yading@10: /*
yading@10:  * DSP utils
yading@10:  * Copyright (c) 2000, 2001 Fabrice Bellard
yading@10:  * Copyright (c) 2002-2004 Michael Niedermayer <michaelni@gmx.at>
yading@10:  *
yading@10:  * gmc & q-pel & 32/64 bit based MC by Michael Niedermayer <michaelni@gmx.at>
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: 
yading@10: /**
yading@10:  * @file
yading@10:  * DSP utils
yading@10:  */
yading@10: 
yading@10: #include "bit_depth_template.c"
yading@10: 
yading@10: #if BIT_DEPTH == 8
yading@10: /* draw the edges of width 'w' of an image of size width, height */
yading@10: //FIXME check that this is ok for mpeg4 interlaced
yading@10: static void FUNCC(draw_edges)(uint8_t *p_buf, int p_wrap, int width, int height, int w, int h, int sides)
yading@10: {
yading@10:     pixel *buf = (pixel*)p_buf;
yading@10:     int wrap = p_wrap / sizeof(pixel);
yading@10:     pixel *ptr, *last_line;
yading@10:     int i;
yading@10: 
yading@10:     /* left and right */
yading@10:     ptr = buf;
yading@10:     for(i=0;i<height;i++) {
yading@10:         memset(ptr - w, ptr[0], w);
yading@10:         memset(ptr + width, ptr[width-1], w);
yading@10:         ptr += wrap;
yading@10:     }
yading@10: 
yading@10:     /* top and bottom + corners */
yading@10:     buf -= w;
yading@10:     last_line = buf + (height - 1) * wrap;
yading@10:     if (sides & EDGE_TOP)
yading@10:         for(i = 0; i < h; i++)
yading@10:             memcpy(buf - (i + 1) * wrap, buf, (width + w + w) * sizeof(pixel)); // top
yading@10:     if (sides & EDGE_BOTTOM)
yading@10:         for (i = 0; i < h; i++)
yading@10:             memcpy(last_line + (i + 1) * wrap, last_line, (width + w + w) * sizeof(pixel)); // bottom
yading@10: }
yading@10: #endif
yading@10: 
yading@10: static void FUNCC(get_pixels)(int16_t *av_restrict block,
yading@10:                               const uint8_t *_pixels,
yading@10:                               int line_size)
yading@10: {
yading@10:     const pixel *pixels = (const pixel *) _pixels;
yading@10:     int i;
yading@10: 
yading@10:     /* read the pixels */
yading@10:     for(i=0;i<8;i++) {
yading@10:         block[0] = pixels[0];
yading@10:         block[1] = pixels[1];
yading@10:         block[2] = pixels[2];
yading@10:         block[3] = pixels[3];
yading@10:         block[4] = pixels[4];
yading@10:         block[5] = pixels[5];
yading@10:         block[6] = pixels[6];
yading@10:         block[7] = pixels[7];
yading@10:         pixels += line_size / sizeof(pixel);
yading@10:         block += 8;
yading@10:     }
yading@10: }
yading@10: 
yading@10: #if BIT_DEPTH == 8
yading@10: static void FUNCC(clear_block)(int16_t *block)
yading@10: {
yading@10:     memset(block, 0, sizeof(int16_t)*64);
yading@10: }
yading@10: 
yading@10: static void FUNCC(clear_blocks)(int16_t *blocks)
yading@10: {
yading@10:     memset(blocks, 0, sizeof(int16_t)*6*64);
yading@10: }
yading@10: #endif
yading@10: 
yading@10: #if BIT_DEPTH == 8
yading@10: #include "hpel_template.c"
