?? deblock_vert_lpf9.c
字號:
#include "postprocess_mmx.h"
const static uint64_t mm_fours = 0x0004000400040004;
INLINE void deblock_vert_lpf9(uint64_t *v_local, uint64_t *p1p2, uint8_t *v, int stride) {
#ifdef PP_SELF_CHECK
int j, k;
uint8_t selfcheck[64], *vv;
int p1, p2, psum;
int l1 = 1 * stride;
int l2 = 2 * stride;
int l3 = 3 * stride;
int l4 = 4 * stride;
int l5 = 5 * stride;
int l6 = 6 * stride;
int l7 = 7 * stride;
int l8 = 8 * stride;
#endif
#ifdef PP_SELF_CHECK
for (j=0; j<8; j++) {
vv = &(v[j]);
p1 = ((uint16_t *)(&(p1p2[0+j/4])))[j%4];
p2 = ((uint16_t *)(&(p1p2[2+j/4])))[j%4];
psum = p1 + p1 + p1 + vv[l1] + vv[l2] + vv[l3] + vv[l4] + 4;
selfcheck[j+8*0] = (((psum + vv[l1]) << 1) - (vv[l4] - vv[l5])) >> 4;
psum += vv[l5] - p1;
selfcheck[j+8*1] = (((psum + vv[l2]) << 1) - (vv[l5] - vv[l6])) >> 4;
psum += vv[l6] - p1;
selfcheck[j+8*2] = (((psum + vv[l3]) << 1) - (vv[l6] - vv[l7])) >> 4;
psum += vv[l7] - p1;
selfcheck[j+8*3] = (((psum + vv[l4]) << 1) + p1 - vv[l1] - (vv[l7] - vv[l8])) >> 4;
psum += vv[l8] - vv[l1];
selfcheck[j+8*4] = (((psum + vv[l5]) << 1) + (vv[l1] - vv[l2]) - vv[l8] + p2) >> 4;
psum += p2 - vv[l2];
selfcheck[j+8*5] = (((psum + vv[l6]) << 1) + (vv[l2] - vv[l3])) >> 4;
psum += p2 - vv[l3];
selfcheck[j+8*6] = (((psum + vv[l7]) << 1) + (vv[l3] - vv[l4])) >> 4;
psum += p2 - vv[l4];
selfcheck[j+8*7] = (((psum + vv[l8]) << 1) + (vv[l4] - vv[l5])) >> 4;
}
#endif
__asm {
push eax
push ebx
push ecx
mov eax, p1p2
mov ebx, v_local
mov ecx, v
movq mm2, [eax] /* mm2 = p1p2[0] 0 1 2w3 4 5 6 7 */
add ecx, stride /* ecx points at v[1*stride] 0 1 2 3 4 5 6 7 */
movq mm3, 8[eax] /* mm3 = p1p2[1] 0 1 2 3w4 5 6 7 */
movq mm4, mm_fours /* mm4 = 0x0004000400040004 0 1 2 3 4w5 6 7 */
*/
movq mm5, mm4 /* mm5 = 0x0004000400040004 0 1 2 3 4 5w6 7 */
paddsw mm4, 16[ebx] /* mm4 += vv[1] left 0 1 2 3 4m5 6 7 */
paddw mm5, mm3 /* mm5 += p2 left 0 1 2 3r4 5m6 7 */
paddsw mm4, 32[ebx] /* mm4 += vv[2] left 0 1 2 3 4m5 6 7 */
paddw mm5, mm3 /* mm5 += p2 left 0 1 2 3r4 5m6 7 */
paddsw mm4, 48[ebx] /* mm4 += vv[3] left 0 1 2 3 4m5 6 7 */
paddw mm5, mm3 /* mm5 += p2 left 0 1 2 3r4 5m6 7 */
paddsw mm5, 24[ebx] /* mm5 += vv[1] right 0 1 2 3 4 5m6 7 */
paddw mm4, mm2 /* mm4 += p1 left 0 1 2r3 4m5 6 7 */
paddsw mm5, 40[ebx] /* mm5 += vv[2] right 0 1 2 3 4 5m6 7 */
paddw mm4, mm2 /* mm4 += p1 left 0 1 2r3 4m5 6 7 */
paddsw mm5, 56[ebx] /* mm5 += vv[3] right 0 1 2 3 4 5m6 7 */
paddw mm4, mm2 /* mm4 += p1 left 0 1 2r3 4m5 6 7 */
paddsw mm4, 64[ebx] /* mm4 += vv[4] left 0 1 2 3 4m5 6 7 */
paddsw mm5, 72[ebx] /* mm5 += vv[4] right 0 1 2 3 4 5m6 7 */
movq mm0, mm4 /* mm0 = psum left 0w1 2 3 4 5 6 7 */
paddsw mm0, 16[ebx] /* mm0 += vv[1] left 0m1 2 3 4 5 6 7 */
movq mm1, mm5 /* mm1 = psum right 0 1w2 3 4 5r6 7 */
paddsw mm1, 24[ebx] /* mm1 += vv[1] right 0 1 2 3 4 5 6 7 */
psllw mm0, 1 /* mm0 <<= 1 0m1 2 3 4 5 6 7 */
psubsw mm0, 64[ebx] /* mm0 -= vv[4] left 0m1 2 3 4 5 6 7 */
psllw mm1, 1 /* mm1 <<= 1 0 1 2 3 4 5 6 7 */
