yading@10: /*
yading@10:  * Blackfin Pixel Operations
yading@10:  * Copyright (C) 2007 Marc Hoffman <marc.hoffman@analog.com>
yading@10:  *
yading@10:  * This file is part of FFmpeg.
yading@10:  *
yading@10:  * FFmpeg is free software; you can redistribute it and/or
yading@10:  * modify it under the terms of the GNU Lesser General Public
yading@10:  * License as published by the Free Software Foundation; either
yading@10:  * version 2.1 of the License, or (at your option) any later version.
yading@10:  *
yading@10:  * FFmpeg is distributed in the hope that it will be useful,
yading@10:  * but WITHOUT ANY WARRANTY; without even the implied warranty of
yading@10:  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
yading@10:  * Lesser General Public License for more details.
yading@10:  *
yading@10:  * You should have received a copy of the GNU Lesser General Public
yading@10:  * License along with FFmpeg; if not, write to the Free Software
yading@10:  * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
yading@10:  */
yading@10: #include "config_bfin.h"
yading@10: 
yading@10: DEFUN(put_pixels_clamped,mL1,
yading@10:         (int16_t *block, uint8_t *dest, int line_size)):
yading@10:     [--SP] = (R7:4);
yading@10:     R4 = 0;
yading@10:     R5.l = 0x00ff;
yading@10:     R5.h = 0x00ff;
yading@10:     I0 = R0;         // block
yading@10:     I1 = R1;         // dest
yading@10:     R2 += -4;        // line_size
yading@10:     M1 = R2;
yading@10:     P0 = 8;
yading@10:     R0 = [I0++];
yading@10:     R1 = [I0++];
yading@10:     R2 = MAX(R0, R4) (V);
yading@10:     LSETUP (ppc$0,ppc$1) LC0=P0;
yading@10: ppc$0: R2 = MIN(R2, R5) (V);
yading@10:        R3 = MAX(R1, R4) (V);
yading@10:        R3 = MIN(R3, R5) (V)      || R0 = [I0++];
yading@10:        R6 = BYTEPACK (R2,R3)     || R1 = [I0++];
yading@10:        R2 = MAX(R0, R4) (V)      || [I1++] = R6;
yading@10:        R2 = MIN(R2, R5) (V);
yading@10:        R3 = MAX(R1, R4) (V);
yading@10:        R3 = MIN(R3, R5) (V)      || R0 = [I0++];
yading@10:        R6 = BYTEPACK (R2,R3)     || R1 = [I0++];
yading@10: ppc$1: R2 = Max(R0, R4) (V)      || [I1++M1] = R6;
yading@10: 
yading@10:     (R7:4) = [SP++];
yading@10:     RTS;
yading@10: DEFUN_END(put_pixels_clamped)
yading@10: 
yading@10: DEFUN(add_pixels_clamped,mL1,
yading@10:         (int16_t *block, uint8_t *dest, int line_size)):
yading@10:     [-- SP] = (R7:4);
yading@10:     R4 = 0;
yading@10:     I0 = 0;
yading@10:     R2 += -4;        // line_size
yading@10:     M0 = R2;
yading@10:     I1 = R1;         // dest
yading@10:     I3 = R0;         // block
yading@10:     I2 = R1;         // dest
yading@10:     P0 = 8;
yading@10:     M3 = 2;
yading@10:     R0 = [I3++]  || R2 = [I1];
yading@10:     R2 = R2 << 8                      || R0.H = W[I3--]  || R3 = [I1++];
yading@10:     R3 = R3 >> 8                      || R1.L = W[I3]    || I3 += 4;
yading@10:     R6 = BYTEOP3P(R1:0, R3:2) (LO)    || R1.H = W[I3++]  || R2 = [I1];
yading@10: 
yading@10:     LSETUP(apc$2,apc$3) LC1 = P0;
yading@10: apc$2: R7 = BYTEOP3P(R1:0, R3:2) (HI, R) || R0 = [I3++]     || R3 = [I1++M0];
yading@10:        R2 = R2 << 8                      || R0.H = W[I3--];
yading@10:        R3 = R3 >> 8                      || R1.L = W[I3]    || I3 += 4;
yading@10:        R6 = R6 + R7 (S)                  || R1.H = W[I3];
yading@10:        R6 = BYTEOP3P(R1:0, R3:2) (LO)    || I3+=M3          || [I2++]=R6;
