mirror of
https://github.com/claunia/flac.git
synced 2025-12-16 18:54:26 +00:00
add miroslav's version of FLAC__lpc_restore_signal, add SSE version of FLAC__lpc_compute_autocorrelation
This commit is contained in:
Josh Coalson
@@ -21,6 +21,7 @@
|
||||
data_section
|
||||
|
||||
cglobal FLAC__lpc_compute_autocorrelation_asm
|
||||
cglobal FLAC__lpc_restore_signal_asm
|
||||
|
||||
code_section
|
||||
|
||||
@@ -28,65 +29,430 @@ cglobal FLAC__lpc_compute_autocorrelation_asm
|
||||
;
|
||||
; void FLAC__lpc_compute_autocorrelation_asm(const real data[], unsigned data_len, unsigned lag, real autoc[])
|
||||
; {
|
||||
; real d;
|
||||
; unsigned i;
|
||||
; real d;
|
||||
; unsigned sample, coeff;
|
||||
; const unsigned limit = data_len - lag;
|
||||
;
|
||||
; assert(lag > 0);
|
||||
; assert(lag <= data_len);
|
||||
;
|
||||
; while(lag--) {
|
||||
; for(i = lag, d = 0.0; i < data_len; i++)
|
||||
; d += data[i] * data[i - lag];
|
||||
; autoc[lag] = d;
|
||||
; }
|
||||
; for(coeff = 0; coeff < lag; coeff++)
|
||||
; autoc[coeff] = 0.0;
|
||||
; for(sample = 0; sample <= limit; sample++) {
|
||||
; d = data[sample];
|
||||
; for(coeff = 0; coeff < lag; coeff++)
|
||||
; autoc[coeff] += d * data[sample+coeff];
|
||||
; }
|
||||
; for(; sample < data_len; sample++) {
|
||||
; d = data[sample];
|
||||
; for(coeff = 0; coeff < data_len - sample; coeff++)
|
||||
; autoc[coeff] += d * data[sample+coeff];
|
||||
; }
|
||||
; }
|
||||
;
|
||||
FLAC__lpc_compute_autocorrelation_asm:
|
||||
|
||||
; esp + 20 == data[]
|
||||
; esp + 24 == data_len
|
||||
; esp + 28 == lag
|
||||
; esp + 32 == autoc[]
|
||||
|
||||
push ebp
|
||||
lea ebp, [esp + 8]
|
||||
push eax
|
||||
push ebx
|
||||
push ecx
|
||||
push edx
|
||||
push esi
|
||||
push edi
|
||||
|
||||
mov esi, [ebp] ; esi == data
|
||||
mov ecx, [ebp + 4] ; ecx == data_len
|
||||
mov edx, [ebp + 8] ; edx == lag
|
||||
mov edi, [ebp + 12] ; edi == autoc
|
||||
mov edx, [esp + 28] ; edx == lag
|
||||
mov edi, [esp + 32] ; edi == autoc
|
||||
|
||||
; for(coeff = 0; coeff < lag; coeff++)
|
||||
; autoc[coeff] = 0.0;
|
||||
mov ecx, edx ; ecx = # of dwords of 0 to write
|
||||
xor eax, eax
|
||||
rep stosd
|
||||
|
||||
; const unsigned limit = data_len - lag;
|
||||
mov esi, [esp + 24] ; esi == data_len
|
||||
mov ecx, esi
|
||||
sub ecx, edx ; ecx == limit
|
||||
mov edi, [esp + 32] ; edi == autoc
|
||||
inc ecx ; we are looping <= limit so we add one to the counter
|
||||
|
||||
; for(sample = 0; sample <= limit; sample++) {
|
||||
; d = data[sample];
|
||||
; for(coeff = 0; coeff < lag; coeff++)
|
||||
; autoc[coeff] += d * data[sample+coeff];
|
||||
; }
|
||||
mov eax, [esp + 20] ; eax == &data[sample] <- &data[0]
|
||||
xor ebp, ebp ; ebp == sample <- 0
|
||||
ALIGN 16
|
||||
.outer_loop:
|
||||
test edx, edx
|
||||
jz .outer_end
|
||||
dec edx ; lag--
|
||||
|
||||
mov ebx, edx ; ebx == i <- lag
|
||||
