subl $-128,%ecx
addl $4,%esi
decl %ebp
jnz .L55
flds (%ecx)
fmuls (%ebx)
flds 8(%ecx)
fmuls 8(%ebx)
flds 16(%ecx)
fmuls 16(%ebx)
fxch %st(2)
faddp %st,%st(1)
flds 24(%ecx)
fmuls 24(%ebx)
fxch %st(2)
faddp %st,%st(1)
flds 32(%ecx)
fmuls 32(%ebx)
fxch %st(2)
faddp %st,%st(1)
flds 40(%ecx)
fmuls 40(%ebx)
fxch %st(2)
faddp %st,%st(1)
flds 48(%ecx)
fmuls 48(%ebx)
fxch %st(2)
faddp %st,%st(1)
flds 56(%ecx)
fmuls 56(%ebx)
fxch %st(2)
subl $4,%esp
faddp %st,%st(1)
fxch %st(1)
faddp %st,%st(1)
fistpl (%esp)
popl %eax
cmpl $32767,%eax
jg 1f
cmpl $-32768,%eax
jl 2f
movw %ax,(%esi)
jmp 4f
1: movw $32767,(%esi)
jmp 3f
2: movw $-32768,(%esi)
3: incl %edi
4:
.L62:
addl $-64,%ebx
addl $4,%esi
movl 16(%esp),%edx
leal -128(%ecx,%edx,8),%ecx
movl $15,%ebp
.L68:
flds -4(%ecx)
fchs
fmuls (%ebx)
flds -8(%ecx)
fmuls 4(%ebx)
fxch %st(1)
flds -12(%ecx)
fmuls 8(%ebx)
fxch %st(2)
fsubrp %st,%st(1)
flds -16(%ecx)
fmuls 12(%ebx)
fxch %st(2)
fsubrp %st,%st(1)
flds -20(%ecx)
fmuls 16(%ebx)
fxch %st(2)
fsubrp %st,%st(1)
flds -24(%ecx)
fmuls 20(%ebx)
fxch %st(2)
fsubrp %st,%st(1)
flds -28(%ecx)
fmuls 24(%ebx)
fxch %st(2)
fsubrp %st,%st(1)
flds -32(%ecx)
fmuls 28(%ebx)
fxch %st(2)
fsubrp %st,%st(1)
flds -36(%ecx)
fmuls 32(%ebx)
fxch %st(2)
fsubrp %st,%st(1)
flds -40(%ecx)
fmuls 36(%ebx)
fxch %st(2)
fsubrp %st,%st(1)
flds -44(%ecx)
fmuls 40(%ebx)
fxch %st(2)
fsubrp %st,%st(1)
flds -48(%ecx)
fmuls 44(%ebx)
fxch %st(2)
fsubrp %st,%st(1)
flds -52(%ecx)
fmuls 48(%ebx)
fxch %st(2)
fsubrp %st,%st(1)
flds -56(%ecx)
fmuls 52(%ebx)
fxch %st(2)
fsubrp %st,%st(1)
flds -60(%ecx)
fmuls 56(%ebx)
fxch %st(2)
fsubrp %st,%st(1)
flds (%ecx)
fmuls 60(%ebx)
fxch %st(2)
subl $4,%esp
fsubrp %st,%st(1)
fxch %st(1)
fsubrp %st,%st(1)
fistpl (%esp)
popl %eax
cmpl $32767,%eax
jg 1f
cmpl $-32768,%eax
jl 2f
movw %ax,(%esi)
jmp 4f
1: movw $32767,(%esi)
jmp 3f
2: movw $-32768,(%esi)
3: incl %edi
4:
.L67:
addl $-64,%ebx
addl $-128,%ecx
addl $4,%esi
decl %ebp
jnz .L68
movl %edi,%eax
popl %ebx
popl %esi
popl %edi
popl %ebp
addl $12,%esp
ret
NONEXEC_STACK