yading@10:        R7 = BYTEOP3P(R1:0, R3:2) (HI, R) || R0 = [I3++]     || R2 = [I1];
yading@10:        R2 = R2 << 8                      || R0.H = W[I3--]  || R3 = [I1++];
yading@10:        R3 = R3 >> 8                      || R1.L = W[I3]    || I3 += 4;
yading@10:        R6 = R6 + R7 (S)                  || R1.H = W[I3++];
yading@10: apc$3: R6 = BYTEOP3P(R1:0, R3:2) (LO)    || [I2++M0] = R6   || R2 = [I1];
yading@10: 
yading@10:     (R7:4) = [SP++];
yading@10:     RTS;
yading@10: DEFUN_END(add_pixels_clamped)
yading@10: 
yading@10: DEFUN(diff_pixels,mL1,
yading@10:        (int16_t *block, uint8_t *s1, uint8_t *s2, int stride)):
yading@10:         link 0;
yading@10:         [--sp] = (r7:4);
yading@10:         p0=8;
yading@10:         i3=r0;        // block
yading@10:         i0=r1;        // s1
yading@10:         i1=r2;        // s2
yading@10:         r2=[fp+20];   // stride
yading@10:         r2+=-8;
yading@10:         m0=r2;
yading@10: 
yading@10: 
yading@10:         LSETUP(.LS0,.LE0) LC0=P0;
yading@10:         DISALGNEXCPT                       || R0 = [I0++]   || R2  =[I1++];
yading@10: 
yading@10: .LS0:   DISALGNEXCPT                       || R1 = [I0++]   || R3 = [I1++];
yading@10:         (R5,R4) = BYTEOP16M (R1:0,R3:2)    || R0 = [I0++M0] || R2 = [I1++M0];
yading@10:         (R7,R6) = BYTEOP16M (R1:0,R3:2) (R)|| R0 = [I0++]   || [I3++] = R4;
yading@10:         DISALGNEXCPT                       || R2 = [I1++]   || [I3++] = R5;
yading@10:         [i3++]=r6;
yading@10: .LE0:  [i3++]=r7;
yading@10: 
yading@10:         (r7:4) = [sp++];
yading@10:         unlink;
yading@10:         rts;
yading@10: DEFUN_END(diff_pixels)
yading@10: 
yading@10: /*
yading@10:     for (i = 0; i < 16; i++) {
yading@10:         for (j = 0; j < 16; j++) {
yading@10:           sum += pix[j];
yading@10:         }
yading@10:         pix += line_size;
yading@10:     }
yading@10: */
yading@10: DEFUN(pix_sum,mL1,
yading@10:         (uint8_t *p, int stride)):
yading@10:         link 0;
yading@10:         [--sp] = (r7:4);
yading@10:         p0=8;
yading@10:         i0=r0;        // s1
yading@10:         i1=r0;
yading@10:         m1=r1;
yading@10:         r1=r1+r1;
yading@10:         r1+=-16;       // stride
yading@10:         m0=r1;
yading@10:         i1+=m1;
yading@10: 
yading@10:         r6=0;
yading@10: 
yading@10:         LSETUP(LS$PS,LE$PS) LC0=P0;
yading@10:         DISALGNEXCPT                       || R0 = [I0++]   || R2  =[I1++];
yading@10: 
yading@10: LS$PS:  DISALGNEXCPT                       || R1 = [I0++]   || R3 = [I1++];
yading@10:         (R5,R4) = BYTEOP16P (R3:2,R1:0)    || R0 = [I0++]   || R2 = [I1++];
yading@10:         r6=r6+|+r5;
yading@10:         r6=r6+|+r4;
yading@10:         (R5,R4) = BYTEOP16P (R3:2,R1:0) (R)|| R1 = [I0++]   || R3 = [I1++];
yading@10:         r6=r6+|+r5;
yading@10:         r6=r6+|+r4;
yading@10:         (R5,R4) = BYTEOP16P (R3:2,R1:0)    || R0 = [I0++m0] || R2 = [I1++m0];
yading@10:         r6=r6+|+r5;
yading@10:         r6=r6+|+r4;
yading@10:         (R5,R4) = BYTEOP16P (R3:2,R1:0) (R)|| R0 = [I0++]   || R2 = [I1++];
yading@10:         r6=r6+|+r5;
yading@10: LE$PS:  r6=r6+|+r4;
yading@10:         r0.l=r6.l+r6.h;
yading@10:         r0.h=0;
yading@10: 
yading@10:         (r7:4) = [sp++];
yading@10:         unlink;
yading@10:         rts;
yading@10: DEFUN_END(pix_sum)
yading@10: 
yading@10: 
yading@10: DEFUN(get_pixels,mL1,
yading@10:         (int16_t *av_restrict block, const uint8_t *pixels, int line_size)):
