ffmpeg/libswscale/bfin/internal_bfin.S

600 lines
19 KiB
ArmAsm

/*
* Copyright (C) 2007 Marc Hoffman <marc.hoffman@analog.com>
* April 20, 2007
*
* Blackfin video color space converter operations
* convert I420 YV12 to RGB in various formats
*
* This file is part of FFmpeg.
*
* FFmpeg is free software; you can redistribute it and/or
* modify it under the terms of the GNU Lesser General Public
* License as published by the Free Software Foundation; either
* version 2.1 of the License, or (at your option) any later version.
*
* FFmpeg is distributed in the hope that it will be useful,
* but WITHOUT ANY WARRANTY; without even the implied warranty of
* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
* Lesser General Public License for more details.
*
* You should have received a copy of the GNU Lesser General Public
* License along with FFmpeg; if not, write to the Free Software
* Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
*/
/*
YUV420 to RGB565 conversion. This routine takes a YUV 420 planar macroblock
and converts it to RGB565. R:5 bits, G:6 bits, B:5 bits.. packed into shorts.
The following calculation is used for the conversion:
r = clipz((y - oy) * cy + crv * (v - 128))
g = clipz((y - oy) * cy + cgv * (v - 128) + cgu * (u - 128))
b = clipz((y - oy) * cy + cbu * (u - 128))
y, u, v are prescaled by a factor of 4 i.e. left-shifted to gain precision.
New factorization to eliminate the truncation error which was
occurring due to the byteop3p.
1) Use the bytop16m to subtract quad bytes we use this in U8 this
then so the offsets need to be renormalized to 8bits.
2) Scale operands up by a factor of 4 not 8 because Blackfin
multiplies include a shift.
3) Compute into the accumulators cy * yx0, cy * yx1.
4) Compute each of the linear equations:
r = clipz((y - oy) * cy + crv * (v - 128))
g = clipz((y - oy) * cy + cgv * (v - 128) + cgu * (u - 128))
b = clipz((y - oy) * cy + cbu * (u - 128))
Reuse of the accumulators requires that we actually multiply
twice once with addition and the second time with a subtraction.
Because of this we need to compute the equations in the order R B
then G saving the writes for B in the case of 24/32 bit color
formats.
API: yuv2rgb_kind (uint8_t *Y, uint8_t *U, uint8_t *V, int *out,
int dW, uint32_t *coeffs);
A B
--- ---
i2 = cb i3 = cr
i1 = coeff i0 = y
Where coeffs have the following layout in memory.
uint32_t oy, oc, zero, cy, crv, rmask, cbu, bmask, cgu, cgv;
coeffs is a pointer to oy.
The {rgb} masks are only utilized by the 565 packing algorithm. Note the data
replication is used to simplify the internal algorithms for the dual Mac
architecture of BlackFin.
All routines are exported with _ff_bfin_ as a symbol prefix.
Rough performance gain compared against -O3:
2779809/1484290 187.28%
which translates to ~33c/pel to ~57c/pel for the reference vs 17.5
c/pel for the optimized implementations. Not sure why there is such a
huge variation on the reference codes on Blackfin I guess it must have
to do with the memory system.
