annotate libswscale/internal_bfin.S @ 23289:786823779b67

Blackfin optimized YUV420 to RGB CSC Color Space Converters. YUV2 -> RGB BGR for 565, 555 and 888 a.k.a. 24bit color. Speed-up compared to C version compiled with -O3 187.28% Patch by Marc Hoffman %mmh A pleasantst P com% Original thread: Date: May 9, 2007 2:46 AM Subject: [FFmpeg-devel] PATCH BlackFin yuv2rgb color space conversion

/*
 * Copyright (C) 2007 Marc Hoffman <[e-mail]>
 *                    April 20, 2007
 *
 * Blackfin Video Color Space Converters 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 pre scaled by a factor of 4 i.e. left shifted to gain precision.


    New factorization to elliminate the truncation error which was
    occuring 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 subtaction.

     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.

*/

#define mL1 .l1.text
#define mL3 .text
#define MEM mL1

#define DEFUN(fname,where,interface) \
        .section where;              \
        .global _ff_bfin_ ## fname;  \
        .type _ff_bfin_ ## fname, STT_FUNC; \
        .align 8;                    \
        _ff_bfin_ ## fname

#define DEFUN_END(fname) \
        .size _ff_bfin_ ## fname, . - _ff_bfin_ ## fname


.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(yuv2rgb888_line)