xor eax, eax ; eax == i-lag
|
||||
fldz ; ST = d <- 0.0
|
||||
push eax ; save &data[sample], for inner_loop:
|
||||
; eax == &data[sample+coeff] <- &data[sample]
|
||||
fld dword [eax] ; ST = d <- data[sample]
|
||||
mov edx, [esp + 32] ; edx <- lag (note the +4 due to the above 'push eax')
|
||||
mov ebx, edi ; ebx == &autoc[coeff] <- &autoc[0]
|
||||
and edx, 3 ; edx <- lag % 4
|
||||
jz .inner_start
|
||||
cmp edx, 1
|
||||
je .warmup_1
|
||||
cmp edx, 2
|
||||
je .warmup_2
|
||||
.warmup_3:
|
||||
fld st0 ; ST = d d
|
||||
fmul dword [eax] ; ST = d*data[sample+coeff] d
|
||||
add eax, byte 4 ; (sample+coeff)++
|
||||
fadd dword [ebx] ; ST = autoc[coeff]+d*data[sample+coeff] d
|
||||
fstp dword [ebx] ; autoc[coeff]+=d*data[sample+coeff] ST = d
|
||||
add ebx, byte 4 ; coeff++
|
||||
.warmup_2:
|
||||
fld st0 ; ST = d d
|
||||
fmul dword [eax] ; ST = d*data[sample+coeff] d
|
||||
add eax, byte 4 ; (sample+coeff)++
|
||||
fadd dword [ebx] ; ST = autoc[coeff]+d*data[sample+coeff] d
|
||||
fstp dword [ebx] ; autoc[coeff]+=d*data[sample+coeff] ST = d
|
||||
add ebx, byte 4 ; coeff++
|
||||
.warmup_1:
|
||||
fld st0 ; ST = d d
|
||||
fmul dword [eax] ; ST = d*data[sample+coeff] d
|
||||
add eax, byte 4 ; (sample+coeff)++
|
||||
fadd dword [ebx] ; ST = autoc[coeff]+d*data[sample+coeff] d
|
||||
fstp dword [ebx] ; autoc[coeff]+=d*data[sample+coeff] ST = d
|
||||
add ebx, byte 4 ; coeff++
|
||||
.inner_start:
|
||||
mov edx, [esp + 32] ; edx <- lag (note the +4 due to the above 'push eax')
|
||||
shr edx, 2 ; edx <- lag / 4
|
||||
jz .inner_end
|
||||
ALIGN 16
|
||||
.inner_loop:
|
||||
cmp ebx, ecx
|
||||
jae short .inner_end
|
||||
fld qword [esi + ebx * 8] ; ST = data[i] d
|
||||
fmul qword [esi + eax * 8] ; ST = data[i]*data[i-lag] d
|
||||
faddp st1 ; d += data[i]*data[i-lag] ST = d
|
||||
inc eax
|
||||
inc ebx
|
||||
jmp .inner_loop
|
||||
fld st0 ; ST = d d
|
||||
fmul dword [eax] ; ST = d*data[sample+coeff] d
|
||||
add eax, byte 4 ; (sample+coeff)++
|
||||
fadd dword [ebx] ; ST = autoc[coeff]+d*data[sample+coeff] d
|
||||
fstp dword [ebx] ; autoc[coeff]+=d*data[sample+coeff] ST = d
|
||||
add ebx, byte 4 ; coeff++
|
||||
fld st0 ; ST = d d
|
||||
fmul dword [eax] ; ST = d*data[sample+coeff] d
|
||||
add eax, byte 4 ; (sample+coeff)++
|
||||
fadd dword [ebx] ; ST = autoc[coeff]+d*data[sample+coeff] d
|
||||
fstp dword [ebx] ; autoc[coeff]+=d*data[sample+coeff] ST = d
|
||||
add ebx, byte 4 ; coeff++
|
||||
fld st0 ; ST = d d
|
||||
fmul dword [eax] ; ST = d*data[sample+coeff] d
|
||||
add eax, byte 4 ; (sample+coeff)++
|
||||
fadd dword [ebx] ; ST = autoc[coeff]+d*data[sample+coeff] d
|
||||
fstp dword [ebx] ; autoc[coeff]+=d*data[sample+coeff] ST = d
|
||||
add ebx, byte 4 ; coeff++
|
||||
fld st0 ; ST = d d
|
||||
fmul dword [eax] ; ST = d*data[sample+coeff] d
|
||||
add eax, byte 4 ; (sample+coeff)++
|
||||
fadd dword [ebx] ; ST = autoc[coeff]+d*data[sample+coeff] d
|