yading@10:         [--sp] = (r7:4);
yading@10:         i3=r0;        // dest
yading@10:         i0=r1;        // src0
yading@10:         p0=8;
yading@10:         r2+=-8;
yading@10:         m0=r2;
yading@10:         LSETUP(gp8$0,gp8$1) LC0=P0;
yading@10: 
yading@10:         DISALGNEXCPT                   || R0 = [I0++];
yading@10:         DISALGNEXCPT                   || R1 = [I0++];
yading@10: 
yading@10: gp8$0:  (R7,R6) = byteunpack R1:0      || R0 = [I0++M0];
yading@10:         (R5,R4) = byteunpack R1:0 (R)  || R0 = [I0++]    || [I3++]=R6;
yading@10:         DISALGNEXCPT                   || R1 = [I0++]    || [I3++]=R7;
yading@10:         [I3++]=R4;
yading@10: gp8$1:  [I3++]=R5
yading@10: 
yading@10: 
yading@10:         (r7:4) = [sp++];
yading@10:         RTS;
yading@10: DEFUN_END(get_pixels)
yading@10: 
yading@10: 
yading@10: /* sad = sad16x16 (ubyte *mb, ubyte *refwin, srcwidth, refwinwidth, h) */
yading@10: /* 91 cycles */
yading@10: DEFUN(z_sad16x16,mL1,
yading@10:         (uint8_t *blk1, uint8_t *blk2, int dsz, int line_size, int h)):
yading@10:         link 0;
yading@10:         I0 = R0;
yading@10:         I1 = R1;
yading@10: 
yading@10:         A1 = A0 = 0;
yading@10:         R0 = [sp+20]; // rwidth
yading@10:         P2 = [sp+24]; // height
yading@10:         R3 = 16;
yading@10:         R0 = R0 - R3;
yading@10:         R3 = R2 - R3;
yading@10:         M1 = R0;
yading@10:         M0 = R3;
yading@10: 
yading@10:         DISALGNEXCPT         || R0 = [I0++]    || R2 = [I1++];
yading@10:         LSETUP (s$16, e$16) LC0=P2;
yading@10: s$16:   DISALGNEXCPT         || R1 = [I0++]    || R3 = [I1++];
yading@10:         SAA (R1:0,R3:2)      || R0 = [I0++]    || R2 = [I1++];
yading@10:         SAA (R1:0,R3:2) (R)  || R1 = [I0++]    || R3 = [I1++];
yading@10:         SAA (R1:0,R3:2)      || R0 = [I0++M0]  || R2 = [I1++M1];
yading@10: e$16:   SAA (R1:0,R3:2) (R)  || R0 = [I0++]    || R2 = [I1++];
yading@10: 
yading@10:         R3=A1.L+A1.H,  R2=A0.L+A0.H ;
yading@10:         R0 = R2 + R3 ;
yading@10:         unlink;
yading@10:         RTS;
yading@10: DEFUN_END(z_sad16x16)
yading@10: 
yading@10: /* sad = sad8x8 (ubyte *mb, ubyte *refwin, int srcwidth, int refwinwidth, int h) */
yading@10: /* 36 cycles */
yading@10: DEFUN(z_sad8x8,mL1,
yading@10:         (uint8_t *blk1, uint8_t *blk2, int dsz, int line_size, int h)):
yading@10:         I0 = R0;
yading@10:         I1 = R1;
yading@10: 
yading@10:         A1 = A0 = 0;
yading@10:         r0 = [sp+12]; // rwidth
yading@10:         P2 = [sp+16]; //height
yading@10:         R3 = 8;
yading@10:         R0 = R0 - R3;
yading@10:         R3 = R2 - R3;
yading@10:         M0 = R3;
yading@10:         M1 = R0;
yading@10: 
yading@10:         LSETUP (s$8, e$8) LC0=P2;
yading@10:         DISALGNEXCPT         || R0 = [I0++]   || R2 = [I1++];
yading@10:         DISALGNEXCPT         || R1 = [I0++]   || R3 = [I1++];
yading@10: s$8:    SAA (R1:0,R3:2)      || R0 = [I0++M0] || R2 = [I1++M1];
yading@10:         SAA (R1:0,R3:2) (R)  || R0 = [I0++]   || R2 = [I1++];
yading@10: e$8:    DISALGNEXCPT         || R1 = [I0++]   || R3 = [I1++];
yading@10: 
yading@10:         R3=A1.L+A1.H,  R2=A0.L+A0.H ;
yading@10:         R0 = R2 + R3 ;
yading@10:         RTS;
yading@10: DEFUN_END(z_sad8x8)
yading@10: 
yading@10: DEFUN(pix_norm1,mL1,
yading@10:         (uint8_t * pix, int line_size)):
yading@10:         [--SP]=(R7:4,P5:3);
yading@10: 
yading@10:         // Fetch the input arguments.
