Index: apps/plugins/mpegplayer/decode.c =================================================================== RCS file: /cvsroot/rockbox/apps/plugins/mpegplayer/decode.c,v retrieving revision 1.1 diff -u -r1.1 decode.c --- apps/plugins/mpegplayer/decode.c 7 Aug 2006 22:11:07 -0000 1.1 +++ apps/plugins/mpegplayer/decode.c 25 Sep 2006 22:15:23 -0000 @@ -416,14 +416,21 @@ } +#if defined(CPU_COLDFIRE) && !defined(SIMULATOR) +static mpeg2dec_t static_mpeg2dec IBSS_ATTR; +#endif + mpeg2dec_t * mpeg2_init (void) { mpeg2dec_t * mpeg2dec; mpeg2_accel (MPEG2_ACCEL_DETECT); - mpeg2dec = (mpeg2dec_t *) mpeg2_malloc (sizeof (mpeg2dec_t), - MPEG2_ALLOC_MPEG2DEC); +#if defined(CPU_COLDFIRE) && !defined(SIMULATOR) + mpeg2dec = &static_mpeg2dec; +#else + mpeg2dec = (mpeg2dec_t *) mpeg2_malloc (sizeof (mpeg2dec_t),MPEG2_ALLOC_MPEG2DEC); +#endif if (mpeg2dec == NULL) return NULL; Index: apps/plugins/mpegplayer/idct.c =================================================================== RCS file: /cvsroot/rockbox/apps/plugins/mpegplayer/idct.c,v retrieving revision 1.2 diff -u -r1.2 idct.c --- apps/plugins/mpegplayer/idct.c 8 Aug 2006 22:56:35 -0000 1.2 +++ apps/plugins/mpegplayer/idct.c 25 Sep 2006 22:15:25 -0000 @@ -28,13 +28,38 @@ #include "mpeg2.h" #include "attributes.h" #include "mpeg2_internal.h" - +#define W0 2048 /* 2048 * sqrt (2) * cos (4 * pi / 16) */ #define W1 2841 /* 2048 * sqrt (2) * cos (1 * pi / 16) */ #define W2 2676 /* 2048 * sqrt (2) * cos (2 * pi / 16) */ #define W3 2408 /* 2048 * sqrt (2) * cos (3 * pi / 16) */ +#define W4 2048 /* 2048 * sqrt (2) * cos (4 * pi / 16) */ #define W5 1609 /* 2048 * sqrt (2) * cos (5 * pi / 16) */ #define W6 1108 /* 2048 * sqrt (2) * cos (6 * pi / 16) */ #define W7 565 /* 2048 * sqrt (2) * cos (7 * pi / 16) */ + +#define SC(x) #x + +#define WA 1448 //=cos(pi/4) 2048*cos(4pi/16) 0,707106781 +#define WB 2009 //=cos(pi/16) 2048*cos(pi/16) 0,98078528 +#define WC 1892 //=cos(pi/8) 2048*cos(2pi/16) 0,923879533 +#define WD 1703 //=cos(3pi/16) 2048*cos(3pi/16) 0,831469612 +#define WE 1448 //=sin(pi/4) 2048*sin(4pi/16) 0,707106781 +#define WF 1138 //=sin(3pi/16) 2048*sin(3pi/16) 0,555570233 +#define WG 784 //=sin(pi/8) 2048*sin(2pi/16) 0,382683432 +#define WH 400 //=sin(pi/16) 2048*sin(pi/16) 0,195090322 + +/* theroic matrix used in the asm 1d_idct */ +/* +{WA, WC, WE, WG, + WA, WG, -WE, -WC, + WA, -WG, -WE, WC, + WA, -WC, WE, -WG}; + +{WB, WD, WF, WH, + WD, -WH, -WB, -WF, + WF, -WB, WH, WD, + WH, -WF, WD, -WB}; +*/ /* idct main entry point */ void (* mpeg2_idct_copy) (int16_t * block, uint8_t * dest, int stride); @@ -51,13 +76,14 @@ static inline unsigned CLIP(int value) { asm ( /* Note: Uses knowledge that only the low byte of the result is used */ - "cmp.l #255,%[v] \n" /* overflow? */ + "cmp.l #255,%[v] \n" /* overflow? */ "bls.b 1f \n" /* no: return value */ "spl.b %[v] \n" /* yes: set low byte to appropriate boundary */ - "1: \n" + "1 : \n" : /* outputs */ [v]"+d"(value) ); + return value; } #elif defined CPU_ARM @@ -76,6 +102,21 @@ #define CLIP(i) ((mpeg2_clip + 3840)[i]) #endif +#define MATRIX_MUL(m00,m01,m02,m03,\ + m10,m11,m12,m13, \ + m20,m21,m22,m23, \ + m30,m31,m32,m33, \ + n0,n1,n2,n3,\ + p0,p1,p2,p3) \ +do{ \ + p0=n0*m00+n1*m01+n2*m02+n3*m03; \ + p1=n0*m10+n1*m11+n2*m12+n3*m13; \ + p2=n0*m20+n1*m21+n2*m22+n3*m23; \ + p3=n0*m30+n1*m31+n2*m32+n3*m33; \ +} while(0) + + + #if 0 #define BUTTERFLY(t0,t1,W0,W1,d0,d1) \ do { \ @@ -91,6 +132,47 @@ } while (0) #endif + +#ifdef C_VERSION_OF_ASM +static inline void idct_row (int16_t * const block) +{ + int f0,f1,f2,f3,f4,f5,f6,f7; + int a0,a1,a2,a3; + int b0,b1,b2,b3; + + f0 = block[0]+1; + f2 = block[1]; + f4 = block[2]; + f6 = block[3]; + f1 = block[4]; + f3 = block[5]; + f5 = block[6]; + f7 = block[7]; + + MATRIX_MUL(W1, W3, W5, W7, + W3, -W7, -W1, -W5, + W5, -W1, W7, W3, + W7, -W5, W3, -W1, + f1,f3,f5,f7, + b0,b1,b2,b3); + + MATRIX_MUL(W0, W2, W4, W6, + W0, W6, -W4, -W2, + W0, -W6, -W4, W2, + W0, -W2, W4, -W6, + f0,f2,f4,f6, + a0,a1,a2,a3); + + block[0]= (a0+b0) >> 12; + block[1]= (a1+b1) >> 12; + block[2]= (a2+b2) >> 12; + block[3]= (a3+b3) >> 12; + block[4]= (a3-b3) >> 12; + block[5]= (a2-b2) >> 12; + block[6]= (a1-b1) >> 12; + block[7]= (a0-b0) >> 12; +} +#else static inline void idct_row (int16_t * const block) { int d0, d1, d2, d3; @@ -143,7 +225,49 @@ block[6] = (a1 - b1) >> 12; block[7] = (a0 - b0) >> 12; } +#endif + +#ifdef C_VERSION_OF_ASM +static inline void idct_col (int16_t * const block) +{ + int f0,f1,f2,f3,f4,f5,f6,f7; + int a0,a1,a2,a3; + int b0,b1,b2,b3; + + f0 = block[0*8] + 32; + f2 = block[1*8]; + f4 = block[2*8]; + f6 = block[3*8]; + f1 = block[4*8]; + f3 = block[5*8]; + f5 = block[6*8]; + f7 = block[7*8]; + + MATRIX_MUL(W1, W3, W5, W7, + W3, -W7, -W1, -W5, + W5, -W1, W7, W3, + W7, -W5, W3, -W1, + f1,f3,f5,f7, + b0,b1,b2,b3); + + MATRIX_MUL(W0, W2, W4, W6, + W0, W6, -W4, -W2, + W0, -W6, -W4, W2, + W0, -W2, W4, -W6, + f0,f2,f4,f6, + a0,a1,a2,a3); + + block[0*8]= (a0+b0) >> 17; + block[1*8]= (a1+b1) >> 17; + block[2*8]= (a2+b2) >> 17; + block[3*8]= (a3+b3) >> 17; + block[4*8]= (a3-b3) >> 17; + block[5*8]= (a2-b2) >> 17; + block[6*8]= (a1-b1) >> 17; + block[7*8]= (a0-b0) >> 17; +} +#else static inline void idct_col (int16_t * const block) { int d0, d1, d2, d3; @@ -184,13 +308,962 @@ block[8*6] = (a1 - b1) >> 17; block[8*7] = (a0 - b0) >> 17; } +#endif + +#if defined(CPU_COLDFIRE) && !defined(SIMULATOR) +static void idct(int16_t * const block) ICODE_ATTR; +static void idct(int16_t * const block) +{ + asm volatile( + + "move.l #0x80,%%macsr \n\t" + /* matrix start */ + + "move.l #2048*65536+2841,%%a0 \n\t" + "move.l #2676*65536+2408,%%a1 \n\t" + "move.l #2048*65536+1609,%%a2 \n\t" + "move.l #1108*65536+565,%%a3 \n\t" + + "move.l %[block],%%a4 \n\t" + "add.l #128,%%a4 \n\t" + /* init loop for i=0 to 7 */ + "move.l #8,%%d3 \n\t" + + "row: \n\t" + /* load P[0] P[2] P[4] P[6] P[1] P[3] P[5] P[7] from the row */ + "movem.l (%[block]),%%d4-%%d7 \n\t" + + "clr.l %%d2 \n\t" + "move.w %%d4,%%d2 \n\t" + "or.l %%d5,%%d2 \n\t" + "or.l %%d6,%%d2 \n\t" + "or.l %%d7,%%d2 \n\t" + "bne continue \n\t" + /* uint32_t tmp = (uint16_t) (block[0] >> 1); tmp |= tmp << 16; */ + "move.l #17,%%d5 \n\t" + "asr.l %%d5,%%d4 \n\t" + + /* ((int32_t *)block)[0] = tmp; ((int32_t *)block)[1] = tmp; + * ((int32_t *)block)[2] = tmp; ((int32_t *)block)[3] = tmp; */ + "move.w %%d4,(%%a4) \n\t" + "move.w %%d4,(1*16,%%a4) \n\t" + "move.w %%d4,(2*16,%%a4) \n\t" + "move.w %%d4,(3*16,%%a4) \n\t" + "move.w %%d4,(4*16,%%a4) \n\t" + "move.w %%d4,(5*16,%%a4) \n\t" + "move.w %%d4,(6*16,%%a4) \n\t" + "move.w %%d4,(7*16,%%a4) \n\t" + + "jmp shortcut \n\t" + "continue: \n\t" + /* compute second matrix * P[1],P[3],P[5],P[7]*/ + "mac.w %%d6u, %%a0l, %%acc0 \n\t" + "mac.w %%d6u, %%a1l, %%acc1 \n\t" + "mac.w %%d6u, %%a2l, %%acc2 \n\t" + "mac.w %%d6u, %%a3l, %%acc3 \n\t" + + "mac.w %%d6l, %%a1l, %%acc0 \n\t" + "msac.w %%d6l, %%a3l, %%acc1 \n\t" + "msac.w %%d6l, %%a0l, %%acc2 \n\t" + "msac.w %%d6l, %%a2l, %%acc3 \n\t" + + "mac.w %%d7u, %%a2l, %%acc0 \n\t" + "msac.w %%d7u, %%a0l, %%acc1 \n\t" + "mac.w %%d7u, %%a3l, %%acc2 \n\t" + "mac.w %%d7u, %%a1l, %%acc3 \n\t" + + "mac.w %%d7l, %%a3l, %%acc0 \n\t" + "msac.w %%d7l, %%a2l, %%acc1 \n\t" + "mac.w %%d7l, %%a1l, %%acc2 \n\t" + "msac.w %%d7l, %%a0l, %%acc3 \n\t" + + /* save second matrix row */ + "movclr.l %%acc0, %%d2 \n\t" + "movclr.l %%acc1, %%d6 \n\t" + "movclr.l %%acc2, %%d7 \n\t" + "movclr.l %%acc3, %%a5 \n\t" + + /* probably some error correction */ + "add.l #65536*1,%%d4 \n\t" + /* compute first matrix * P[0],P[2],P[4],P[6]*/ + "mac.w %%d4u, %%a0u, %%acc0 \n\t" + "mac.w %%d4u, %%a0u, %%acc1 \n\t" + "mac.w %%d4u, %%a0u, %%acc2 \n\t" + "mac.w %%d4u, %%a0u, %%acc3 \n\t" + + "mac.w %%d4l, %%a1u, %%acc0 \n\t" + "mac.w %%d4l, %%a3u, %%acc1 \n\t" + "msac.w %%d4l, %%a3u, %%acc2 \n\t" + "msac.w %%d4l, %%a1u, %%acc3 \n\t" + + "mac.w %%d5u, %%a2u, %%acc0 \n\t" + "msac.w %%d5u, %%a2u, %%acc1 \n\t" + "msac.w %%d5u, %%a2u, %%acc2 \n\t" + "mac.w %%d5u, %%a2u, %%acc3 \n\t" + + "mac.w %%d5l, %%a3u, %%acc0 \n\t" + "msac.w %%d5l, %%a1u, %%acc1 \n\t" + "mac.w %%d5l, %%a1u, %%acc2 \n\t" + "msac.w %%d5l, %%a3u, %%acc3 \n\t" + + /* divide by */ + "move.l #12, %%d4 \n\t" + + /* mat 1 + mat 2 */ + "move.l %%acc0, %%d5 \n\t" + "add.l %%d2,%%d5 \n\t" + "asr.l %%d4, %%d5 \n\t" + "move.w %%d5,(%%a4) \n\t" + + "move.l %%acc1, %%d5 \n\t" + "add.l %%d6,%%d5 \n\t" + "asr.l %%d4, %%d5 \n\t" + "move.w %%d5,(1*16,%%a4) \n\t" + + "move.l %%acc2, %%d5 \n\t" + "add.l %%d7,%%d5 \n\t" + "asr.l %%d4, %%d5 \n\t" + "move.w %%d5,(2*16,%%a4) \n\t" + + "move.l %%acc3, %%d5 \n\t" + "add.l %%a5,%%d5 \n\t" + "asr.l %%d4, %%d5 \n\t" + "move.w %%d5,(3*16,%%a4) \n\t" + + /* mat 1 - mat2*/ + "movclr.l %%acc0, %%d5 \n\t" + "sub.l %%d2,%%d5 \n\t" + "asr.l %%d4, %%d5 \n\t" + "move.w %%d5,(7*16,%%a4) \n\t" + + "movclr.l %%acc1, %%d5 \n\t" + "sub.l %%d6,%%d5 \n\t" + "asr.l %%d4, %%d5 \n\t" + "move.w %%d5,(6*16,%%a4) \n\t" + + "movclr.l %%acc2, %%d5 \n\t" + "sub.l %%d7,%%d5 \n\t" + "asr.l %%d4, %%d5 \n\t" + "move.w %%d5,(5*16,%%a4) \n\t" + + "movclr.l %%acc3, %%d5 \n\t" + "sub.l %%a5,%%d5 \n\t" + "asr.l %%d4, %%d5 \n\t" + "move.w %%d5,(4*16,%%a4) \n\t" + + "shortcut: \n\t" + /* next line, address +16 */ + "add.l #16,%[block] \n\t" + /* next dest column */ + "add.l #2,%%a4 \n\t" + "sub.l #1,%%d3 \n\t" + "bne row \n\t" + /* for end */ + + "sub.l #16+128,%%a4 \n\t" /* return to intial block value */ + "move.l #8,%%d3 \n\t" +// columns + "col: \n\t" + + /* load P[0] P[2] P[4] P[6] P[1] P[3] P[5] P[7] from the row */ + "movem.l (%[block]),%%d4-%%d7 \n\t" + + /* compute second matrix * P[1],P[3],P[5],P[7]*/ + "mac.w %%d6u, %%a0l, %%acc0 \n\t" + "mac.w %%d6u, %%a1l, %%acc1 \n\t" + "mac.w %%d6u, %%a2l, %%acc2 \n\t" + "mac.w %%d6u, %%a3l, %%acc3 \n\t" + + "mac.w %%d6l, %%a1l, %%acc0 \n\t" + "msac.w %%d6l, %%a3l, %%acc1 \n\t" + "msac.w %%d6l, %%a0l, %%acc2 \n\t" + "msac.w %%d6l, %%a2l, %%acc3 \n\t" + + "mac.w %%d7u, %%a2l, %%acc0 \n\t" + "msac.w %%d7u, %%a0l, %%acc1 \n\t" + "mac.w %%d7u, %%a3l, %%acc2 \n\t" + "mac.w %%d7u, %%a1l, %%acc3 \n\t" + + "mac.w %%d7l, %%a3l, %%acc0 \n\t" + "msac.w %%d7l, %%a2l, %%acc1 \n\t" + "mac.w %%d7l, %%a1l, %%acc2 \n\t" + "msac.w %%d7l, %%a0l, %%acc3 \n\t" + + /* save second matrix row */ + "movclr.l %%acc0, %%d2 \n\t" + "movclr.l %%acc1, %%d6 \n\t" + "movclr.l %%acc2, %%d7 \n\t" + "movclr.l %%acc3, %%a5 \n\t" + + /* load P[0] P[2] P[4] P[6] from the row */ + /* probably some error correction */ + "add.l #32*65536,%%d4 \n\t" + /* compute first matrix * P[0],P[2],P[4],P[6]*/ + "mac.w %%d4u, %%a0u, %%acc0 \n\t" + "mac.w %%d4u, %%a0u, %%acc1 \n\t" + "mac.w %%d4u, %%a0u, %%acc2 \n\t" + "mac.w %%d4u, %%a0u, %%acc3 \n\t" + + "mac.w %%d4l, %%a1u, %%acc0 \n\t" + "mac.w %%d4l, %%a3u, %%acc1 \n\t" + "msac.w %%d4l, %%a3u, %%acc2 \n\t" + "msac.w %%d4l, %%a1u, %%acc3 \n\t" + + "mac.w %%d5u, %%a2u, %%acc0 \n\t" + "msac.w %%d5u, %%a2u, %%acc1 \n\t" + "msac.w %%d5u, %%a2u, %%acc2 \n\t" + "mac.w %%d5u, %%a2u, %%acc3 \n\t" + + "mac.w %%d5l, %%a3u, %%acc0 \n\t" + "msac.w %%d5l, %%a1u, %%acc1 \n\t" + "mac.w %%d5l, %%a1u, %%acc2 \n\t" + "msac.w %%d5l, %%a3u, %%acc3 \n\t" + + /* divide by */ + "move.l #17, %%d4 \n\t" + + /* mat 1 + mat 2 */ + "move.l %%acc0, %%d5 \n\t" + "add.l %%d2,%%d5 \n\t" + "asr.l %%d4, %%d5 \n\t" + "move.w %%d5,(%%a4) \n\t" + + "move.l %%acc1, %%d5 \n\t" + "add.l %%d6,%%d5 \n\t" + "asr.l %%d4, %%d5 \n\t" + "move.w %%d5,(1*16,%%a4) \n\t" + + "move.l %%acc2, %%d5 \n\t" + "add.l %%d7,%%d5 \n\t" + "asr.l %%d4, %%d5 \n\t" + "move.w %%d5,(2*16,%%a4) \n\t" + + "move.l %%acc3, %%d5 \n\t" + "add.l %%a5,%%d5 \n\t" + "asr.l %%d4, %%d5 \n\t" + "move.w %%d5,(3*16,%%a4) \n\t" + + /* mat 1 - mat2*/ + "movclr.l %%acc0, %%d5 \n\t" + "sub.l %%d2,%%d5 \n\t" + "asr.l %%d4, %%d5 \n\t" + "move.w %%d5,(7*16,%%a4) \n\t" + + "movclr.l %%acc1, %%d5 \n\t" + "sub.l %%d6,%%d5 \n\t" + "asr.l %%d4, %%d5 \n\t" + "move.w %%d5,(6*16,%%a4) \n\t" + + "movclr.l %%acc2, %%d5 \n\t" + "sub.l %%d7,%%d5 \n\t" + "asr.l %%d4, %%d5 \n\t" + "move.w %%d5,(5*16,%%a4) \n\t" + + "movclr.l %%acc3, %%d5 \n\t" + "sub.l %%a5,%%d5 \n\t" + "asr.l %%d4, %%d5 \n\t" + "move.w %%d5,(4*16,%%a4) \n\t" + /****************************/ + + /* next line, address +16 */ + "add.l #16,%[block] \n\t" + /* next dest column */ + "add.l #2,%%a4 \n\t" + + "sub.l #1,%%d3 \n\t" + "bne col \n\t" + + "sub.l #16*16,%[block] \n\t" + : + : [block]"a"(block) + : "%a0","%a1","%a2","%a3","%a4","%a5","%d2","%d3","%d4","%d5","%d6","%d7" + ); + +} +#endif + +#if defined(CPU_COLDFIRE) && !defined(SIMULATOR) +static void clip_block_to_dest(int16_t * block, uint8_t * dest,const int stride) +{ + asm volatile( + + "move.l %[dest],%%a0 \n\t" + /* init loop for i=0 to 7 */ + "move.l #8,%%d1 \n\t" + "clr.l %%d0 \n\t" + "clr.l %%d3 \n\t" + "foradd: \n\t" + + "movem.l (%[block]),%%d4-%%d7 \n\t" + + /* dot 2*/ + "move.w %%d4,%%d3 \n\t" + "cmp.l #255,%%d3 \n\t" + "bls.b 2f \n\t" + "spl.b %%d3 \n\t" + "2 : \n\t" + + /* dot 1*/ + "swap.w %%d4 \n\t" + "move.w %%d4,%%d0 \n\t" + "cmp.l #255,%%d0 \n\t" /* overflow? */ + "bls.b 1f \n\t" /* no: return value */ + "spl.b %%d0 \n\t" /* yes: set low byte to appropriate boundary */ + "1 : \n\t" + + /* save dot 1*/ + "move.b %%d0,%%d2 \n\t" + /* save dot 2*/ + "lsl.l #8,%%d2 \n\t" + "move.b %%d3,%%d2 \n\t" + + /* dot 4*/ + "move.w %%d5,%%d3 \n\t" + "cmp.l #255,%%d3 \n\t" + "bls.b 4f \n\t" + "spl.b %%d3 \n\t" + "4 : \n\t" + + /* dot 3*/ + "swap.w %%d5 \n\t" + "move.w %%d5,%%d0 \n\t" + "cmp.l #255,%%d0 \n\t" /* overflow? */ + "bls.b 3f \n\t" /* no: return value */ + "spl.b %%d0 \n\t" /* yes: set low byte to appropriate boundary */ + "3 : \n\t" + + /* save dot 3*/ + "lsl.l #8,%%d2 \n\t" + "move.b %%d0,%%d2 \n\t" + /* save dot 4*/ + "lsl.l #8,%%d2 \n\t" + "move.b %%d3,%%d2 \n\t" + + + /* writes 4 bytes */ + "move.l %%d2,(%%a0) \n\t" + + /* dot 6 */ + "move.w %%d6,%%d3 \n\t" + "cmp.l #255,%%d3 \n\t" + "bls.b 6f \n\t" + "spl.b %%d3 \n\t" + "6 : \n\t" + + "move.l #0,%%d4 \n\t" /* */ + /* dot 5 */ + "move.l %%d4,(%[block])+ \n\t" /* */ + "swap.w %%d6 \n\t" + "move.w %%d6,%%d0 \n\t" + "move.l %%d4,(%[block])+ \n\t" /* */ + "cmp.l #255,%%d0 \n\t" + "bls.b 5f \n\t" + "spl.b %%d0 \n\t" + "5 : \n\t" + + /* save dot 5*/ + "move.b %%d0,%%d2 \n\t" + /* save dot 6*/ + "lsl.l #8,%%d2 \n\t" + "move.b %%d3,%%d2 \n\t" + + + /* dot 8*/ + "move.w %%d7,%%d3 \n\t" + "cmp.l #255,%%d3 \n\t" + "bls.b 8f \n\t" + "spl.b %%d3 \n\t" + "8 : \n\t" + + "move.l %%d4,(%[block])+ \n\t" /* */ + /* dot 7*/ + "swap.w %%d7 \n\t" + "move.w %%d7,%%d0 \n\t" + "cmp.l #255,%%d0 \n\t" + "bls.b 