void mlib_v_ImageAffineTableLine_8nw_3_2_1( mlib_d64 *buff, const mlib_d64 *filterX, const mlib_d64 *filterY, const mlib_u8 **lineAddr, mlib_affine_workspace *ws) { DECLAREVAR; DECLAREVAR2; mlib_d64 yFilter2; mlib_d64 yFilter3; mlib_d64 row20, row30; mlib_d64 *dpSrc; mlib_d64 data0, data1, zero; vis_write_gsr64((((mlib_u64)0x0145ABEF) << 32) + 4); dstPixelPtr = (mlib_s16 *)buff; zero = vis_to_double_dup(0); #pragma pipeloop(0) for (i = 0; i <= size - 2; i += 2) { CALC_2_SRC_PTR; LOAD_3x2; FILTER_MERGE_4x2; MAKE_4x2; *buff1 = res1; buff1++; } dstPixelPtr = (mlib_s16 *)buff1; for (; i < size; i++) { CALC_SRC_PTR(sPtr); LOAD_FILTERS(fx0, yFilter); xFilter = vis_write_hi(xFilter, fx0); LOAD_PIXEL_3; v0 = vis_fmul8x16au(vis_read_hi(row00), vis_read_hi(yFilter)); v1 = vis_fmul8x16al(vis_read_hi(row10), vis_read_hi(yFilter)); sum = vis_fpadd16(v0, v1); v0 = vis_fmul8x16au(vis_read_hi(row20), vis_read_lo(yFilter)); sum = vis_fpadd16(v0, sum); v0 = vis_fmul8sux16(sum, xFilter); v1 = vis_fmul8ulx16(sum, xFilter); v3 = vis_fpadd16(v1, v0); v2 = vis_fmuld8ulx16(vis_scale, vis_read_hi(v3)); res = vis_write_lo(res, vis_fpadd32s(vis_read_hi(v2), vis_read_lo(v2))); vis_st_u16(res, dstPixelPtr++); } }
mlib_status __mlib_VideoDownSample422( mlib_u8 *dst, const mlib_u8 *src, mlib_s32 n) { mlib_d64 *sp0 = (mlib_d64 *)src; mlib_f32 *pd = (mlib_f32 *)dst; mlib_d64 d0; mlib_d64 tmp0, tmp1, data; mlib_d64 acc0_hi, acc0_lo; mlib_d64 round = vis_to_double_dup(0x1); mlib_f32 fone = vis_to_float(0x1000000); mlib_s32 i, bias = 0; if (n <= 0) return (MLIB_FAILURE); vis_write_gsr(6 << 3); #pragma pipeloop(0) for (i = 0; i <= n - 8; i += 8) { d0 = (*sp0++); tmp0 = vis_fpmerge(vis_read_hi(d0), vis_read_lo(d0)); tmp1 = vis_fpmerge(vis_read_hi(tmp0), vis_read_lo(tmp0)); acc0_hi = vis_fmul8x16au(vis_read_hi(tmp1), fone); acc0_lo = vis_fmul8x16au(vis_read_lo(tmp1), fone); acc0_hi = vis_fpadd16(acc0_hi, acc0_lo); data = vis_fpadd16(acc0_hi, round); (*pd++) = vis_fpack16(data); } dst = (mlib_u8 *)pd; for (; i < n; i += 2) { (*dst++) = (src[i] + src[i + 1] + bias) >> 1; /* 1=>2, 2=>1 */ bias ^= 1; } return (MLIB_SUCCESS); }
void mlib_v_ImageAffineTableLine_8nw_2_2_1( mlib_d64 *buff, const mlib_d64 *filterX, const mlib_d64 *filterY, const mlib_u8 **lineAddr, mlib_affine_workspace *ws) { DECLAREVAR; DECLAREVAR2; vis_write_gsr64((((mlib_u64)0x0145ABEF) << 32) + 4); dstPixelPtr = (mlib_s16 *)buff; #pragma pipeloop(0) for (i = 0; i <= size - 2; i += 2) { CALC_2_SRC_PTR; LOAD_2x2(row00, row10); FILTER_MERGE; MAKE_2x2; *buff1 = res1; buff1++; } dstPixelPtr = (mlib_s16 *)buff1; #pragma pipeloop(0) for (; i < size; i++) { CALC_SRC_PTR(sPtr); LOAD_FILTERS(fx0, fy0); xFilter = vis_write_lo(xFilter, fx0); row00 = vis_fpmerge(LD_U8(sPtr, 0), LD_U8(sPtr, 1)); row10 = vis_fpmerge(LD_U8(sPtr, srcStride), LD_U8(sPtr, srcStride + 1)); v0 = vis_fmul8x16au(vis_read_lo(row00), fy0); v1 = vis_fmul8x16al(vis_read_lo(row10), fy0); sum = vis_fpadd16(v0, v1); v0 = vis_fmul8sux16(sum, xFilter); v1 = vis_fmul8ulx16(sum, xFilter); v3 = vis_fpadd16(v1, v0); v2 = vis_fmuld8ulx16(vis_scale, vis_read_lo(v3)); res = vis_write_lo(res, vis_fpadd32s(vis_read_hi(v2), vis_read_lo(v2))); vis_st_u16(res, dstPixelPtr++); } }
static mlib_status mlib_v_VideoColorYUV2RGB444_nonalign( mlib_u8 *rgb, const mlib_u8 *y, const mlib_u8 *u, const mlib_u8 *v, mlib_s32 width, mlib_s32 height, mlib_s32 rgb_stride, mlib_s32 yuv_stride) { /* all. pointer to y, u, v */ mlib_d64 *spy, *dfu, *dfv; /* y data */ mlib_d64 dy0, dy1, dy3; mlib_d64 du, dv, du0, du1, dv0, dv1; /* (1.1644, 1.5966)*8192 */ mlib_f32 k12 = vis_to_float(0x25433317); /* (-.3920, -.8132)*8192 */ mlib_f32 k34 = vis_to_float(0xf375e5fa); /* 2.0184*8192 */ mlib_f32 k5 = vis_to_float(0x1004097); mlib_d64 k_222_9952 = vis_to_double_dup(0x1be01be0); mlib_d64 k_135_6352 = vis_to_double_dup(0x10f410f4); mlib_d64 k_276_9856 = vis_to_double_dup(0x22a022a0); mlib_d64 u_3920_hi, u_20184_hi, v_15966_hi, v_8132_hi; mlib_d64 u_3920_lo, u_20184_lo, v_15966_lo, v_8132_lo; mlib_d64 y_11644_hi, y_11644_lo; mlib_d64 r_hi, r_lo, g_hi, g_lo, b_hi, b_lo; mlib_d64 red, green, blue, *ddp, dd0, dd1, dd2; /* loop variable */ mlib_s32 i, j; mlib_d64 *buf, BUFF[16 * 1024]; mlib_u8 *tmp, *dp; if (width * 3 > 16 * 1024) { tmp = __mlib_malloc(width * 3 * sizeof (mlib_u8) + 7); if (tmp == NULL) return (MLIB_FAILURE); buf = (mlib_d64 *)((mlib_addr)(tmp + 7) & ~7); } else { buf = (mlib_d64 *)BUFF; } dp = (mlib_u8 *)buf; ddp = (mlib_d64 *)dp; for (j = 0; j < height; j++) { dfu = (mlib_d64 *)vis_alignaddr((void *)u, 0); du0 = (*dfu++); du1 = vis_ld_d64_nf(dfu); dfu++; du = vis_faligndata(du0, du1); du0 = du1; dfv = (mlib_d64 *)vis_alignaddr((void *)v, 0); dv0 = (*dfv++); dv1 = vis_ld_d64_nf(dfv); dfv++; dv = vis_faligndata(dv0, dv1); dv0 = dv1; /* U*(-0.3920); */ u_3920_hi = vis_fmul8x16au(vis_read_hi(du), k34); /* V*(-0.8132); */ v_8132_hi = vis_fmul8x16al(vis_read_hi(dv), k34); /* U*(-0.3920); */ u_3920_lo = vis_fmul8x16au(vis_read_lo(du), k34); /* V*(-0.8132); */ v_8132_lo = vis_fmul8x16al(vis_read_lo(dv), k34); spy = (mlib_d64 *)vis_alignaddr((void *)y, 0); dy0 = (*spy++); dy3 = vis_ld_d64_nf(spy); spy++; dy1 = vis_faligndata(dy0, dy3); dy0 = dy3; /* U*2.0184 */ u_20184_hi = vis_fmul8x16al(vis_read_hi(du), k5); g_hi = vis_fpadd16(u_3920_hi, v_8132_hi); u_20184_lo = vis_fmul8x16al(vis_read_lo(du), k5); g_hi = vis_fpadd16(g_hi, k_135_6352); /* V*1.5966 */ v_15966_hi = vis_fmul8x16al(vis_read_hi(dv), k12); g_lo = vis_fpadd16(u_3920_lo, v_8132_lo); v_15966_lo = vis_fmul8x16al(vis_read_lo(dv), k12); g_lo = vis_fpadd16(g_lo, k_135_6352); vis_alignaddr((void *)u, 0); du1 = vis_ld_d64_nf(dfu); dfu++; du = vis_faligndata(du0, du1); du0 = du1; /* Y*1.1644 */ y_11644_hi = vis_fmul8x16au(vis_read_hi(dy1), k12); b_hi = vis_fpsub16(u_20184_hi, k_276_9856); vis_alignaddr((void *)v, 0); dv1 = vis_ld_d64_nf(dfv); dfv++; dv = vis_faligndata(dv0, dv1); dv0 = dv1; /* Y*1.1644 */ y_11644_lo = vis_fmul8x16au(vis_read_lo(dy1), k12); b_lo = vis_fpsub16(u_20184_lo, k_276_9856); /* U*(-0.3920); */ u_3920_hi = vis_fmul8x16au(vis_read_hi(du), k34); r_hi = vis_fpsub16(v_15966_hi, k_222_9952); /* V*(-0.8132); */ v_8132_hi = vis_fmul8x16al(vis_read_hi(dv), k34); r_lo = vis_fpsub16(v_15966_lo, k_222_9952); u_3920_lo = vis_fmul8x16au(vis_read_lo(du), k34); g_hi = vis_fpadd16(g_hi, y_11644_hi); v_8132_lo = vis_fmul8x16al(vis_read_lo(dv), k34); g_lo = vis_fpadd16(g_lo, y_11644_lo); green = vis_fpack16_pair(g_hi, g_lo); b_hi = vis_fpadd16(b_hi, y_11644_hi); b_lo = vis_fpadd16(b_lo, y_11644_lo); blue = vis_fpack16_pair(b_hi, b_lo); r_hi = vis_fpadd16(r_hi, y_11644_hi); r_lo = vis_fpadd16(r_lo, y_11644_lo); red = vis_fpack16_pair(r_hi, r_lo); vis_alignaddr((void *)y, 0); dy3 = vis_ld_d64_nf(spy); spy++; dy1 = vis_faligndata(dy0, dy3); dy0 = dy3; #pragma pipeloop(0) for (i = 0; i <= width - 8; i += 8) { vis_write_bmask(0x0801902A, 0); dd0 = vis_bshuffle(red, green); vis_write_bmask(0x03B04C05, 0); dd1 = vis_bshuffle(red, green); vis_write_bmask(0xD06E07F0, 0); dd2 = vis_bshuffle(red, green); vis_write_bmask(0x01834967, 0); ddp[0] = vis_bshuffle(dd0, blue); vis_write_bmask(0xA12B45C7, 0); ddp[1] = vis_bshuffle(dd1, blue); vis_write_bmask(0x0D23E56F, 0); ddp[2] = vis_bshuffle(dd2, blue); /* U*2.0184 */ u_20184_hi = vis_fmul8x16al(vis_read_hi(du), k5); g_hi = vis_fpadd16(u_3920_hi, v_8132_hi); u_20184_lo = vis_fmul8x16al(vis_read_lo(du), k5); g_hi = vis_fpadd16(g_hi, k_135_6352); /* V*1.5966 */ v_15966_hi = vis_fmul8x16al(vis_read_hi(dv), k12); g_lo = vis_fpadd16(u_3920_lo, v_8132_lo); v_15966_lo = vis_fmul8x16al(vis_read_lo(dv), k12); g_lo = vis_fpadd16(g_lo, k_135_6352); vis_alignaddr((void *)u, 0); du1 = vis_ld_d64_nf(dfu); dfu++; du = vis_faligndata(du0, du1); du0 = du1; /* Y*1.1644 */ y_11644_hi = vis_fmul8x16au(vis_read_hi(dy1), k12); b_hi = vis_fpsub16(u_20184_hi, k_276_9856); vis_alignaddr((void *)v, 0); dv1 = vis_ld_d64_nf(dfv); dfv++; dv = vis_faligndata(dv0, dv1); dv0 = dv1; /* Y*1.1644 */ y_11644_lo = vis_fmul8x16au(vis_read_lo(dy1), k12); b_lo = vis_fpsub16(u_20184_lo, k_276_9856); /* U*(-0.3920); */ u_3920_hi = vis_fmul8x16au(vis_read_hi(du), k34); r_hi = vis_fpsub16(v_15966_hi, k_222_9952); /* V*(-0.8132); */ v_8132_hi = vis_fmul8x16al(vis_read_hi(dv), k34); r_lo = vis_fpsub16(v_15966_lo, k_222_9952); u_3920_lo = vis_fmul8x16au(vis_read_lo(du), k34); g_hi = vis_fpadd16(g_hi, y_11644_hi); v_8132_lo = vis_fmul8x16al(vis_read_lo(dv), k34); g_lo = vis_fpadd16(g_lo, y_11644_lo); green = vis_fpack16_pair(g_hi, g_lo); b_hi = vis_fpadd16(b_hi, y_11644_hi); b_lo = vis_fpadd16(b_lo, y_11644_lo); blue = vis_fpack16_pair(b_hi, b_lo); r_hi = vis_fpadd16(r_hi, y_11644_hi); r_lo = vis_fpadd16(r_lo, y_11644_lo); red = vis_fpack16_pair(r_hi, r_lo); vis_alignaddr((void *)y, 0); dy3 = vis_ld_d64_nf(spy); spy++; dy1 = vis_faligndata(dy0, dy3); dy0 = dy3; ddp += 3; } dp = (mlib_u8 *)ddp; vis_alignaddr((void *)(width - i), 0); blue = vis_faligndata(blue, blue); green = vis_faligndata(green, green); red = vis_faligndata(red, red); dp += ((width - i - 1) * 3); vis_alignaddr((void *)spy, 7); for (; i < width; i++) { STORE_PIXEL(0, 1, 2); dp -= 3; } __mlib_VectorCopy_U8(rgb, (mlib_u8 *)buf, width * 3); rgb += rgb_stride; dp = (mlib_u8 *)buf; ddp = (mlib_d64 *)dp; y += yuv_stride; u += yuv_stride; v += yuv_stride; } if (width * 3 > 16 * 1024) __mlib_free(tmp); return (MLIB_SUCCESS); }
mlib_status __mlib_VideoDownSample422( mlib_u8 *dst, const mlib_u8 *src, mlib_s32 n) { mlib_d64 *sp0 = (mlib_d64 *)src; mlib_d64 *pd = (mlib_d64 *)dst; mlib_d64 d0; mlib_d64 tmp, data0, data1; mlib_d64 acc0_hi, acc0_lo; mlib_d64 round = vis_to_double_dup(0x1); mlib_f32 fone = vis_to_float(0x1000000); mlib_s32 i, edge; if (n <= 0) return (MLIB_FAILURE); vis_write_gsr(6 << 3); vis_write_bmask(0x02461357, 0); #pragma pipeloop(0) for (i = 0; i <= n - 16; i += 16) { d0 = (*sp0++); tmp = vis_bshuffle(d0, d0); acc0_hi = vis_fmul8x16au(vis_read_hi(tmp), fone); acc0_lo = vis_fmul8x16au(vis_read_lo(tmp), fone); acc0_hi = vis_fpadd16(acc0_hi, acc0_lo); data0 = vis_fpadd16(acc0_hi, round); d0 = (*sp0++); tmp = vis_bshuffle(d0, d0); acc0_hi = vis_fmul8x16au(vis_read_hi(tmp), fone); acc0_lo = vis_fmul8x16au(vis_read_lo(tmp), fone); acc0_hi = vis_fpadd16(acc0_hi, acc0_lo); data1 = vis_fpadd16(acc0_hi, round); (*pd++) = vis_fpack16_pair(data0, data1); } if (i < n) { d0 = (*sp0++); tmp = vis_bshuffle(d0, d0); acc0_hi = vis_fmul8x16au(vis_read_hi(tmp), fone); acc0_lo = vis_fmul8x16au(vis_read_lo(tmp), fone); acc0_hi = vis_fpadd16(acc0_hi, acc0_lo); data0 = vis_fpadd16(acc0_hi, round); d0 = vis_ld_d64_nf(sp0); tmp = vis_bshuffle(d0, d0); acc0_hi = vis_fmul8x16au(vis_read_hi(tmp), fone); acc0_lo = vis_fmul8x16au(vis_read_lo(tmp), fone); acc0_hi = vis_fpadd16(acc0_hi, acc0_lo); data1 = vis_fpadd16(acc0_hi, round); edge = vis_edge8(pd, (dst + (n / 2) - 1)); vis_pst_8(vis_fpack16_pair(data0, data1), pd, edge); } return (MLIB_SUCCESS); }
