--- interpolate8x8.h 2003/02/15 15:22:18 1.10 +++ interpolate8x8.h 2005/09/13 12:12:15 1.16 @@ -1,3 +1,28 @@ +/***************************************************************************** + * + * XVID MPEG-4 VIDEO CODEC + * - Interpolation related header - + * + * Copyright(C) 2001-2003 Peter Ross + * + * This program is free software ; you can redistribute it and/or modify + * it under the terms of the GNU General Public License as published by + * the Free Software Foundation ; either version 2 of the License, or + * (at your option) any later version. + * + * This program 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 General Public License for more details. + * + * You should have received a copy of the GNU General Public License + * along with this program ; if not, write to the Free Software + * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA + * + * $Id: interpolate8x8.h,v 1.16 2005/09/13 12:12:15 suxen_drol Exp $ + * + ****************************************************************************/ + #ifndef _INTERPOLATE8X8_H_ #define _INTERPOLATE8X8_H_ @@ -9,6 +34,12 @@ const uint32_t rounding); typedef INTERPOLATE8X8 *INTERPOLATE8X8_PTR; +typedef void (INTERPOLATE8X4) (uint8_t * const dst, + const uint8_t * const src, + const uint32_t stride, + const uint32_t rounding); +typedef INTERPOLATE8X4 *INTERPOLATE8X4_PTR; + typedef void (INTERPOLATE8X8_AVG2) (uint8_t *dst, const uint8_t *src1, const uint8_t *src2, @@ -48,10 +79,26 @@ typedef INTERPOLATE8X8_6TAP_LOWPASS *INTERPOLATE8X8_6TAP_LOWPASS_PTR; +/* These function do: dst = interpolate(src) */ extern INTERPOLATE8X8_PTR interpolate8x8_halfpel_h; extern INTERPOLATE8X8_PTR interpolate8x8_halfpel_v; extern INTERPOLATE8X8_PTR interpolate8x8_halfpel_hv; +extern INTERPOLATE8X4_PTR interpolate8x4_halfpel_h; +extern INTERPOLATE8X4_PTR interpolate8x4_halfpel_v; +extern INTERPOLATE8X4_PTR interpolate8x4_halfpel_hv; + +/* These functions do: dst = (dst+interpolate(src) + 1)/2 + * Suitable for direct/interpolated bvop prediction block + * building w/o the need for intermediate interpolated result + * storing/reading + * NB: the rounding applies to the interpolation, but not + * the averaging step which will always use rounding=0 */ +extern INTERPOLATE8X8_PTR interpolate8x8_halfpel_add; +extern INTERPOLATE8X8_PTR interpolate8x8_halfpel_h_add; +extern INTERPOLATE8X8_PTR interpolate8x8_halfpel_v_add; +extern INTERPOLATE8X8_PTR interpolate8x8_halfpel_hv_add; + extern INTERPOLATE8X8_AVG2_PTR interpolate8x8_avg2; extern INTERPOLATE8X8_AVG4_PTR interpolate8x8_avg4; @@ -71,31 +118,104 @@ INTERPOLATE8X8 interpolate8x8_halfpel_v_c; INTERPOLATE8X8 interpolate8x8_halfpel_hv_c; +INTERPOLATE8X4 interpolate8x4_halfpel_h_c; +INTERPOLATE8X4 interpolate8x4_halfpel_v_c; +INTERPOLATE8X4 interpolate8x4_halfpel_hv_c; + +INTERPOLATE8X8 interpolate8x8_halfpel_add_c; +INTERPOLATE8X8 interpolate8x8_halfpel_h_add_c; +INTERPOLATE8X8 interpolate8x8_halfpel_v_add_c; +INTERPOLATE8X8 interpolate8x8_halfpel_hv_add_c; + +#ifdef ARCH_IS_IA32 INTERPOLATE8X8 