*/
#include "libavutil/bfin/asm.h"
#define MEM mL1
.text
#define COEFF_LEN 11*4
#define COEFF_REL_CY_OFF 4*4
#define ARG_OUT 20
#define ARG_W 24
#define ARG_COEFF 28
DEFUN(yuv2rgb565_line,MEM,
(uint8_t *Y, uint8_t *U, uint8_t *V, int *out, int dW, uint32_t *coeffs)):
link 0;
[--sp] = (r7:4);
p1 = [fp+ARG_OUT];
r3 = [fp+ARG_W];
i0 = r0;
i2 = r1;
i3 = r2;
r0 = [fp+ARG_COEFF];
i1 = r0;
b1 = i1;
l1 = COEFF_LEN;
m0 = COEFF_REL_CY_OFF;
p0 = r3;
r0 = [i0++]; // 2Y
r1.l = w[i2++]; // 2u
r1.h = w[i3++]; // 2v
p0 = p0>>2;
lsetup (.L0565, .L1565) lc0 = p0;
/*
uint32_t oy,oc,zero,cy,crv,rmask,cbu,bmask,cgu,cgv
r0 -- used to load 4ys
r1 -- used to load 2us,2vs
r4 -- y3,y2
r5 -- y1,y0
r6 -- u1,u0
r7 -- v1,v0
*/
r2=[i1++]; // oy
.L0565:
/*
rrrrrrrr gggggggg bbbbbbbb
5432109876543210
bbbbb >>3
gggggggg <<3
rrrrrrrr <<8
rrrrrggggggbbbbb
*/
(r4,r5) = byteop16m (r1:0, r3:2) || r3=[i1++]; // oc
(r7,r6) = byteop16m (r1:0, r3:2) (r);
r5 = r5 << 2 (v); // y1,y0
r4 = r4 << 2 (v); // y3,y2
r6 = r6 << 2 (v) || r0=[i1++]; // u1,u0, r0=zero
r7 = r7 << 2 (v) || r1=[i1++]; // v1,v0 r1=cy
/* Y' = y*cy */
a1 = r1.h*r5.h, a0 = r1.l*r5.l || r1=[i1++]; // crv
/* R = Y+ crv*(Cr-128) */
r2.h = (a1 += r1.h*r7.l), r2.l = (a0 += r1.l*r7.l);
a1 -= r1.h*r7.l, a0 -= r1.l*r7.l || r5=[i1++]; // rmask
r2 = byteop3p(r3:2, r1:0)(LO) || r1=[i1++]; // cbu
r2 = r2 >> 3 (v);
r3 = r2 & r5;
/* B = Y+ cbu*(Cb-128) */
r2.h = (a1 += r1.h*r6.l), r2.l = (a0 += r1.l*r6.l);
a1 -= r1.h*r6.l, a0 -= r1.l*r6.l || r5=[i1++]; // bmask
r2 = byteop3p(r3:2, r1:0)(LO) || r1=[i1++]; // cgu
r2 = r2 << 8 (v);
r2 = r2 & r5;
r3 = r3 | r2;
/* G = Y+ cgu*(Cb-128)+cgv*(Cr-128) */
a1 += r1.h*r6.l, a0 += r1.l*r6.l || r1=[i1++]; // cgv
r2.h = (a1 += r1.h*r7.l), r2.l = (a0 += r1.l*r7.l);
r2 = byteop3p(r3:2, r1:0)(LO) || r5=[i1++m0]; // gmask
r2 = r2 << 3 (v);
r2 = r2 & r5;
r3 = r3 | r2;
[p1++]=r3 || r1=[i1++]; // cy
/* Y' = y*cy */
a1 = r1.h*r4.h, a0 = r1.l*r4.l || r1=[i1++]; // crv
/* R = Y+ crv*(Cr-128) */
r2.h = (a1 += r1.h*r7.h), r2.l = (a0 += r1.l*r7.h);
a1 -= r1.h*r7.h, a0 -= r1.l*r7.h || r5=[i1++]; // rmask
r2 = byteop3p(r3:2, r1:0)(LO) || r1=[i1++]; // cbu
r2 = r2 >> 3 (v);
r3 = r2 & r5;
/* B = Y+ cbu*(Cb-128) */
r2.h = (a1 += r1.h*r6.h), r2.l = (a0 += r1.l*r6.h);