static mlib_status mlib_v_VideoColorYUV2RGB411_nonalign( mlib_u8 *rgb, const mlib_u8 *y, const mlib_u8 *u, const mlib_u8 *v, mlib_s32 width, mlib_s32 height, mlib_s32 rgb_stride, mlib_s32 y_stride, mlib_s32 uv_stride) { /* pointers to src address */ mlib_u8 *sp1, *sp2, *sp3, *sl1, *sl2, *sl3; /* pointers to dst address */ mlib_u8 *dp, *dl; /* all. pointer to y */ mlib_d64 *spy; /* all. pointers to u, v */ mlib_d64 *dfu, *dfv; /* u, v data */ mlib_f32 fu, fv; /* y data */ mlib_d64 dy0, dy1, dy2, dy3; mlib_d64 ddy1, ddy2, ddy3, ddy4; mlib_d64 du0, du1, fu0, fu1; mlib_d64 dv1, dv2, fv0, fv1; mlib_d64 dr, dr1, dr2, dr3, dr4; mlib_d64 dg, dg1, dg2, dg3, dg4; mlib_d64 db, db1, db2, db3, db4; mlib_d64 dtmp; /* 1.1644 * 4096 */ mlib_f32 f0 = vis_to_float(0x12a1); /* 2.0184 * 8192 */ mlib_f32 f1 = vis_to_float(0x4097); /* -0.3920 * 8192 */ mlib_f32 f4 = vis_to_float(0xf375); /* -0.8132 * 8192 */ mlib_f32 f5 = vis_to_float(0xe5fa); /* 1.5966 * 8192 */ mlib_f32 f8 = vis_to_float(0x3317); /* -276.9856 * 32 */ mlib_d64 doff0 = vis_to_double_dup(0xdd60dd60); /* 135.6352 * 32 */ mlib_d64 doff1 = vis_to_double_dup(0x10f410f4); /* -222.9952 * 32 */ mlib_d64 doff2 = vis_to_double_dup(0xe420e420); mlib_f32 fscale = vis_to_float(0x80808080); /* loop variable */ mlib_s32 i, j; mlib_d64 *buf, BUFF[16 * 1024]; mlib_d64 *ddp, dd01, dd11, dd21, dd02, dd12, dd22; mlib_u8 *tmp; if (width * 3 > 16 * 1024) { tmp = __mlib_malloc(width * 3 * sizeof (mlib_u8) + 7); buf = (mlib_d64 *)((mlib_addr)(tmp + 7) & ~7); } else { buf = (mlib_d64 *)BUFF; } /* * initialize GSR scale factor */ vis_write_gsr(3 << 3); sp1 = sl1 = (mlib_u8 *)y; sp2 = sl2 = (mlib_u8 *)u; sp3 = sl3 = (mlib_u8 *)v; dp = (mlib_u8 *)buf; dl = rgb; ddp = (mlib_d64 *)dp; /* * row loop */ for (j = 0; j < height; j++) { spy = (mlib_d64 *)vis_alignaddr(sp1, 0); dfu = (mlib_d64 *)vis_alignaddr(sp2, 0); fu0 = (*dfu++); fu1 = vis_ld_d64_nf(dfu); dfu++; fu = vis_read_hi(vis_faligndata(fu0, fu1)); sp2 += 4; dfv = (mlib_d64 *)vis_alignaddr(sp3, 0); fv0 = (*dfv++); fv1 = vis_ld_d64_nf(dfv); dfv++; fv = vis_read_hi(vis_faligndata(fv0, fv1)); sp3 += 4; dy0 = (*spy++); dy3 = vis_ld_d64_nf(spy); spy++; vis_alignaddr(sp1, 0); dy1 = vis_faligndata(dy0, dy3); dy0 = vis_ld_d64_nf(spy); spy++; dy2 = vis_faligndata(dy3, dy0); du0 = vis_fmul8x16al(fu, f1); db = vis_fpadd16(du0, doff0); du1 = vis_fmul8x16al(fu, f4); dv1 = vis_fmul8x16al(fv, f5); dtmp = vis_fpadd16(du1, dv1); dg = vis_fpadd16(dtmp, doff1); dv2 = vis_fmul8x16al(fv, f8); dr = vis_fpadd16(dv2, doff2); ddy1 = vis_fmul8x16al(vis_read_hi(dy1), f0); ddy2 = vis_fmul8x16al(vis_read_lo(dy1), f0); ddy3 = vis_fmul8x16al(vis_read_hi(dy2), f0); ddy4 = vis_fmul8x16al(vis_read_lo(dy2), f0); db1 = vis_fmul8x16au(fscale, vis_read_hi(db)); db1 = vis_fpadd16(ddy1, db1); db2 = vis_fmul8x16al(fscale, vis_read_hi(db)); db2 = vis_fpadd16(ddy2, db2); db3 = vis_fmul8x16au(fscale, vis_read_lo(db)); db3 = vis_fpadd16(ddy3, db3); db4 = vis_fmul8x16al(fscale, vis_read_lo(db)); db4 = vis_fpadd16(ddy4, db4); dg1 = vis_fmul8x16au(fscale, vis_read_hi(dg)); dg1 = vis_fpadd16(ddy1, dg1); dg2 = vis_fmul8x16al(fscale, vis_read_hi(dg)); dg2 = vis_fpadd16(ddy2, dg2); dg3 = vis_fmul8x16au(fscale, vis_read_lo(dg)); dg3 = vis_fpadd16(ddy3, dg3); dg4 = vis_fmul8x16al(fscale, vis_read_lo(dg)); dg4 = vis_fpadd16(ddy4, dg4); dr1 = vis_fmul8x16au(fscale, vis_read_hi(dr)); dr1 = vis_fpadd16(ddy1, dr1); dr2 = vis_fmul8x16al(fscale, vis_read_hi(dr)); dr2 = vis_fpadd16(ddy2, dr2); dr3 = vis_fmul8x16au(fscale, vis_read_lo(dr)); dr3 = vis_fpadd16(ddy3, dr3); dr4 = vis_fmul8x16al(fscale, vis_read_lo(dr)); dr4 = vis_fpadd16(ddy4, dr4); db = vis_fpack16_pair(db1, db2); db1 = vis_fpack16_pair(db3, db4); dr = vis_fpack16_pair(dr1, dr2); dr1 = vis_fpack16_pair(dr3, dr4); dg = vis_fpack16_pair(dg1, dg2); dg1 = vis_fpack16_pair(dg3, dg4); dfu = (mlib_d64 *)vis_alignaddr(sp2, 0); fu0 = vis_ld_d64_nf(dfu); dfu++; fu1 = vis_ld_d64_nf(dfu); dfu++; fu = vis_read_hi(vis_faligndata(fu0, fu1)); sp2 += 4; dfv = (mlib_d64 *)vis_alignaddr(sp3, 0); fv0 = vis_ld_d64_nf(dfv); dfv++; fv1 = vis_ld_d64_nf(dfv); dfv++; fv = vis_read_hi(vis_faligndata(fv0, fv1)); sp3 += 4; /* * 16-pixel column loop */ #pragma pipeloop(0) for (i = 0; i <= width - 16; i += 16) { vis_write_bmask(0x0801902A, 0); dd01 = vis_bshuffle(dr, dg); dd02 = vis_bshuffle(dr1, dg1); vis_write_bmask(0x03B04C05, 0); dd11 = vis_bshuffle(dr, dg); dd12 = vis_bshuffle(dr1, dg1); vis_write_bmask(0xD06E07F0, 0); dd21 = vis_bshuffle(dr, dg); dd22 = vis_bshuffle(dr1, dg1); vis_write_bmask(0x01834967, 0); ddp[0] = vis_bshuffle(dd01, db); ddp[3] = vis_bshuffle(dd02, db1); vis_write_bmask(0xA12B45C7, 0); ddp[1] = vis_bshuffle(dd11, db); ddp[4] = vis_bshuffle(dd12, db1); vis_write_bmask(0x0D23E56F, 0); ddp[2] = vis_bshuffle(dd21, db); ddp[5] = vis_bshuffle(dd22, db1); dy3 = vis_ld_d64_nf(spy); spy++; vis_alignaddr(sp1, 0); dy1 = vis_faligndata(dy0, dy3); dy0 = vis_ld_d64_nf(spy); spy++; dy2 = vis_faligndata(dy3, dy0); du0 = vis_fmul8x16al(fu, f1); db = vis_fpadd16(du0, doff0); du1 = vis_fmul8x16al(fu, f4); dv1 = vis_fmul8x16al(fv, f5); dtmp = vis_fpadd16(du1, dv1); dg = vis_fpadd16(dtmp, doff1); dv2 = vis_fmul8x16al(fv, f8); dr = vis_fpadd16(dv2, doff2); ddy1 = vis_fmul8x16al(vis_read_hi(dy1), f0); ddy2 = vis_fmul8x16al(vis_read_lo(dy1), f0); ddy3 = vis_fmul8x16al(vis_read_hi(dy2), f0); ddy4 = vis_fmul8x16al(vis_read_lo(dy2), f0); db1 = vis_fmul8x16au(fscale, vis_read_hi(db)); db1 = vis_fpadd16(ddy1, db1); db2 = vis_fmul8x16al(fscale, vis_read_hi(db)); db2 = vis_fpadd16(ddy2, db2); db3 = vis_fmul8x16au(fscale, vis_read_lo(db)); db3 = vis_fpadd16(ddy3, db3); db4 = vis_fmul8x16al(fscale, vis_read_lo(db)); db4 = vis_fpadd16(ddy4, db4); dg1 = vis_fmul8x16au(fscale, vis_read_hi(dg)); dg1 = vis_fpadd16(ddy1, dg1); dg2 = vis_fmul8x16al(fscale, vis_read_hi(dg)); dg2 = vis_fpadd16(ddy2, dg2); dg3 = vis_fmul8x16au(fscale, vis_read_lo(dg)); dg3 = vis_fpadd16(ddy3, dg3); dg4 = vis_fmul8x16al(fscale, vis_read_lo(dg)); dg4 = vis_fpadd16(ddy4, dg4); dr1 = vis_fmul8x16au(fscale, vis_read_hi(dr)); dr1 = vis_fpadd16(ddy1, dr1); dr2 = vis_fmul8x16al(fscale, vis_read_hi(dr)); dr2 = vis_fpadd16(ddy2, dr2); dr3 = vis_fmul8x16au(fscale, vis_read_lo(dr)); dr3 = vis_fpadd16(ddy3, dr3); dr4 = vis_fmul8x16al(fscale, vis_read_lo(dr)); dr4 = vis_fpadd16(ddy4, dr4); db = vis_fpack16_pair(db1, db2); db1 = vis_fpack16_pair(db3, db4); dr = vis_fpack16_pair(dr1, dr2); dr1 = vis_fpack16_pair(dr3, dr4); dg = vis_fpack16_pair(dg1, dg2); dg1 = vis_fpack16_pair(dg3, dg4); dfu = (mlib_d64 *)vis_alignaddr(sp2, 0); fu0 = vis_ld_d64_nf(dfu); dfu++; fu1 = vis_ld_d64_nf(dfu); dfu++; fu = vis_read_hi(vis_faligndata(fu0, fu1)); sp2 += 4; dfv = (mlib_d64 *)vis_alignaddr(sp3, 0); fv0 = vis_ld_d64_nf(dfv); dfv++; fv1 = vis_ld_d64_nf(dfv); dfv++; fv = vis_read_hi(vis_faligndata(fv0, fv1)); sp3 += 4; ddp += 6; } if (i <= width - 8) { vis_write_bmask(0x0801902A, 0); dd01 = vis_bshuffle(dr, dg); vis_write_bmask(0x03B04C05, 0); dd11 = vis_bshuffle(dr, dg); vis_write_bmask(0xD06E07F0, 0); dd21 = vis_bshuffle(dr, dg); vis_write_bmask(0x01834967, 0); ddp[0] = vis_bshuffle(dd01, db); vis_write_bmask(0xA12B45C7, 0); ddp[1] = vis_bshuffle(dd11, db); vis_write_bmask(0x0D23E56F, 0); ddp[2] = vis_bshuffle(dd21, db); db = db1; dr = dr1; dg = dg1; ddp += 3; i += 8; } dp = (mlib_u8 *)ddp; vis_alignaddr((void *)(width - i), 0); db = vis_faligndata(db, db); dg = vis_faligndata(dg, dg); dr = vis_faligndata(dr, dr); dp += ((width - i - 1) * 3); vis_alignaddr((void *)7, 0); for (; i < width; i++) { STORE_PIXEL(0, 1, 2); dp -= 3; } sp1 = sl1 = sl1 + y_stride; sp2 = sl2 = sl2 + uv_stride; sp3 = sl3 = sl3 + uv_stride; __mlib_VectorCopy_U8(dl, (mlib_u8 *)buf, width * 3); dl = dp = dl + rgb_stride; dp = (mlib_u8 *)buf; ddp = (mlib_d64 *)dp; } if (width * 3 > 16 * 1024) __mlib_free(tmp); return (MLIB_SUCCESS); }
static mlib_status mlib_v_VideoColorYUV2RGB420_nonalign( mlib_u8 *rgb, const mlib_u8 *y, const mlib_u8 *u, const mlib_u8 *v, mlib_s32 width, mlib_s32 height, mlib_s32 rgb_stride, mlib_s32 y_stride, mlib_s32 uv_stride) { /* pointers to src address */ mlib_u8 *sp2, *sp3, *sl2, *sl3; /* pointers to src address */ mlib_u8 *sp11, *sp12, *sl11, *sl12; /* pointers to dst address */ mlib_u8 *dp1, *dl1; /* pointers to dst address */ mlib_u8 *dp2, *dl2; /* all. pointer to y */ mlib_d64 *spy1, *spy2; /* all. pointers to u, v */ mlib_f32 *dfu, *dfv; /* y data */ mlib_d64 dy0, dy1, dy2, dy3, dy4, dy5; /* u, v data */ mlib_f32 fu0, fu1, fv0, fv1; mlib_d64 du, dv, du0, du1, dv0, dv1; /* (1.1644, 1.5966)*8192 */ mlib_f32 k12 = vis_to_float(0x25433317); /* (-.3920, -.8132)*8192 */ mlib_f32 k34 = vis_to_float(0xf375e5fa); /* 2.0184*8192 */ mlib_f32 k5 = vis_to_float(0x1004097); mlib_d64 k_222_9952 = vis_to_double(0x1be01be0, 0x1be01be0); mlib_d64 k_135_6352 = vis_to_double(0x10f410f4, 0x10f410f4); mlib_d64 k_276_9856 = vis_to_double(0x22a022a0, 0x22a022a0); mlib_d64 u_3920_hi, u_20184_hi, v_15966_hi, v_8132_hi; mlib_d64 u_3920_lo, u_20184_lo, v_15966_lo, v_8132_lo; mlib_d64 y_11644_hi, y_11644_lo; mlib_d64 z_11644_hi, z_11644_lo; mlib_d64 r_hi, r_lo, g_hi, g_lo, b_hi, b_lo; mlib_d64 temp_r_hi, temp_r_lo, temp_g_hi, temp_g_lo, temp_b_hi, temp_b_lo; /* loop variables */ mlib_s32 i, j; mlib_s32 y_stride2 = 2 * y_stride; mlib_s32 rgb_stride2 = 2 * rgb_stride; mlib_s32 off2, off3; mlib_d64 red1, green1, blue1, *ddp1, dd01, dd11, dd21; mlib_d64 red2, green2, blue2, *ddp2, dd02, dd12, dd22; mlib_d64 *buf1, BUFF1[16 * 1024]; mlib_d64 *buf2, BUFF2[16 * 1024]; mlib_u8 *tmp1, *tmp2; if (width * 3 > 16 * 1024) { tmp1 = __mlib_malloc(width * 3 * sizeof (mlib_u8) + 7); tmp2 = __mlib_malloc(width * 3 * sizeof (mlib_u8) + 7); buf1 = (mlib_d64 *)((mlib_addr)(tmp1 + 7) & ~7); buf2 = (mlib_d64 *)((mlib_addr)(tmp2 + 7) & ~7); } else { buf1 = (mlib_d64 *)BUFF1; buf2 = (mlib_d64 *)BUFF2; } /* * initialize GSR scale factor */ vis_write_gsr(2 << 3); sp11 = sl11 = (mlib_u8 *)y; sp12 = sl12 = (mlib_u8 *)y + y_stride; sp2 = sl2 = (mlib_u8 *)u; sp3 = sl3 = (mlib_u8 *)v; dp1 = (mlib_u8 *)buf1; dp2 = (mlib_u8 *)buf2; dl1 = (mlib_u8 *)rgb; dl2 = (mlib_u8 *)(rgb + rgb_stride); ddp1 = (mlib_d64 *)dp1; ddp2 = (mlib_d64 *)dp2; /* * row loop */ for (j = 0; j < height / 2; j++) { spy1 = (mlib_d64 *)vis_alignaddr(sp11, 0); spy2 = (mlib_d64 *)vis_alignaddr(sp12, 0); dfu = (mlib_f32 *)((mlib_addr)sp2 & ~3); off2 = (sp2 - (mlib_u8 *)dfu) * 2; dfv = (mlib_f32 *)((mlib_addr)sp3 & ~3); off3 = (sp3 - (mlib_u8 *)dfv) * 2; vis_alignaddr((void *)off2, 0); fu0 = (*dfu++); fu1 = vis_ld_f32_nf(dfu); dfu++; du0 = vis_fpmerge(fu0, fu0); du1 = vis_fpmerge(fu1, fu1); du = vis_faligndata(du0, du1); du0 = du1; vis_alignaddr((void *)off3, 0); fv0 = (*dfv++); fv1 = vis_ld_f32_nf(dfv); dfv++; dv0 = vis_fpmerge(fv0, fv0); dv1 = vis_fpmerge(fv1, fv1); dv = vis_faligndata(dv0, dv1); dv0 = dv1; /* U*(-0.3920); */ u_3920_hi = vis_fmul8x16au(vis_read_hi(du), k34); /* V*(-0.8132); */ v_8132_hi = vis_fmul8x16al(vis_read_hi(dv), k34); /* U*(-0.3920); */ u_3920_lo = vis_fmul8x16au(vis_read_lo(du), k34); /* V*(-0.8132); */ v_8132_lo = vis_fmul8x16al(vis_read_lo(dv), k34); dy0 = (*spy1++); dy4 = (*spy2++); dy3 = vis_ld_d64_nf(spy1); spy1++; vis_alignaddr(sp11, 0); dy1 = vis_faligndata(dy0, dy3); dy0 = dy3; dy5 = vis_ld_d64_nf(spy2); spy2++; vis_alignaddr(sp12, 0); dy2 = vis_faligndata(dy4, dy5); dy4 = dy5; /* U*2.0184 */ u_20184_hi = vis_fmul8x16al(vis_read_hi(du), k5); g_hi = vis_fpadd16(u_3920_hi, v_8132_hi); u_20184_lo = vis_fmul8x16al(vis_read_lo(du), k5); g_hi = vis_fpadd16(g_hi, k_135_6352); /* V*1.5966 */ v_15966_hi = vis_fmul8x16al(vis_read_hi(dv), k12); g_lo = vis_fpadd16(u_3920_lo, v_8132_lo); v_15966_lo = vis_fmul8x16al(vis_read_lo(dv), k12); g_lo = vis_fpadd16(g_lo, k_135_6352); /* Y*1.1644 */ y_11644_hi = vis_fmul8x16au(vis_read_hi(dy1), k12); b_hi = vis_fpsub16(u_20184_hi, k_276_9856); /* Y*1.1644 */ y_11644_lo = vis_fmul8x16au(vis_read_lo(dy1), k12); b_lo = vis_fpsub16(u_20184_lo, k_276_9856); /* Z*1.1644 */ z_11644_hi = vis_fmul8x16au(vis_read_hi(dy2), k12); r_hi = vis_fpsub16(v_15966_hi, k_222_9952); /* Z*1.1644 */ z_11644_lo = vis_fmul8x16au(vis_read_lo(dy2), k12); r_lo = vis_fpsub16(v_15966_lo, k_222_9952); temp_g_hi = vis_fpadd16(g_hi, y_11644_hi); temp_b_hi = vis_fpadd16(b_hi, y_11644_hi); green1 = vis_fpack16_to_hi(green1, temp_g_hi); temp_r_hi = vis_fpadd16(r_hi, y_11644_hi); blue1 = vis_fpack16_to_hi(blue1, temp_b_hi); temp_g_lo = vis_fpadd16(g_lo, y_11644_lo); vis_alignaddr((void *)off2, 0); fu1 = vis_ld_f32_nf(dfu); dfu++; du1 = vis_fpmerge(fu1, fu1); du = vis_faligndata(du0, du1); du0 = du1; red1 = vis_fpack16_to_hi(red1, temp_r_hi); temp_b_lo = vis_fpadd16(b_lo, y_11644_lo); vis_alignaddr((void *)off3, 0); fv1 = vis_ld_f32_nf(dfv); dfv++; dv1 = vis_fpmerge(fv1, fv1); dv = vis_faligndata(dv0, dv1); dv0 = dv1; green1 = vis_fpack16_to_lo(green1, temp_g_lo); temp_r_lo = vis_fpadd16(r_lo, y_11644_lo); blue1 = vis_fpack16_to_lo(blue1, temp_b_lo); red1 = vis_fpack16_to_lo(red1, temp_r_lo); /* U*(-0.3920); */ u_3920_hi = vis_fmul8x16au(vis_read_hi(du), k34); temp_g_hi = vis_fpadd16(g_hi, z_11644_hi); /* V*(-0.8132); */ v_8132_hi = vis_fmul8x16al(vis_read_hi(dv), k34); temp_b_hi = vis_fpadd16(b_hi, z_11644_hi); green2 = vis_fpack16_to_hi(green2, temp_g_hi); temp_r_hi = vis_fpadd16(r_hi, z_11644_hi); blue2 = vis_fpack16_to_hi(blue2, temp_b_hi); temp_g_lo = vis_fpadd16(g_lo, z_11644_lo); red2 = vis_fpack16_to_hi(red2, temp_r_hi); temp_b_lo = vis_fpadd16(b_lo, z_11644_lo); green2 = vis_fpack16_to_lo(green2, temp_g_lo); temp_r_lo = vis_fpadd16(r_lo, z_11644_lo); blue2 = vis_fpack16_to_lo(blue2, temp_b_lo); red2 = vis_fpack16_to_lo(red2, temp_r_lo); u_3920_lo = vis_fmul8x16au(vis_read_lo(du), k34); v_8132_lo = vis_fmul8x16al(vis_read_lo(dv), k34); /* * 16-pixel column loop */ #pragma pipeloop(0) for (i = 0; i <= width - 8; i += 8) { vis_write_bmask(0x0801902A, 0); dd01 = vis_bshuffle(red1, green1); dd02 = vis_bshuffle(red2, green2); vis_write_bmask(0x03B04C05, 0); dd11 = vis_bshuffle(red1, green1); dd12 = vis_bshuffle(red2, green2); vis_write_bmask(0xD06E07F0, 0); dd21 = vis_bshuffle(red1, green1); dd22 = vis_bshuffle(red2, green2); vis_write_bmask(0x01834967, 0); ddp1[0] = vis_bshuffle(dd01, blue1); ddp2[0] = vis_bshuffle(dd02, blue2); vis_write_bmask(0xA12B45C7, 0); ddp1[1] = vis_bshuffle(dd11, blue1); ddp2[1] = vis_bshuffle(dd12, blue2); vis_write_bmask(0x0D23E56F, 0); ddp1[2] = vis_bshuffle(dd21, blue1); ddp2[2] = vis_bshuffle(dd22, blue2); dy3 = vis_ld_d64_nf(spy1); spy1++; vis_alignaddr(sp11, 0); dy1 = vis_faligndata(dy0, dy3); dy0 = dy3; dy5 = vis_ld_d64_nf(spy2); spy2++; vis_alignaddr(sp12, 0); dy2 = vis_faligndata(dy4, dy5); dy4 = dy5; /* U*2.0184 */ u_20184_hi = vis_fmul8x16al(vis_read_hi(du), k5); g_hi = vis_fpadd16(u_3920_hi, v_8132_hi); u_20184_lo = vis_fmul8x16al(vis_read_lo(du), k5); g_hi = vis_fpadd16(g_hi, k_135_6352); /* V*1.5966 */ v_15966_hi = vis_fmul8x16al(vis_read_hi(dv), k12); g_lo = vis_fpadd16(u_3920_lo, v_8132_lo); v_15966_lo = vis_fmul8x16al(vis_read_lo(dv), k12); g_lo = vis_fpadd16(g_lo, k_135_6352); /* Y*1.1644 */ y_11644_hi = vis_fmul8x16au(vis_read_hi(dy1), k12); b_hi = vis_fpsub16(u_20184_hi, k_276_9856); /* Y*1.1644 */ y_11644_lo = vis_fmul8x16au(vis_read_lo(dy1), k12); b_lo = vis_fpsub16(u_20184_lo, k_276_9856); /* Z*1.1644 */ z_11644_hi = vis_fmul8x16au(vis_read_hi(dy2), k12); r_hi = vis_fpsub16(v_15966_hi, k_222_9952); /* Z*1.1644 */ z_11644_lo = vis_fmul8x16au(vis_read_lo(dy2), k12); r_lo = vis_fpsub16(v_15966_lo, k_222_9952); temp_g_hi = vis_fpadd16(g_hi, y_11644_hi); temp_b_hi = vis_fpadd16(b_hi, y_11644_hi); green1 = vis_fpack16_to_hi(green1, temp_g_hi); temp_r_hi = vis_fpadd16(r_hi, y_11644_hi); blue1 = vis_fpack16_to_hi(blue1, temp_b_hi); temp_g_lo = vis_fpadd16(g_lo, y_11644_lo); vis_alignaddr((void *)off2, 0); fu1 = vis_ld_f32_nf(dfu); dfu++; du1 = vis_fpmerge(fu1, fu1); du = vis_faligndata(du0, du1); du0 = du1; red1 = vis_fpack16_to_hi(red1, temp_r_hi); temp_b_lo = vis_fpadd16(b_lo, y_11644_lo); vis_alignaddr((void *)off3, 0); fv1 = vis_ld_f32_nf(dfv); dfv++; dv1 = vis_fpmerge(fv1, fv1); dv = vis_faligndata(dv0, dv1); dv0 = dv1; green1 = vis_fpack16_to_lo(green1, temp_g_lo); temp_r_lo = vis_fpadd16(r_lo, y_11644_lo); blue1 = vis_fpack16_to_lo(blue1, temp_b_lo); red1 = vis_fpack16_to_lo(red1, temp_r_lo); /* U*(-0.3920); */ u_3920_hi = vis_fmul8x16au(vis_read_hi(du), k34); temp_g_hi = vis_fpadd16(g_hi, z_11644_hi); /* V*(-0.8132); */ v_8132_hi = vis_fmul8x16al(vis_read_hi(dv), k34); temp_b_hi = vis_fpadd16(b_hi, z_11644_hi); green2 = vis_fpack16_to_hi(green2, temp_g_hi); temp_r_hi = vis_fpadd16(r_hi, z_11644_hi); blue2 = vis_fpack16_to_hi(blue2, temp_b_hi); temp_g_lo = vis_fpadd16(g_lo, z_11644_lo); red2 = vis_fpack16_to_hi(red2, temp_r_hi); temp_b_lo = vis_fpadd16(b_lo, z_11644_lo); green2 = vis_fpack16_to_lo(green2, temp_g_lo); temp_r_lo = vis_fpadd16(r_lo, z_11644_lo); blue2 = vis_fpack16_to_lo(blue2, temp_b_lo); red2 = vis_fpack16_to_lo(red2, temp_r_lo); u_3920_lo = vis_fmul8x16au(vis_read_lo(du), k34); v_8132_lo = vis_fmul8x16al(vis_read_lo(dv), k34); ddp1 += 3; ddp2 += 3; } dp1 = (mlib_u8 *)ddp1; dp2 = (mlib_u8 *)ddp2; vis_alignaddr((void *)(width - i), 0); blue1 = vis_faligndata(blue1, blue1); green1 = vis_faligndata(green1, green1); red1 = vis_faligndata(red1, red1); dp1 += ((width - i - 1) * 3); blue2 = vis_faligndata(blue2, blue2); green2 = vis_faligndata(green2, green2); red2 = vis_faligndata(red2, red2); dp2 += ((width - i - 1) * 3); vis_alignaddr((void *)7, 0); for (; i < width; i++) { STORE_PIXEL1(0, 1, 2); STORE_PIXEL2(0, 1, 2); dp1 -= 3; dp2 -= 3; } sp11 = sl11 = sl11 + y_stride2; sp12 = sl12 = sl12 + y_stride2; sp2 = sl2 = sl2 + uv_stride; sp3 = sl3 = sl3 + uv_stride; __mlib_VectorCopy_U8(dl1, (mlib_u8 *)buf1, width * 3); __mlib_VectorCopy_U8(dl2, (mlib_u8 *)buf2, width * 3); dl1 = dp1 = dl1 + rgb_stride2; dl2 = dp2 = dl2 + rgb_stride2; dp1 = (mlib_u8 *)buf1; dp2 = (mlib_u8 *)buf2; ddp1 = (mlib_d64 *)dp1; ddp2 = (mlib_d64 *)dp2; } if (width * 3 > 16 * 1024) { __mlib_free(tmp1); __mlib_free(tmp2); } return (MLIB_SUCCESS); }
mlib_status __mlib_VideoColorJFIFYCC2RGB420_Nearest( mlib_u8 *rgb0, mlib_u8 *rgb1, const mlib_u8 *y0, const mlib_u8 *y1, const mlib_u8 *cb, const mlib_u8 *cr, mlib_s32 n) { /* pointers to dst address */ mlib_u8 *dp1, *dp2; /* all. pointer to y */ mlib_d64 *spy1, *spy2; /* all. pointers to u, v */ mlib_f32 *dfu, *dfv; /* u, v data */ mlib_f32 fu, fv; /* y data */ mlib_d64 dy1, dy2; mlib_d64 du, dv; /* (1.00000, 1.40200)*8192 */ mlib_f32 k12 = vis_to_float(0x20002cdd); /* (-.34414, -.71414)*8192 */ mlib_f32 k34 = vis_to_float(0xf4fde926); /* 1.77200*8192 */ mlib_f32 k5 = vis_to_float(0x10038b4); /* (179.45600 - 0.5)*32 */ mlib_d64 k_179_456 = vis_to_double(0x165f165f, 0x165f165f); /* (135.45984 + 0.5)*32 */ mlib_d64 k_135_45984 = vis_to_double(0x10ff10ff, 0x10ff10ff); /* (226.81600 - 0.5)*32 */ mlib_d64 k_226_816 = vis_to_double(0x1c4a1c4a, 0x1c4a1c4a); mlib_d64 u_3920_hi, u_20184_hi, v_15966_hi, v_8132_hi; mlib_d64 u_3920_lo, u_20184_lo, v_15966_lo, v_8132_lo; mlib_d64 y_11644_hi, y_11644_lo; mlib_d64 z_11644_hi, z_11644_lo; mlib_d64 r_hi, r_lo, g_hi, g_lo, b_hi, b_lo; mlib_d64 temp_r_hi, temp_r_lo, temp_g_hi, temp_g_lo, temp_b_hi, temp_b_lo; /* loop variable */ mlib_s32 i; mlib_d64 red1, green1, blue1, *ddp1, dd01, dd11, dd21; mlib_d64 red2, green2, blue2, *ddp2, dd02, dd12, dd22; if (n <= 0) return (MLIB_FAILURE); /* * initialize GSR scale factor */ vis_write_gsr((2 << 3) + 7); dp1 = (mlib_u8 *)rgb0; dp2 = (mlib_u8 *)rgb1; ddp1 = (mlib_d64 *)dp1; ddp2 = (mlib_d64 *)dp2; spy1 = (mlib_d64 *)y0; spy2 = (mlib_d64 *)y1; dfu = (mlib_f32 *)cb; dfv = (mlib_f32 *)cr; fu = vis_ld_f32_nf(dfu); dfu++; fv = vis_ld_f32_nf(dfv); dfv++; du = vis_fpmerge(fu, fu); dv = vis_fpmerge(fv, fv); /* U*(-0.3920); */ u_3920_hi = vis_fmul8x16au(vis_read_hi(du), k34); /* V*(-0.8132); */ v_8132_hi = vis_fmul8x16al(vis_read_hi(dv), k34); /* U*(-0.3920); */ u_3920_lo = vis_fmul8x16au(vis_read_lo(du), k34); /* V*(-0.8132); */ v_8132_lo = vis_fmul8x16al(vis_read_lo(dv), k34); dy1 = vis_ld_d64_nf(spy1); spy1++; dy2 = vis_ld_d64_nf(spy2); spy2++; /* U*2.0184 */ u_20184_hi = vis_fmul8x16al(vis_read_hi(du), k5); g_hi = vis_fpadd16(u_3920_hi, v_8132_hi); u_20184_lo = vis_fmul8x16al(vis_read_lo(du), k5); g_hi = vis_fpadd16(g_hi, k_135_45984); /* V*1.5966 */ v_15966_hi = vis_fmul8x16al(vis_read_hi(dv), k12); g_lo = vis_fpadd16(u_3920_lo, v_8132_lo); v_15966_lo = vis_fmul8x16al(vis_read_lo(dv), k12); g_lo = vis_fpadd16(g_lo, k_135_45984); /* Y*1.1644 */ y_11644_hi = vis_fmul8x16au(vis_read_hi(dy1), k12); b_hi = vis_fpsub16(u_20184_hi, k_226_816); /* Y*1.1644 */ y_11644_lo = vis_fmul8x16au(vis_read_lo(dy1), k12); b_lo = vis_fpsub16(u_20184_lo, k_226_816); /* Z*1.1644 */ z_11644_hi = vis_fmul8x16au(vis_read_hi(dy2), k12); r_hi = vis_fpsub16(v_15966_hi, k_179_456); /* Z*1.1644 */ z_11644_lo = vis_fmul8x16au(vis_read_lo(dy2), k12); r_lo = vis_fpsub16(v_15966_lo, k_179_456); temp_g_hi = vis_fpadd16(g_hi, y_11644_hi); temp_b_hi = vis_fpadd16(b_hi, y_11644_hi); green1 = vis_fpack16_to_hi(green1, temp_g_hi); temp_r_hi = vis_fpadd16(r_hi, y_11644_hi); blue1 = vis_fpack16_to_hi(blue1, temp_b_hi); temp_g_lo = vis_fpadd16(g_lo, y_11644_lo); fu = vis_ld_f32_nf(dfu); dfu++; red1 = vis_fpack16_to_hi(red1, temp_r_hi); temp_b_lo = vis_fpadd16(b_lo, y_11644_lo); fv = vis_ld_f32_nf(dfv); dfv++; green1 = vis_fpack16_to_lo(green1, temp_g_lo); temp_r_lo = vis_fpadd16(r_lo, y_11644_lo); blue1 = vis_fpack16_to_lo(blue1, temp_b_lo); du = vis_fpmerge(fu, fu); red1 = vis_fpack16_to_lo(red1, temp_r_lo); dv = vis_fpmerge(fv, fv); /* U*(-0.3920); */ u_3920_hi = vis_fmul8x16au(vis_read_hi(du), k34); temp_g_hi = vis_fpadd16(g_hi, z_11644_hi); /* V*(-0.8132); */ v_8132_hi = vis_fmul8x16al(vis_read_hi(dv), k34); temp_b_hi = vis_fpadd16(b_hi, z_11644_hi); green2 = vis_fpack16_to_hi(green2, temp_g_hi); temp_r_hi = vis_fpadd16(r_hi, z_11644_hi); blue2 = vis_fpack16_to_hi(blue2, temp_b_hi); temp_g_lo = vis_fpadd16(g_lo, z_11644_lo); red2 = vis_fpack16_to_hi(red2, temp_r_hi); temp_b_lo = vis_fpadd16(b_lo, z_11644_lo); green2 = vis_fpack16_to_lo(green2, temp_g_lo); temp_r_lo = vis_fpadd16(r_lo, z_11644_lo); blue2 = vis_fpack16_to_lo(blue2, temp_b_lo); red2 = vis_fpack16_to_lo(red2, temp_r_lo); u_3920_lo = vis_fmul8x16au(vis_read_lo(du), k34); v_8132_lo = vis_fmul8x16al(vis_read_lo(dv), k34); dy1 = vis_ld_d64_nf(spy1); spy1++; dy2 = vis_ld_d64_nf(spy2); spy2++; /* * 16-pixel column loop */ #pragma pipeloop(0) for (i = 0; i <= n - 8; i += 8) { vis_write_bmask(0x0801902A, 0); dd01 = vis_bshuffle(red1, green1); dd02 = vis_bshuffle(red2, green2); vis_write_bmask(0x03B04C05, 0); dd11 = vis_bshuffle(red1, green1); dd12 = vis_bshuffle(red2, green2); vis_write_bmask(0xD06E07F0, 0); dd21 = vis_bshuffle(red1, green1); dd22 = vis_bshuffle(red2, green2); vis_write_bmask(0x01834967, 0); ddp1[0] = vis_bshuffle(dd01, blue1); ddp2[0] = vis_bshuffle(dd02, blue2); vis_write_bmask(0xA12B45C7, 0); ddp1[1] = vis_bshuffle(dd11, blue1); ddp2[1] = vis_bshuffle(dd12, blue2); vis_write_bmask(0x0D23E56F, 0); ddp1[2] = vis_bshuffle(dd21, blue1); ddp2[2] = vis_bshuffle(dd22, blue2); /* U*2.0184 */ u_20184_hi = vis_fmul8x16al(vis_read_hi(du), k5); g_hi = vis_fpadd16(u_3920_hi, v_8132_hi); u_20184_lo = vis_fmul8x16al(vis_read_lo(du), k5); g_hi = vis_fpadd16(g_hi, k_135_45984); /* V*1.5966 */ v_15966_hi = vis_fmul8x16al(vis_read_hi(dv), k12); g_lo = vis_fpadd16(u_3920_lo, v_8132_lo); v_15966_lo = vis_fmul8x16al(vis_read_lo(dv), k12); g_lo = vis_fpadd16(g_lo, k_135_45984); /* Y*1.1644 */ y_11644_hi = vis_fmul8x16au(vis_read_hi(dy1), k12); b_hi = vis_fpsub16(u_20184_hi, k_226_816); /* Y*1.1644 */ y_11644_lo = vis_fmul8x16au(vis_read_lo(dy1), k12); b_lo = vis_fpsub16(u_20184_lo, k_226_816); /* Z*1.1644 */ z_11644_hi = vis_fmul8x16au(vis_read_hi(dy2), k12); r_hi = vis_fpsub16(v_15966_hi, k_179_456); /* Z*1.1644 */ z_11644_lo = vis_fmul8x16au(vis_read_lo(dy2), k12); r_lo = vis_fpsub16(v_15966_lo, k_179_456); temp_g_hi = vis_fpadd16(g_hi, y_11644_hi); temp_b_hi = vis_fpadd16(b_hi, y_11644_hi); green1 = vis_fpack16_to_hi(green1, temp_g_hi); temp_r_hi = vis_fpadd16(r_hi, y_11644_hi); blue1 = vis_fpack16_to_hi(blue1, temp_b_hi); temp_g_lo = vis_fpadd16(g_lo, y_11644_lo); fu = vis_ld_f32_nf(dfu); dfu++; red1 = vis_fpack16_to_hi(red1, temp_r_hi); temp_b_lo = vis_fpadd16(b_lo, y_11644_lo); fv = vis_ld_f32_nf(dfv); dfv++; green1 = vis_fpack16_to_lo(green1, temp_g_lo); temp_r_lo = vis_fpadd16(r_lo, y_11644_lo); blue1 = vis_fpack16_to_lo(blue1, temp_b_lo); du = vis_fpmerge(fu, fu); red1 = vis_fpack16_to_lo(red1, temp_r_lo); dv = vis_fpmerge(fv, fv); /* U*(-0.3920); */ u_3920_hi = vis_fmul8x16au(vis_read_hi(du), k34); temp_g_hi = vis_fpadd16(g_hi, z_11644_hi); /* V*(-0.8132); */ v_8132_hi = vis_fmul8x16al(vis_read_hi(dv), k34); temp_b_hi = vis_fpadd16(b_hi, z_11644_hi); green2 = vis_fpack16_to_hi(green2, temp_g_hi); temp_r_hi = vis_fpadd16(r_hi, z_11644_hi); blue2 = vis_fpack16_to_hi(blue2, temp_b_hi); temp_g_lo = vis_fpadd16(g_lo, z_11644_lo); red2 = vis_fpack16_to_hi(red2, temp_r_hi); temp_b_lo = vis_fpadd16(b_lo, z_11644_lo); green2 = vis_fpack16_to_lo(green2, temp_g_lo); temp_r_lo = vis_fpadd16(r_lo, z_11644_lo); blue2 = vis_fpack16_to_lo(blue2, temp_b_lo); red2 = vis_fpack16_to_lo(red2, temp_r_lo); u_3920_lo = vis_fmul8x16au(vis_read_lo(du), k34); v_8132_lo = vis_fmul8x16al(vis_read_lo(dv), k34); dy1 = vis_ld_d64_nf(spy1); spy1++; dy2 = vis_ld_d64_nf(spy2); spy2++; ddp1 += 3; ddp2 += 3; } dp1 = (mlib_u8 *)ddp1; dp2 = (mlib_u8 *)ddp2; vis_alignaddr((void *)(n - i), 0); blue1 = vis_faligndata(blue1, blue1); green1 = vis_faligndata(green1, green1); red1 = vis_faligndata(red1, red1); dp1 += ((n - i - 1) * 3); blue2 = vis_faligndata(blue2, blue2); green2 = vis_faligndata(green2, green2); red2 = vis_faligndata(red2, red2); dp2 += ((n - i - 1) * 3); vis_alignaddr((void *)7, 0); for (; i < n; i++) { STORE_PIXEL1(0, 1, 2); STORE_PIXEL2(0, 1, 2); dp1 -= 3; dp2 -= 3; } return (MLIB_SUCCESS); }
static void mlib_v_VideoYUV2ABGR_aarray_411( mlib_u32 *abgr, const mlib_d64 *y, const mlib_f32 *u, const mlib_f32 *v, const mlib_d64 *a_array, mlib_s32 count, mlib_s32 left, mlib_s32 isrgb) { /* all. pointer to dst */ mlib_d64 *dpp = (mlib_d64 *)abgr; /* u, v data */ mlib_f32 fu, fv; /* y data */ mlib_d64 dy1, dy2; mlib_d64 ddy1, ddy2, ddy3, ddy4; mlib_d64 du0, du1; mlib_d64 dv1, dv2; mlib_d64 dr, dr1, dr2, dr3, dr4; mlib_d64 dg, dg1, dg2, dg3, dg4; mlib_d64 db, db1, db2, db3, db4; mlib_d64 *dpa, da0, da1, da2, da3, da4; mlib_d64 dtmp; /* 1.1644 * 4096 */ mlib_f32 f0 = vis_to_float(0x12a1); /* 2.0184 * 8192 */ mlib_f32 f1 = vis_to_float(0x4097); /* -0.3920 * 8192 */ mlib_f32 f4 = vis_to_float(0xf375); /* -0.8132 * 8192 */ mlib_f32 f5 = vis_to_float(0xe5fa); /* 1.5966 * 8192 */ mlib_f32 f8 = vis_to_float(0x3317); /* -276.9856 * 32 */ mlib_d64 doff0 = vis_to_double_dup(0xdd60dd60); /* 135.6352 * 32 */ mlib_d64 doff1 = vis_to_double_dup(0x10f410f4); /* -222.9952 * 32 */ mlib_d64 doff2 = vis_to_double_dup(0xe420e420); mlib_f32 fscale = vis_to_float(0x80808080); /* loop variables */ mlib_s32 i; if (isrgb) { f0 = vis_to_float(0x12a1); f1 = vis_to_float(0x3317); f4 = vis_to_float(0xe5fa); f5 = vis_to_float(0xf375); f8 = vis_to_float(0x4097); doff0 = vis_to_double_dup(0xe420e420); doff1 = vis_to_double_dup(0x10f410f4); doff2 = vis_to_double_dup(0xdd60dd60); } dpa = vis_alignaddr((void *)a_array, 0); dy1 = (*y++); dy2 = vis_ld_d64_nf((mlib_d64 *)y); y++; fu = (*u++); fv = (*v++); da2 = (*dpa++); da3 = vis_ld_d64_nf(dpa); dpa++; da4 = vis_ld_d64_nf(dpa); dpa++; du0 = vis_fmul8x16al(fu, f1); du1 = vis_fmul8x16al(fu, f4); dv1 = vis_fmul8x16al(fv, f5); dv2 = vis_fmul8x16al(fv, f8); if (!((mlib_addr)abgr & 7)) { #pragma pipeloop(0) for (i = 0; i < count; i++) { da0 = vis_faligndata(da2, da3); da1 = vis_faligndata(da3, da4); ddy1 = vis_fmul8x16al(vis_read_hi(dy1), f0); ddy2 = vis_fmul8x16al(vis_read_lo(dy1), f0); ddy3 = vis_fmul8x16al(vis_read_hi(dy2), f0); ddy4 = vis_fmul8x16al(vis_read_lo(dy2), f0); db = vis_fpadd16(du0, doff0); dtmp = vis_fpadd16(du1, dv1); dg = vis_fpadd16(dtmp, doff1); dr = vis_fpadd16(dv2, doff2); db1 = vis_fmul8x16au(fscale, vis_read_hi(db)); db1 = vis_fpadd16(ddy1, db1); db2 = vis_fmul8x16al(fscale, vis_read_hi(db)); db2 = vis_fpadd16(ddy2, db2); db3 = vis_fmul8x16au(fscale, vis_read_lo(db)); db3 = vis_fpadd16(ddy3, db3); db4 = vis_fmul8x16al(fscale, vis_read_lo(db)); db4 = vis_fpadd16(ddy4, db4); dg1 = vis_fmul8x16au(fscale, vis_read_hi(dg)); dg1 = vis_fpadd16(ddy1, dg1); dg2 = vis_fmul8x16al(fscale, vis_read_hi(dg)); dg2 = vis_fpadd16(ddy2, dg2); dg3 = vis_fmul8x16au(fscale, vis_read_lo(dg)); dg3 = vis_fpadd16(ddy3, dg3); dg4 = vis_fmul8x16al(fscale, vis_read_lo(dg)); dg4 = vis_fpadd16(ddy4, dg4); dr1 = vis_fmul8x16au(fscale, vis_read_hi(dr)); dr1 = vis_fpadd16(ddy1, dr1); dr2 = vis_fmul8x16al(fscale, vis_read_hi(dr)); dr2 = vis_fpadd16(ddy2, dr2); dr3 = vis_fmul8x16au(fscale, vis_read_lo(dr)); dr3 = vis_fpadd16(ddy3, dr3); dr4 = vis_fmul8x16al(fscale, vis_read_lo(dr)); dr4 = vis_fpadd16(ddy4, dr4); dr = vis_fpack16_pair(dr1, dr2); dr1 = vis_fpack16_pair(dr3, dr4); dg = vis_fpack16_pair(dg1, dg2); dg1 = vis_fpack16_pair(dg3, dg4); db = vis_fpack16_pair(db1, db2); db1 = vis_fpack16_pair(db3, db4); dg2 = vis_fpmerge(vis_read_hi(da0), vis_read_hi(dg)); dg3 = vis_fpmerge(vis_read_hi(db), vis_read_hi(dr)); dy1 = vis_ld_d64_nf((mlib_d64 *)y + 2 * i); dy2 = vis_ld_d64_nf((mlib_d64 *)y + 2 * i + 1); fu = vis_ld_f32_nf((mlib_f32 *)u + i); fv = vis_ld_f32_nf((mlib_f32 *)v + i); da2 = da4; da3 = vis_ld_d64_nf(dpa + 2 * i); da4 = vis_ld_d64_nf(dpa + 2 * i + 1); dpp[8 * i] = vis_fpmerge(vis_read_hi(dg2), vis_read_hi(dg3)); dpp[8 * i + 1] = vis_fpmerge(vis_read_lo(dg2), vis_read_lo(dg3)); dg2 = vis_fpmerge(vis_read_lo(da0), vis_read_lo(dg)); dg3 = vis_fpmerge(vis_read_lo(db), vis_read_lo(dr)); dpp[8 * i + 2] = vis_fpmerge(vis_read_hi(dg2), vis_read_hi(dg3)); dpp[8 * i + 3] = vis_fpmerge(vis_read_lo(dg2), vis_read_lo(dg3)); dg2 = vis_fpmerge(vis_read_hi(da1), vis_read_hi(dg1)); dg3 = vis_fpmerge(vis_read_hi(db1), vis_read_hi(dr1)); dpp[8 * i + 4] = vis_fpmerge(vis_read_hi(dg2), vis_read_hi(dg3)); dpp[8 * i + 5] = vis_fpmerge(vis_read_lo(dg2), vis_read_lo(dg3)); dg2 = vis_fpmerge(vis_read_lo(da1), vis_read_lo(dg1)); dg3 = vis_fpmerge(vis_read_lo(db1), vis_read_lo(dr1)); dpp[8 * i + 6] = vis_fpmerge(vis_read_hi(dg2), vis_read_hi(dg3)); dpp[8 * i + 7] = vis_fpmerge(vis_read_lo(dg2), vis_read_lo(dg3)); du0 = vis_fmul8x16al(fu, f1); du1 = vis_fmul8x16al(fu, f4); dv1 = vis_fmul8x16al(fv, f5); dv2 = vis_fmul8x16al(fv, f8); } } else { mlib_d64 dd; #pragma pipeloop(0) for (i = 0; i < count; i++) { da0 = vis_faligndata(da2, da3); da1 = vis_faligndata(da3, da4); ddy1 = vis_fmul8x16al(vis_read_hi(dy1), f0); ddy2 = vis_fmul8x16al(vis_read_lo(dy1), f0); ddy3 = vis_fmul8x16al(vis_read_hi(dy2), f0); ddy4 = vis_fmul8x16al(vis_read_lo(dy2), f0); db = vis_fpadd16(du0, doff0); dtmp = vis_fpadd16(du1, dv1); dg = vis_fpadd16(dtmp, doff1); dr = vis_fpadd16(dv2, doff2); db1 = vis_fmul8x16au(fscale, vis_read_hi(db)); db1 = vis_fpadd16(ddy1, db1); db2 = vis_fmul8x16al(fscale, vis_read_hi(db)); db2 = vis_fpadd16(ddy2, db2); db3 = vis_fmul8x16au(fscale, vis_read_lo(db)); db3 = vis_fpadd16(ddy3, db3); db4 = vis_fmul8x16al(fscale, vis_read_lo(db)); db4 = vis_fpadd16(ddy4, db4); dg1 = vis_fmul8x16au(fscale, vis_read_hi(dg)); dg1 = vis_fpadd16(ddy1, dg1); dg2 = vis_fmul8x16al(fscale, vis_read_hi(dg)); dg2 = vis_fpadd16(ddy2, dg2); dg3 = vis_fmul8x16au(fscale, vis_read_lo(dg)); dg3 = vis_fpadd16(ddy3, dg3); dg4 = vis_fmul8x16al(fscale, vis_read_lo(dg)); dg4 = vis_fpadd16(ddy4, dg4); dr1 = vis_fmul8x16au(fscale, vis_read_hi(dr)); dr1 = vis_fpadd16(ddy1, dr1); dr2 = vis_fmul8x16al(fscale, vis_read_hi(dr)); dr2 = vis_fpadd16(ddy2, dr2); dr3 = vis_fmul8x16au(fscale, vis_read_lo(dr)); dr3 = vis_fpadd16(ddy3, dr3); dr4 = vis_fmul8x16al(fscale, vis_read_lo(dr)); dr4 = vis_fpadd16(ddy4, dr4); dr = vis_fpack16_pair(dr1, dr2); dr1 = vis_fpack16_pair(dr3, dr4); dg = vis_fpack16_pair(dg1, dg2); dg1 = vis_fpack16_pair(dg3, dg4); db = vis_fpack16_pair(db1, db2); db1 = vis_fpack16_pair(db3, db4); dg2 = vis_fpmerge(vis_read_hi(da0), vis_read_hi(dg)); dg3 = vis_fpmerge(vis_read_hi(db), vis_read_hi(dr)); dy1 = vis_ld_d64_nf((mlib_d64 *)y + 2 * i); dy2 = vis_ld_d64_nf((mlib_d64 *)y + 2 * i + 1); fu = vis_ld_f32_nf((mlib_f32 *)u + i); fv = vis_ld_f32_nf((mlib_f32 *)v + i); da2 = da4; da3 = vis_ld_d64_nf(dpa + 2 * i); da4 = vis_ld_d64_nf(dpa + 2 * i + 1); dd = vis_fpmerge(vis_read_hi(dg2), vis_read_hi(dg3)); ((mlib_f32 *)dpp)[16 * i] = vis_read_hi(dd); ((mlib_f32 *)dpp)[16 * i + 1] = vis_read_lo(dd); dd = vis_fpmerge(vis_read_lo(dg2), vis_read_lo(dg3)); ((mlib_f32 *)dpp)[16 * i + 2] = vis_read_hi(dd); ((mlib_f32 *)dpp)[16 * i + 3] = vis_read_lo(dd); dg2 = vis_fpmerge(vis_read_lo(da0), vis_read_lo(dg)); dg3 = vis_fpmerge(vis_read_lo(db), vis_read_lo(dr)); dd = vis_fpmerge(vis_read_hi(dg2), vis_read_hi(dg3)); ((mlib_f32 *)dpp)[16 * i + 4] = vis_read_hi(dd); ((mlib_f32 *)dpp)[16 * i + 5] = vis_read_lo(dd); dd = vis_fpmerge(vis_read_lo(dg2), vis_read_lo(dg3)); ((mlib_f32 *)dpp)[16 * i + 6] = vis_read_hi(dd); ((mlib_f32 *)dpp)[16 * i + 7] = vis_read_lo(dd); dg2 = vis_fpmerge(vis_read_hi(da1), vis_read_hi(dg1)); dg3 = vis_fpmerge(vis_read_hi(db1), vis_read_hi(dr1)); dd = vis_fpmerge(vis_read_hi(dg2), vis_read_hi(dg3)); ((mlib_f32 *)dpp)[16 * i + 8] = vis_read_hi(dd); ((mlib_f32 *)dpp)[16 * i + 9] = vis_read_lo(dd); dd = vis_fpmerge(vis_read_lo(dg2), vis_read_lo(dg3)); ((mlib_f32 *)dpp)[16 * i + 10] = vis_read_hi(dd); ((mlib_f32 *)dpp)[16 * i + 11] = vis_read_lo(dd); dg2 = vis_fpmerge(vis_read_lo(da1), vis_read_lo(dg1)); dg3 = vis_fpmerge(vis_read_lo(db1), vis_read_lo(dr1)); dd = vis_fpmerge(vis_read_hi(dg2), vis_read_hi(dg3)); ((mlib_f32 *)dpp)[16 * i + 12] = vis_read_hi(dd); ((mlib_f32 *)dpp)[16 * i + 13] = vis_read_lo(dd); dd = vis_fpmerge(vis_read_lo(dg2), vis_read_lo(dg3)); ((mlib_f32 *)dpp)[16 * i + 14] = vis_read_hi(dd); ((mlib_f32 *)dpp)[16 * i + 15] = vis_read_lo(dd); du0 = vis_fmul8x16al(fu, f1); du1 = vis_fmul8x16al(fu, f4); dv1 = vis_fmul8x16al(fv, f5); dv2 = vis_fmul8x16al(fv, f8); } } if (left) { mlib_d64 res_buf[8]; da0 = vis_faligndata(da2, da3); da1 = vis_faligndata(da3, da4); ddy1 = vis_fmul8x16al(vis_read_hi(dy1), f0); ddy2 = vis_fmul8x16al(vis_read_lo(dy1), f0); ddy3 = vis_fmul8x16al(vis_read_hi(dy2), f0); ddy4 = vis_fmul8x16al(vis_read_lo(dy2), f0); db = vis_fpadd16(du0, doff0); dtmp = vis_fpadd16(du1, dv1); dg = vis_fpadd16(dtmp, doff1); dr = vis_fpadd16(dv2, doff2); db1 = vis_fmul8x16au(fscale, vis_read_hi(db)); db1 = vis_fpadd16(ddy1, db1); db2 = vis_fmul8x16al(fscale, vis_read_hi(db)); db2 = vis_fpadd16(ddy2, db2); db3 = vis_fmul8x16au(fscale, vis_read_lo(db)); db3 = vis_fpadd16(ddy3, db3); db4 = vis_fmul8x16al(fscale, vis_read_lo(db)); db4 = vis_fpadd16(ddy4, db4); dg1 = vis_fmul8x16au(fscale, vis_read_hi(dg)); dg1 = vis_fpadd16(ddy1, dg1); dg2 = vis_fmul8x16al(fscale, vis_read_hi(dg)); dg2 = vis_fpadd16(ddy2, dg2); dg3 = vis_fmul8x16au(fscale, vis_read_lo(dg)); dg3 = vis_fpadd16(ddy3, dg3); dg4 = vis_fmul8x16al(fscale, vis_read_lo(dg)); dg4 = vis_fpadd16(ddy4, dg4); dr1 = vis_fmul8x16au(fscale, vis_read_hi(dr)); dr1 = vis_fpadd16(ddy1, dr1); dr2 = vis_fmul8x16al(fscale, vis_read_hi(dr)); dr2 = vis_fpadd16(ddy2, dr2); dr3 = vis_fmul8x16au(fscale, vis_read_lo(dr)); dr3 = vis_fpadd16(ddy3, dr3); dr4 = vis_fmul8x16al(fscale, vis_read_lo(dr)); dr4 = vis_fpadd16(ddy4, dr4); dr = vis_fpack16_pair(dr1, dr2); dr1 = vis_fpack16_pair(dr3, dr4); dg = vis_fpack16_pair(dg1, dg2); dg1 = vis_fpack16_pair(dg3, dg4); db = vis_fpack16_pair(db1, db2); db1 = vis_fpack16_pair(db3, db4); dg2 = vis_fpmerge(vis_read_hi(da0), vis_read_hi(dg)); dg3 = vis_fpmerge(vis_read_hi(db), vis_read_hi(dr)); res_buf[0] = vis_fpmerge(vis_read_hi(dg2), vis_read_hi(dg3)); res_buf[1] = vis_fpmerge(vis_read_lo(dg2), vis_read_lo(dg3)); dg2 = vis_fpmerge(vis_read_lo(da0), vis_read_lo(dg)); dg3 = vis_fpmerge(vis_read_lo(db), vis_read_lo(dr)); res_buf[2] = vis_fpmerge(vis_read_hi(dg2), vis_read_hi(dg3)); res_buf[3] = vis_fpmerge(vis_read_lo(dg2), vis_read_lo(dg3)); dg2 = vis_fpmerge(vis_read_hi(da1), vis_read_hi(dg1)); dg3 = vis_fpmerge(vis_read_hi(db1), vis_read_hi(dr1)); res_buf[4] = vis_fpmerge(vis_read_hi(dg2), vis_read_hi(dg3)); res_buf[5] = vis_fpmerge(vis_read_lo(dg2), vis_read_lo(dg3)); dg2 = vis_fpmerge(vis_read_lo(da1), vis_read_lo(dg1)); dg3 = vis_fpmerge(vis_read_lo(db1), vis_read_lo(dr1)); res_buf[6] = vis_fpmerge(vis_read_hi(dg2), vis_read_hi(dg3)); res_buf[7] = vis_fpmerge(vis_read_lo(dg2), vis_read_lo(dg3)); for (i = 0; i < left; i++) ((mlib_f32 *)dpp)[16 * count + i] = ((mlib_f32 *)res_buf)[i]; } }
