diff options
Diffstat (limited to 'libavcodec/x86/dsputil_mmx.c')
| -rw-r--r-- | libavcodec/x86/dsputil_mmx.c | 123 |
1 files changed, 0 insertions, 123 deletions
diff --git a/libavcodec/x86/dsputil_mmx.c b/libavcodec/x86/dsputil_mmx.c index 310039957f..19c03d8586 100644 --- a/libavcodec/x86/dsputil_mmx.c +++ b/libavcodec/x86/dsputil_mmx.c @@ -247,127 +247,4 @@ void ff_draw_edges_mmx(uint8_t *buf, int wrap, int width, int height, } } -#if CONFIG_VIDEODSP -void ff_gmc_mmx(uint8_t *dst, uint8_t *src, - int stride, int h, int ox, int oy, - int dxx, int dxy, int dyx, int dyy, - int shift, int r, int width, int height) -{ - const int w = 8; - const int ix = ox >> (16 + shift); - const int iy = oy >> (16 + shift); - const int oxs = ox >> 4; - const int oys = oy >> 4; - const int dxxs = dxx >> 4; - const int dxys = dxy >> 4; - const int dyxs = dyx >> 4; - const int dyys = dyy >> 4; - const uint16_t r4[4] = { r, r, r, r }; - const uint16_t dxy4[4] = { dxys, dxys, dxys, dxys }; - const uint16_t dyy4[4] = { dyys, dyys, dyys, dyys }; - const uint64_t shift2 = 2 * shift; -#define MAX_STRIDE 4096U -#define MAX_H 8U - uint8_t edge_buf[(MAX_H + 1) * MAX_STRIDE]; - int x, y; - - const int dxw = (dxx - (1 << (16 + shift))) * (w - 1); - const int dyh = (dyy - (1 << (16 + shift))) * (h - 1); - const int dxh = dxy * (h - 1); - const int dyw = dyx * (w - 1); - int need_emu = (unsigned) ix >= width - w || - (unsigned) iy >= height - h; - - if ( // non-constant fullpel offset (3% of blocks) - ((ox ^ (ox + dxw)) | (ox ^ (ox + dxh)) | (ox ^ (ox + dxw + dxh)) | - (oy ^ (oy + dyw)) | (oy ^ (oy + dyh)) | (oy ^ (oy + dyw + dyh))) >> (16 + shift) || - // uses more than 16 bits of subpel mv (only at huge resolution) - (dxx | dxy | dyx | dyy) & 15 || - (need_emu && (h > MAX_H || stride > MAX_STRIDE))) { - // FIXME could still use mmx for some of the rows - ff_gmc_c(dst, src, stride, h, ox, oy, dxx, dxy, dyx, dyy, - shift, r, width, height); - return; - } - - src += ix + iy * stride; - if (need_emu) { - ff_emulated_edge_mc_8(edge_buf, src, stride, stride, w + 1, h + 1, ix, iy, width, height); - src = edge_buf; - } - - __asm__ volatile ( - "movd %0, %%mm6 \n\t" - "pxor %%mm7, %%mm7 \n\t" - "punpcklwd %%mm6, %%mm6 \n\t" - "punpcklwd %%mm6, %%mm6 \n\t" - :: "r" (1 << shift)); - - for (x = 0; x < w; x += 4) { - uint16_t dx4[4] = { oxs - dxys + dxxs * (x + 0), - oxs - dxys + dxxs * (x + 1), - oxs - dxys + dxxs * (x + 2), - oxs - dxys + dxxs * (x + 3) }; - uint16_t dy4[4] = { oys - dyys + dyxs * (x + 0), - oys - dyys + dyxs * (x + 1), - oys - dyys + dyxs * (x + 2), - oys - dyys + dyxs * (x + 3) }; - - for (y = 0; y < h; y++) { - __asm__ volatile ( - "movq %0, %%mm4 \n\t" - "movq %1, %%mm5 \n\t" - "paddw %2, %%mm4 \n\t" - "paddw %3, %%mm5 \n\t" - "movq %%mm4, %0 \n\t" - "movq %%mm5, %1 \n\t" - "psrlw $12, %%mm4 \n\t" - "psrlw $12, %%mm5 \n\t" - : "+m" (*dx4), "+m" (*dy4) - : "m" (*dxy4), "m" (*dyy4)); - - __asm__ volatile ( - "movq %%mm6, %%mm2 \n\t" - "movq %%mm6, %%mm1 \n\t" - "psubw %%mm4, %%mm2 \n\t" - "psubw %%mm5, %%mm1 \n\t" - "movq %%mm2, %%mm0 \n\t" - "movq %%mm4, %%mm3 \n\t" - "pmullw %%mm1, %%mm0 \n\t" // (s - dx) * (s - dy) - "pmullw %%mm5, %%mm3 \n\t" // dx * dy - "pmullw %%mm5, %%mm2 \n\t" // (s - dx) * dy - "pmullw %%mm4, %%mm1 \n\t" // dx * (s - dy) - - "movd %4, %%mm5 \n\t" - "movd %3, %%mm4 \n\t" - "punpcklbw %%mm7, %%mm5 \n\t" - "punpcklbw %%mm7, %%mm4 \n\t" - "pmullw %%mm5, %%mm3 \n\t" // src[1, 1] * dx * dy - "pmullw %%mm4, %%mm2 \n\t" // src[0, 1] * (s - dx) * dy - - "movd %2, %%mm5 \n\t" - "movd %1, %%mm4 \n\t" - "punpcklbw %%mm7, %%mm5 \n\t" - "punpcklbw %%mm7, %%mm4 \n\t" - "pmullw %%mm5, %%mm1 \n\t" // src[1, 0] * dx * (s - dy) - "pmullw %%mm4, %%mm0 \n\t" // src[0, 0] * (s - dx) * (s - dy) - "paddw %5, %%mm1 \n\t" - "paddw %%mm3, %%mm2 \n\t" - "paddw %%mm1, %%mm0 \n\t" - "paddw %%mm2, %%mm0 \n\t" - - "psrlw %6, %%mm0 \n\t" - "packuswb %%mm0, %%mm0 \n\t" - "movd %%mm0, %0 \n\t" - - : "=m" (dst[x + y * stride]) - : "m" (src[0]), "m" (src[1]), - "m" (src[stride]), "m" (src[stride + 1]), - "m" (*r4), "m" (shift2)); - src += stride; - } - src += 4 - h * stride; - } -} -#endif #endif /* HAVE_INLINE_ASM */ |