a1 -= r1.h*r6.h, a0 -= r1.l*r6.h || r5=[i1++]; // bmask
r2 = byteop3p(r3:2, r1:0)(LO) || r1=[i1++]; // cgu
r2 = r2 << 8 (v);
r2 = r2 & r5;
r3 = r3 | r2;
/* G = Y+ cgu*(Cb-128)+cgv*(Cr-128) */
a1 += r1.h*r6.h, a0 += r1.l*r6.h || r1=[i1++]; // cgv
r2.h = (a1 += r1.h*r7.h), r2.l = (a0 += r1.l*r7.h) || r5=[i1++]; // gmask
r2 = byteop3p(r3:2, r1:0)(LO) || r0 = [i0++]; // 2Y
r2 = r2 << 3 (v) || r1.l = w[i2++]; // 2u
r2 = r2 & r5;
r3 = r3 | r2;
[p1++]=r3 || r1.h = w[i3++]; // 2v
.L1565: r2=[i1++]; // oy
l1 = 0;
(r7:4) = [sp++];
unlink;
rts;
DEFUN_END(yuv2rgb565_line)
DEFUN(yuv2rgb555_line,MEM,
(uint8_t *Y, uint8_t *U, uint8_t *V, int *out, int dW, uint32_t *coeffs)):
link 0;
[--sp] = (r7:4);
p1 = [fp+ARG_OUT];
r3 = [fp+ARG_W];
i0 = r0;
i2 = r1;
i3 = r2;
r0 = [fp+ARG_COEFF];
i1 = r0;
b1 = i1;
l1 = COEFF_LEN;
m0 = COEFF_REL_CY_OFF;
p0 = r3;
r0 = [i0++]; // 2Y
r1.l = w[i2++]; // 2u
r1.h = w[i3++]; // 2v
p0 = p0>>2;
lsetup (.L0555, .L1555) lc0 = p0;
/*
uint32_t oy,oc,zero,cy,crv,rmask,cbu,bmask,cgu,cgv
r0 -- used to load 4ys
r1 -- used to load 2us,2vs
r4 -- y3,y2
r5 -- y1,y0
r6 -- u1,u0
r7 -- v1,v0
*/
r2=[i1++]; // oy
.L0555:
/*
rrrrrrrr gggggggg bbbbbbbb
5432109876543210
bbbbb >>3
gggggggg <<2
rrrrrrrr <<7
xrrrrrgggggbbbbb
*/
(r4,r5) = byteop16m (r1:0, r3:2) || r3=[i1++]; // oc
(r7,r6) = byteop16m (r1:0, r3:2) (r);
r5 = r5 << 2 (v); // y1,y0
r4 = r4 << 2 (v); // y3,y2
r6 = r6 << 2 (v) || r0=[i1++]; // u1,u0, r0=zero
r7 = r7 << 2 (v) || r1=[i1++]; // v1,v0 r1=cy
/* Y' = y*cy */
a1 = r1.h*r5.h, a0 = r1.l*r5.l || r1=[i1++]; // crv
/* R = Y+ crv*(Cr-128) */
r2.h = (a1 += r1.h*r7.l), r2.l = (a0 += r1.l*r7.l);
a1 -= r1.h*r7.l, a0 -= r1.l*r7.l || r5=[i1++]; // rmask
r2 = byteop3p(r3:2, r1:0)(LO) || r1=[i1++]; // cbu
r2 = r2 >> 3 (v);
r3 = r2 & r5;
/* B = Y+ cbu*(Cb-128) */
r2.h = (a1 += r1.h*r6.l), r2.l = (a0 += r1.l*r6.l);
a1 -= r1.h*r6.l, a0 -= r1.l*r6.l || r5=[i1++]; // bmask
r2 = byteop3p(r3:2, r1:0)(LO) || r1=[i1++]; // cgu
r2 = r2 << 7 (v);
r2 = r2 & r5;
r3 = r3 | r2;
/* G = Y+ cgu*(Cb-128)+cgv*(Cr-128) */
a1 += r1.h*r6.l, a0 += r1.l*r6.l || r1=[i1++]; // cgv
r2.h = (a1 += r1.h*r7.l), r2.l = (a0 += r1.l*r7.l);
r2 = byteop3p(r3:2, r1:0)(LO) || r5=[i1++m0]; // gmask
r2 = r2 << 2 (v);
r2 = r2 & r5;