yading@10:         P1 = R0;  // pix
yading@10:         P0 = R1;  // line_size
yading@10:         P5 = 16;  // loop ctr.
yading@10:         P0 -= P5;
yading@10:         M0 = P0;  // M0 = line_size-16;
yading@10:         // Now for the real work.
yading@10:         A1 = A0 = 0;
yading@10:         lsetup(_pix_norm1_blkfn_loopStart, _pix_norm1_blkfn_loopEnd) LC1 = P5;
yading@10:         I0 = P1;
yading@10:         DISALGNEXCPT || r0 = [i0++];
yading@10: 
yading@10: _pix_norm1_blkfn_loopStart:
yading@10:         // following unpacks pix1[0..15] pix1+line_size[0..15]
yading@10:         DISALGNEXCPT || r1 = [i0++];
yading@10: 
yading@10:         (r5, r4) = byteunpack r1:0 || r0 = [i0++];
yading@10:         a1 += r5.h * r5.h, a0 += r5.l * r5.l (is);
yading@10:         a1 += r4.h * r4.h, a0 += r4.l * r4.l (is);
yading@10:         (r5, r4) = byteunpack r1:0(r) || r1 = [i0++];
yading@10:         a1 += r5.h * r5.h, a0 += r5.l * r5.l (is);
yading@10:         a1 += r4.h * r4.h, a0 += r4.l * r4.l (is);
yading@10:         (r5, r4) = byteunpack r1:0 || r0 = [i0++M0];
yading@10:         a1 += r5.h * r5.h, a0 += r5.l * r5.l (is);
yading@10:         a1 += r4.h * r4.h, a0 += r4.l * r4.l (is);
yading@10:         (r5, r4) = byteunpack r1:0(r) || r0 = [i0++];
yading@10:         a1 += r5.h * r5.h, a0 += r5.l * r5.l (is);
yading@10: _pix_norm1_blkfn_loopEnd:
yading@10:         a1 += r4.h * r4.h, a0 += r4.l * r4.l (is);
yading@10: 
yading@10: 
yading@10: // Clean up at the end:
yading@10:         R2 = A0, R3 = A1;
yading@10:         R0 = R2 + R3 (S);
yading@10: 
yading@10:         (R7:4,P5:3)=[SP++];
yading@10: 
yading@10:         RTS;
yading@10: DEFUN_END(pix_norm1)
yading@10: 
yading@10: DEFUN(sse4,mL1,
yading@10:         (void *v, uint8_t *pix1, uint8_t *pix2, int line_size, int h)):
yading@10:         link 0;
yading@10:         [--sp] = (r7:6);
yading@10:         p0=[fp+24];   // h
yading@10:         i0=r1;        // pix1
yading@10:         i1=r2;        // pix2
yading@10:         r2=[fp+20];   // line_size
yading@10:         r2+=-4;
yading@10:         m0=r2;
yading@10: 
yading@10:         a0=a1=0;
yading@10:         LSETUP(.S40,.E40) LC0=P0;
yading@10:         DISALGNEXCPT                       || R0 = [I0++]   || R2  =[I1++];
yading@10: 
yading@10: .S40:   DISALGNEXCPT                       || R1 = [I0++M0] || R3 = [I1++M0];
yading@10:         (R7,R6) = BYTEOP16M (R1:0,R3:2);
yading@10:         a0 += r7.l * r7.l, a1 += r7.h * r7.h (is);
yading@10: .E40:   a0 += r6.l * r6.l, a1 += r6.h * r6.h (is);
yading@10:         a0 += a1;
yading@10:         r0 = a0;
yading@10: 
yading@10:         (r7:6) = [sp++];
yading@10:         unlink;
yading@10:         rts;
yading@10: DEFUN_END(sse4)
yading@10: 
yading@10: DEFUN(sse8,mL1,
yading@10:         (void *v, uint8_t *pix1, uint8_t *pix2, int line_size, int h)):
yading@10:         link 0;