interpolate8x8_halfpel_h_mmx; INTERPOLATE8X8 interpolate8x8_halfpel_v_mmx; INTERPOLATE8X8 interpolate8x8_halfpel_hv_mmx; +INTERPOLATE8X4 interpolate8x4_halfpel_h_mmx; +INTERPOLATE8X4 interpolate8x4_halfpel_v_mmx; +INTERPOLATE8X4 interpolate8x4_halfpel_hv_mmx; + +INTERPOLATE8X8 interpolate8x8_halfpel_add_mmx; +INTERPOLATE8X8 interpolate8x8_halfpel_h_add_mmx; +INTERPOLATE8X8 interpolate8x8_halfpel_v_add_mmx; +INTERPOLATE8X8 interpolate8x8_halfpel_hv_add_mmx; + INTERPOLATE8X8 interpolate8x8_halfpel_h_xmm; INTERPOLATE8X8 interpolate8x8_halfpel_v_xmm; INTERPOLATE8X8 interpolate8x8_halfpel_hv_xmm; +INTERPOLATE8X4 interpolate8x4_halfpel_h_xmm; +INTERPOLATE8X4 interpolate8x4_halfpel_v_xmm; +INTERPOLATE8X4 interpolate8x4_halfpel_hv_xmm; + +INTERPOLATE8X8 interpolate8x8_halfpel_add_xmm; +INTERPOLATE8X8 interpolate8x8_halfpel_h_add_xmm; +INTERPOLATE8X8 interpolate8x8_halfpel_v_add_xmm; +INTERPOLATE8X8 interpolate8x8_halfpel_hv_add_xmm; + INTERPOLATE8X8 interpolate8x8_halfpel_h_3dn; INTERPOLATE8X8 interpolate8x8_halfpel_v_3dn; INTERPOLATE8X8 interpolate8x8_halfpel_hv_3dn; +INTERPOLATE8X4 interpolate8x4_halfpel_h_3dn; +INTERPOLATE8X4 interpolate8x4_halfpel_v_3dn; +INTERPOLATE8X4 interpolate8x4_halfpel_hv_3dn; + INTERPOLATE8X8 interpolate8x8_halfpel_h_3dne; INTERPOLATE8X8 interpolate8x8_halfpel_v_3dne; INTERPOLATE8X8 interpolate8x8_halfpel_hv_3dne; +INTERPOLATE8X4 interpolate8x4_halfpel_h_3dne; +INTERPOLATE8X4 interpolate8x4_halfpel_v_3dne; +INTERPOLATE8X4 interpolate8x4_halfpel_hv_3dne; +#endif + +#ifdef ARCH_IS_IA64 INTERPOLATE8X8 interpolate8x8_halfpel_h_ia64; INTERPOLATE8X8 interpolate8x8_halfpel_v_ia64; INTERPOLATE8X8 interpolate8x8_halfpel_hv_ia64; +#endif + +#ifdef ARCH_IS_PPC +INTERPOLATE8X8 interpolate8x8_halfpel_h_altivec_c; +INTERPOLATE8X8 interpolate8x8_halfpel_v_altivec_c; +INTERPOLATE8X8 interpolate8x8_halfpel_hv_altivec_c; + +INTERPOLATE8X8 interpolate8x8_halfpel_add_altivec_c; +INTERPOLATE8X8 interpolate8x8_halfpel_h_add_altivec_c; +INTERPOLATE8X8 interpolate8x8_halfpel_v_add_altivec_c; +INTERPOLATE8X8 interpolate8x8_halfpel_hv_add_altivec_c; +#endif + +#ifdef ARCH_IS_X86_64 +INTERPOLATE8X8 interpolate8x8_halfpel_h_x86_64; +INTERPOLATE8X8 interpolate8x8_halfpel_v_x86_64; +INTERPOLATE8X8 interpolate8x8_halfpel_hv_x86_64; + +INTERPOLATE8X8 interpolate8x8_halfpel_add_x86_64; +INTERPOLATE8X8 interpolate8x8_halfpel_h_add_x86_64; +INTERPOLATE8X8 interpolate8x8_halfpel_v_add_x86_64; +INTERPOLATE8X8 interpolate8x8_halfpel_hv_add_x86_64; +#endif INTERPOLATE8X8_AVG2 interpolate8x8_avg2_c; INTERPOLATE8X8_AVG4 interpolate8x8_avg4_c; +#ifdef ARCH_IS_IA32 INTERPOLATE8X8_AVG2 interpolate8x8_avg2_mmx; INTERPOLATE8X8_AVG4 interpolate8x8_avg4_mmx; +#endif + +#ifdef ARCH_IS_PPC +INTERPOLATE8X8_AVG2 interpolate8x8_avg2_altivec_c; +INTERPOLATE8X8_AVG4 interpolate8x8_avg4_altivec_c; +#endif + +#ifdef ARCH_IS_X86_64 +INTERPOLATE8X8_AVG2 interpolate8x8_avg2_x86_64; +INTERPOLATE8X8_AVG4 interpolate8x8_avg4_x86_64; +#endif INTERPOLATE_LOWPASS interpolate8x8_lowpass_h_c; INTERPOLATE_LOWPASS interpolate8x8_lowpass_v_c; @@ -109,11 +229,22 @@ INTERPOLATE8X8_6TAP_LOWPASS interpolate8x8_6tap_lowpass_h_c; INTERPOLATE8X8_6TAP_LOWPASS interpolate8x8_6tap_lowpass_v_c; +#ifdef ARCH_IS_IA32 INTERPOLATE8X8_6TAP_LOWPASS interpolate8x8_6tap_lowpass_h_mmx; INTERPOLATE8X8_6TAP_LOWPASS interpolate8x8_6tap_lowpass_v_mmx; +#endif + +#ifdef ARCH_IS_PPC +INTERPOLATE8X8_6TAP_LOWPASS interpolate8x8_6tap_lowpass_h_altivec_c; +#endif + +#ifdef ARCH_IS_X86_64 +INTERPOLATE8X8_6TAP_LOWPASS interpolate8x8_6tap_lowpass_h_x86_64; +INTERPOLATE8X8_6TAP_LOWPASS interpolate8x8_6tap_lowpass_v_x86_64; +#endif static __inline void -interpolate8x8_switch(uint8_t * const cur, +interpolate8x4_switch(uint8_t * const cur, const uint8_t * const refn, const uint32_t x, const uint32_t y, @@ -122,43 +253,86 @@ const uint32_t stride, const uint32_t rounding) { - int32_t ddx, ddy; - switch (((dx & 1) << 1) + (dy & 1)) // ((dx%2)?2:0)+((dy%2)?1:0) - { + const uint8_t * const src = refn + (int)((y + (dy>>1)) * stride + x + (dx>>1)); + uint8_t * const dst = cur + (int)(y * stride + x); + + switch (((dx & 1) << 1) + (dy & 1)) + { /* ((dx%2)?2:0)+((dy%2)?1:0) */ case 0: - ddx = dx / 2; - ddy = dy / 2; - transfer8x8_copy(cur + y * stride + x, - refn + (int)((y + ddy) * stride + x + ddx), stride); + transfer8x4_copy(dst, src, stride); break; - case 1: - ddx = dx / 2; - ddy = (dy - 1) / 2; - interpolate8x8_halfpel_v(cur + y * stride + x, - refn + (int)((y + ddy) * stride + x + ddx), stride, - rounding); + interpolate8x4_halfpel_v(dst, src, stride, rounding); break; - case 2: - ddx = (dx - 1) / 2; - ddy = dy / 2; - interpolate8x8_halfpel_h(cur + y * stride + x, - refn + (int)((y + ddy) * stride + x + ddx), stride, - rounding); + interpolate8x4_halfpel_h(dst, src, stride, rounding); + break; + default: + interpolate8x4_halfpel_hv(dst, src, stride, rounding); break; + } +} +static __inline void +interpolate8x8_switch(uint8_t * const cur, + const uint8_t * const refn, + const uint32_t x, + const uint32_t y, + const int32_t dx, + const int dy, + const uint32_t stride, + const uint32_t rounding) +{ + + const uint8_t * const src = refn + (int)((y + (dy>>1)) * stride + x + (dx>>1)); + uint8_t * const dst = cur + (int)(y * stride + x); + + switch (((dx & 1) << 1) + (dy & 1)) { /* ((dx%2)?2:0)+((dy%2)?1:0) */ + case 0: + transfer8x8_copy(dst, src, stride); + break; + case 1: + interpolate8x8_halfpel_v(dst, src, stride, rounding); + break; + case 2: + interpolate8x8_halfpel_h(dst, src, stride, rounding); + break; default: - ddx = (dx - 1) / 2; - ddy = (dy - 1) / 2; - interpolate8x8_halfpel_hv(cur + y * stride + x, - refn + (int)((y + ddy) * stride + x + ddx), stride, - rounding); + interpolate8x8_halfpel_hv(dst, src, stride, rounding); break; } } +static __inline void +interpolate8x8_add_switch(uint8_t * const cur, + const uint8_t * const refn, + const uint32_t x, + const uint32_t y, + const int32_t dx, + const int dy, + const uint32_t stride, + const uint32_t rounding) +{ + + const uint8_t * const src = refn + (int)((y + (dy>>1)) * stride + x + (dx>>1)); + uint8_t * const