r3 = r3 | r2;
[p1++]=r3 || r1=[i1++]; // cy
/* Y' = y*cy */
a1 = r1.h*r4.h, a0 = r1.l*r4.l || r1=[i1++]; // crv
/* R = Y+ crv*(Cr-128) */
r2.h = (a1 += r1.h*r7.h), r2.l = (a0 += r1.l*r7.h);
a1 -= r1.h*r7.h, a0 -= r1.l*r7.h || r5=[i1++]; // rmask
r2 = byteop3p(r3:2, r1:0)(LO) || r1=[i1++]; // cbu
r2 = r2 >> 3 (v);
r3 = r2 & r5;
/* B = Y+ cbu*(Cb-128) */
r2.h = (a1 += r1.h*r6.h), r2.l = (a0 += r1.l*r6.h);
a1 -= r1.h*r6.h, a0 -= r1.l*r6.h || r5=[i1++]; // bmask
r2 = byteop3p(r3:2, r1:0)(LO) || r1=[i1++]; // cgu
r2 = r2 << 7 (v);
r2 = r2 & r5;
r3 = r3 | r2;
/* G = Y+ cgu*(Cb-128)+cgv*(Cr-128) */
a1 += r1.h*r6.h, a0 += r1.l*r6.h || r1=[i1++]; // cgv
r2.h = (a1 += r1.h*r7.h), r2.l = (a0 += r1.l*r7.h) || r5=[i1++]; // gmask
r2 = byteop3p(r3:2, r1:0)(LO) || r0=[i0++]; // 4Y
r2 = r2 << 2 (v) || r1.l=w[i2++]; // 2u
r2 = r2 & r5;
r3 = r3 | r2;
[p1++]=r3 || r1.h=w[i3++]; // 2v
.L1555: r2=[i1++]; // oy
l1 = 0;
(r7:4) = [sp++];
unlink;
rts;
DEFUN_END(yuv2rgb555_line)
DEFUN(yuv2rgb24_line,MEM,
(uint8_t *Y, uint8_t *U, uint8_t *V, int *out, int dW, uint32_t *coeffs)):
link 0;
[--sp] = (r7:4);
p1 = [fp+ARG_OUT];
r3 = [fp+ARG_W];
p2 = p1;
p2 += 3;
i0 = r0;
i2 = r1;
i3 = r2;
r0 = [fp+ARG_COEFF]; // coeff buffer
i1 = r0;
b1 = i1;
l1 = COEFF_LEN;
m0 = COEFF_REL_CY_OFF;
p0 = r3;
r0 = [i0++]; // 2Y
r1.l = w[i2++]; // 2u
r1.h = w[i3++]; // 2v
p0 = p0>>2;
lsetup (.L0888, .L1888) lc0 = p0;
/*
uint32_t oy,oc,zero,cy,crv,rmask,cbu,bmask,cgu,cgv
r0 -- used to load 4ys
r1 -- used to load 2us,2vs
r4 -- y3,y2
r5 -- y1,y0
r6 -- u1,u0
r7 -- v1,v0
*/
r2=[i1++]; // oy
.L0888:
(r4,r5) = byteop16m (r1:0, r3:2) || r3=[i1++]; // oc
(r7,r6) = byteop16m (r1:0, r3:2) (r);
r5 = r5 << 2 (v); // y1,y0
r4 = r4 << 2 (v); // y3,y2
r6 = r6 << 2 (v) || r0=[i1++]; // u1,u0, r0=zero
r7 = r7 << 2 (v) || r1=[i1++]; // v1,v0 r1=cy
/* Y' = y*cy */
a1 = r1.h*r5.h, a0 = r1.l*r5.l || r1=[i1++]; // crv
/* R = Y+ crv*(Cr-128) */
r2.h = (a1 += r1.h*r7.l), r2.l = (a0 += r1.l*r7.l);
a1 -= r1.h*r7.l, a0 -= r1.l*r7.l || r5=[i1++]; // rmask
r2 = byteop3p(r3:2, r1:0)(LO) || r1=[i1++]; // cbu
r2=r2>>16 || B[p1++]=r2;
B[p2++]=r2;
/* B = Y+ cbu*(Cb-128) */
r2.h = (a1 += r1.h*r6.l), r2.l = (a0 += r1.l*r6.l);
a1 -= r1.h*r6.l, a0 -= r1.l*r6.l || r5=[i1++]; // bmask