||||
fstp dword [ebx] ; autoc[coeff]+=d*data[sample+coeff] ST = d
|
||||
add ebx, byte 4 ; coeff++
|
||||
dec edx
|
||||
jnz .inner_loop
|
||||
.inner_end:
|
||||
pop eax ; restore &data[sample]
|
||||
fstp st0 ; pop d, ST = empty
|
||||
inc ebp ; sample++
|
||||
add eax, byte 4 ; &data[sample++]
|
||||
dec ecx
|
||||
jnz .outer_loop
|
||||
|
||||
fstp qword [edi + edx * 8]
|
||||
|
||||
jmp .outer_loop
|
||||
.outer_end:
|
||||
; for(; sample < data_len; sample++) {
|
||||
; d = data[sample];
|
||||
; for(coeff = 0; coeff < data_len - sample; coeff++)
|
||||
; autoc[coeff] += d * data[sample+coeff];
|
||||
; }
|
||||
mov ecx, [esp + 28] ; ecx <- lag
|
||||
dec ecx ; ecx <- lag - 1
|
||||
jz .end ; skip loop if 0
|
||||
ALIGN 16
|
||||
.outer_loop2:
|
||||
push eax ; save &data[sample], for inner_loop2:
|
||||
; eax == &data[sample+coeff] <- &data[sample]
|
||||
fld dword [eax] ; ST = d <- data[sample]
|
||||
mov edx, esi ; edx <- data_len
|
||||
sub edx, ebp ; edx <- data_len-sample
|
||||
mov ebx, edi ; ebx == &autoc[coeff] <- &autoc[0]
|
||||
ALIGN 16
|
||||
.inner_loop2:
|
||||
fld st0 ; ST = d d
|
||||
fmul dword [eax] ; ST = d*data[sample+coeff] d
|
||||
add eax, byte 4 ; (sample+coeff)++
|
||||
fadd dword [ebx] ; ST = autoc[coeff]+d*data[sample+coeff] d
|
||||
fstp dword [ebx] ; autoc[coeff]+=d*data[sample+coeff] ST = d
|
||||
add ebx, byte 4 ; coeff++
|
||||
dec edx
|
||||
jnz .inner_loop2
|
||||
pop eax ; restore &data[sample]
|
||||
fstp st0 ; pop d, ST = empty
|
||||
inc ebp ; sample++
|
||||
add eax, byte 4 ; &data[sample++]
|
||||
dec ecx
|
||||
jnz .outer_loop2
|
||||
|
||||
.end:
|
||||
pop edi
|
||||
pop esi
|
||||
pop ebx
|
||||
pop ebp
|
||||
ret
|
||||
|
||||
; NOTE: this SSE version is not even tested yet and only works for lag == 8
|
||||
FLAC__lpc_compute_autocorrelation_sse:
|
||||
|
||||
; esp + 4 == data[]
|
||||
; esp + 8 == data_len
|
||||
; esp + 12 == lag
|
||||
; esp + 16 == autoc[]
|
||||
|
||||
; for(coeff = 0; coeff < lag; coeff++)
|
||||
; autoc[coeff] = 0.0;
|
||||
xorps xmm6, xmm6
|
||||
xorps xmm7, xmm7
|
||||
|
||||
mov edx, [esp + 8] ; edx == data_len
|
||||
mov eax, [esp + 4] ; eax == &data[sample] <- &data[0]
|
||||
|
||||
movss xmm0, [eax] ; xmm0 = 0,0,0,data[0]
|
||||
add eax, 4
|
||||
movaps xmm2, xmm0 ; xmm2 = 0,0,0,data[0]
|
||||
shufps xmm0, xmm0, 0 ; xmm0 = data[0],data[0],data[0],data[0]
|
||||
movaps xmm1, xmm0 ; xmm1 = data[0],data[0],data[0],data[0]
|
||||
xorps xmm3, xmm3 ; xmm3 = 0,0,0,0
|
||||
.warmup: ; xmm3:xmm2 = data[sample-[7..0]]
|
||||
movps xmm4, xmm0
|
||||
movps xmm5, xmm1 ; xmm5:xmm4 = xmm1:xmm0 = data[sample]*8
|
||||
mulps xmm4, xmm2
|
||||
mulps xmm5, xmm3 ; xmm5:xmm4 = xmm1:xmm0 * xmm3:xmm2
|
||||
addps xmm6, xmm4
|
||||
addps xmm7, xmm5 ; xmm7:xmm6 += xmm1:xmm0 * xmm3:xmm2
|
||||
dec edx
|
||||
;* there's no need to even check for this because we know that lag == 8
|
||||