yading@10: #endif
yading@10: 
yading@10: #define PIXOP2(OPNAME, OP) \
yading@10: static inline void FUNC(OPNAME ## _no_rnd_pixels8_l2)(uint8_t *dst, const uint8_t *src1, const uint8_t *src2, int dst_stride, \
yading@10:                                                 int src_stride1, int src_stride2, int h){\
yading@10:     int i;\
yading@10:     for(i=0; i<h; i++){\
yading@10:         pixel4 a,b;\
yading@10:         a= AV_RN4P(&src1[i*src_stride1  ]);\
yading@10:         b= AV_RN4P(&src2[i*src_stride2  ]);\
yading@10:         OP(*((pixel4*)&dst[i*dst_stride  ]), no_rnd_avg_pixel4(a, b));\
yading@10:         a= AV_RN4P(&src1[i*src_stride1+4*sizeof(pixel)]);\
yading@10:         b= AV_RN4P(&src2[i*src_stride2+4*sizeof(pixel)]);\
yading@10:         OP(*((pixel4*)&dst[i*dst_stride+4*sizeof(pixel)]), no_rnd_avg_pixel4(a, b));\
yading@10:     }\
yading@10: }\
yading@10: \
yading@10: static inline void FUNC(OPNAME ## _no_rnd_pixels16_l2)(uint8_t *dst, const uint8_t *src1, const uint8_t *src2, int dst_stride, \
yading@10:                                                 int src_stride1, int src_stride2, int h){\
yading@10:     FUNC(OPNAME ## _no_rnd_pixels8_l2)(dst  , src1  , src2  , dst_stride, src_stride1, src_stride2, h);\
yading@10:     FUNC(OPNAME ## _no_rnd_pixels8_l2)(dst+8*sizeof(pixel), src1+8*sizeof(pixel), src2+8*sizeof(pixel), dst_stride, src_stride1, src_stride2, h);\
yading@10: }\
yading@10: \
yading@10: static inline void FUNC(OPNAME ## _pixels8_l4)(uint8_t *dst, const uint8_t *src1, const uint8_t *src2, const uint8_t *src3, const uint8_t *src4,\
yading@10:                  int dst_stride, int src_stride1, int src_stride2,int src_stride3,int src_stride4, int h){\
yading@10:     /* FIXME HIGH BIT DEPTH */\
yading@10:     int i;\
yading@10:     for(i=0; i<h; i++){\
yading@10:         uint32_t a, b, c, d, l0, l1, h0, h1;\
yading@10:         a= AV_RN32(&src1[i*src_stride1]);\
yading@10:         b= AV_RN32(&src2[i*src_stride2]);\
yading@10:         c= AV_RN32(&src3[i*src_stride3]);\
yading@10:         d= AV_RN32(&src4[i*src_stride4]);\
yading@10:         l0=  (a&0x03030303UL)\
yading@10:            + (b&0x03030303UL)\
yading@10:            + 0x02020202UL;\
yading@10:         h0= ((a&0xFCFCFCFCUL)>>2)\
yading@10:           + ((b&0xFCFCFCFCUL)>>2);\
yading@10:         l1=  (c&0x03030303UL)\
yading@10:            + (d&0x03030303UL);\
yading@10:         h1= ((c&0xFCFCFCFCUL)>>2)\
yading@10:           + ((d&0xFCFCFCFCUL)>>2);\
yading@10:         OP(*((uint32_t*)&dst[i*dst_stride]), h0+h1+(((l0+l1)>>2)&0x0F0F0F0FUL));\
yading@10:         a= AV_RN32(&src1[i*src_stride1+4]);\
yading@10:         b= AV_RN32(&src2[i*src_stride2+4]);\
yading@10:         c= AV_RN32(&src3[i*src_stride3+4]);\
yading@10:         d= AV_RN32(&src4[i*src_stride4+4]);\
yading@10:         l0=  (a&0x03030303UL)\
yading@10:            + (b&0x03030303UL)\
yading@10:            + 0x02020202UL;\
yading@10:         h0= ((a&0xFCFCFCFCUL)>>2)\