r3 = byteop3p(r3:2, r1:0)(LO) || r1=[i1++]; // cgu
/* G = Y+ cgu*(Cb-128)+cgv*(Cr-128) */
a1 += r1.h*r6.l, a0 += r1.l*r6.l || r1=[i1++]; // cgv
r2.h = (a1 += r1.h*r7.l), r2.l = (a0 += r1.l*r7.l);
r2 = byteop3p(r3:2, r1:0)(LO) || r5=[i1++m0]; // gmask, oy,cy,zero
r2=r2>>16 || B[p1++]=r2;
B[p2++]=r2;
r3=r3>>16 || B[p1++]=r3;
B[p2++]=r3 || r1=[i1++]; // cy
p1+=3;
p2+=3;
/* Y' = y*cy */
a1 = r1.h*r4.h, a0 = r1.l*r4.l || r1=[i1++]; // crv
/* R = Y+ crv*(Cr-128) */
r2.h = (a1 += r1.h*r7.h), r2.l = (a0 += r1.l*r7.h);
a1 -= r1.h*r7.h, a0 -= r1.l*r7.h || r5=[i1++]; // rmask
r2 = byteop3p(r3:2, r1:0)(LO) || r1=[i1++]; // cbu
r2=r2>>16 || B[p1++]=r2;
B[p2++]=r2;
/* B = Y+ cbu*(Cb-128) */
r2.h = (a1 += r1.h*r6.h), r2.l = (a0 += r1.l*r6.h);
a1 -= r1.h*r6.h, a0 -= r1.l*r6.h || r5=[i1++]; // bmask
r3 = byteop3p(r3:2, r1:0)(LO) || r1=[i1++]; // cgu
/* G = Y+ cgu*(Cb-128)+cgv*(Cr-128) */
a1 += r1.h*r6.h, a0 += r1.l*r6.h || r1=[i1++]; // cgv
r2.h = (a1 += r1.h*r7.h), r2.l = (a0 += r1.l*r7.h);
r2 = byteop3p(r3:2, r1:0)(LO) || r5=[i1++]; // gmask
r2=r2>>16 || B[p1++]=r2 || r0 = [i0++]; // 4y
B[p2++]=r2 || r1.l = w[i2++]; // 2u
r3=r3>>16 || B[p1++]=r3 || r1.h = w[i3++]; // 2v
B[p2++]=r3 || r2=[i1++]; // oy
p1+=3;
.L1888: p2+=3;
l1 = 0;
(r7:4) = [sp++];
unlink;
rts;
DEFUN_END(yuv2rgb24_line)
#define ARG_vdst 20
#define ARG_width 24
#define ARG_height 28
#define ARG_lumStride 32
#define ARG_chromStride 36
#define ARG_srcStride 40
DEFUN(uyvytoyv12, mL3, (const uint8_t *src, uint8_t *ydst, uint8_t *udst, uint8_t *vdst,
int width, int height,
int lumStride, int chromStride, int srcStride)):
link 0;
[--sp] = (r7:4,p5:4);
p0 = r1; // Y top even
i2 = r2; // *u
r2 = [fp + ARG_vdst];
i3 = r2; // *v
r1 = [fp + ARG_srcStride];
r2 = r0 + r1;
i0 = r0; // uyvy_T even
i1 = r2; // uyvy_B odd
p2 = [fp + ARG_lumStride];
p1 = p0 + p2; // Y bot odd
p5 = [fp + ARG_width];
p4 = [fp + ARG_height];
r0 = p5;
p4 = p4 >> 1;
p5 = p5 >> 2;
r2 = r0 << 1;
r1 = r1 << 1;
r1 = r1 - r2; // srcStride + (srcStride - 2*width)
r1 += -8; // i0,i1 is pre read need to correct
m0 = r1;
r2 = [fp + ARG_chromStride];
r0 = r0 >> 1;
r2 = r2 - r0;
m1 = r2;
/* I0,I1 - src input line pointers
* p0,p1 - luma output line pointers
* I2 - dstU
