len0rd/rockbox
1
0
Fork 1
mirror of https://github.com/Rockbox/rockbox.git synced 2025-12-10 13:45:10 -05:00
Code Issues Projects Releases Packages Wiki Activity Actions

tta codec: optimize for arm target. pp502x: ~1% speed up.

Browse source 
git-svn-id: svn://svn.rockbox.org/rockbox/trunk@26478 a1c6a512-1295-4272-9138-f99709370657
This commit is contained in:
Yoshihisa Uchida 2010-06-02 10:54:11 +00:00
parent e8a662432a
commit 6ebe76c147
1 changed files with 36 additions and 41 deletions
Download patch file Download diff file Expand all files Collapse all files

77
apps/codecs/libtta/filter_arm.S
View file

@ -47,17 +47,17 @@ hybrid_filter:
@ r6 fs->error @ r6 fs->error
@ lr sum := fs->round @ lr sum := fs->round
add r2, r0, #148 @ r2 = fs->dl
add r3, r0, #52 @ r3 = fs->dx
add r4, r0, #20 @ r4 = fs->qm
ldmia r0, {r5, r6, lr} @ r5 = fs->index ldmia r0, {r5, r6, lr} @ r5 = fs->index
@ r6 = fs->error @ r6 = fs->error
@ lr = fs->round @ lr = fs->round
mov r5, r5, asl #2 add r2, r0, #148 @ r2 = fs->dl
add r2, r2, r5 @ r2 = fs->dl + fs->index add r3, r0, #52 @ r3 = fs->dx
add r3, r3, r5 @ r3 = fs->dx + fs->index add r4, r0, #20 @ r4 = fs->qm
add r2, r2, r5 @ r2 = (unsigned char*)fs->dl + fs->index
add r3, r3, r5 @ r3 = (unsigned char*)fs->dx + fs->index
cmp r6, #0 cmp r6, #0
bmi .hf_negative
bne .hf_positive bne .hf_positive
@ case fs->error == 0 @ case fs->error == 0
@ -72,32 +72,6 @@ hybrid_filter:
ldmia r4!, {r5, r6, r7, r8 } ldmia r4!, {r5, r6, r7, r8 }
b .hf2 b .hf2
.hf_positive:
blt .hf_negative
@ case fs->error > 0
ldmia r4, {r5, r6, r7, r8 }
ldmia r3!, {r9, r10, r11, r12}
add r5, r5, r9
add r6, r6, r10
add r7, r7, r11
add r8, r8, r12
stmia r4!, {r5, r6, r7, r8 } @ update fs->qm[0], ..., fs->qm[3]
ldmia r2!, {r9, r10, r11, r12}
mla lr, r5, r9, lr
mla lr, r6, r10, lr
mla lr, r7, r11, lr
mla lr, r8, r12, lr
ldmia r4, {r5, r6, r7, r8 }
ldmia r3!, {r9, r10, r11, r12}
add r5, r5, r9
add r6, r6, r10
add r7, r7, r11
add r8, r8, r12
stmia r4!, {r5, r6, r7, r8 } @ update fs->qm[4], ..., fs->qm[7]
b .hf2
.hf_negative: .hf_negative:
@ case fs->error < 0 @ case fs->error < 0
@ -120,6 +94,30 @@ hybrid_filter:
sub r7, r7, r11 sub r7, r7, r11
sub r8, r8, r12 sub r8, r8, r12
stmia r4!, {r5, r6, r7, r8 } @ update fs->qm[4], ..., fs->qm[7] stmia r4!, {r5, r6, r7, r8 } @ update fs->qm[4], ..., fs->qm[7]
b .hf2
.hf_positive:
@ case fs->error > 0
ldmia r4, {r5, r6, r7, r8 }
ldmia r3!, {r9, r10, r11, r12}
add r5, r5, r9
add r6, r6, r10
add r7, r7, r11
add r8, r8, r12
stmia r4!, {r5, r6, r7, r8 } @ update fs->qm[0], ..., fs->qm[3]
ldmia r2!, {r9, r10, r11, r12}
mla lr, r5, r9, lr
mla lr, r6, r10, lr
mla lr, r7, r11, lr
mla lr, r8, r12, lr
ldmia r4, {r5, r6, r7, r8 }
ldmia r3!, {r9, r10, r11, r12}
add r5, r5, r9
add r6, r6, r10
add r7, r7, r11
add r8, r8, r12
stmia r4!, {r5, r6, r7, r8 } @ update fs->qm[4], ..., fs->qm[7]
.hf2: .hf2:
ldmia r2!, {r9, r10, r11, r12} ldmia r2!, {r9, r10, r11, r12}
@ -140,9 +138,9 @@ hybrid_filter:
@ update fs->index @ update fs->index
ldr r1, [r0] @ r1 = fs->index ldr r1, [r0] @ r1 = fs->index
add r1, r1, #1 add r1, r1, #4
ands r1, r1, #15 @ set Z flag (after this, CPSR must keep !!) ands r1, r1, #63 @ set Z flag (after this, CPSR must keep !!)
stmia r0, {r1, r5} @ fs->index = (++fs->index & 15) stmia r0, {r1, r5} @ fs->index = (fs->index + 4) & 63
@ fs->error = (original) *in @ fs->error = (original) *in
@ change *pM, *(pM-1), *(pM-2), *(pM-3) @ change *pM, *(pM-1), *(pM-2), *(pM-3)
@ -166,13 +164,10 @@ hybrid_filter:
sub r11, r12, r11 sub r11, r12, r11
sub r10, r11, r10 sub r10, r11, r10
@ check fs->index is zero
beq .hf_memshl
@ set to the memory: *pA, *(pA-1), *(pA-2), *(pA-3), *pM, *(pM-1), *(pM-2), *(pM-3) @ set to the memory: *pA, *(pA-1), *(pA-2), *(pA-3), *pM, *(pM-1), *(pM-2), *(pM-3)
stmda r2, {r10, r11, r12, lr} stmneda r2, {r10, r11, r12, lr}
stmda r3, {r5, r6, r7, r8} stmneda r3, {r5, r6, r7, r8}
ldmfd sp!, {r4-r12, pc} @ hybrid_filter end (when fs->index != 0) ldmnefd sp!, {r4-r12, pc} @ hybrid_filter end (when fs->index != 0)
.hf_memshl: .hf_memshl:
@ memshl (fs->dl) @ memshl (fs->dl)