dst = cur + (int)(y * stride + x); + + switch (((dx & 1) << 1) + (dy & 1)) { /* ((dx%2)?2:0)+((dy%2)?1:0) */ + case 0: + interpolate8x8_halfpel_add(dst, src, stride, rounding); + break; + case 1: + interpolate8x8_halfpel_v_add(dst, src, stride, rounding); + break; + case 2: + interpolate8x8_halfpel_h_add(dst, src, stride, rounding); + break; + default: + interpolate8x8_halfpel_hv_add(dst, src, stride, rounding); + break; + } +} static __inline void interpolate16x16_switch(uint8_t * const cur, @@ -176,6 +350,21 @@ interpolate8x8_switch(cur, refn, x+8, y+8, dx, dy, stride, rounding); } +static __inline void +interpolate16x16_add_switch(uint8_t * const cur, + const uint8_t * const refn, + const uint32_t x, + const uint32_t y, + const int32_t dx, + const int dy, + const uint32_t stride, + const uint32_t rounding) +{ + interpolate8x8_add_switch(cur, refn, x, y, dx, dy, stride, rounding); + interpolate8x8_add_switch(cur, refn, x+8, y, dx, dy, stride, rounding); + interpolate8x8_add_switch(cur, refn, x, y+8, dx, dy, stride, rounding); + interpolate8x8_add_switch(cur, refn, x+8, y+8, dx, dy, stride, rounding); +} static __inline void interpolate32x32_switch(uint8_t * const cur, @@ -193,9 +382,8 @@ interpolate16x16_switch(cur, refn, x+16, y+16, dx, dy, stride, rounding); } - -static __inline uint8_t * -interpolate8x8_switch2(uint8_t * const buffer, +static __inline void +interpolate32x32_add_switch(uint8_t * const cur, const uint8_t * const refn, const uint32_t x, const uint32_t y, @@ -204,342 +392,39 @@ const uint32_t stride, const uint32_t rounding) { - int32_t ddx, ddy; - - switch (((dx & 1) << 1) + (dy & 1)) // ((dx%2)?2:0)+((dy%2)?1:0) - { - case 0: - return (uint8_t *)refn + (int)((y + dy/2) * stride + x + dx/2); - - case 1: - ddx = dx / 2; - ddy = (dy - 1) / 2; - interpolate8x8_halfpel_v(buffer, - refn + (int)((y + ddy) * stride + x + ddx), stride, - rounding); - break; - - case 2: - ddx = (dx - 1) / 2; - ddy = dy / 2; - interpolate8x8_halfpel_h(buffer, - refn + (int)((y + ddy) * stride + x + ddx), stride, - rounding); - break; - - default: - ddx = (dx - 1) / 2; - ddy = (dy - 1) / 2; - interpolate8x8_halfpel_hv(buffer, - refn + (int)((y + ddy) * stride + x + ddx), stride, - rounding); - break; - } - return buffer; -} - -static __inline void interpolate8x8_quarterpel(uint8_t * const cur, - uint8_t * const refn, - uint8_t * const refh, - uint8_t * const refv, - uint8_t * const refhv, - const uint32_t x, const uint32_t y, - const int32_t dx, const int dy, - const uint32_t stride, - const uint32_t rounding) -{ - const int32_t xRef = x*4 + dx; - const int32_t yRef = y*4 + dy; - - uint8_t *src, *dst; - uint8_t *halfpel_h, *halfpel_v, *halfpel_hv; - int32_t x_int, y_int, x_frac, y_frac; - - x_int = xRef/4; - if (xRef < 0 && xRef % 4) - x_int--; - - x_frac = xRef - (4*x_int); - - y_int = yRef/4; - if (yRef < 0 && yRef % 4) - y_int--; - - y_frac = yRef - (4*y_int); - - src = refn + y_int * stride + x_int; - halfpel_h = refh; - halfpel_v = refv; - halfpel_hv = refhv; - - dst = cur + y * stride + x; - - switch((y_frac << 2) | (x_frac)) { - - case 0: - transfer8x8_copy(dst, src, stride); - break; - - case 1: - interpolate8x8_lowpass_h(halfpel_h, src, stride, rounding); - interpolate8x8_avg2(dst, src, halfpel_h, stride, rounding, 8); - break; - - case 2: - interpolate8x8_lowpass_h(dst, src, stride, rounding); - break; - - case 3: - interpolate8x8_lowpass_h(halfpel_h, src, stride, rounding); - interpolate8x8_avg2(dst, src + 1, halfpel_h, stride, rounding, 8); - break; - - case 4: - interpolate8x8_lowpass_v(halfpel_v, src, stride, rounding); - interpolate8x8_avg2(dst, src, halfpel_v, stride, rounding, 8); - break; - - case 5: - interpolate8x8_lowpass_h(halfpel_h, src, stride, rounding); - interpolate8x8_avg2(halfpel_v, src, halfpel_h, stride, rounding, 9); - interpolate8x8_lowpass_v(halfpel_hv, halfpel_v, stride, rounding); - interpolate8x8_avg2(dst, halfpel_v, halfpel_hv, stride, rounding, 8); - break; - - case 6: - interpolate8x8_lowpass_hv(halfpel_hv, halfpel_h, src, stride, rounding); - interpolate8x8_avg2(dst, halfpel_h, halfpel_hv, stride, rounding, 8); - break; - - case 7: - interpolate8x8_lowpass_h(halfpel_h, src, stride, rounding); - interpolate8x8_avg2(halfpel_v, src + 1, halfpel_h, stride, rounding, 9); - interpolate8x8_lowpass_v(halfpel_hv, halfpel_v, stride, rounding); - interpolate8x8_avg2(dst, halfpel_v, halfpel_hv, stride, rounding, 8); - break; - - case 8: - interpolate8x8_lowpass_v(dst, src, stride, rounding); - break; - - case 9: - interpolate8x8_lowpass_h(halfpel_h, src, stride, rounding); - interpolate8x8_avg2(halfpel_v, src, halfpel_h, stride, rounding, 9); - interpolate8x8_lowpass_v(dst, halfpel_v, stride, rounding); - break; - - case 10: - interpolate8x8_lowpass_hv(dst, halfpel_h, src, stride, rounding); - break; - - case 11: - interpolate8x8_lowpass_h(halfpel_h, src, stride, rounding); - interpolate8x8_avg2(halfpel_v, src + 1, halfpel_h, stride, rounding, 9); - interpolate8x8_lowpass_v(dst, halfpel_v, stride, rounding); - break; - - case 12: - interpolate8x8_lowpass_v(halfpel_v, src, stride, rounding); - interpolate8x8_avg2(dst, src+stride, halfpel_v, stride, rounding, 8); - break; - - case 13: - interpolate8x8_lowpass_h(halfpel_h, src, stride, rounding); - interpolate8x8_avg2(halfpel_v, src, halfpel_h, stride, rounding, 9); - interpolate8x8_lowpass_v(halfpel_hv, halfpel_v, stride, rounding); - interpolate8x8_avg2(dst, halfpel_v+stride, halfpel_hv, stride, rounding, 8); - break; - - case 14: - interpolate8x8_lowpass_hv(halfpel_hv, halfpel_h, src, stride, rounding); - interpolate8x8_avg2(dst, halfpel_h+stride, halfpel_hv, stride, rounding, 8); - break; - - case 15: - interpolate8x8_lowpass_h(halfpel_h, src, stride, rounding); - interpolate8x8_avg2(halfpel_v, src + 1, halfpel_h, stride, rounding, 9); - interpolate8x8_lowpass_v(halfpel_hv, halfpel_v, stride, rounding); - interpolate8x8_avg2(dst, halfpel_hv, halfpel_v + stride, stride, rounding, 