* I3 - dstV
*/
lsetup (0f, 1f) lc1 = p4; // H/2
0: r0 = [i0++] || r2 = [i1++];
r1 = [i0++] || r3 = [i1++];
r4 = byteop1p(r1:0, r3:2);
r5 = byteop1p(r1:0, r3:2) (r);
lsetup (2f, 3f) lc0 = p5; // W/4
2: r0 = r0 >> 8(v);
r1 = r1 >> 8(v);
r2 = r2 >> 8(v);
r3 = r3 >> 8(v);
r0 = bytepack(r0, r1);
r2 = bytepack(r2, r3) || [p0++] = r0; // yyyy
r6 = pack(r5.l, r4.l) || [p1++] = r2; // yyyy
r7 = pack(r5.h, r4.h) || r0 = [i0++] || r2 = [i1++];
r6 = bytepack(r6, r7) || r1 = [i0++] || r3 = [i1++];
r4 = byteop1p(r1:0, r3:2) || w[i2++] = r6.l; // uu
3: r5 = byteop1p(r1:0, r3:2) (r) || w[i3++] = r6.h; // vv
i0 += m0;
i1 += m0;
i2 += m1;
i3 += m1;
p0 = p0 + p2;
1: p1 = p1 + p2;
(r7:4,p5:4) = [sp++];
unlink;
rts;
DEFUN_END(uyvytoyv12)
DEFUN(yuyvtoyv12, mL3, (const uint8_t *src, uint8_t *ydst, uint8_t *udst, uint8_t *vdst,
int width, int height,
int lumStride, int chromStride, int srcStride)):
link 0;
[--sp] = (r7:4,p5:4);
p0 = r1; // Y top even
i2 = r2; // *u
r2 = [fp + ARG_vdst];
i3 = r2; // *v
r1 = [fp + ARG_srcStride];
r2 = r0 + r1;
i0 = r0; // uyvy_T even
i1 = r2; // uyvy_B odd
p2 = [fp + ARG_lumStride];
p1 = p0 + p2; // Y bot odd
p5 = [fp + ARG_width];
p4 = [fp + ARG_height];
r0 = p5;
p4 = p4 >> 1;
p5 = p5 >> 2;
r2 = r0 << 1;
r1 = r1 << 1;
r1 = r1 - r2; // srcStride + (srcStride - 2*width)
r1 += -8; // i0,i1 is pre read need to correct
m0 = r1;
r2 = [fp + ARG_chromStride];
r0 = r0 >> 1;
r2 = r2 - r0;
m1 = r2;
/* I0,I1 - src input line pointers
* p0,p1 - luma output line pointers
* I2 - dstU
* I3 - dstV
*/
lsetup (0f, 1f) lc1 = p4; // H/2
0: r0 = [i0++] || r2 = [i1++];
r1 = [i0++] || r3 = [i1++];
r4 = bytepack(r0, r1);
r5 = bytepack(r2, r3);
lsetup (2f, 3f) lc0 = p5; // W/4
2: r0 = r0 >> 8(v) || [p0++] = r4; // yyyy-even
r1 = r1 >> 8(v) || [p1++] = r5; // yyyy-odd
r2 = r2 >> 8(v);
r3 = r3 >> 8(v);
r4 = byteop1p(r1:0, r3:2);
r5 = byteop1p(r1:0, r3:2) (r);
r6 = pack(r5.l, r4.l);
r7 = pack(r5.h, r4.h) || r0 = [i0++] || r2 = [i1++];
r6 = bytepack(r6, r7) || r1 = [i0++] || r3 = [i1++];
r4 = bytepack(r0, r1) || w[i2++] = r6.l; // uu
3: r5 = bytepack(r2, r3) || w[i3++] = r6.h; // vv
i0 += m0;
i1 += m0;
i2 += m1;
i3 += m1;
p0 = p0 + p2;
1: p1 = p1 + p2;
(r7:4,p5:4) = [sp++];
unlink;
rts;
DEFUN_END(yuyvtoyv12)