7f \n\t" + "spl.b %%d0 \n\t" + "7 : \n\t" + + "move.l %%d4,(%[block])+ \n\t" /* */ + /* save dot 7*/ + "lsl.l #8,%%d2 \n\t" + "move.b %%d0,%%d2 \n\t" + /* save dot 8*/ + "lsl.l #8,%%d2 \n\t" + "move.b %%d3,%%d2 \n\t" + + + /* write to dest */ + "move.l %%d2,(4,%%a0) \n\t" + + /* clear row, address +16 */ + + /* next line, address +stride */ + "add.l %[stride],%%a0 \n\t" + "sub.l #1,%%d1 \n\t" + "bne foradd \n\t" + /* for end */ + "sub.l #128,%[block] \n\t" + + : + : [block]"a"(block),[dest]"a"(dest),[stride]"a"(stride) + : "%a0","%d0","%d1","%d2","%d3","%d4","%d5","%d6","%d7" + ); + +} +#endif + + + +#ifdef CPU_COLDFIRE__ +static void clip_block_to_dest_add(int16_t * block, uint8_t * dest,const int stride) +{ + asm volatile ( + + "move.l %[dest],%%a1 \n\t" + /* init loop for i=0 to 7 */ + "move.l #8,%%d1 \n\t" + "clr.l %%d2 \n\t" + "foradd2: \n\t" + + "move.w (%[block]),%%d3 \n\t" + "move.b (%%a1),%%d2 \n\t" + "ext.l %%d3 \n\t" + "add.l %%d2,%%d3 \n\t" + "cmp.l #255,%%d3 \n\t" /* overflow? */ + "bls.b 1f \n\t" /* no: return value */ + "spl.b %%d3 \n\t" /* yes: set low byte to appropriate boundary */ + "1 : \n\t" + + + "move.w (2,%[block]),%%d0 \n\t" + "move.b (1,%%a1),%%d2 \n\t" + "ext.l %%d0 \n\t" + "add.l %%d2,%%d0 \n\t" + "cmp.l #255,%%d0 \n\t" + "bls.b 2f \n\t" + "spl.b %%d0 \n\t" + "2 : \n\t" + "lsl.l #8,%%d3 \n\t" + "move.b %%d0,%%d3 \n\t" + + + "move.w (4,%[block]),%%d0 \n\t" + "move.b (2,%%a1),%%d2 \n\t" + "ext.l %%d0 \n\t" + "add.l %%d2,%%d0 \n\t" + "cmp.l #255,%%d0 \n\t" + "bls.b 3f \n\t" + "spl.b %%d0 \n\t" + "3 : \n\t" + "lsl.l #8,%%d3 \n\t" + "move.b %%d0,%%d3 \n\t" + + + "move.w (6,%[block]),%%d0 \n\t" + "move.b (3,%%a1),%%d2 \n\t" + "ext.l %%d0 \n\t" + "add.l %%d2,%%d0 \n\t" + "cmp.l #255,%%d0 \n\t" + "bls.b 4f \n\t" + "spl.b %%d0 \n\t" + "4 : \n\t" + "lsl.l #8,%%d3 \n\t" + "move.b %%d0,%%d3 \n\t" + + /* writes 4 bytes */ + "move.l %%d3,(%%a1) \n\t" + + + "move.w (8,%[block]),%%d3 \n\t" + "move.b (4,%%a1),%%d2 \n\t" + "ext.l %%d3 \n\t" + "add.l %%d2,%%d3 \n\t" + "cmp.l #255,%%d3 \n\t" + "bls.b 5f \n\t" + "spl.b %%d3 \n\t" + "5 : \n\t" + + "move.w (10,%[block]),%%d0 \n\t" + "move.b (5,%%a1),%%d2 \n\t" + "ext.l %%d0 \n\t" + "add.l %%d2,%%d0 \n\t" + "cmp.l #255,%%d0 \n\t" + "bls.b 6f \n\t" + "spl.b %%d0 \n\t" + "6 : \n\t" + "lsl.l #8,%%d3 \n\t" + "move.b %%d0,%%d3 \n\t" + + "move.w (12,%[block]),%%d0 \n\t" + "move.b (6,%%a1),%%d2 \n\t" + "ext.l %%d0 \n\t" + "add.l %%d2,%%d0 \n\t" + "cmp.l #255,%%d0 \n\t" + "bls.b 7f \n\t" + "spl.b %%d0 \n\t" + "7 : \n\t" + "lsl.l #8,%%d3 \n\t" + "move.b %%d0,%%d3 \n\t" + + "move.w (14,%[block]),%%d0 \n\t" + "move.b (7,%%a1),%%d2 \n\t" + "ext.l %%d0 \n\t" + "add.l %%d2,%%d0 \n\t" + "cmp.l #255,%%d0 \n\t" + "bls.b 8f \n\t" + "spl.b %%d0 \n\t" + "8 : \n\t" + "lsl.l #8,%%d3 \n\t" + "move.b %%d0,%%d3 \n\t" + + /* write to dest */ + "move.l %%d3,(4,%%a1) \n\t" + + /* clear row, address +16 */ + "move.l #0,%%d0 \n\t" + "move.l %%d0,(%[block])+ \n\t" + "move.l %%d0,(%[block])+ \n\t" + "move.l %%d0,(%[block])+ \n\t" + "move.l %%d0,(%[block])+ \n\t" + + /* next line, address +stride */ + "add.l %[stride],%%a1 \n\t" + "sub.l #1,%%d1 \n\t" + "bne foradd2 \n\t" + /* for end */ + "sub.l #128,%[block] \n\t" + : + : [block]"a"(block),[dest]"a"(dest),[stride]"a"(stride) + : "%a1","%d0","%d1","%d2","%d3" + ); +} +#endif + +#if defined(CPU_COLDFIRE) && !defined(SIMULATOR) +static void clip_block_to_dest_add(int16_t * block, uint8_t * dest,const int stride) +{ + asm volatile ( + + "move.l %[dest],%%a0 \n\t" + /* init loop for i=0 to 7 */ + "move.l #8,%%d1 \n\t" + "clr.l %%d2 \n\t" + "foradd2: \n\t" + + "movem.l (%%a0),%%d4-%%d5 \n\t" + "movem.l (%[block]),%%d6-%%d7/%%a1-%%a2 \n\t" + + /* dot 4 */ + "move.w %%d7,%%d0 \n\t" + "ext.l %%d0 \n\t" + "move.b %%d4,%%d2 \n\t" + "add.l %%d2,%%d0 \n\t" + "cmp.l #255,%%d0 \n\t" + "bls.b 4f \n\t" + "spl.b %%d0 \n\t" + "4 : \n\t" + "move.b %%d0,%%d3 \n\t" + + /* dot 3 */ + "swap.w %%d7 \n\t" + "ext.l %%d7 \n\t" + "lsr.l #8,%%d4 \n\t" + "move.b %%d4,%%d2 \n\t" + "add.l %%d2,%%d7 \n\t" + "cmp.l #255,%%d7 \n\t" + "bls.b 3f \n\t" + "spl.b %%d7 \n\t" + "3 : \n\t" + "lsl.l #8,%%d3 \n\t" + "move.b %%d7,%%d3 \n\t" + + /* dot 2 */ + "move.w %%d6,%%d0 \n\t" + "ext.l %%d0 \n\t" + "lsr.l #8,%%d4 \n\t" + "move.b %%d4,%%d2 \n\t" + "add.l %%d2,%%d0 \n\t" + "cmp.l #255,%%d0 \n\t" + "bls.b 2f \n\t" + "spl.b %%d0 \n\t" + "2 : \n\t" + "lsl.l #8,%%d3 \n\t" + "move.b %%d0,%%d3 \n\t" + + /* dot 1 */ + "swap.w %%d6 \n\t" + "ext.l %%d6 \n\t" + "lsr.l #8,%%d4 \n\t" + "move.b %%d4,%%d2 \n\t" + "add.l %%d2,%%d6 \n\t" + "cmp.l #255,%%d6 \n\t" + "bls.b 1f \n\t" + "spl.b %%d6 \n\t" + "1 : \n\t" + "lsl.l #8,%%d3 \n\t" + "move.b %%d6,%%d3 \n\t" + + "move.l #0x00FF00FF,%%d7 \n\t" /* val = ABCD */ + "and.l %%d3,%%d7 \n\t" /* mask = .B.D */ + "eor.l %%d7,%%d3 \n\t" /* val = A.C. */ + "lsl.l #8,%%d7 \n\t" /* mask = B.D. */ + "lsr.l #8,%%d3 \n\t" /* val = .A.C */ + "or.l %%d7,%%d3 \n\t" /* val = BADC */ + "swap %%d3 \n\t" /* val = DCBA */ + + /* writes 4 bytes */ + "move.l %%d3,(%%a0) \n\t" + + /* dot 8 */ + "move.l %%a2,%%d0 \n\t" + "ext.l %%d0 \n\t" + "move.b %%d5,%%d2 \n\t" + "add.l %%d2,%%d0 \n\t" + "cmp.l #255,%%d0 \n\t" + "bls.b 8f \n\t" + "spl.b %%d0 \n\t" + "8 : \n\t" + "move.b %%d0,%%d3 \n\t" + + /* dot 7 */ + "move.l %%a2,%%d0 \n\t" + "swap %%d0 \n\t" + "ext.l %%d0 \n\t" + "lsr.l #8,%%d5 \n\t" + "move.b %%d5,%%d2 \n\t" + "add.l %%d2,%%d0 \n\t" + "cmp.l #255,%%d0 \n\t" + "bls.b 7f \n\t" + "spl.b %%d0 \n\t" + "7 : \n\t" + "lsl.l #8,%%d3 \n\t" + "move.b %%d0,%%d3 \n\t" + + /* dot 6 */ + "move.l %%a1,%%d0 \n\t" + "ext.l %%d0 \n\t" + "lsr.l #8,%%d5 \n\t" + "move.b %%d5,%%d2 \n\t" + "add.l %%d2,%%d0 \n\t" + "cmp.l #255,%%d0 \n\t" + "bls.b 6f \n\t" + "spl.b %%d0 \n\t" + "6 : \n\t" + "lsl.l #8,%%d3 \n\t" + "move.b %%d0,%%d3 \n\t" + + /* dot 5 */ + "move.l %%a1,%%d0 \n\t" + "swap %%d0 \n\t" + "ext.l %%d0 \n\t" + "lsr.l #8,%%d5 \n\t" + "move.b %%d5,%%d2 \n\t" + "add.l %%d2,%%d0 \n\t" + "cmp.l #255,%%d0 \n\t" + "bls.b 5f \n\t" + "spl.b %%d0 \n\t" + "5 : \n\t" + "lsl.l #8,%%d3 \n\t" + "move.b %%d0,%%d3 \n\t" + + "move.l #0x00FF00FF,%%d7 \n\t" /* val = ABCD */ + + /* clear row, address +16 */ /* (pipelined below)*/ + "move.l #0,%%d0 \n\t" + "move.l %%d0,(%[block])+ \n\t" /* */ + "and.l %%d3,%%d7 \n\t" /* mask = .B.D */ + "move.l %%d0,(%[block])+ \n\t" /* */ + "eor.l %%d7,%%d3 \n\t" /* val = A.C. */ + "move.l %%d0,(%[block])+ \n\t" /* */ + "lsl.l #8,%%d7 \n\t" /* mask = B.D. */ + "move.l %%d0,(%[block])+ \n\t" /* */ + "lsr.l #8,%%d3 \n\t" /* val = .A.C */ + + "or.l %%d7,%%d3 \n\t" /* val = BADC */ + "swap %%d3 \n\t" /* val = DCBA */ + + + /* writes 4 bytes */ + "move.l %%d3,(4,%%a0) \n\t" + + + + /* next line, address +stride */ + "add.l %[stride],%%a0 \n\t" + "sub.l #1,%%d1 \n\t" + "bne foradd2 \n\t" + /* for end */ + "sub.l #128,%[block] \n\t" + : + : [block]"a"(block),[dest]"a"(dest),[stride]"a"(stride) + : "%a0","%a1","%a2","%d0","%d1","%d2","%d3","%d4","%d5","%d6","%d7" + ); + +} +#endif + +#if defined(CPU_COLDFIRE) && !defined(SIMULATOR) +static void clip_block_to_dest_add_DC(int16_t * block, uint8_t * dest,const int stride) +{ + asm volatile ( + + "move.l %[dest],%%a1 \n\t" + + /* DC = (block[0] + 64) >> 7; */ + "move.w (%[block]),%%d0 \n\t" + "ext.l %%d0 \n\t" + "add.l #64,%%d0 \n\t" + "asr.l #7,%%d0 \n\t" + + /*block[0] = block[63] = 0;*/ + "move.w #0,%%d2 \n\t" + "move.w %%d2,(%[block]) \n\t" + /* init loop for i=0 to 7 */ + "move.l #8,%%d1 \n\t" /* pipeline on store */ + "move.w %%d2,(63*2,%[block]) \n\t" + + + "clr.l %%d2 \n\t" + "foraddDC: \n\t" + "movem.l (%%a1),%%d4-%%d5 \n\t" + + /* read dot 3, add DC, save dot */ + "move.b %%d4,%%d2 \n\t" + "add.l %%d0,%%d2 \n\t" + "cmp.l #255,%%d2 \n\t" + "bls.b 4f \n\t" + "spl.b %%d2 \n\t" + "4 : \n\t" + "move.b %%d2,%%d3 \n\t" + + /* read dot 2, add DC, save dot */ + "lsr.l #8,%%d4 \n\t" + "move.b %%d4,%%d2 \n\t" + "add.l %%d0,%%d2 \n\t" + "cmp.l #255,%%d2 \n\t" + "bls.b 3f \n\t" + "spl.b %%d2 \n\t" + "3 : \n\t" + "move.b %%d2,%%d6 \n\t" + + + /* read dot 1, add DC, save dot */ + "lsr.l #8,%%d4 \n\t" + "move.b %%d4,%%d2 \n\t" + "add.l %%d0,%%d2 \n\t" + "cmp.l #255,%%d2 \n\t" + "bls.b 2f \n\t" + "spl.b %%d2 \n\t" + "2 : \n\t" + "move.b %%d2,%%d7 \n\t" + + /* read dot 0, add DC, save dot */ + "lsr.l #8,%%d4 \n\t" + "move.b %%d4,%%d2 \n\t" + "add.l %%d0,%%d2 \n\t" + "cmp.l #255,%%d2 \n\t" /* overflow? */ + "bls.b 1f \n\t" /* no: return value */ + "spl.b %%d2 \n\t" /* yes: set low byte to appropriate boundary */ + "1 : \n\t" + + /* save dot 0 1 2 3 */ + "move.b %%d2,%%d4 \n\t" + "lsl.l #8,%%d4 \n\t" + "move.b %%d7,%%d4 \n\t" + "lsl.l #8,%%d4 \n\t" + "move.b %%d6,%%d4 \n\t" + "lsl.l #8,%%d4 \n\t" + "move.b %%d3,%%d4 \n\t" + + + /* read dot 7, add DC, save dot */ + "move.b %%d5,%%d2 \n\t" + "add.l %%d0,%%d2 \n\t" + "cmp.l #255,%%d2 \n\t" + "bls.b 8f \n\t" + "spl.b %%d2 \n\t" + "8 : \n\t" + "move.b %%d2,%%d3 \n\t" + + /* read dot 6, add DC, save dot */ + "lsr.l #8,%%d5 \n\t" + "move.b %%d5,%%d2 \n\t" + "add.l %%d0,%%d2 \n\t" + "cmp.l #255,%%d2 \n\t" + "bls.b 7f \n\t" + "spl.b %%d2 \n\t" + "7 : \n\t" + "move.b %%d2,%%d6 \n\t" + + + /* read dot 5, add DC, save dot */ + "lsr.l #8,%%d5 \n\t" + "move.b %%d5,%%d2 \n\t" + "add.l %%d0,%%d2 \n\t" + "cmp.l #255,%%d2 \n\t" + "bls.b 6f \n\t" + "spl.b %%d2 \n\t" + "6 : \n\t" + "move.b %%d2,%%d7 \n\t" + + /* read dot 4, add DC, save dot */ + "lsr.l #8,%%d5 \n\t" + "move.b %%d5,%%d2 \n\t" + "add.l %%d0,%%d2 \n\t" + "cmp.l #255,%%d2 \n\t" /* overflow? */ + "bls.b 5f \n\t" /* no: return value */ + "spl.b %%d2 \n\t" /* yes: set low byte to appropriate boundary */ + "5 : \n\t" + /* save dot 4 5 6 7 */ + "move.b %%d2,%%d5 \n\t" + "lsl.l #8,%%d5 \n\t" + "move.b %%d7,%%d5 \n\t" + "lsl.l #8,%%d5 \n\t" + "move.b %%d6,%%d5 \n\t" + "lsl.l #8,%%d5 \n\t" + "move.b %%d3,%%d5 \n\t" + + /* write to dest */ + "movem.l %%d4-%%d5,(%%a1) \n\t" + + "add.l %[stride],%%a1 \n\t" + "sub.l #1,%%d1 \n\t" + "bne foraddDC \n\t" + /* for end */ + + : + : [block]"a"(block),[dest]"a"(dest),[stride]"a"(stride) + : "%a0","%a1","%d0","%d1","%d2","%d3","%d4","%d5","%d6","%d7" + ); + +} +#endif + +#ifdef CPU_COLDFIRE__ +static void clip_block_to_dest_add_DC(int16_t * block, uint8_t * dest,const int stride) +{ + asm volatile ( + + "move.l %[block],%%a0 \n\t" + "move.l %[dest],%%a1 \n\t" + + /* DC */ + "move.w (%%a0),%%d0 \n\t" + "ext.l %%d0 \n\t" + "add.l #64,%%d0 \n\t" + "asr.l #7,%%d0 \n\t" + "move.w #0,(%%a0) \n\t" + "move.w #0,%%d2 \n\t" + "move.w %%d2,(126,%%a0) \n\t" + + /* init loop for i=0 to 7 */ + "move.l #8,%%d1 \n\t" + "clr.l %%d2 \n\t" + "foraddDC: \n\t" + + "move.b (%%a1),%%d2 \n\t" + "add.l %%d0,%%d2 \n\t" + "cmp.l #255,%%d2 \n\t" /* overflow? */ + "bls.b 1f \n\t" /* no: return value */ + "spl.b %%d2 \n\t" /* yes: set low byte to appropriate boundary */ + "1 : \n\t" + "move.b %%d2,%%d3 \n\t" + "lsl.l #8,%%d3 \n\t" + + "move.b (1,%%a1),%%d2 \n\t" + "add.l %%d0,%%d2 \n\t" + "cmp.l #255,%%d2 \n\t" + "bls.b 