mlib_status mlib_convMxN_8nw_mask(mlib_image *dst, const mlib_image *src, mlib_s32 m, mlib_s32 n, mlib_s32 dm, mlib_s32 dn, const mlib_s32 *kern, mlib_s32 scale, mlib_s32 cmask) { mlib_d64 *buffs_local[3 * (MAX_N + 1)], **buffs = buffs_local, **buff; mlib_d64 *buff0, *buff1, *buff2, *buff3, *buffn, *buffd, *buffe; mlib_d64 s00, s01, s10, s11, s20, s21, s30, s31, s0, s1, s2, s3; mlib_d64 d00, d01, d10, d11, d20, d21, d30, d31; mlib_d64 dd, d0, d1; mlib_s32 ik, jk, ik_last, jk_size, coff, off, doff; mlib_u8 *sl, *sp, *dl; mlib_s32 hgt = mlib_ImageGetHeight(src); mlib_s32 wid = mlib_ImageGetWidth(src); mlib_s32 sll = mlib_ImageGetStride(src); mlib_s32 dll = mlib_ImageGetStride(dst); mlib_u8 *adr_src = (mlib_u8 *) mlib_ImageGetData(src); mlib_u8 *adr_dst = (mlib_u8 *) mlib_ImageGetData(dst); mlib_s32 ssize, xsize, dsize, esize, buff_ind; mlib_d64 *pbuff, *dp; mlib_f32 *karr = (mlib_f32 *) kern; mlib_s32 gsr_scale = (31 - scale) << 3; mlib_d64 drnd = vis_to_double_dup(mlib_round_8[31 - scale]); mlib_s32 i, j, l, chan, testchan; mlib_s32 nchan = mlib_ImageGetChannels(dst); void (*p_proc_load) (const mlib_u8 *, mlib_u8 *, mlib_s32, mlib_s32); void (*p_proc_store) (const mlib_u8 *, mlib_u8 *, mlib_s32, mlib_s32); if (n > MAX_N) { buffs = mlib_malloc(3 * (n + 1) * sizeof(mlib_d64 *)); if (buffs == NULL) return MLIB_FAILURE; } buff = buffs + 2 * (n + 1); adr_dst += dn * dll + dm * nchan; ssize = wid; dsize = (ssize + 7) / 8; esize = dsize + 4; pbuff = mlib_malloc((n + 4) * esize * sizeof(mlib_d64)); if (pbuff == NULL) { if (buffs != buffs_local) mlib_free(buffs); return MLIB_FAILURE; } for (i = 0; i < (n + 1); i++) buffs[i] = pbuff + i * esize; for (i = 0; i < (n + 1); i++) buffs[(n + 1) + i] = buffs[i]; buffd = buffs[n] + esize; buffe = buffd + 2 * esize; hgt -= (n - 1); xsize = ssize - (m - 1); vis_write_gsr(gsr_scale + 7); if (nchan == 2) { p_proc_load = &mlib_v_ImageChannelExtract_U8_21_D1; p_proc_store = &mlib_v_ImageChannelInsert_U8_12_D1; } else if (nchan == 3) { p_proc_load = &mlib_v_ImageChannelExtract_U8_31_D1; p_proc_store = &mlib_v_ImageChannelInsert_U8_13_D1; } else { p_proc_load = &mlib_v_ImageChannelExtract_U8_41_D1; p_proc_store = &mlib_v_ImageChannelInsert_U8_14_D1; } testchan = 1; for (chan = 0; chan < nchan; chan++) { buff_ind = 0; sl = adr_src; dl = adr_dst; if ((cmask & testchan) == 0) { testchan <<= 1; continue; } for (l = 0; l < n; l++) { mlib_d64 *buffn = buffs[l]; sp = sl + l * sll; (*p_proc_load) ((mlib_u8 *) sp, (mlib_u8 *) buffn, ssize, testchan); } /* init buffer */ #pragma pipeloop(0) for (i = 0; i < (xsize + 7) / 8; i++) { buffd[2 * i] = drnd; buffd[2 * i + 1] = drnd; } for (j = 0; j < hgt; j++) { mlib_d64 **buffc = buffs + buff_ind; mlib_f32 *pk = karr, k0, k1, k2, k3; sp = sl + n * sll; for (l = 0; l < n; l++) { buff[l] = buffc[l]; } buffn = buffc[n]; (*p_proc_load) ((mlib_u8 *) sp, (mlib_u8 *) buffn, ssize, testchan); ik_last = (m - 1); for (jk = 0; jk < n; jk += jk_size) { jk_size = n - jk; if (jk_size >= 6) jk_size = 4; if (jk_size == 5) jk_size = 3; coff = 0; if (jk_size == 1) { for (ik = 0; ik < m; ik++, coff++) { if (!jk && ik == ik_last) continue; k0 = pk[ik]; doff = coff / 8; buff0 = buff[jk] + doff; off = coff & 7; vis_write_gsr(gsr_scale + off); s01 = buff0[0]; #pragma pipeloop(0) for (i = 0; i < (xsize + 7) / 8; i++) { s00 = s01; s01 = buff0[i + 1]; s0 = vis_faligndata(s00, s01); d00 = vis_fmul8x16au(vis_read_hi(s0), k0); d01 = vis_fmul8x16au(vis_read_lo(s0), k0); d0 = buffd[2 * i]; d1 = buffd[2 * i + 1]; d0 = vis_fpadd16(d00, d0); d1 = vis_fpadd16(d01, d1); buffd[2 * i] = d0; buffd[2 * i + 1] = d1; } } pk += m; } else if (jk_size == 2) { for (ik = 0; ik < m; ik++, coff++) { if (!jk && ik == ik_last) continue; k0 = pk[ik]; k1 = pk[ik + m]; doff = coff / 8; buff0 = buff[jk] + doff; buff1 = buff[jk + 1] + doff; off = coff & 7; vis_write_gsr(gsr_scale + off); s01 = buff0[0]; s11 = buff1[0]; #pragma pipeloop(0) for (i = 0; i < (xsize + 7) / 8; i++) { s00 = s01; s10 = s11; s01 = buff0[i + 1]; s11 = buff1[i + 1]; s0 = vis_faligndata(s00, s01); s1 = vis_faligndata(s10, s11); d00 = vis_fmul8x16au(vis_read_hi(s0), k0); d01 = vis_fmul8x16au(vis_read_lo(s0), k0); d10 = vis_fmul8x16au(vis_read_hi(s1), k1); d11 = vis_fmul8x16au(vis_read_lo(s1), k1); d0 = buffd[2 * i]; d1 = buffd[2 * i + 1]; d0 = vis_fpadd16(d00, d0); d0 = vis_fpadd16(d10, d0); d1 = vis_fpadd16(d01, d1); d1 = vis_fpadd16(d11, d1); buffd[2 * i] = d0; buffd[2 * i + 1] = d1; } } pk += 2 * m; } else if (jk_size == 3) { for (ik = 0; ik < m; ik++, coff++) { if (!jk && ik == ik_last) continue; k0 = pk[ik]; k1 = pk[ik + m]; k2 = pk[ik + 2 * m]; doff = coff / 8; buff0 = buff[jk] + doff; buff1 = buff[jk + 1] + doff; buff2 = buff[jk + 2] + doff; off = coff & 7; vis_write_gsr(gsr_scale + off); if (off == 0) { #pragma pipeloop(0) for (i = 0; i < (xsize + 7) / 8; i++) { d0 = buffd[2 * i]; d1 = buffd[2 * i + 1]; s0 = buff0[i]; s1 = buff1[i]; s2 = buff2[i]; d00 = vis_fmul8x16au(vis_read_hi(s0), k0); d01 = vis_fmul8x16au(vis_read_lo(s0), k0); d10 = vis_fmul8x16au(vis_read_hi(s1), k1); d11 = vis_fmul8x16au(vis_read_lo(s1), k1); d20 = vis_fmul8x16au(vis_read_hi(s2), k2); d21 = vis_fmul8x16au(vis_read_lo(s2), k2); d00 = vis_fpadd16(d00, d10); d0 = vis_fpadd16(d20, d0); d0 = vis_fpadd16(d00, d0); d01 = vis_fpadd16(d01, d11); d1 = vis_fpadd16(d21, d1); d1 = vis_fpadd16(d01, d1); buffd[2 * i] = d0; buffd[2 * i + 1] = d1; } } else if (off == 4) { s01 = buff0[0]; s11 = buff1[0]; s21 = buff2[0]; #pragma pipeloop(0) for (i = 0; i < (xsize + 7) / 8; i++) { d0 = buffd[2 * i]; d1 = buffd[2 * i + 1]; s00 = s01; s10 = s11; s20 = s21; s01 = buff0[i + 1]; s11 = buff1[i + 1]; s21 = buff2[i + 1]; d00 = vis_fmul8x16au(vis_read_lo(s00), k0); d01 = vis_fmul8x16au(vis_read_hi(s01), k0); d10 = vis_fmul8x16au(vis_read_lo(s10), k1); d11 = vis_fmul8x16au(vis_read_hi(s11), k1); d20 = vis_fmul8x16au(vis_read_lo(s20), k2); d21 = vis_fmul8x16au(vis_read_hi(s21), k2); d00 = vis_fpadd16(d00, d10); d0 = vis_fpadd16(d20, d0); d0 = vis_fpadd16(d00, d0); d01 = vis_fpadd16(d01, d11); d1 = vis_fpadd16(d21, d1); d1 = vis_fpadd16(d01, d1); buffd[2 * i] = d0; buffd[2 * i + 1] = d1; } } else { s01 = buff0[0]; s11 = buff1[0]; s21 = buff2[0]; #pragma pipeloop(0) for (i = 0; i < (xsize + 7) / 8; i++) { d0 = buffd[2 * i]; d1 = buffd[2 * i + 1]; s00 = s01; s10 = s11; s20 = s21; s01 = buff0[i + 1]; s11 = buff1[i + 1]; s21 = buff2[i + 1]; s0 = vis_faligndata(s00, s01); s1 = vis_faligndata(s10, s11); s2 = vis_faligndata(s20, s21); d00 = vis_fmul8x16au(vis_read_hi(s0), k0); d01 = vis_fmul8x16au(vis_read_lo(s0), k0); d10 = vis_fmul8x16au(vis_read_hi(s1), k1); d11 = vis_fmul8x16au(vis_read_lo(s1), k1); d20 = vis_fmul8x16au(vis_read_hi(s2), k2); d21 = vis_fmul8x16au(vis_read_lo(s2), k2); d00 = vis_fpadd16(d00, d10); d0 = vis_fpadd16(d20, d0); d0 = vis_fpadd16(d00, d0); d01 = vis_fpadd16(d01, d11); d1 = vis_fpadd16(d21, d1); d1 = vis_fpadd16(d01, d1); buffd[2 * i] = d0; buffd[2 * i + 1] = d1; } } } pk += 3 * m; } else { /* jk_size == 4 */ for (ik = 0; ik < m; ik++, coff++) { if (!jk && ik == ik_last) continue; k0 = pk[ik]; k1 = pk[ik + m]; k2 = pk[ik + 2 * m]; k3 = pk[ik + 3 * m]; doff = coff / 8; buff0 = buff[jk] + doff; buff1 = buff[jk + 1] + doff; buff2 = buff[jk + 2] + doff; buff3 = buff[jk + 3] + doff; off = coff & 7; vis_write_gsr(gsr_scale + off); if (off == 0) { #pragma pipeloop(0) for (i = 0; i < (xsize + 7) / 8; i++) { d0 = buffd[2 * i]; d1 = buffd[2 * i + 1]; s0 = buff0[i]; s1 = buff1[i]; s2 = buff2[i]; s3 = buff3[i]; d00 = vis_fmul8x16au(vis_read_hi(s0), k0); d01 = vis_fmul8x16au(vis_read_lo(s0), k0); d10 = vis_fmul8x16au(vis_read_hi(s1), k1); d11 = vis_fmul8x16au(vis_read_lo(s1), k1); d20 = vis_fmul8x16au(vis_read_hi(s2), k2); d21 = vis_fmul8x16au(vis_read_lo(s2), k2); d30 = vis_fmul8x16au(vis_read_hi(s3), k3); d31 = vis_fmul8x16au(vis_read_lo(s3), k3); d00 = vis_fpadd16(d00, d10); d20 = vis_fpadd16(d20, d30); d0 = vis_fpadd16(d0, d00); d0 = vis_fpadd16(d0, d20); d01 = vis_fpadd16(d01, d11); d21 = vis_fpadd16(d21, d31); d1 = vis_fpadd16(d1, d01); d1 = vis_fpadd16(d1, d21); buffd[2 * i] = d0; buffd[2 * i + 1] = d1; } } else if (off == 4) { s01 = buff0[0]; s11 = buff1[0]; s21 = buff2[0]; s31 = buff3[0]; #pragma pipeloop(0) for (i = 0; i < (xsize + 7) / 8; i++) { d0 = buffd[2 * i]; d1 = buffd[2 * i + 1]; s00 = s01; s10 = s11; s20 = s21; s30 = s31; s01 = buff0[i + 1]; s11 = buff1[i + 1]; s21 = buff2[i + 1]; s31 = buff3[i + 1]; d00 = vis_fmul8x16au(vis_read_lo(s00), k0); d01 = vis_fmul8x16au(vis_read_hi(s01), k0); d10 = vis_fmul8x16au(vis_read_lo(s10), k1); d11 = vis_fmul8x16au(vis_read_hi(s11), k1); d20 = vis_fmul8x16au(vis_read_lo(s20), k2); d21 = vis_fmul8x16au(vis_read_hi(s21), k2); d30 = vis_fmul8x16au(vis_read_lo(s30), k3); d31 = vis_fmul8x16au(vis_read_hi(s31), k3); d00 = vis_fpadd16(d00, d10); d20 = vis_fpadd16(d20, d30); d0 = vis_fpadd16(d0, d00); d0 = vis_fpadd16(d0, d20); d01 = vis_fpadd16(d01, d11); d21 = vis_fpadd16(d21, d31); d1 = vis_fpadd16(d1, d01); d1 = vis_fpadd16(d1, d21); buffd[2 * i] = d0; buffd[2 * i + 1] = d1; } } else { s01 = buff0[0]; s11 = buff1[0]; s21 = buff2[0]; s31 = buff3[0]; #pragma pipeloop(0) for (i = 0; i < (xsize + 7) / 8; i++) { d0 = buffd[2 * i]; d1 = buffd[2 * i + 1]; s00 = s01; s10 = s11; s20 = s21; s30 = s31; s01 = buff0[i + 1]; s11 = buff1[i + 1]; s21 = buff2[i + 1]; s31 = buff3[i + 1]; s0 = vis_faligndata(s00, s01); s1 = vis_faligndata(s10, s11); s2 = vis_faligndata(s20, s21); s3 = vis_faligndata(s30, s31); d00 = vis_fmul8x16au(vis_read_hi(s0), k0); d01 = vis_fmul8x16au(vis_read_lo(s0), k0); d10 = vis_fmul8x16au(vis_read_hi(s1), k1); d11 = vis_fmul8x16au(vis_read_lo(s1), k1); d20 = vis_fmul8x16au(vis_read_hi(s2), k2); d21 = vis_fmul8x16au(vis_read_lo(s2), k2); d30 = vis_fmul8x16au(vis_read_hi(s3), k3); d31 = vis_fmul8x16au(vis_read_lo(s3), k3); d00 = vis_fpadd16(d00, d10); d20 = vis_fpadd16(d20, d30); d0 = vis_fpadd16(d0, d00); d0 = vis_fpadd16(d0, d20); d01 = vis_fpadd16(d01, d11); d21 = vis_fpadd16(d21, d31); d1 = vis_fpadd16(d1, d01); d1 = vis_fpadd16(d1, d21); buffd[2 * i] = d0; buffd[2 * i + 1] = d1; } } } pk += 4 * m; } } /***************************************** ***************************************** ** Final iteration ** ***************************************** *****************************************/ jk_size = n; if (jk_size >= 6) jk_size = 4; if (jk_size == 5) jk_size = 3; k0 = karr[ik_last]; k1 = karr[ik_last + m]; k2 = karr[ik_last + 2 * m]; k3 = karr[ik_last + 3 * m]; off = ik_last; doff = off / 8; off &= 7; buff0 = buff[0] + doff; buff1 = buff[1] + doff; buff2 = buff[2] + doff; buff3 = buff[3] + doff; vis_write_gsr(gsr_scale + off); if (jk_size == 1) { dp = buffe; s01 = buff0[0]; #pragma pipeloop(0) for (i = 0; i < (xsize + 7) / 8; i++) { s00 = s01; s01 = buff0[i + 1]; s0 = vis_faligndata(s00, s01); d00 = vis_fmul8x16au(vis_read_hi(s0), k0); d01 = vis_fmul8x16au(vis_read_lo(s0), k0); d0 = buffd[2 * i]; d1 = buffd[2 * i + 1]; d0 = vis_fpadd16(d0, d00); d1 = vis_fpadd16(d1, d01); dd = vis_fpack16_pair(d0, d1); dp[i] = dd; buffd[2 * i] = drnd; buffd[2 * i + 1] = drnd; } } else if (jk_size == 2) { dp = buffe; s01 = buff0[0]; s11 = buff1[0]; #pragma pipeloop(0) for (i = 0; i < (xsize + 7) / 8; i++) { s00 = s01; s10 = s11; s01 = buff0[i + 1]; s11 = buff1[i + 1]; s0 = vis_faligndata(s00, s01); s1 = vis_faligndata(s10, s11); d00 = vis_fmul8x16au(vis_read_hi(s0), k0); d01 = vis_fmul8x16au(vis_read_lo(s0), k0); d10 = vis_fmul8x16au(vis_read_hi(s1), k1); d11 = vis_fmul8x16au(vis_read_lo(s1), k1); d0 = buffd[2 * i]; d1 = buffd[2 * i + 1]; d0 = vis_fpadd16(d0, d00); d0 = vis_fpadd16(d0, d10); d1 = vis_fpadd16(d1, d01); d1 = vis_fpadd16(d1, d11); dd = vis_fpack16_pair(d0, d1); dp[i] = dd; buffd[2 * i] = drnd; buffd[2 * i + 1] = drnd; } } else if (jk_size == 3) { dp = buffe; s01 = buff0[0]; s11 = buff1[0]; s21 = buff2[0]; #pragma pipeloop(0) for (i = 0; i < (xsize + 7) / 8; i++) { s00 = s01; s10 = s11; s20 = s21; s01 = buff0[i + 1]; s11 = buff1[i + 1]; s21 = buff2[i + 1]; s0 = vis_faligndata(s00, s01); s1 = vis_faligndata(s10, s11); s2 = vis_faligndata(s20, s21); d00 = vis_fmul8x16au(vis_read_hi(s0), k0); d01 = vis_fmul8x16au(vis_read_lo(s0), k0); d10 = vis_fmul8x16au(vis_read_hi(s1), k1); d11 = vis_fmul8x16au(vis_read_lo(s1), k1); d20 = vis_fmul8x16au(vis_read_hi(s2), k2); d21 = vis_fmul8x16au(vis_read_lo(s2), k2); d0 = buffd[2 * i]; d1 = buffd[2 * i + 1]; d0 = vis_fpadd16(d0, d00); d0 = vis_fpadd16(d0, d10); d0 = vis_fpadd16(d0, d20); d1 = vis_fpadd16(d1, d01); d1 = vis_fpadd16(d1, d11); d1 = vis_fpadd16(d1, d21); dd = vis_fpack16_pair(d0, d1); dp[i] = dd; buffd[2 * i] = drnd; buffd[2 * i + 1] = drnd; } } else { /* if (jk_size == 4) */ dp = buffe; s01 = buff0[0]; s11 = buff1[0]; s21 = buff2[0]; s31 = buff3[0]; #pragma pipeloop(0) for (i = 0; i < (xsize + 7) / 8; i++) { s00 = s01; s10 = s11; s20 = s21; s30 = s31; s01 = buff0[i + 1]; s11 = buff1[i + 1]; s21 = buff2[i + 1]; s31 = buff3[i + 1]; s0 = vis_faligndata(s00, s01); s1 = vis_faligndata(s10, s11); s2 = vis_faligndata(s20, s21); s3 = vis_faligndata(s30, s31); d00 = vis_fmul8x16au(vis_read_hi(s0), k0); d01 = vis_fmul8x16au(vis_read_lo(s0), k0); d10 = vis_fmul8x16au(vis_read_hi(s1), k1); d11 = vis_fmul8x16au(vis_read_lo(s1), k1); d20 = vis_fmul8x16au(vis_read_hi(s2), k2); d21 = vis_fmul8x16au(vis_read_lo(s2), k2); d30 = vis_fmul8x16au(vis_read_hi(s3), k3); d31 = vis_fmul8x16au(vis_read_lo(s3), k3); d0 = buffd[2 * i]; d1 = buffd[2 * i + 1]; d0 = vis_fpadd16(d0, d00); d0 = vis_fpadd16(d0, d10); d0 = vis_fpadd16(d0, d20); d0 = vis_fpadd16(d0, d30); d1 = vis_fpadd16(d1, d01); d1 = vis_fpadd16(d1, d11); d1 = vis_fpadd16(d1, d21); d1 = vis_fpadd16(d1, d31); dd = vis_fpack16_pair(d0, d1); dp[i] = dd; buffd[2 * i] = drnd; buffd[2 * i + 1] = drnd; } } (*p_proc_store) ((mlib_u8 *) buffe, (mlib_u8 *) dl, xsize, testchan); sl += sll; dl += dll; buff_ind++; if (buff_ind >= (n + 1)) buff_ind = 0; } testchan <<= 1; } mlib_free(pbuff); if (buffs != buffs_local) mlib_free(buffs); return MLIB_SUCCESS; }
static mlib_status mlib_v_VideoColorYUV2ABGR422_nonalign( mlib_u8 *abgr, const mlib_u8 *y, const mlib_u8 *u, const mlib_u8 *v, mlib_s32 width, mlib_s32 height, mlib_s32 abgr_stride, mlib_s32 y_stride, mlib_s32 uv_stride) { /* pointers to src address */ mlib_u8 *sp2, *sp3, *sl2, *sl3; /* pointers to src address */ mlib_u8 *sp1, *sl1; /* pointers to dst address */ mlib_u8 *dp, *dl, *dend; /* all. pointer to y */ mlib_d64 *spy; /* all. pointer to dst */ mlib_d64 *dpp; /* u, v data */ mlib_f32 fu0, fu1, fv0, fv1; /* y data */ mlib_d64 dy0, dy1, dy3; mlib_d64 du, dv; /* (1.1644, 1.5966)*8192 */ mlib_f32 k12 = vis_to_float(0x25433317); /* (-.3920, -.8132)*8192 */ mlib_f32 k34 = vis_to_float(0xf375e5fa); /* 2.0184*8192 */ mlib_f32 k5 = vis_to_float(0x1004097); mlib_d64 k_222_9952 = vis_to_double(0x1be01be0, 0x1be01be0); mlib_d64 k_135_6352 = vis_to_double(0x10f410f4, 0x10f410f4); mlib_d64 k_276_9856 = vis_to_double(0x22a022a0, 0x22a022a0); mlib_d64 u_3920_hi, u_20184_hi, v_15966_hi, v_8132_hi; mlib_d64 u_3920_lo, u_20184_lo, v_15966_lo, v_8132_lo; mlib_d64 y_11644_hi, y_11644_lo; mlib_d64 r_hi, r_lo, g_hi, g_lo, b_hi, b_lo; mlib_d64 temp_r_hi, temp_r_lo, temp_g_hi, temp_g_lo, temp_b_hi, temp_b_lo; mlib_f32 red_hi, red_lo, green_hi, green_lo, blue_hi, blue_lo; mlib_d64 blue_red_hi, x_green_hi, blue_red_lo, x_green_lo; mlib_d64 dd, dd0, dd1; /* loop variable */ mlib_s32 i, j; /* alpha_ch. is not written */ mlib_s32 emask = 0x7777; mlib_s32 emask1; mlib_s32 off; mlib_f32 *dfu, *dfv; mlib_d64 du0, du1, dv0, dv1; mlib_s32 off2, off3; mlib_s32 inc; /* * initialize GSR scale factor */ vis_write_gsr(2 << 3); sp1 = sl1 = (mlib_u8 *)y; sp2 = sl2 = (mlib_u8 *)u; sp3 = sl3 = (mlib_u8 *)v; dl = dp = (mlib_u8 *)abgr; /* * row loop */ for (j = 0; j < height; j++) { spy = (mlib_d64 *)vis_alignaddr(sp1, 0); dpp = (mlib_d64 *)vis_alignaddr(dp, 0); dfu = (mlib_f32 *)((mlib_addr)sp2 & ~3); off2 = (sp2 - (mlib_u8 *)dfu) * 2; dfv = (mlib_f32 *)((mlib_addr)sp3 & ~3); off3 = (sp3 - (mlib_u8 *)dfv) * 2; dend = dp + width * 4 - 1; emask1 = vis_edge8(dp, dend); i = dp - (mlib_u8 *)dpp; emask >>= i; inc = (emask1 != 0xff); emask1 &= emask; off = 8 - i; vis_alignaddr((void *)off2, 0); fu0 = vis_ld_f32_nf(dfu); dfu++; fu1 = vis_ld_f32_nf(dfu); dfu++; du0 = vis_fpmerge(fu0, fu0); du1 = vis_fpmerge(fu1, fu1); du = vis_faligndata(du0, du1); du0 = du1; vis_alignaddr((void *)off3, 0); fv0 = vis_ld_f32_nf(dfv); dfv++; fv1 = vis_ld_f32_nf(dfv); dfv++; dv0 = vis_fpmerge(fv0, fv0); dv1 = vis_fpmerge(fv1, fv1); dv = vis_faligndata(dv0, dv1); dv0 = dv1; /* U*(-0.3920); */ u_3920_hi = vis_fmul8x16au(vis_read_hi(du), k34); /* V*(-0.8132); */ v_8132_hi = vis_fmul8x16al(vis_read_hi(dv), k34); /* U*(-0.3920); */ u_3920_lo = vis_fmul8x16au(vis_read_lo(du), k34); /* V*(-0.8132); */ v_8132_lo = vis_fmul8x16al(vis_read_lo(dv), k34); vis_alignaddr(sp1, 0); dy0 = vis_ld_d64_nf(spy); spy++; dy3 = vis_ld_d64_nf(spy); spy++; dy1 = vis_faligndata(dy0, dy3); dy0 = dy3; /* * 16-pixel column loop */ #pragma pipeloop(0) for (i = 0; i <= width - 8; i += 8) { /* U*2.0184 */ u_20184_hi = vis_fmul8x16al(vis_read_hi(du), k5); g_hi = vis_fpadd16(u_3920_hi, v_8132_hi); u_20184_lo = vis_fmul8x16al(vis_read_lo(du), k5); g_hi = vis_fpadd16(g_hi, k_135_6352); /* V*1.5966 */ v_15966_hi = vis_fmul8x16al(vis_read_hi(dv), k12); g_lo = vis_fpadd16(u_3920_lo, v_8132_lo); v_15966_lo = vis_fmul8x16al(vis_read_lo(dv), k12); g_lo = vis_fpadd16(g_lo, k_135_6352); /* Y*1.1644 */ y_11644_hi = vis_fmul8x16au(vis_read_hi(dy1), k12); b_hi = vis_fpsub16(u_20184_hi, k_276_9856); /* Y*1.1644 */ y_11644_lo = vis_fmul8x16au(vis_read_lo(dy1), k12); b_lo = vis_fpsub16(u_20184_lo, k_276_9856); r_hi = vis_fpsub16(v_15966_hi, k_222_9952); r_lo = vis_fpsub16(v_15966_lo, k_222_9952); temp_g_hi = vis_fpadd16(g_hi, y_11644_hi); temp_b_hi = vis_fpadd16(b_hi, y_11644_hi); green_hi = vis_fpack16(temp_g_hi); temp_r_hi = vis_fpadd16(r_hi, y_11644_hi); blue_hi = vis_fpack16(temp_b_hi); temp_g_lo = vis_fpadd16(g_lo, y_11644_lo); red_hi = vis_fpack16(temp_r_hi); temp_b_lo = vis_fpadd16(b_lo, y_11644_lo); vis_alignaddr((void *)off2, 0); fu1 = vis_ld_f32_nf(dfu); dfu++; du1 = vis_fpmerge(fu1, fu1); du = vis_faligndata(du0, du1); du0 = du1; green_lo = vis_fpack16(temp_g_lo); temp_r_lo = vis_fpadd16(r_lo, y_11644_lo); blue_lo = vis_fpack16(temp_b_lo); x_green_hi = vis_fmul8x16au(green_hi, k5); red_lo = vis_fpack16(temp_r_lo); blue_red_hi = vis_fpmerge(blue_hi, red_hi); x_green_lo = vis_fmul8x16au(green_lo, k5); blue_red_lo = vis_fpmerge(blue_lo, red_lo); vis_alignaddr((void *)off3, 0); fv1 = vis_ld_f32_nf(dfv); dfv++; dv1 = vis_fpmerge(fv1, fv1); dv = vis_faligndata(dv0, dv1); dv0 = dv1; vis_alignaddr((void *)off, 0); /* U*(-0.3920); */ u_3920_hi = vis_fmul8x16au(vis_read_hi(du), k34); dd1 = vis_fpmerge(vis_read_hi(x_green_hi), vis_read_hi(blue_red_hi)); dd = vis_faligndata(dd0, dd1); vis_pst_8(dd, dpp, emask1); dpp += inc; inc = 1; /* V*(-0.8132); */ v_8132_hi = vis_fmul8x16al(vis_read_hi(dv), k34); dd0 = vis_fpmerge(vis_read_lo(x_green_hi), vis_read_lo(blue_red_hi)); dd = vis_faligndata(dd1, dd0); vis_pst_8(dd, dpp++, emask); u_3920_lo = vis_fmul8x16au(vis_read_lo(du), k34); dd1 = vis_fpmerge(vis_read_hi(x_green_lo), vis_read_hi(blue_red_lo)); dd = vis_faligndata(dd0, dd1); vis_pst_8(dd, dpp++, emask); v_8132_lo = vis_fmul8x16al(vis_read_lo(dv), k34); dd0 = vis_fpmerge(vis_read_lo(x_green_lo), vis_read_lo(blue_red_lo)); dd = vis_faligndata(dd1, dd0); vis_pst_8(dd, dpp++, emask); vis_alignaddr(sp1, 0); dy3 = vis_ld_d64_nf(spy); spy++; dy1 = vis_faligndata(dy0, dy3); dy0 = dy3; emask1 = emask; } if (i < width) { vis_alignaddr((void *)off, 0); /* U*2.0184 */ u_20184_hi = vis_fmul8x16al(vis_read_hi(du), k5); g_hi = vis_fpadd16(u_3920_hi, v_8132_hi); u_20184_lo = vis_fmul8x16al(vis_read_lo(du), k5); g_hi = vis_fpadd16(g_hi, k_135_6352); /* V*1.5966 */ v_15966_hi = vis_fmul8x16al(vis_read_hi(dv), k12); g_lo = vis_fpadd16(u_3920_lo, v_8132_lo); v_15966_lo = vis_fmul8x16al(vis_read_lo(dv), k12); g_lo = vis_fpadd16(g_lo, k_135_6352); /* Y*1.1644 */ y_11644_hi = vis_fmul8x16au(vis_read_hi(dy1), k12); b_hi = vis_fpsub16(u_20184_hi, k_276_9856); /* Y*1.1644 */ y_11644_lo = vis_fmul8x16au(vis_read_lo(dy1), k12); b_lo = vis_fpsub16(u_20184_lo, k_276_9856); r_hi = vis_fpsub16(v_15966_hi, k_222_9952); r_lo = vis_fpsub16(v_15966_lo, k_222_9952); temp_g_hi = vis_fpadd16(g_hi, y_11644_hi); temp_b_hi = vis_fpadd16(b_hi, y_11644_hi); green_hi = vis_fpack16(temp_g_hi); temp_r_hi = vis_fpadd16(r_hi, y_11644_hi); blue_hi = vis_fpack16(temp_b_hi); temp_g_lo = vis_fpadd16(g_lo, y_11644_lo); red_hi = vis_fpack16(temp_r_hi); temp_b_lo = vis_fpadd16(b_lo, y_11644_lo); green_lo = vis_fpack16(temp_g_lo); temp_r_lo = vis_fpadd16(r_lo, y_11644_lo); blue_lo = vis_fpack16(temp_b_lo); x_green_hi = vis_fmul8x16au(green_hi, k5); red_lo = vis_fpack16(temp_r_lo); blue_red_hi = vis_fpmerge(blue_hi, red_hi); x_green_lo = vis_fmul8x16au(green_lo, k5); blue_red_lo = vis_fpmerge(blue_lo, red_lo); dd1 = vis_fpmerge(vis_read_hi(x_green_hi), vis_read_hi(blue_red_hi)); dd = vis_faligndata(dd0, dd1); vis_pst_8(dd, dpp, emask1); dd0 = dd1; dpp += inc; i += 2; if (i < width) { dd1 = vis_fpmerge(vis_read_lo(x_green_hi), vis_read_lo(blue_red_hi)); dd = vis_faligndata(dd0, dd1); vis_pst_8(dd, dpp++, emask); dd0 = dd1; i += 2; if (i < width) { dd1 = vis_fpmerge(vis_read_hi (x_green_lo), vis_read_hi(blue_red_lo)); dd = vis_faligndata(dd0, dd1); vis_pst_8(dd, dpp++, emask); dd0 = dd1; } } } vis_alignaddr((void *)off, 0); emask1 = vis_edge8(dpp, dend); emask1 &= emask; dd = vis_faligndata(dd0, dd1); vis_pst_8(dd, dpp, emask1); sp1 = sl1 = sl1 + y_stride; sp2 = sl2 = sl2 + uv_stride; sp3 = sl3 = sl3 + uv_stride; dl = dp = dl + abgr_stride; emask = 0x7777; } return (MLIB_SUCCESS); }