yading@10:           + ((b&0xFCFCFCFCUL)>>2);\
yading@10:         l1=  (c&0x03030303UL)\
yading@10:            + (d&0x03030303UL);\
yading@10:         h1= ((c&0xFCFCFCFCUL)>>2)\
yading@10:           + ((d&0xFCFCFCFCUL)>>2);\
yading@10:         OP(*((uint32_t*)&dst[i*dst_stride+4]), h0+h1+(((l0+l1)>>2)&0x0F0F0F0FUL));\
yading@10:     }\
yading@10: }\
yading@10: \
yading@10: static inline void FUNC(OPNAME ## _no_rnd_pixels8_l4)(uint8_t *dst, const uint8_t *src1, const uint8_t *src2, const uint8_t *src3, const uint8_t *src4,\
yading@10:                  int dst_stride, int src_stride1, int src_stride2,int src_stride3,int src_stride4, int h){\
yading@10:     /* FIXME HIGH BIT DEPTH*/\
yading@10:     int i;\
yading@10:     for(i=0; i<h; i++){\
yading@10:         uint32_t a, b, c, d, l0, l1, h0, h1;\
yading@10:         a= AV_RN32(&src1[i*src_stride1]);\
yading@10:         b= AV_RN32(&src2[i*src_stride2]);\
yading@10:         c= AV_RN32(&src3[i*src_stride3]);\
yading@10:         d= AV_RN32(&src4[i*src_stride4]);\
yading@10:         l0=  (a&0x03030303UL)\
yading@10:            + (b&0x03030303UL)\
yading@10:            + 0x01010101UL;\
yading@10:         h0= ((a&0xFCFCFCFCUL)>>2)\
yading@10:           + ((b&0xFCFCFCFCUL)>>2);\
yading@10:         l1=  (c&0x03030303UL)\
yading@10:            + (d&0x03030303UL);\
yading@10:         h1= ((c&0xFCFCFCFCUL)>>2)\
yading@10:           + ((d&0xFCFCFCFCUL)>>2);\
yading@10:         OP(*((uint32_t*)&dst[i*dst_stride]), h0+h1+(((l0+l1)>>2)&0x0F0F0F0FUL));\
yading@10:         a= AV_RN32(&src1[i*src_stride1+4]);\
yading@10:         b= AV_RN32(&src2[i*src_stride2+4]);\
yading@10:         c= AV_RN32(&src3[i*src_stride3+4]);\
yading@10:         d= AV_RN32(&src4[i*src_stride4+4]);\
yading@10:         l0=  (a&0x03030303UL)\
yading@10:            + (b&0x03030303UL)\
yading@10:            + 0x01010101UL;\
yading@10:         h0= ((a&0xFCFCFCFCUL)>>2)\
yading@10:           + ((b&0xFCFCFCFCUL)>>2);\
yading@10:         l1=  (c&0x03030303UL)\
yading@10:            + (d&0x03030303UL);\
yading@10:         h1= ((c&0xFCFCFCFCUL)>>2)\
yading@10:           + ((d&0xFCFCFCFCUL)>>2);\
yading@10:         OP(*((uint32_t*)&dst[i*dst_stride+4]), h0+h1+(((l0+l1)>>2)&0x0F0F0F0FUL));\
yading@10:     }\
yading@10: }\
yading@10: static inline void FUNC(OPNAME ## _pixels16_l4)(uint8_t *dst, const uint8_t *src1, const uint8_t *src2, const uint8_t *src3, const uint8_t *src4,\
yading@10:                  int dst_stride, int src_stride1, int src_stride2,int src_stride3,int src_stride4, int h){\
yading@10:     FUNC(OPNAME ## _pixels8_l4)(dst  , src1  , src2  , src3  , src4  , dst_stride, src_stride1, src_stride2, src_stride3, src_stride4, h);\
yading@10:     FUNC(OPNAME ## _pixels8_l4)(dst+8*sizeof(pixel), src1+8*sizeof(pixel), src2+8*sizeof(pixel), src3+8*sizeof(pixel), src4+8*sizeof(pixel), dst_stride, src_stride1, src_stride2, src_stride3, src_stride4, h);\
yading@10: }\