2f \n\t" + "spl.b %%d2 \n\t" + "2 : \n\t" + "move.b %%d2,%%d3 \n\t" + "lsl.l #8,%%d3 \n\t" + + "move.b (2,%%a1),%%d2 \n\t" + "add.l %%d0,%%d2 \n\t" + "cmp.l #255,%%d2 \n\t" + "bls.b 3f \n\t" + "spl.b %%d2 \n\t" + "3 : \n\t" + "move.b %%d2,%%d3 \n\t" + "lsl.l #8,%%d3 \n\t" + + "move.b (3,%%a1),%%d2 \n\t" + "add.l %%d0,%%d2 \n\t" + "cmp.l #255,%%d2 \n\t" + "bls.b 4f \n\t" + "spl.b %%d2 \n\t" + "4 : \n\t" + "move.b %%d2,%%d3 \n\t" + + "move.b (4,%%a1),%%d2 \n\t" + "add.l %%d0,%%d2 \n\t" + "cmp.l #255,%%d2 \n\t" + "bls.b 5f \n\t" + "spl.b %%d2 \n\t" + "5 : \n\t" + "move.b %%d2,%%d4 \n\t" + "lsl.l #8,%%d4 \n\t" + + "move.b (5,%%a1),%%d2 \n\t" + "add.l %%d0,%%d2 \n\t" + "cmp.l #255,%%d2 \n\t" + "bls.b 6f \n\t" + "spl.b %%d2 \n\t" + "6 : \n\t" + "move.b %%d2,%%d4 \n\t" + "lsl.l #8,%%d4 \n\t" + + "move.b (6,%%a1),%%d2 \n\t" + "add.l %%d0,%%d2 \n\t" + "cmp.l #255,%%d2 \n\t" + "bls.b 7f \n\t" + "spl.b %%d2 \n\t" + "7 : \n\t" + "move.b %%d2,%%d4 \n\t" + "lsl.l #8,%%d4 \n\t" + + "move.b (7,%%a1),%%d2 \n\t" + "add.l %%d0,%%d2 \n\t" + "cmp.l #255,%%d2 \n\t" + "bls.b 8f \n\t" + "spl.b %%d2 \n\t" + "8 : \n\t" + "move.b %%d2,%%d4 \n\t" + + /* write to dest */ + "movem.l %%d3-%%d4,(%%a1) \n\t" + + /* next line, address +16 */ + "add.l #16,%%a0 \n\t" + "add.l %[stride],%%a1 \n\t" + "sub.l #1,%%d1 \n\t" + "bne foraddDC \n\t" + /* for end */ + + : + : [block]"a"(block),[dest]"a"(dest),[stride]"a"(stride) + : "%a0","%a1","%d0","%d1","%d2","%d3","%d4" + ); + +} +#endif + +#if defined(CPU_COLDFIRE) && !defined(SIMULATOR) static void mpeg2_idct_copy_c (int16_t * block, uint8_t * dest, const int stride) { - int i; - - for (i = 0; i < 8; i++) + idct (block); + clip_block_to_dest(block,dest,stride); +} +#else +static void mpeg2_idct_copy_c (int16_t * block, uint8_t * dest, + const int stride) +{ + int i; + for (i = 0; i < 8; i++) idct_row (block + 8 * i); for (i = 0; i < 8; i++) idct_col (block + i); @@ -210,18 +1283,31 @@ dest += stride; block += 8; } while (--i); -} +} +#endif +#if defined(CPU_COLDFIRE) && !defined(SIMULATOR) +static void mpeg2_idct_add_c (const int last, int16_t * block, + uint8_t * dest, const int stride) +{ + if (last != 129 || (block[0] & (7 << 4)) == (4 << 4)) { + idct (block); + clip_block_to_dest_add(block,dest,stride); + } else { + clip_block_to_dest_add_DC(block,dest,stride); + } +} +#else static void mpeg2_idct_add_c (const int last, int16_t * block, uint8_t * dest, const int stride) { int i; if (last != 129 || (block[0] & (7 << 4)) == (4 << 4)) { - for (i = 0; i < 8; i++) - idct_row (block + 8 * i); - for (i = 0; i < 8; i++) - idct_col (block + i); + for (i = 0; i < 8; i++) + idct_row (block + 8 * i); + for (i = 0; i < 8; i++) + idct_col (block + i); do { dest[0] = CLIP (block[0] + dest[0]); dest[1] = CLIP (block[1] + dest[1]); @@ -257,7 +1343,8 @@ } while (--i); } } - +#endif + void mpeg2_idct_init (uint32_t accel) { (void)accel; Index: apps/plugins/mpegplayer/mpeg2_internal.h =================================================================== RCS file: /cvsroot/rockbox/apps/plugins/mpegplayer/mpeg2_internal.h,v retrieving revision 1.1 diff -u -r1.1 mpeg2_internal.h --- apps/plugins/mpegplayer/mpeg2_internal.h 7 Aug 2006 22:11:07 -0000 1.1 +++ apps/plugins/mpegplayer/mpeg2_internal.h 25 Sep 2006 22:15:26 -0000 @@ -92,8 +92,12 @@ int16_t dc_dct_pred[3]; /* DCT coefficients */ +#if defined(CPU_COLDFIRE) && !defined(SIMULATOR) + // we need another temp buffer + int16_t DCTblock[128] ATTR_ALIGN(64); +#else int16_t DCTblock[64] ATTR_ALIGN(64); - +#endif uint8_t * picture_dest[3]; void (* convert) (void * convert_id, uint8_t * const * src, unsigned int v_offset);