devs/foxbox
1
0
Fork 0
forked from len0rd/rockbox
Code Pull requests Activity

Add ASM optimised subband filtering loop for ARM targets. Restructure

Browse source 
other loops a little bit.


git-svn-id: svn://svn.rockbox.org/rockbox/trunk@8754 a1c6a512-1295-4272-9138-f99709370657
This commit is contained in:
Thom Johansen 2006-02-20 20:40:29 +00:00
parent ec93f4685c
commit 831cfb8e44
1 changed files with 60 additions and 11 deletions
Download patch file Download diff file Expand all files Collapse all files

51
apps/codecs/libmusepack/synth_filter.c
View file

@ -342,8 +342,8 @@ static void Synthese_Filter_float_internal(MPC_SAMPLE_FORMAT * OutData,MPC_SAMPL
for ( k = 0; k < 32; k++, D += 16, V++ ) {
#if defined(CPU_COLDFIRE) && !defined(SIMULATOR)
for ( k = 0; k < 32; k++, D += 16, V++ ) {
asm volatile (
"movem.l (%[D]), %%d0-%%d3 \n\t"
"move.l (%[V]), %%a5 \n\t"
@ -371,7 +371,56 @@ static void Synthese_Filter_float_internal(MPC_SAMPLE_FORMAT * OutData,MPC_SAMPL
: [Data] "+a" (Data)
: [V] "a" (V), [D] "a" (D)
: "d0", "d1", "d2", "d3", "a5");
#elif defined(CPU_ARM) && !defined(SIMULATOR)
for ( k = 0; k < 32; k++, V++ ) {
asm volatile (
"ldmia %[D]!, { r0-r3 } \n\t"
"ldr r4, [%[V]] \n\t"
"smull r5, r6, r0, r4 \n\t"
"ldr r4, [%[V], #96*4] \n\t"
"smlal r5, r6, r1, r4 \n\t"
"ldr r4, [%[V], #128*4] \n\t"
"smlal r5, r6, r2, r4 \n\t"
"ldr r4, [%[V], #224*4] \n\t"
"smlal r5, r6, r3, r4 \n\t"
"ldmia %[D]!, { r0-r3 } \n\t"
"ldr r4, [%[V], #256*4] \n\t"
"smlal r5, r6, r0, r4 \n\t"
"ldr r4, [%[V], #352*4] \n\t"
"smlal r5, r6, r1, r4 \n\t"
"ldr r4, [%[V], #384*4] \n\t"
"smlal r5, r6, r2, r4 \n\t"
"ldr r4, [%[V], #480*4] \n\t"
"smlal r5, r6, r3, r4 \n\t"
"ldmia %[D]!, { r0-r3 } \n\t"
"ldr r4, [%[V], #512*4] \n\t"
"smlal r5, r6, r0, r4 \n\t"
"ldr r4, [%[V], #608*4] \n\t"
"smlal r5, r6, r1, r4 \n\t"
"ldr r4, [%[V], #640*4] \n\t"
"smlal r5, r6, r2, r4 \n\t"
"ldr r4, [%[V], #736*4] \n\t"
"smlal r5, r6, r3, r4 \n\t"
"ldmia %[D]!, { r0-r3 } \n\t"
"ldr r4, [%[V], #768*4] \n\t"
"smlal r5, r6, r0, r4 \n\t"
"ldr r4, [%[V], #864*4] \n\t"
"smlal r5, r6, r1, r4 \n\t"
"ldr r4, [%[V], #896*4] \n\t"
"smlal r5, r6, r2, r4 \n\t"
"ldr r4, [%[V], #992*4] \n\t"
"smlal r5, r6, r3, r4 \n\t"
"mov r4, r6, lsl #1 \n\t"
"orr r4, r4, r5, lsr #31\n\t"
"str r4, [%[Data]], #4 \n"
: [Data] "+r" (Data), [D] "+r" (D)
: [V] "r" (V)
: "r0", "r1", "r2", "r3", "r4", "r5", "r6");
#else
for ( k = 0; k < 32; k++, D += 16, V++ ) {
*Data = MPC_SHL(
MPC_MULTIPLY_FRACT(V[ 0],D[ 0]) + MPC_MULTIPLY_FRACT(V[ 96],D[ 1]) + MPC_MULTIPLY_FRACT(V[128],D[ 2]) + MPC_MULTIPLY_FRACT(V[224],D[ 3])
+ MPC_MULTIPLY_FRACT(V[256],D[ 4]) + MPC_MULTIPLY_FRACT(V[352],D[ 5]) + MPC_MULTIPLY_FRACT(V[384],D[ 6]) + MPC_MULTIPLY_FRACT(V[480],D[ 7])