;* and data_len >= lag, so our 1-sample warmup cannot finish the loop
|
||||
; jz .loop_end
|
||||
ALIGN 16
|
||||
.loop_8:
|
||||
; read the next sample
|
||||
movss xmm0, [eax] ; xmm0 = 0,0,0,data[sample]
|
||||
add eax, 4
|
||||
shufps xmm0, xmm0, 0 ; xmm0 = data[sample],data[sample],data[sample],data[sample]
|
||||
movaps xmm1, xmm0 ; xmm1 = data[sample],data[sample],data[sample],data[sample]
|
||||
; now shift the lagged samples
|
||||
movaps xmm4, xmm2
|
||||
movaps xmm5, xmm3
|
||||
shufps xmm2, xmm4, 2103h ; xmm2
|
||||
shufps xmm3, xmm5, 2103h ; xmm2
|
||||
movss xmm3, xmm2
|
||||
movss xmm2, xmm0
|
||||
|
||||
movps xmm4, xmm0
|
||||
movps xmm5, xmm1 ; xmm5:xmm4 = xmm1:xmm0 = data[sample]*8
|
||||
mulps xmm4, xmm2
|
||||
mulps xmm5, xmm3 ; xmm5:xmm4 = xmm1:xmm0 * xmm3:xmm2
|
||||
addps xmm6, xmm4
|
||||
addps xmm7, xmm5 ; xmm7:xmm6 += xmm1:xmm0 * xmm3:xmm2
|
||||
dec edx
|
||||
jnz .loop_8
|
||||
.loop_end:
|
||||
; store autoc
|
||||
mov edx, [esp + 16] ; edx == autoc
|
||||
movups xmm6, [edx]
|
||||
movups xmm7, [edx + 4]
|
||||
|
||||
.end:
|
||||
ret
|
||||
|
||||
; **********************************************************************
|
||||
;
|
||||
; void FLAC__lpc_restore_signal(const int32 residual[], unsigned data_len, const int32 qlp_coeff[], unsigned order, int lp_quantization, int32 data[])
|
||||
; {
|
||||
; unsigned i, j;
|
||||
; int32 sum;
|
||||
;
|
||||
; assert(order > 0);
|
||||
;
|
||||
; for(i = 0; i < data_len; i++) {
|
||||
; sum = 0;
|
||||
; for(j = 0; j < order; j++)
|
||||
; sum += qlp_coeff[j] * data[i-j-1];
|
||||
; data[i] = residual[i] + (sum >> lp_quantization);
|
||||
; }
|
||||
; }
|
||||
FLAC__lpc_restore_signal_asm:
|
||||
; [esp + 20] == residual[]
|
||||
; [esp + 24] == data_len
|
||||
; [esp + 28] == qlp_coeff[]
|
||||
; [esp + 32] == order
|
||||
; [esp + 36] == lp_quantization
|
||||
; [esp + 40] == data[]
|
||||
|
||||
push ebp
|
||||
push ebx
|
||||
push esi
|
||||
push edi
|
||||
|
||||
mov esi, [esp + 20]
|
||||
mov ebx, [esp + 24]
|
||||
mov eax, [esp + 32]
|
||||
mov edi, [esp + 40]
|
||||
|
||||
cmp eax, byte 32 ; for order <= 32 there is a faster routine
|
||||
jbe short .b
|
||||
|
||||
; This version is here just for completeness, since FLAC__MAX_LPC_ORDER == 32
|
||||
ALIGN 16
|
||||
.loop_i_a
|
||||
xor ebp, ebp
|
||||
mov ecx, [esp + 32]
|
||||
mov edx, ecx
|
||||
shl edx, 2
|
||||
add edx, [esp + 28]
|
||||
neg ecx
|
||||
ALIGN 16
|
||||
.loop_j_a
|
||||
sub edx, 4
|
||||
mov eax, [edx]
|
||||
imul eax, [edi + 4 * ecx]
|
||||
add ebp, eax
|
||||
inc ecx
|
||||
jnz .loop_j_a
|
||||
|
||||
mov cl, [esp + 36]
|
||||
sar ebp, cl
|
||||
add ebp, [esi]
|
||||
mov [edi], ebp
|
||||
add edi, byte 4
|
||||
add esi, byte 4
|
||||
|
||||
dec ebx
|
||||
jnz .loop_i_a
|
||||
|
||||
jmp .end
|
||||
|
||||
.b
|
||||
sub esi, edi
|
||||
neg eax
|
||||
lea edx, [eax + eax * 8 + .jumper_0]
|
||||
inc edx
|
||||
mov eax, [esp + 28]
|
||||