8); - break; - } + interpolate16x16_add_switch(cur, refn, x, y, dx, dy, stride, rounding); + interpolate16x16_add_switch(cur, refn, x+16, y, dx, dy, stride, rounding); + interpolate16x16_add_switch(cur, refn, x, y+16, dx, dy, stride, rounding); + interpolate16x16_add_switch(cur, refn, x+16, y+16, dx, dy, stride, rounding); } -static __inline void interpolate16x16_quarterpel(uint8_t * const cur, - uint8_t * const refn, - uint8_t * const refh, - uint8_t * const refv, - uint8_t * const refhv, - const uint32_t x, const uint32_t y, - const int32_t dx, const int dy, - const uint32_t stride, - const uint32_t rounding) +static __inline uint8_t * +interpolate8x8_switch2(uint8_t * const buffer, + const uint8_t * const refn, + const int x, + const int y, + const int dx, + const int dy, + const uint32_t stride, + const uint32_t rounding) { - const int32_t xRef = x*4 + dx; - const int32_t yRef = y*4 + dy; - - uint8_t *src, *dst; - uint8_t *halfpel_h, *halfpel_v, *halfpel_hv; - int32_t x_int, y_int, x_frac, y_frac; - - x_int = xRef/4; - if (xRef < 0 && xRef % 4) - x_int--; - - x_frac = xRef - (4*x_int); - - y_int = yRef/4; - if (yRef < 0 && yRef % 4) - y_int--; - y_frac = yRef - (4*y_int); - - src = refn + y_int * stride + x_int; - halfpel_h = refh; - halfpel_v = refv; - halfpel_hv = refhv; - - dst = cur + y * stride + x; - - switch((y_frac << 2) | (x_frac)) { + const uint8_t * const src = refn + (int)((y + (dy>>1)) * stride + x + (dx>>1)); + switch (((dx & 1) << 1) + (dy & 1)) { /* ((dx%2)?2:0)+((dy%2)?1:0) */ case 0: - transfer16x16_copy(dst, src, stride); - break; - + return (uint8_t *)src; case 1: - interpolate16x16_lowpass_h(halfpel_h, src, stride, rounding); - interpolate8x8_avg2(dst, src, halfpel_h, stride, rounding, 8); - interpolate8x8_avg2(dst+8, src+8, halfpel_h+8, stride, rounding, 8); - interpolate8x8_avg2(dst+8*stride, src+8*stride, halfpel_h+8*stride, stride, rounding, 8); - interpolate8x8_avg2(dst+8*stride+8, src+8*stride+8, halfpel_h+8*stride+8, stride, rounding, 8); + interpolate8x8_halfpel_v(buffer, src, stride, rounding); break; - case 2: - interpolate16x16_lowpass_h(dst, src, stride, rounding); - break; - - case 3: - interpolate16x16_lowpass_h(halfpel_h, src, stride, rounding); - interpolate8x8_avg2(dst, src + 1, halfpel_h, stride, rounding, 8); - interpolate8x8_avg2(dst+8, src + 8 + 1, halfpel_h+8, stride, rounding, 8); - interpolate8x8_avg2(dst+8*stride, src + 8*stride + 1, halfpel_h+8*stride, stride, rounding, 8); - interpolate8x8_avg2(dst+8*stride+8, src+8*stride+8 + 1, halfpel_h+8*stride+8, stride, rounding, 8); - break; - - case 4: - interpolate16x16_lowpass_v(halfpel_v, src, stride, rounding); - interpolate8x8_avg2(dst, src, halfpel_v, stride, rounding, 8); - interpolate8x8_avg2(dst+8, src+8, halfpel_v+8, stride, rounding, 8); - interpolate8x8_avg2(dst+8*stride, src+8*stride, halfpel_v+8*stride, stride, rounding, 8); - interpolate8x8_avg2(dst+8*stride+8, src+8*stride+8, halfpel_v+8*stride+8, stride, rounding, 8); - break; - - case 5: - interpolate16x16_lowpass_h(halfpel_h, src, stride, rounding); - interpolate8x8_avg2(halfpel_v, src, halfpel_h, stride, rounding, 9); - interpolate8x8_avg2(halfpel_v+8, src + 8, halfpel_h+8, stride, rounding, 9); - interpolate8x8_avg2(halfpel_v+8*stride, src + 8*stride, halfpel_h+8*stride, stride, rounding, 9); - interpolate8x8_avg2(halfpel_v+8*stride+8, src+8*stride+8, halfpel_h+8*stride+8, stride, rounding, 9); - - interpolate16x16_lowpass_v(halfpel_hv, halfpel_v, stride, rounding); - interpolate8x8_avg2(dst, halfpel_hv, halfpel_v, stride, rounding, 8); - interpolate8x8_avg2(dst+8, halfpel_hv+8, halfpel_v+8, stride, rounding, 8); - interpolate8x8_avg2(dst+8*stride, halfpel_hv+8*stride, halfpel_v+8*stride, stride, rounding, 8); - interpolate8x8_avg2(dst+8*stride+8, halfpel_hv+8*stride+8, halfpel_v+8*stride+8, stride, rounding, 8); - break; - - case 6: - interpolate16x16_lowpass_hv(halfpel_hv, halfpel_h, src, stride, rounding); - interpolate8x8_avg2(dst, halfpel_h, halfpel_hv, stride, rounding, 8); - interpolate8x8_avg2(dst+8, halfpel_h+8, halfpel_hv+8, stride, rounding, 8); - interpolate8x8_avg2(dst+8*stride, halfpel_h+8*stride, halfpel_hv+8*stride, stride, rounding, 8); - interpolate8x8_avg2(dst+8*stride+8, halfpel_h+8*stride+8, halfpel_hv+8*stride+8, stride, rounding, 8); - break; - - case 7: - interpolate16x16_lowpass_h(halfpel_h, src, stride, rounding); - interpolate8x8_avg2(halfpel_v, src+1, halfpel_h, stride, rounding, 9); - interpolate8x8_avg2(halfpel_v+8, src+1 + 8, halfpel_h+8, stride, rounding, 9); - interpolate8x8_avg2(halfpel_v+8*stride, src+1 + 8*stride, halfpel_h+8*stride, stride, rounding, 9); - interpolate8x8_avg2(halfpel_v+8*stride+8, src+1+8*stride+8, halfpel_h+8*stride+8, stride, rounding, 9); - - interpolate16x16_lowpass_v(halfpel_hv, halfpel_v, stride, rounding); - interpolate8x8_avg2(dst, halfpel_hv, halfpel_v, stride, rounding, 8); - interpolate8x8_avg2(dst+8, halfpel_hv+8, halfpel_v+8, stride, rounding, 8); - interpolate8x8_avg2(dst+8*stride, halfpel_hv+8*stride, halfpel_v+8*stride, stride, rounding, 8); - interpolate8x8_avg2(dst+8*stride+8, halfpel_hv+8*stride+8, halfpel_v+8*stride+8, stride, rounding, 8); - break; - - case 8: - interpolate16x16_lowpass_v(dst, src, stride, rounding); - break; - - case 9: - interpolate16x16_lowpass_h(halfpel_h, src, stride, rounding); - interpolate8x8_avg2(halfpel_v, src, halfpel_h, stride, rounding, 9); - interpolate8x8_avg2(halfpel_v+8, src + 8, halfpel_h+8, stride, rounding, 9); - interpolate8x8_avg2(halfpel_v+8*stride, src + 8*stride, halfpel_h+8*stride, stride, rounding, 9); - interpolate8x8_avg2(halfpel_v+8*stride+8, src+8*stride+8, halfpel_h+8*stride+8, stride, rounding, 9); - interpolate16x16_lowpass_v(dst, halfpel_v, stride, rounding); - break; - - case 10: - interpolate16x16_lowpass_hv(dst, halfpel_h, src, stride, rounding); - break; - - case 11: - interpolate16x16_lowpass_h(halfpel_h, src, stride, rounding); - interpolate8x8_avg2(halfpel_v, src+1, halfpel_h, stride, rounding, 9); - interpolate8x8_avg2(halfpel_v+8, src+1 + 8, halfpel_h+8, stride, rounding, 9); - interpolate8x8_avg2(halfpel_v+8*stride, src+1 + 8*stride, halfpel_h+8*stride, stride, rounding, 