yading@10: static inline void FUNC(OPNAME ## _no_rnd_pixels16_l4)(uint8_t *dst, const uint8_t *src1, const uint8_t *src2, const uint8_t *src3, const uint8_t *src4,\
yading@10:                  int dst_stride, int src_stride1, int src_stride2,int src_stride3,int src_stride4, int h){\
yading@10:     FUNC(OPNAME ## _no_rnd_pixels8_l4)(dst  , src1  , src2  , src3  , src4  , dst_stride, src_stride1, src_stride2, src_stride3, src_stride4, h);\
yading@10:     FUNC(OPNAME ## _no_rnd_pixels8_l4)(dst+8*sizeof(pixel), src1+8*sizeof(pixel), src2+8*sizeof(pixel), src3+8*sizeof(pixel), src4+8*sizeof(pixel), dst_stride, src_stride1, src_stride2, src_stride3, src_stride4, h);\
yading@10: }\
yading@10: \
yading@10: static inline void FUNCC(OPNAME ## _pixels8_xy2)(uint8_t *block, const uint8_t *pixels, ptrdiff_t line_size, int h)\
yading@10: {\
yading@10:     /* FIXME HIGH BIT DEPTH */\
yading@10:     int j;\
yading@10:     for(j=0; j<2; j++){\
yading@10:         int i;\
yading@10:         const uint32_t a= AV_RN32(pixels  );\
yading@10:         const uint32_t b= AV_RN32(pixels+1);\
yading@10:         uint32_t l0=  (a&0x03030303UL)\
yading@10:                     + (b&0x03030303UL)\
yading@10:                     + 0x02020202UL;\
yading@10:         uint32_t h0= ((a&0xFCFCFCFCUL)>>2)\
yading@10:                    + ((b&0xFCFCFCFCUL)>>2);\
yading@10:         uint32_t l1,h1;\
yading@10: \
yading@10:         pixels+=line_size;\
yading@10:         for(i=0; i<h; i+=2){\
yading@10:             uint32_t a= AV_RN32(pixels  );\
yading@10:             uint32_t b= AV_RN32(pixels+1);\
yading@10:             l1=  (a&0x03030303UL)\
yading@10:                + (b&0x03030303UL);\
yading@10:             h1= ((a&0xFCFCFCFCUL)>>2)\
yading@10:               + ((b&0xFCFCFCFCUL)>>2);\
yading@10:             OP(*((uint32_t*)block), h0+h1+(((l0+l1)>>2)&0x0F0F0F0FUL));\
yading@10:             pixels+=line_size;\
yading@10:             block +=line_size;\
yading@10:             a= AV_RN32(pixels  );\
yading@10:             b= AV_RN32(pixels+1);\
yading@10:             l0=  (a&0x03030303UL)\
yading@10:                + (b&0x03030303UL)\
yading@10:                + 0x02020202UL;\
yading@10:             h0= ((a&0xFCFCFCFCUL)>>2)\
yading@10:               + ((b&0xFCFCFCFCUL)>>2);\
yading@10:             OP(*((uint32_t*)block), h0+h1+(((l0+l1)>>2)&0x0F0F0F0FUL));\
yading@10:             pixels+=line_size;\
yading@10:             block +=line_size;\
yading@10:         }\
yading@10:         pixels+=4-line_size*(h+1);\
yading@10:         block +=4-line_size*h;\
yading@10:     }\
yading@10: }\
yading@10: \
yading@10: CALL_2X_PIXELS(FUNCC(OPNAME ## _pixels16_xy2), FUNCC(OPNAME ## _pixels8_xy2), 8*sizeof(pixel))\
yading@10: 
yading@10: #define op_avg(a, b) a = rnd_avg_pixel4(a, b)
yading@10: #define op_put(a, b) a = b
yading@10: #if BIT_DEPTH == 8
yading@10: #define put_no_rnd_pixels8_8_c put_pixels8_8_c
yading@10: PIXOP2(avg, op_avg)
yading@10: PIXOP2(put, op_put)
yading@10: #endif
yading@10: #undef op_avg
yading@10: #undef op_put