xor ebp, ebp
|
||||
jmp edx
|
||||
;.jumper_32
|
||||
mov ecx, [eax + 124] ;32
|
||||
imul ecx, [edi - 128]
|
||||
add ebp, ecx
|
||||
mov ecx, [eax + 120]
|
||||
imul ecx, [edi - 124]
|
||||
add ebp, ecx
|
||||
mov ecx, [eax + 116] ;30
|
||||
imul ecx, [edi - 120]
|
||||
add ebp, ecx
|
||||
mov ecx, [eax + 112]
|
||||
imul ecx, [edi - 116]
|
||||
add ebp, ecx
|
||||
mov ecx, [eax + 108]
|
||||
imul ecx, [edi - 112]
|
||||
add ebp, ecx
|
||||
mov ecx, [eax + 104]
|
||||
imul ecx, [edi - 108]
|
||||
add ebp, ecx
|
||||
mov ecx, [eax + 100]
|
||||
imul ecx, [edi - 104]
|
||||
add ebp, ecx
|
||||
mov ecx, [eax + 96] ;25
|
||||
imul ecx, [edi - 100]
|
||||
add ebp, ecx
|
||||
mov ecx, [eax + 92]
|
||||
imul ecx, [edi - 96]
|
||||
add ebp, ecx
|
||||
mov ecx, [eax + 88]
|
||||
imul ecx, [edi - 92]
|
||||
add ebp, ecx
|
||||
mov ecx, [eax + 84]
|
||||
imul ecx, [edi - 88]
|
||||
add ebp, ecx
|
||||
mov ecx, [eax + 80]
|
||||
imul ecx, [edi - 84]
|
||||
add ebp, ecx
|
||||
mov ecx, [eax + 76] ;20
|
||||
imul ecx, [edi - 80]
|
||||
add ebp, ecx
|
||||
mov ecx, [eax + 72]
|
||||
imul ecx, [edi - 76]
|
||||
add ebp, ecx
|
||||
mov ecx, [eax + 68]
|
||||
imul ecx, [edi - 72]
|
||||
add ebp, ecx
|
||||
mov ecx, [eax + 64]
|
||||
imul ecx, [edi - 68]
|
||||
add ebp, ecx
|
||||
mov ecx, [eax + 60]
|
||||
imul ecx, [edi - 64]
|
||||
add ebp, ecx
|
||||
mov ecx, [eax + 56] ;15
|
||||
imul ecx, [edi - 60]
|
||||
add ebp, ecx
|
||||
mov ecx, [eax + 52]
|
||||
imul ecx, [edi - 56]
|
||||
add ebp, ecx
|
||||
mov ecx, [eax + 48]
|
||||
imul ecx, [edi - 52]
|
||||
add ebp, ecx
|
||||
mov ecx, [eax + 44]
|
||||
imul ecx, [edi - 48]
|
||||
add ebp, ecx
|
||||
mov ecx, [eax + 40]
|
||||
imul ecx, [edi - 44]
|
||||
add ebp, ecx
|
||||
mov ecx, [eax + 36] ;10
|
||||
imul ecx, [edi - 40]
|
||||
add ebp, ecx
|
||||
mov ecx, [eax + 32]
|
||||
imul ecx, [edi - 36]
|
||||
add ebp, ecx
|
||||
mov ecx, [eax + 28]
|
||||
imul ecx, [edi - 32]
|
||||
add ebp, ecx
|
||||
mov ecx, [eax + 24]
|
||||
imul ecx, [edi - 28]
|
||||
add ebp, ecx
|
||||
mov ecx, [eax + 20]
|
||||
imul ecx, [edi - 24]
|
||||
add ebp, ecx
|
||||
mov ecx, [eax + 16] ;5
|
||||
imul ecx, [edi - 20]
|
||||
add ebp, ecx
|
||||
mov ecx, [eax + 12]
|
||||
imul ecx, [edi - 16]
|
||||
add ebp, ecx
|
||||
mov ecx, [eax + 8]
|
||||
imul ecx, [edi - 12]
|
||||
add ebp, ecx
|
||||
mov ecx, [eax + 4]
|
||||
imul ecx, [edi - 8]
|
||||
add ebp, ecx
|
||||
mov ecx, [eax] ;there is one byte missing
|
||||
imul ecx, [edi - 4]
|
||||
add ebp, ecx
|
||||
.jumper_0
|
||||
|
||||
mov cl, [esp + 36]
|
||||
sar ebp, cl
|
||||
add ebp, [esi + edi]
|
||||
mov [edi], ebp
|
||||
add edi, byte 4
|
||||
|
||||
dec ebx
|
||||
jz short .end
|
||||
xor ebp, ebp
|
||||
jmp edx
|
||||
|
||||
.end
|
||||
pop edi
|
||||
pop esi
|
||||
pop edx
|
||||
pop ecx
|
||||
pop ebx
|
||||
pop eax
|
||||
pop ebp
|
||||
ret
|
||||
|
||||
|
||||
Reference in New Issue
Block a user