9); - interpolate8x8_avg2(halfpel_v+8*stride+8, src+1+8*stride+8, halfpel_h+8*stride+8, stride, rounding, 9); - interpolate16x16_lowpass_v(dst, halfpel_v, stride, rounding); - break; - - case 12: - interpolate16x16_lowpass_v(halfpel_v, src, stride, rounding); - interpolate8x8_avg2(dst, src+stride, halfpel_v, stride, rounding, 8); - interpolate8x8_avg2(dst+8, src+stride+8, halfpel_v+8, stride, rounding, 8); - interpolate8x8_avg2(dst+8*stride, src+stride+8*stride, halfpel_v+8*stride, stride, rounding, 8); - interpolate8x8_avg2(dst+8*stride+8, src+stride+8*stride+8, halfpel_v+8*stride+8, stride, rounding, 8); - break; - - case 13: - interpolate16x16_lowpass_h(halfpel_h, src, stride, rounding); - interpolate8x8_avg2(halfpel_v, src, halfpel_h, stride, rounding, 9); - interpolate8x8_avg2(halfpel_v+8, src + 8, halfpel_h+8, stride, rounding, 9); - interpolate8x8_avg2(halfpel_v+8*stride, src + 8*stride, halfpel_h+8*stride, stride, rounding, 9); - interpolate8x8_avg2(halfpel_v+8*stride+8, src+8*stride+8, halfpel_h+8*stride+8, stride, rounding, 9); - - interpolate16x16_lowpass_v(halfpel_hv, halfpel_v, stride, rounding); - interpolate8x8_avg2(dst, halfpel_hv, halfpel_v+stride, stride, rounding, 8); - interpolate8x8_avg2(dst+8, halfpel_hv+8, halfpel_v+stride+8, stride, rounding, 8); - interpolate8x8_avg2(dst+8*stride, halfpel_hv+8*stride, halfpel_v+stride+8*stride, stride, rounding, 8); - interpolate8x8_avg2(dst+8*stride+8, halfpel_hv+8*stride+8, halfpel_v+stride+8*stride+8, stride, rounding, 8); - break; - - case 14: - interpolate16x16_lowpass_hv(halfpel_hv, halfpel_h, src, stride, rounding); - interpolate8x8_avg2(dst, halfpel_h+stride, halfpel_hv, stride, rounding, 8); - interpolate8x8_avg2(dst+8, halfpel_h+stride+8, halfpel_hv+8, stride, rounding, 8); - interpolate8x8_avg2(dst+8*stride, halfpel_h+stride+8*stride, halfpel_hv+8*stride, stride, rounding, 8); - interpolate8x8_avg2(dst+8*stride+8, halfpel_h+stride+8*stride+8, halfpel_hv+8*stride+8, stride, rounding, 8); - break; - - case 15: - interpolate16x16_lowpass_h(halfpel_h, src, stride, rounding); - interpolate8x8_avg2(halfpel_v, src+1, halfpel_h, stride, rounding, 9); - interpolate8x8_avg2(halfpel_v+8, src+1 + 8, halfpel_h+8, stride, rounding, 9); - interpolate8x8_avg2(halfpel_v+8*stride, src+1 + 8*stride, halfpel_h+8*stride, stride, rounding, 9); - interpolate8x8_avg2(halfpel_v+8*stride+8, src+1+8*stride+8, halfpel_h+8*stride+8, stride, rounding, 9); - - interpolate16x16_lowpass_v(halfpel_hv, halfpel_v, stride, rounding); - interpolate8x8_avg2(dst, halfpel_hv, halfpel_v+stride, stride, rounding, 8); - interpolate8x8_avg2(dst+8, halfpel_hv+8, halfpel_v+stride+8, stride, rounding, 8); - interpolate8x8_avg2(dst+8*stride, halfpel_hv+8*stride, halfpel_v+stride+8*stride, stride, rounding, 8); - interpolate8x8_avg2(dst+8*stride+8, halfpel_hv+8*stride+8, halfpel_v+stride+8*stride+8, stride, rounding, 8); + interpolate8x8_halfpel_h(buffer, src, stride, rounding); + break; + default: + interpolate8x8_halfpel_hv(buffer, src, stride, rounding); break; } + return buffer; } #endif