yading@10:         [--sp] = (r7:6);
yading@10:         p0=[fp+24];   // h
yading@10:         i0=r1;        // pix1
yading@10:         i1=r2;        // pix2
yading@10:         r2=[fp+20];   // line_size
yading@10:         r2+=-8;
yading@10:         m0=r2;
yading@10: 
yading@10:         a0=a1=0;
yading@10:         LSETUP(.S80,.E80) LC0=P0;
yading@10:         DISALGNEXCPT                       || R0 = [I0++]   || R2  =[I1++];
yading@10: 
yading@10: .S80:   DISALGNEXCPT                       || R1 = [I0++]   || R3 = [I1++];
yading@10:         (R7,R6) = BYTEOP16M (R1:0,R3:2)    || R0 = [I0++M0] || R2 = [I1++M0];
yading@10:         a0 += r7.l * r7.l, a1 += r7.h * r7.h (is);
yading@10:         a0 += r6.l * r6.l, a1 += r6.h * r6.h (is);
yading@10:         (R7,R6) = BYTEOP16M (R1:0,R3:2) (R)|| R0 = [I0++]   || R2 = [I1++];
yading@10:         a0 += r7.l * r7.l, a1 += r7.h * r7.h (is);
yading@10: .E80:   a0 += r6.l * r6.l, a1 += r6.h * r6.h (is);
yading@10:         a0 += a1;
yading@10:         r0 = a0;
yading@10: 
yading@10:         (r7:6) = [sp++];
yading@10:         unlink;
yading@10:         rts;
yading@10: DEFUN_END(sse8)
yading@10: 
yading@10: DEFUN(sse16,mL1,
yading@10:         (void *v, uint8_t *pix1, uint8_t *pix2, int line_size, int h)):
yading@10:         link 0;
yading@10:         [--sp] = (r7:6);
yading@10:         p0=[fp+24];   // h
yading@10:         i0=r1;        // pix1
yading@10:         i1=r2;        // pix2
yading@10:         r2=[fp+20];   // line_size
yading@10:         r2+=-16;
yading@10:         m0=r2;
yading@10: 
yading@10:         a0=a1=0;
yading@10:         DISALGNEXCPT                       || R0 = [I0++]   || R2  =[I1++];
yading@10:         LSETUP(.S160,.E160) LC0=P0;
yading@10: 
yading@10: .S160:  DISALGNEXCPT                       || R1 = [I0++]   || R3 = [I1++];
yading@10:         (R7,R6) = BYTEOP16M (R1:0,R3:2)    || R0 = [I0++]   || R2 = [I1++];
yading@10:         a0 += r7.l * r7.l, a1 += r7.h * r7.h (is);
yading@10:         a0 += r6.l * r6.l, a1 += r6.h * r6.h (is);
yading@10:         (R7,R6) = BYTEOP16M (R1:0,R3:2) (R)|| R1 = [I0++]   || R3 = [I1++];
yading@10:         a0 += r7.l * r7.l, a1 += r7.h * r7.h (is);
yading@10:         a0 += r6.l * r6.l, a1 += r6.h * r6.h (is);
yading@10:         (R7,R6) = BYTEOP16M (R1:0,R3:2)    || R0 = [I0++M0] || R2 = [I1++M0];
yading@10:         a0 += r7.l * r7.l, a1 += r7.h * r7.h (is);
yading@10:         a0 += r6.l * r6.l, a1 += r6.h * r6.h (is);
yading@10:         (R7,R6) = BYTEOP16M (R1:0,R3:2) (R)|| R0 = [I0++]   || R2 = [I1++];
yading@10:         a0 += r7.l * r7.l, a1 += r7.h * r7.h (is);
yading@10: .E160:  a0 += r6.l * r6.l, a1 += r6.h * r6.h (is);
yading@10:         a0 += a1;
yading@10:         r0 = a0;
yading@10: 
yading@10:         (r7:6) = [sp++];
yading@10:         unlink;
yading@10:         rts;
yading@10: DEFUN_END(sse16)