psubsw mm1, 72[ebx] /* mm1 -= vv[4] right 0 1m2 3 4 5 6 7 */
paddsw mm0, 80[ebx] /* mm0 += vv[5] left 0m1 2 3 4 5 6 7 */
paddsw mm1, 88[ebx] /* mm1 += vv[5] right 0 1m2 3 4 5 6 7 */
psrlw mm0, 4 /* mm0 >>= 4 0m1 2 3 4 5 6 7 */
paddsw mm4, 80[ebx] /* mm4 += vv[5] left 0 1 2 3 4m5 6 7 */
psrlw mm1, 4 /* mm1 >>= 4 0 1m2 3 4 5 6 7 */
paddsw mm5, 88[ebx] /* mm5 += vv[5] right 0 1 2 3 4 5 6 7 */
psubsw mm4, [eax] /* mm4 -= p1 left 0 1 2 3 4 5 6 7 */
packuswb mm0, mm1 /* pack mm1, mm0 to mm0 0m1 2 3 4 5 6 7 */
psubsw mm5, 8[eax] /* mm5 -= p1 right 0 1 2 3 4 5 6 7 */
movq mm6, mm4 /* mm6 = psum left 0 1 2 3 4 5 6 7 */
paddsw mm6, 32[ebx] /* mm6 += vv[2] left 0 1 2 3 4 5 6 7 */
movq mm7, mm5 /* mm7 = psum right 0 1 2 3 4 5 6 7 */
paddsw mm7, 40[ebx] /* mm7 += vv[2] right 0 1 2 3 4 5 6 7 */
psllw mm6, 1 /* mm6 <<= 1 0 1 2 3 4 5 6 7 */
psubsw mm6, 80[ebx] /* mm6 -= vv[5] left 0 1 2 3 4 5 6 7 */
psllw mm7, 1 /* mm7 <<= 1 0 1 2 3 4 5 6 7 */
psubsw mm7, 88[ebx] /* mm7 -= vv[5] right 0 1 2 3 4 5 6 7 */
movq [ecx], mm0 /* v[1*stride] = mm0 0 1 2 3 4 5 6 7 */
paddsw mm6, 96[ebx] /* mm6 += vv[6] left 0 1 2 3 4 5 6 7 */
add ecx, stride /* ecx points at v[2*stride] 0 1 2 3 4 5 6 7 */
paddsw mm7, 104[ebx] /* mm7 += vv[6] right 0 1 2 3 4 5 6 7 */
paddsw mm4, 96[ebx] /* mm4 += vv[6] left 0 1 2 3 4 5 6 7 */
psrlw mm6, 4 /* mm6 >>= 4 0 1 2 3 4 5 6 7 */
paddsw mm5, 104[ebx] /* mm5 += vv[6] right 0 1 2 3 4 5 6 7 */
psrlw mm7, 4 /* mm7 >>= 4 0 1 2 3 4 5 6 7 */
psubsw mm4, [eax] /* mm4 -= p1 left 0 1 2 3 4 5 6 7 */
packuswb mm6, mm7 /* pack mm7, mm6 to mm6 0 1 2 3 4 5 6 7 */
psubsw mm5, 8[eax] /* mm5 -= p1 right 0 1 2 3 4 5 6 7 */
movq mm0, mm4 /* mm0 = psum left 0 1 2 3 4 5 6 7 */
paddsw mm0, 48[ebx] /* mm0 += vv[3] left 0 1 2 3 4 5 6 7 */
movq mm1, mm5 /* mm1 = psum right 0 1 2 3 4 5 6 7 */
paddsw mm1, 56[ebx] /* mm1 += vv[3] right 0 1 2 3 4 5 6 7 */
psllw mm0, 1 /* mm0 <<= 1 0 1 2 3 4 5 6 7 */
psubsw mm0, 96[ebx] /* mm0 -= vv[6] left 0 1 2 3 4 5 6 7 */
psllw mm1, 1 /* mm1 <<= 1 0 1 2 3 4 5 6 7 */
psubsw mm1, 104[ebx] /* mm1 -= vv[6] right 0 1 2 3 4 5 6 7 */
movq [ecx], mm6 /* v[2*stride] = mm6 0 1 2 3 4 5 6 7 */
paddsw mm0, 112[ebx] /* mm0 += vv[7] left 0 1 2 3 4 5 6 7 */
paddsw mm1, 120[ebx] /* mm1 += vv[7] right 0 1 2 3 4 5 6 7 */
add ecx, stride /* ecx points at v[3*stride] 0 1 2 3 4 5 6 7 */
paddsw mm4, 112[ebx] /* mm4 += vv[5] left 0 1 2 3 4 5 6 7 */
psrlw mm0, 4 /* mm0 >>= 4 0 1 2 3 4 5 6 7 */
paddsw mm5, 120[ebx] /* mm5 += vv[5] right 0 1 2 3 4 5 6 7 */
psrlw mm1, 4 /* mm1 >>= 4 0 1 2 3 4 5 6 7 */
psubsw mm4, [eax] /* mm4 -= p1 left 0 1 2 3 4 5 6 7 */
packuswb mm0, mm1 /* pack mm1, mm0 to mm0 0 1 2 3 4 5 6 7 */
psubsw mm5, 8[eax] /* mm5 -= p1 right 0 1 2 3 4 5 6 7 */
movq [ecx], mm0 /* v[3*stride] = mm0 0 1 2 3 4 5 6 7 */
movq mm6, mm4 /* mm6 = psum left 0 1 2 3 4 5 6 7 */
paddsw mm6, 64[ebx] /* mm6 += vv[4] left 0 1 2 3 4 5 6 7 */
?? 快捷鍵說明
復制代碼
Ctrl + C
搜索代碼
Ctrl + F
全屏模式
F11
切換主題
Ctrl + Shift + D
顯示快捷鍵
?
增大字號
Ctrl + =
減小字號
Ctrl + -