mlib_status __mlib_VideoUpSample420( mlib_u8 *dst0, mlib_u8 *dst1, const mlib_u8 *src0, const mlib_u8 *src1, const mlib_u8 *src2, mlib_s32 n) { mlib_u8 *dend0 = dst0 + 2 * n - 1; mlib_d64 *dp0 = (mlib_d64 *)dst0; mlib_d64 *dp1 = (mlib_d64 *)dst1; mlib_d64 *sp0 = (mlib_d64 *)src0; mlib_d64 *sp1 = (mlib_d64 *)src1; mlib_d64 *sp2 = (mlib_d64 *)src2; mlib_d64 d00, d01, d10, d11, d20, d21; mlib_d64 thiscolsum0_hi, thiscolsum0_lo, lastcolsum0_hi, lastcolsum0_lo; mlib_d64 shiftcolsum0_hi, shiftcolsum0_lo; mlib_d64 thiscolsum1_hi, thiscolsum1_lo, lastcolsum1_hi, lastcolsum1_lo; mlib_d64 shiftcolsum1_hi, shiftcolsum1_lo; mlib_d64 acc0, acc1, acc2, acc3, acc4, acc5, acc6, acc7; mlib_d64 ac0, ac1, ac2, ac3, ac4, ac5, ac6, ac7; mlib_d64 data0, data1, data2, data3, tmp0, tmp1; mlib_f32 fone = vis_to_float(0x4000000); mlib_f32 fthree = vis_to_float(0xC000000); mlib_f32 fone1 = vis_to_float(0x40404040); mlib_f32 fthree1 = vis_to_float(0xC0C0C0C0); mlib_d64 dseven = vis_to_double_dup(0x70007); mlib_d64 deight = vis_to_double_dup(0x80008); mlib_s32 i, emask; if (n <= 0) return (MLIB_FAILURE); vis_write_gsr((3 << 3) + 2); d00 = vis_ld_d64_nf(sp0); d10 = vis_ld_d64_nf(sp1); d20 = vis_ld_d64_nf(sp2); sp0++; sp1++; sp2++; lastcolsum0_hi = vis_fmul8x16au(vis_read_hi(d00), fone); lastcolsum0_lo = vis_fmul8x16au(vis_read_lo(d00), fone); lastcolsum1_hi = vis_fmul8x16au(vis_read_hi(d20), fone); lastcolsum1_lo = vis_fmul8x16au(vis_read_lo(d20), fone); tmp0 = vis_fmul8x16au(vis_read_hi(d10), fthree); tmp1 = vis_fmul8x16au(vis_read_lo(d10), fthree); lastcolsum0_hi = vis_fpadd16(lastcolsum0_hi, tmp0); lastcolsum0_lo = vis_fpadd16(lastcolsum0_lo, tmp1); lastcolsum1_hi = vis_fpadd16(lastcolsum1_hi, tmp0); lastcolsum1_lo = vis_fpadd16(lastcolsum1_lo, tmp1); #pragma pipeloop(0) for (i = 0; i < n - 8; i += 8) { d01 = *sp0; d11 = *sp1; d21 = *sp2; sp0++; sp1++; sp2++; thiscolsum0_hi = vis_fmul8x16au(vis_read_hi(d01), fone); thiscolsum0_lo = vis_fmul8x16au(vis_read_lo(d01), fone); thiscolsum1_hi = vis_fmul8x16au(vis_read_hi(d21), fone); thiscolsum1_lo = vis_fmul8x16au(vis_read_lo(d21), fone); tmp0 = vis_fmul8x16au(vis_read_hi(d11), fthree); tmp1 = vis_fmul8x16au(vis_read_lo(d11), fthree); thiscolsum0_hi = vis_fpadd16(thiscolsum0_hi, tmp0); thiscolsum0_lo = vis_fpadd16(thiscolsum0_lo, tmp1); thiscolsum1_hi = vis_fpadd16(thiscolsum1_hi, tmp0); thiscolsum1_lo = vis_fpadd16(thiscolsum1_lo, tmp1); acc0 = vis_fmul8x16(fone1, lastcolsum0_hi); acc1 = vis_fmul8x16(fone1, lastcolsum0_lo); acc2 = vis_fmul8x16(fthree1, lastcolsum0_hi); acc3 = vis_fmul8x16(fthree1, lastcolsum0_lo); acc4 = vis_fmul8x16(fone1, lastcolsum1_hi); acc5 = vis_fmul8x16(fone1, lastcolsum1_lo); acc6 = vis_fmul8x16(fthree1, lastcolsum1_hi); acc7 = vis_fmul8x16(fthree1, lastcolsum1_lo); shiftcolsum0_hi = vis_faligndata(lastcolsum0_hi, lastcolsum0_lo); shiftcolsum0_lo = vis_faligndata(lastcolsum0_lo, thiscolsum0_hi); shiftcolsum1_hi = vis_faligndata(lastcolsum1_hi, lastcolsum1_lo); shiftcolsum1_lo = vis_faligndata(lastcolsum1_lo, thiscolsum1_hi); acc0 = vis_fpadd16(acc0, deight); acc1 = vis_fpadd16(acc1, deight); acc2 = vis_fpadd16(acc2, dseven); acc3 = vis_fpadd16(acc3, dseven); acc4 = vis_fpadd16(acc4, deight); acc5 = vis_fpadd16(acc5, deight); acc6 = vis_fpadd16(acc6, dseven); acc7 = vis_fpadd16(acc7, dseven); ac0 = vis_fmul8x16(fthree1, shiftcolsum0_hi); ac1 = vis_fmul8x16(fthree1, shiftcolsum0_lo); ac2 = vis_fmul8x16(fone1, shiftcolsum0_hi); ac3 = vis_fmul8x16(fone1, shiftcolsum0_lo); ac4 = vis_fmul8x16(fthree1, shiftcolsum1_hi); ac5 = vis_fmul8x16(fthree1, shiftcolsum1_lo); ac6 = vis_fmul8x16(fone1, shiftcolsum1_hi); ac7 = vis_fmul8x16(fone1, shiftcolsum1_lo); acc0 = vis_fpadd16(acc0, ac0); acc1 = vis_fpadd16(acc1, ac1); acc2 = vis_fpadd16(acc2, ac2); acc3 = vis_fpadd16(acc3, ac3); acc4 = vis_fpadd16(acc4, ac4); acc5 = vis_fpadd16(acc5, ac5); acc6 = vis_fpadd16(acc6, ac6); acc7 = vis_fpadd16(acc7, ac7); data0 = vis_fpack16_pair(acc0, acc1); data1 = vis_fpack16_pair(acc2, acc3); data2 = vis_fpack16_pair(acc4, acc5); data3 = vis_fpack16_pair(acc6, acc7); dp0[0] = vis_fpmerge(vis_read_hi(data1), vis_read_hi(data0)); dp0[1] = vis_fpmerge(vis_read_lo(data1), vis_read_lo(data0)); dp1[0] = vis_fpmerge(vis_read_hi(data3), vis_read_hi(data2)); dp1[1] = vis_fpmerge(vis_read_lo(data3), vis_read_lo(data2)); dp0 += 2; dp1 += 2; lastcolsum0_hi = thiscolsum0_hi; lastcolsum0_lo = thiscolsum0_lo; lastcolsum1_hi = thiscolsum1_hi; lastcolsum1_lo = thiscolsum1_lo; } if (i < n) { acc0 = vis_fmul8x16(fone1, lastcolsum0_hi); acc1 = vis_fmul8x16(fone1, lastcolsum0_lo); acc2 = vis_fmul8x16(fthree1, lastcolsum0_hi); acc3 = vis_fmul8x16(fthree1, lastcolsum0_lo); acc4 = vis_fmul8x16(fone1, lastcolsum1_hi); acc5 = vis_fmul8x16(fone1, lastcolsum1_lo); acc6 = vis_fmul8x16(fthree1, lastcolsum1_hi); acc7 = vis_fmul8x16(fthree1, lastcolsum1_lo); shiftcolsum0_hi = vis_faligndata(lastcolsum0_hi, lastcolsum0_lo); shiftcolsum0_lo = vis_faligndata(lastcolsum0_lo, lastcolsum0_lo); shiftcolsum1_hi = vis_faligndata(lastcolsum1_hi, lastcolsum1_lo); shiftcolsum1_lo = vis_faligndata(lastcolsum1_lo, lastcolsum1_lo); acc0 = vis_fpadd16(acc0, deight); acc1 = vis_fpadd16(acc1, deight); acc2 = vis_fpadd16(acc2, dseven); acc3 = vis_fpadd16(acc3, dseven); acc4 = vis_fpadd16(acc4, deight); acc5 = vis_fpadd16(acc5, deight); acc6 = vis_fpadd16(acc6, dseven); acc7 = vis_fpadd16(acc7, dseven); ac0 = vis_fmul8x16(fthree1, shiftcolsum0_hi); ac1 = vis_fmul8x16(fthree1, shiftcolsum0_lo); ac2 = vis_fmul8x16(fone1, shiftcolsum0_hi); ac3 = vis_fmul8x16(fone1, shiftcolsum0_lo); ac4 = vis_fmul8x16(fthree1, shiftcolsum1_hi); ac5 = vis_fmul8x16(fthree1, shiftcolsum1_lo); ac6 = vis_fmul8x16(fone1, shiftcolsum1_hi); ac7 = vis_fmul8x16(fone1, shiftcolsum1_lo); acc0 = vis_fpadd16(acc0, ac0); acc1 = vis_fpadd16(acc1, ac1); acc2 = vis_fpadd16(acc2, ac2); acc3 = vis_fpadd16(acc3, ac3); acc4 = vis_fpadd16(acc4, ac4); acc5 = vis_fpadd16(acc5, ac5); acc6 = vis_fpadd16(acc6, ac6); acc7 = vis_fpadd16(acc7, ac7); data0 = vis_fpack16_pair(acc0, acc1); data1 = vis_fpack16_pair(acc2, acc3); data2 = vis_fpack16_pair(acc4, acc5); data3 = vis_fpack16_pair(acc6, acc7); acc0 = vis_fpmerge(vis_read_hi(data1), vis_read_hi(data0)); acc1 = vis_fpmerge(vis_read_hi(data3), vis_read_hi(data2)); emask = vis_edge8(dp0, dend0); vis_pst_8(acc0, dp0, emask); vis_pst_8(acc1, dp1, emask); i += 4; dp0++; dp1++; if (i < n) { acc0 = vis_fpmerge(vis_read_lo(data1), vis_read_lo(data0)); acc1 = vis_fpmerge(vis_read_lo(data3), vis_read_lo(data2)); emask = vis_edge8(dp0, dend0); vis_pst_8(acc0, dp0, emask); vis_pst_8(acc1, dp1, emask); } } vis_write_gsr(7); dp0 = (mlib_d64 *)dst0; dp1 = (mlib_d64 *)dst1; ac0 = *dp0; ac2 = *dp1; #pragma pipeloop(0) for (i = 0; i < 2 * n - 8; i += 8) { ac1 = *dp0; ac3 = *dp1; *dp0 = vis_faligndata(ac0, ac1); *dp1 = vis_faligndata(ac2, ac3); dp0++; dp1++; ac0 = ac1; ac2 = ac3; } if (i < 2 * n) { ac1 = vis_ld_d64_nf(dp0); ac3 = vis_ld_d64_nf(dp1); emask = vis_edge8(dp0, dend0); acc0 = vis_faligndata(ac0, ac1); acc1 = vis_faligndata(ac2, ac3); vis_pst_8(acc0, dp0, emask); vis_pst_8(acc1, dp1, emask); } dst0[0] = (4 * (3 * src1[0] + src0[0]) + 8) >> 4; dst1[0] = (4 * (3 * src1[0] + src2[0]) + 8) >> 4; dst0[2 * n - 1] = (4 * (3 * src1[n - 1] + src0[n - 1]) + 7) >> 4; dst1[2 * n - 1] = (4 * (3 * src1[n - 1] + src2[n - 1]) + 7) >> 4; return (MLIB_SUCCESS); }
static mlib_status mlib_v_VideoColorYUV2ABGR411_dst_nonalign( mlib_u8 *abgr, const mlib_u8 *y, const mlib_u8 *u, const mlib_u8 *v, mlib_s32 width, mlib_s32 height, mlib_s32 abgr_stride, mlib_s32 y_stride, mlib_s32 uv_stride) { /* pointers to src address */ mlib_u8 *sp1, *sp2, *sp3, *sl1, *sl2, *sl3; /* pointers to dst address */ mlib_u8 *dp, *dl, *dend; /* all. pointer to y */ mlib_d64 *spy; /* all. pointer to dst */ mlib_d64 *dpp; /* u, v data */ mlib_f32 fu, fv; /* y data */ mlib_d64 dy0, dy1, dy2; mlib_d64 ddy1, ddy2, ddy3, ddy4; mlib_d64 du0, du1; mlib_d64 dv1, dv2; mlib_d64 dr, dr1, dr2, dr3, dr4; mlib_d64 dg, dg1, dg2, dg3, dg4; mlib_d64 db, db1, db2, db3, db4; mlib_d64 dd, dd0, dd1, dtmp; /* used to load u, v into mlib_f32 */ mlib_f32 ffu[1], ffv[1]; /* used to load u, v into mlib_f32 */ mlib_u8 *ufu, *vfu; /* 1.1644 * 4096 */ mlib_f32 f0 = vis_to_float(0x12a1); /* 2.0184 * 8192 */ mlib_f32 f1 = vis_to_float(0x4097); /* -0.3920 * 8192 */ mlib_f32 f4 = vis_to_float(0xf375); /* -0.8132 * 8192 */ mlib_f32 f5 = vis_to_float(0xe5fa); /* 1.5966 * 8192 */ mlib_f32 f8 = vis_to_float(0x3317); /* -276.9856 * 32 */ mlib_d64 doff0 = vis_to_double_dup(0xdd60dd60); /* 135.6352 * 32 */ mlib_d64 doff1 = vis_to_double_dup(0x10f410f4); /* -222.9952 * 32 */ mlib_d64 doff2 = vis_to_double_dup(0xe420e420); mlib_f32 fscale = vis_to_float(0x80808080); /* loop variables */ mlib_s32 i, j; /* alpha_ch. is not written */ mlib_s32 emask = 0x7777; mlib_s32 emask1; mlib_d64 *buf; mlib_s32 inc; ufu = (mlib_u8 *)ffu; vfu = (mlib_u8 *)ffv; /* * initialize GSR scale factor */ vis_write_gsr(3 << 3); buf = (mlib_d64 *)__mlib_malloc((width / 8 + 1) * sizeof (mlib_d64)); if (buf == NULL) return (MLIB_FAILURE); sp1 = sl1 = (mlib_u8 *)y; sp2 = sl2 = (mlib_u8 *)u; sp3 = sl3 = (mlib_u8 *)v; dl = dp = (mlib_u8 *)abgr; /* * row loop */ for (j = 0; j < height; j++) { spy = (mlib_d64 *)vis_alignaddr(sp1, 0); dpp = buf; dy0 = vis_ld_d64_nf(spy); spy++; #pragma pipeloop(0) for (i = 0; i < width; i += 8) { dy1 = vis_ld_d64_nf(spy); spy++; (*dpp++) = vis_faligndata(dy0, dy1); dy0 = dy1; } spy = buf; dend = dp + width * 4 - 1; emask1 = vis_edge8(dp, dend); dpp = (mlib_d64 *)vis_alignaddr(dp, 0); i = dp - (mlib_u8 *)dpp; emask >>= i; vis_alignaddr((void *)(8 - i), 0); inc = (emask1 != 0xff); emask1 &= emask; ufu[0] = vis_ld_u8_nf(sp2); ufu[1] = vis_ld_u8_nf(sp2 + 1); ufu[2] = vis_ld_u8_nf(sp2 + 2); ufu[3] = vis_ld_u8_nf(sp2 + 3); vfu[0] = vis_ld_u8_nf(sp3); vfu[1] = vis_ld_u8_nf(sp3 + 1); vfu[2] = vis_ld_u8_nf(sp3 + 2); vfu[3] = vis_ld_u8_nf(sp3 + 3); sp2 += 4; sp3 += 4; fu = ffu[0]; fv = ffv[0]; /* * 16-pixel column loop */ #pragma pipeloop(0) for (i = 0; i <= width - 16; i += 16) { dy1 = (*spy++); dy2 = (*spy++); du0 = vis_fmul8x16al(fu, f1); db = vis_fpadd16(du0, doff0); du1 = vis_fmul8x16al(fu, f4); dv1 = vis_fmul8x16al(fv, f5); dtmp = vis_fpadd16(du1, dv1); dg = vis_fpadd16(dtmp, doff1); dv2 = vis_fmul8x16al(fv, f8); dr = vis_fpadd16(dv2, doff2); ddy1 = vis_fmul8x16al(vis_read_hi(dy1), f0); ddy2 = vis_fmul8x16al(vis_read_lo(dy1), f0); ufu[0] = vis_ld_u8_nf(sp2); ddy3 = vis_fmul8x16al(vis_read_hi(dy2), f0); ddy4 = vis_fmul8x16al(vis_read_lo(dy2), f0); ufu[1] = vis_ld_u8_nf(sp2 + 1); db1 = vis_fmul8x16au(fscale, vis_read_hi(db)); db1 = vis_fpadd16(ddy1, db1); ufu[2] = vis_ld_u8_nf(sp2 + 2); db2 = vis_fmul8x16al(fscale, vis_read_hi(db)); db2 = vis_fpadd16(ddy2, db2); ufu[3] = vis_ld_u8_nf(sp2 + 3); db3 = vis_fmul8x16au(fscale, vis_read_lo(db)); db3 = vis_fpadd16(ddy3, db3); vfu[0] = vis_ld_u8_nf(sp3); db4 = vis_fmul8x16al(fscale, vis_read_lo(db)); db4 = vis_fpadd16(ddy4, db4); vfu[1] = vis_ld_u8_nf(sp3 + 1); dg1 = vis_fmul8x16au(fscale, vis_read_hi(dg)); dg1 = vis_fpadd16(ddy1, dg1); vfu[2] = vis_ld_u8_nf(sp3 + 2); dg2 = vis_fmul8x16al(fscale, vis_read_hi(dg)); dg2 = vis_fpadd16(ddy2, dg2); vfu[3] = vis_ld_u8_nf(sp3 + 3); dg3 = vis_fmul8x16au(fscale, vis_read_lo(dg)); dg3 = vis_fpadd16(ddy3, dg3); dg4 = vis_fmul8x16al(fscale, vis_read_lo(dg)); dg4 = vis_fpadd16(ddy4, dg4); dr1 = vis_fmul8x16au(fscale, vis_read_hi(dr)); dr1 = vis_fpadd16(ddy1, dr1); dr2 = vis_fmul8x16al(fscale, vis_read_hi(dr)); dr2 = vis_fpadd16(ddy2, dr2); dr3 = vis_fmul8x16au(fscale, vis_read_lo(dr)); dr3 = vis_fpadd16(ddy3, dr3); dr4 = vis_fmul8x16al(fscale, vis_read_lo(dr)); dr4 = vis_fpadd16(ddy4, dr4); dr = vis_fpack16_pair(dr1, dr2); dr1 = vis_fpack16_pair(dr3, dr4); dg = vis_fpack16_pair(dg1, dg2); dg1 = vis_fpack16_pair(dg3, dg4); db = vis_fpack16_pair(db1, db2); db1 = vis_fpack16_pair(db3, db4); dg2 = vis_fpmerge(vis_read_hi(db), vis_read_hi(dg)); dg3 = vis_fpmerge(vis_read_hi(db), vis_read_hi(dr)); dd1 = vis_fpmerge(vis_read_hi(dg2), vis_read_hi(dg3)); dd = vis_faligndata(dd0, dd1); vis_pst_8(dd, dpp, emask1); dpp += inc; inc = 1; dd0 = vis_fpmerge(vis_read_lo(dg2), vis_read_lo(dg3)); dd = vis_faligndata(dd1, dd0); vis_pst_8(dd, dpp++, emask); dg2 = vis_fpmerge(vis_read_lo(db), vis_read_lo(dg)); dg3 = vis_fpmerge(vis_read_lo(db), vis_read_lo(dr)); dd1 = vis_fpmerge(vis_read_hi(dg2), vis_read_hi(dg3)); dd = vis_faligndata(dd0, dd1); vis_pst_8(dd, dpp++, emask); dd0 = vis_fpmerge(vis_read_lo(dg2), vis_read_lo(dg3)); dd = vis_faligndata(dd1, dd0); vis_pst_8(dd, dpp++, emask); dg2 = vis_fpmerge(vis_read_hi(db1), vis_read_hi(dg1)); dg3 = vis_fpmerge(vis_read_hi(db1), vis_read_hi(dr1)); dd1 = vis_fpmerge(vis_read_hi(dg2), vis_read_hi(dg3)); dd = vis_faligndata(dd0, dd1); vis_pst_8(dd, dpp++, emask); dd0 = vis_fpmerge(vis_read_lo(dg2), vis_read_lo(dg3)); dd = vis_faligndata(dd1, dd0); vis_pst_8(dd, dpp++, emask); dg2 = vis_fpmerge(vis_read_lo(db1), vis_read_lo(dg1)); dg3 = vis_fpmerge(vis_read_lo(db1), vis_read_lo(dr1)); dd1 = vis_fpmerge(vis_read_hi(dg2), vis_read_hi(dg3)); dd = vis_faligndata(dd0, dd1); vis_pst_8(dd, dpp++, emask); dd0 = vis_fpmerge(vis_read_lo(dg2), vis_read_lo(dg3)); dd = vis_faligndata(dd1, dd0); vis_pst_8(dd, dpp++, emask); fu = ffu[0]; fv = ffv[0]; sp2 += 4; sp3 += 4; emask1 = emask; } if (i <= width - 8) { dy1 = (*spy++); du0 = vis_fmul8x16al(fu, f1); db = vis_fpadd16(du0, doff0); du1 = vis_fmul8x16al(fu, f4); dv1 = vis_fmul8x16al(fv, f5); dtmp = vis_fpadd16(du1, dv1); dg = vis_fpadd16(dtmp, doff1); dv2 = vis_fmul8x16al(fv, f8); dr = vis_fpadd16(dv2, doff2); ddy1 = vis_fmul8x16al(vis_read_hi(dy1), f0); ddy2 = vis_fmul8x16al(vis_read_lo(dy1), f0); ufu[0] = ufu[2]; db1 = vis_fmul8x16au(fscale, vis_read_hi(db)); db1 = vis_fpadd16(ddy1, db1); vfu[0] = vfu[2]; db2 = vis_fmul8x16al(fscale, vis_read_hi(db)); db2 = vis_fpadd16(ddy2, db2); dg1 = vis_fmul8x16au(fscale, vis_read_hi(dg)); dg1 = vis_fpadd16(ddy1, dg1); dg2 = vis_fmul8x16al(fscale, vis_read_hi(dg)); dg2 = vis_fpadd16(ddy2, dg2); dr1 = vis_fmul8x16au(fscale, vis_read_hi(dr)); dr1 = vis_fpadd16(ddy1, dr1); dr2 = vis_fmul8x16al(fscale, vis_read_hi(dr)); dr2 = vis_fpadd16(ddy2, dr2); dr = vis_fpack16_pair(dr1, dr2); dg = vis_fpack16_pair(dg1, dg2); db = vis_fpack16_pair(db1, db2); dg2 = vis_fpmerge(vis_read_hi(db), vis_read_hi(dg)); dg3 = vis_fpmerge(vis_read_hi(db), vis_read_hi(dr)); dd1 = vis_fpmerge(vis_read_hi(dg2), vis_read_hi(dg3)); dd = vis_faligndata(dd0, dd1); vis_pst_8(dd, dpp, emask1); dpp += inc; inc = 1; dd0 = vis_fpmerge(vis_read_lo(dg2), vis_read_lo(dg3)); dd = vis_faligndata(dd1, dd0); vis_pst_8(dd, dpp++, emask); dg2 = vis_fpmerge(vis_read_lo(db), vis_read_lo(dg)); dg3 = vis_fpmerge(vis_read_lo(db), vis_read_lo(dr)); dd1 = vis_fpmerge(vis_read_hi(dg2), vis_read_hi(dg3)); dd = vis_faligndata(dd0, dd1); vis_pst_8(dd, dpp++, emask); dd0 = vis_fpmerge(vis_read_lo(dg2), vis_read_lo(dg3)); dd = vis_faligndata(dd1, dd0); vis_pst_8(dd, dpp++, emask); fu = ffu[0]; fv = ffv[0]; i += 8; emask1 = emask; } if (i < width) { dy1 = vis_ld_d64_nf(spy); du0 = vis_fmul8x16al(fu, f1); db = vis_fpadd16(du0, doff0); du1 = vis_fmul8x16al(fu, f4); dv1 = vis_fmul8x16al(fv, f5); dtmp = vis_fpadd16(du1, dv1); dg = vis_fpadd16(dtmp, doff1); dv2 = vis_fmul8x16al(fv, f8); dr = vis_fpadd16(dv2, doff2); ddy1 = vis_fmul8x16al(vis_read_hi(dy1), f0); db1 = vis_fmul8x16au(fscale, vis_read_hi(db)); db1 = vis_fpadd16(ddy1, db1); dg1 = vis_fmul8x16au(fscale, vis_read_hi(dg)); dg1 = vis_fpadd16(ddy1, dg1); dr1 = vis_fmul8x16au(fscale, vis_read_hi(dr)); dr1 = vis_fpadd16(ddy1, dr1); fu = vis_fpack16(db1); dg2 = vis_fpmerge(fu, vis_fpack16(dg1)); dg3 = vis_fpmerge(fu, vis_fpack16(dr1)); dd1 = vis_fpmerge(vis_read_hi(dg2), vis_read_hi(dg3)); dd = vis_faligndata(dd0, dd1); vis_pst_8(dd, dpp, emask1); dpp += inc; dd0 = vis_fpmerge(vis_read_lo(dg2), vis_read_lo(dg3)); dd = vis_faligndata(dd1, dd0); vis_pst_8(dd, dpp++, emask); } emask1 = vis_edge8(dpp, dend); emask1 &= emask; dd = vis_faligndata(dd0, dd1); vis_pst_8(dd, dpp, emask1); sp1 = sl1 = sl1 + y_stride; sp2 = sl2 = sl2 + uv_stride; sp3 = sl3 = sl3 + uv_stride; dl = dp = dl + abgr_stride; emask = 0x7777; } __mlib_free(buf); return (MLIB_SUCCESS); }
void __mlib_VideoColorYUV444int_to_UYVY422int( mlib_u32 *uyvy, const mlib_u8 *yuv, mlib_s32 w, mlib_s32 h, mlib_s32 dlb, mlib_s32 slb) { mlib_s32 i, val_y0, val_y1, val_u0, val_v0, count, left; dlb >>= 2; w >>= 1; count = w >> 2; left = w - (count << 2); if (w == 0 || h == 0) return; vis_write_gsr(6 << 3); for (i = 0; i < h; i++, yuv += slb, uyvy += dlb) { if ((((mlib_addr)yuv | (mlib_addr)uyvy) & 7) == 0) { mlib_d64 w_y, w_u, w_v, w_uv, w_tmp0, w_tmp1, w_acc0, w_acc1; mlib_d64 w_ld0, w_ld1, w_ld2; mlib_f32 v_one = vis_to_float(0x1000000); mlib_f32 v_u, v_v; mlib_s32 j; #pragma pipeloop(0) for (j = 0; j < count; j++) { w_ld0 = ((mlib_d64 *)yuv)[3 * j]; w_ld1 = ((mlib_d64 *)yuv)[3 * j + 1]; w_ld2 = ((mlib_d64 *)yuv)[3 * j + 2]; MLIB_SPLIT3_U8(w_y, w_u, w_v, w_ld0, w_ld1, w_ld2); w_tmp0 = vis_fpmerge(vis_read_hi(w_u), vis_read_lo(w_u)); w_tmp1 = vis_fpmerge(vis_read_hi(w_tmp0), vis_read_lo(w_tmp0)); w_acc0 = vis_fmul8x16au(vis_read_hi(w_tmp1), v_one); w_acc1 = vis_fmul8x16au(vis_read_lo(w_tmp1), v_one); v_u = vis_fpack16(vis_fpadd16(w_acc0, w_acc1)); w_tmp0 = vis_fpmerge(vis_read_hi(w_v), vis_read_lo(w_v)); w_tmp1 = vis_fpmerge(vis_read_hi(w_tmp0), vis_read_lo(w_tmp0)); w_acc0 = vis_fmul8x16au(vis_read_hi(w_tmp1), v_one); w_acc1 = vis_fmul8x16au(vis_read_lo(w_tmp1), v_one); v_v = vis_fpack16(vis_fpadd16(w_acc0, w_acc1)); w_uv = vis_fpmerge(v_u, v_v); ((mlib_d64 *)uyvy)[2 * j] = VIS_FPMERGE_HI(w_uv, w_y); ((mlib_d64 *)uyvy)[2 * j + 1] = VIS_FPMERGE_LO(w_uv, w_y); } if (left) { mlib_d64 res_buf[2]; w_ld0 = vis_ld_d64_nf((mlib_d64 *)yuv + 3 * count); w_ld1 = vis_ld_d64_nf((mlib_d64 *)yuv + 3 * count + 1); w_ld2 = vis_ld_d64_nf((mlib_d64 *)yuv + 3 * count + 2); MLIB_SPLIT3_U8(w_y, w_u, w_v, w_ld0, w_ld1, w_ld2); w_tmp0 = vis_fpmerge(vis_read_hi(w_u), vis_read_lo(w_u)); w_tmp1 = vis_fpmerge(vis_read_hi(w_tmp0), vis_read_lo(w_tmp0)); w_acc0 = vis_fmul8x16au(vis_read_hi(w_tmp1), v_one); w_acc1 = vis_fmul8x16au(vis_read_lo(w_tmp1), v_one); v_u = vis_fpack16(vis_fpadd16(w_acc0, w_acc1)); w_tmp0 = vis_fpmerge(vis_read_hi(w_v), vis_read_lo(w_v)); w_tmp1 = vis_fpmerge(vis_read_hi(w_tmp0), vis_read_lo(w_tmp0)); w_acc0 = vis_fmul8x16au(vis_read_hi(w_tmp1), v_one); w_acc1 = vis_fmul8x16au(vis_read_lo(w_tmp1), v_one); v_v = vis_fpack16(vis_fpadd16(w_acc0, w_acc1)); w_uv = vis_fpmerge(v_u, v_v); res_buf[0] = VIS_FPMERGE_HI(w_uv, w_y); res_buf[1] = VIS_FPMERGE_LO(w_uv, w_y); for (j = 0; j < left; j++) { ((mlib_f32 *)uyvy)[4 * count + j] = ((mlib_f32 *)res_buf)[j]; } } } else { mlib_d64 w_y, w_u, w_v, w_uv, w_tmp0, w_tmp1, w_acc0, w_acc1; mlib_d64 w_ld0, w_ld1, w_ld2; mlib_f32 v_one = vis_to_float(0x1000000); mlib_f32 v_u, v_v; mlib_s32 j; mlib_d64 *al_addr; mlib_d64 l0, l1, l2, l3; const mlib_u8 *pyuv = yuv; mlib_u32 *puyvy = uyvy; if ((mlib_addr)puyvy & 7) { val_y0 = yuv[0]; val_y1 = yuv[3]; val_u0 = (yuv[1] + yuv[4]) >> 1; val_v0 = (yuv[2] + yuv[5]) >> 1; puyvy[0] = (val_u0 << 24) | (val_y0 << 16) | (val_v0 << 8) | val_y1; pyuv += 6; puyvy++; count = (w - 1) >> 2; left = (w - 1) - (count << 2); } else { count = w >> 2; left = w - (count << 2); } al_addr = vis_alignaddr((void *)pyuv, 0); l0 = vis_ld_d64_nf(al_addr); al_addr++; #pragma pipeloop(0) for (j = 0; j < count; j++) { l1 = (*al_addr++); l2 = (*al_addr++); l3 = vis_ld_d64_nf(al_addr); al_addr++; w_ld0 = vis_faligndata(l0, l1); w_ld1 = vis_faligndata(l1, l2); w_ld2 = vis_faligndata(l2, l3); l0 = l3; MLIB_SPLIT3_U8(w_y, w_u, w_v, w_ld0, w_ld1, w_ld2); w_tmp0 = vis_fpmerge(vis_read_hi(w_u), vis_read_lo(w_u)); w_tmp1 = vis_fpmerge(vis_read_hi(w_tmp0), vis_read_lo(w_tmp0)); w_acc0 = vis_fmul8x16au(vis_read_hi(w_tmp1), v_one); w_acc1 = vis_fmul8x16au(vis_read_lo(w_tmp1), v_one); v_u = vis_fpack16(vis_fpadd16(w_acc0, w_acc1)); w_tmp0 = vis_fpmerge(vis_read_hi(w_v), vis_read_lo(w_v)); w_tmp1 = vis_fpmerge(vis_read_hi(w_tmp0), vis_read_lo(w_tmp0)); w_acc0 = vis_fmul8x16au(vis_read_hi(w_tmp1), v_one); w_acc1 = vis_fmul8x16au(vis_read_lo(w_tmp1), v_one); v_v = vis_fpack16(vis_fpadd16(w_acc0, w_acc1)); w_uv = vis_fpmerge(v_u, v_v); ((mlib_d64 *)puyvy)[2 * j] = VIS_FPMERGE_HI(w_uv, w_y); ((mlib_d64 *)puyvy)[2 * j + 1] = VIS_FPMERGE_LO(w_uv, w_y); } if (left) { mlib_d64 res_buf[2]; l1 = vis_ld_d64_nf(al_addr); al_addr++; l2 = vis_ld_d64_nf(al_addr); al_addr++; l3 = vis_ld_d64_nf(al_addr); w_ld0 = vis_faligndata(l0, l1); w_ld1 = vis_faligndata(l1, l2); w_ld2 = vis_faligndata(l2, l3); MLIB_SPLIT3_U8(w_y, w_u, w_v, w_ld0, w_ld1, w_ld2); w_tmp0 = vis_fpmerge(vis_read_hi(w_u), vis_read_lo(w_u)); w_tmp1 = vis_fpmerge(vis_read_hi(w_tmp0), vis_read_lo(w_tmp0)); w_acc0 = vis_fmul8x16au(vis_read_hi(w_tmp1), v_one); w_acc1 = vis_fmul8x16au(vis_read_lo(w_tmp1), v_one); v_u = vis_fpack16(vis_fpadd16(w_acc0, w_acc1)); w_tmp0 = vis_fpmerge(vis_read_hi(w_v), vis_read_lo(w_v)); w_tmp1 = vis_fpmerge(vis_read_hi(w_tmp0), vis_read_lo(w_tmp0)); w_acc0 = vis_fmul8x16au(vis_read_hi(w_tmp1), v_one); w_acc1 = vis_fmul8x16au(vis_read_lo(w_tmp1), v_one); v_v = vis_fpack16(vis_fpadd16(w_acc0, w_acc1)); w_uv = vis_fpmerge(v_u, v_v); res_buf[0] = VIS_FPMERGE_HI(w_uv, w_y); res_buf[1] = VIS_FPMERGE_LO(w_uv, w_y); for (j = 0; j < left; j++) { ((mlib_f32 *)puyvy)[4 * count + j] = ((mlib_f32 *)res_buf)[j]; } } count = w >> 2; left = w - (count << 2); }
void __mlib_VideoColorYUV444seq_to_UYVY422int( mlib_u32 *uyvy, const mlib_u8 *y, const mlib_u8 *u, const mlib_u8 *v, mlib_s32 w, mlib_s32 h, mlib_s32 dlb, mlib_s32 slb) { mlib_s32 i, j2, val_y0, val_y1, val_u0, val_v0, count, left; dlb >>= 2; w >>= 1; if (w == 0 || h == 0) return; count = w >> 2; left = w - (count << 2); vis_write_gsr(6 << 3); for (i = 0; i < h; i++, y += slb, u += slb, v += slb, uyvy += dlb) { if ((((mlib_addr)u | (mlib_addr)v | (mlib_addr)y | (mlib_addr) uyvy) & 7) == 0) { mlib_d64 w_y, w_u, w_v, w_uv, w_tmp0, w_tmp1, w_acc0, w_acc1; mlib_f32 v_one = vis_to_float(0x1000000); mlib_f32 v_u, v_v; mlib_s32 j; #pragma pipeloop(0) for (j = 0; j < count; j++) { w_y = ((mlib_d64 *)y)[j]; w_u = ((mlib_d64 *)u)[j]; w_v = ((mlib_d64 *)v)[j]; w_tmp0 = vis_fpmerge(vis_read_hi(w_u), vis_read_lo(w_u)); w_tmp1 = vis_fpmerge(vis_read_hi(w_tmp0), vis_read_lo(w_tmp0)); w_acc0 = vis_fmul8x16au(vis_read_hi(w_tmp1), v_one); w_acc1 = vis_fmul8x16au(vis_read_lo(w_tmp1), v_one); v_u = vis_fpack16(vis_fpadd16(w_acc0, w_acc1)); w_tmp0 = vis_fpmerge(vis_read_hi(w_v), vis_read_lo(w_v)); w_tmp1 = vis_fpmerge(vis_read_hi(w_tmp0), vis_read_lo(w_tmp0)); w_acc0 = vis_fmul8x16au(vis_read_hi(w_tmp1), v_one); w_acc1 = vis_fmul8x16au(vis_read_lo(w_tmp1), v_one); v_v = vis_fpack16(vis_fpadd16(w_acc0, w_acc1)); w_uv = vis_fpmerge(v_u, v_v); ((mlib_d64 *)uyvy)[2 * j] = VIS_FPMERGE_HI(w_uv, w_y); ((mlib_d64 *)uyvy)[2 * j + 1] = VIS_FPMERGE_LO(w_uv, w_y); } if (left) { mlib_d64 res_buf[2]; w_y = ((mlib_d64 *)y)[count]; w_u = ((mlib_d64 *)u)[count]; w_v = ((mlib_d64 *)v)[count]; w_tmp0 = vis_fpmerge(vis_read_hi(w_u), vis_read_lo(w_u)); w_tmp1 = vis_fpmerge(vis_read_hi(w_tmp0), vis_read_lo(w_tmp0)); w_acc0 = vis_fmul8x16au(vis_read_hi(w_tmp1), v_one); w_acc1 = vis_fmul8x16au(vis_read_lo(w_tmp1), v_one); v_u = vis_fpack16(vis_fpadd16(w_acc0, w_acc1)); w_tmp0 = vis_fpmerge(vis_read_hi(w_v), vis_read_lo(w_v)); w_tmp1 = vis_fpmerge(vis_read_hi(w_tmp0), vis_read_lo(w_tmp0)); w_acc0 = vis_fmul8x16au(vis_read_hi(w_tmp1), v_one); w_acc1 = vis_fmul8x16au(vis_read_lo(w_tmp1), v_one); v_v = vis_fpack16(vis_fpadd16(w_acc0, w_acc1)); w_uv = vis_fpmerge(v_u, v_v); res_buf[0] = VIS_FPMERGE_HI(w_uv, w_y); res_buf[1] = VIS_FPMERGE_LO(w_uv, w_y); for (j = 0; j < left; j++) { ((mlib_f32 *)uyvy)[4 * count + j] = ((mlib_f32 *)res_buf)[j]; } } } else { #pragma pipeloop(0) for (j2 = 0; j2 < w; j2++) { mlib_s32 j = 2 * j2; mlib_s32 j1 = j + 1; val_y0 = y[j]; val_y1 = y[j1]; val_u0 = (u[j] + u[j1]) >> 1; val_v0 = (v[j] + v[j1]) >> 1; uyvy[j2] = (val_u0 << 24) | (val_y0 << 16) | (val_v0 << 8) | val_y1; } } } }