自学教程：C++ vis_read_hi函数代码示例

mlib_status__mlib_VideoUpSample420_Nearest(	mlib_u8 *dst0,	mlib_u8 *dst1,	const mlib_u8 *src,	mlib_s32 n){	mlib_d64 *sp = (mlib_d64 *)src;	mlib_d64 *dp0 = (mlib_d64 *)dst0;	mlib_d64 *dp1 = (mlib_d64 *)dst1;	mlib_u8 *dend0 = dst0 + 2 * n - 1;	mlib_d64 sa, da;	mlib_s32 emask, i;	if (n <= 0)		return (MLIB_FAILURE);#pragma pipeloop(0)	for (i = 0; i <= (n - 8); i += 8) {		sa = *sp;		*dp0 = *dp1 = vis_fpmerge(vis_read_hi(sa), vis_read_hi(sa));		*(dp0 + 1) = *(dp1 + 1) =			vis_fpmerge(vis_read_lo(sa), vis_read_lo(sa));		sp++;		dp0 += 2;		dp1 += 2;	}	if (i < n) {		sa = vis_ld_d64_nf(sp);		da = vis_fpmerge(vis_read_hi(sa), vis_read_hi(sa));		emask = vis_edge8(dp0, dend0);		vis_pst_8(da, dp0, emask);		vis_pst_8(da, dp1, emask);		i += 4;		dp0++;		dp1++;		if (i < n) {			da = vis_fpmerge(vis_read_lo(sa), vis_read_lo(sa));			emask = vis_edge8(dp0, dend0);			vis_pst_8(da, dp0, emask);			vis_pst_8(da, dp1, emask);		}	}	return (MLIB_SUCCESS);}

mlib_status__mlib_VideoInterpX_S16_U8_16x16(	mlib_s16 *mc_block,	const mlib_u8 *ref_block,	mlib_s32 frame_stride,	mlib_s32 field_stride){	mlib_s32 y;	mlib_d64 *dd, ss0[16 * 2], *sp1, *sp2, s1hi, s1lo, s2hi, s2lo, s2, s3;	mlib_f32 strunc = vis_read_hi(*(mlib_d64 *)mlib_IX16const);	mlib_f32 fexpd = vis_read_lo(*(mlib_d64 *)mlib_IX16const);	dd = (mlib_d64 *)mc_block;	sp1 = (mlib_d64 *)vis_alignaddr((void *)ref_block, 0);#pragma pipeloop(0)	MLIB_V_VIDEOCOPY16(16);	sp2 = (mlib_d64 *)vis_alignaddr((void *)(ref_block + 1), 0);#pragma pipeloop(0)	MLIB_V_VIDEOINTERP16(16);	return (MLIB_SUCCESS);}

voidmlib_v_ImageAffineTableLine_8nw_4_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;	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_4x2;		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_4;		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);		v1 = vis_fmul8x16al(vis_read_hi(row30), vis_read_lo(yFilter));		sum = vis_fpadd16(v1, 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);}

voidmlib_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++);	}}

mlib_status__mlib_VideoColorSplit3_S16(	mlib_s16 *color1,	mlib_s16 *color2,	mlib_s16 *color3,	const mlib_s16 *colors,	mlib_s32 n){	mlib_d64 *sp = (mlib_d64 *)colors;	mlib_d64 *dp0 = (mlib_d64 *)color1;	mlib_d64 *dp1 = (mlib_d64 *)color2;	mlib_d64 *dp2 = (mlib_d64 *)color3;	mlib_d64 sd0, sd1, sd2, dd0, dd1, dd2, dd3;	mlib_s32 i;	vis_write_gsr(4);	vis_write_bmask(0x02CE13DF, 0);#pragma pipeloop(0)#pragma unroll(4)	for (i = 0; i <= (n - 4); i += 4) {		sd0 = sp[0];		sd1 = sp[1];		sd2 = sp[2];		dd1 = vis_fpmerge(vis_read_lo(sd0), vis_read_hi(sd2));		dd0 = vis_fpmerge(vis_read_hi(sd0), vis_read_lo(sd1));		(*dp0++) = vis_bshuffle(dd0, dd1);		dd2 = vis_fpmerge(vis_read_hi(sd1), vis_read_lo(sd2));		dd3 = vis_faligndata(dd0, dd2);		(*dp1++) = vis_bshuffle(dd3, dd3);		(*dp2++) = vis_bshuffle(dd1, dd2);		sp += 3;	}/* * last 4 pixels */	if (i < n) {		mlib_s32 emask = 0xF0 >> (n & 3);		mlib_d64 st0, st1, st2;		sd0 = sp[0];		sd1 = vis_ld_d64_nf(sp + 1);		sd2 = vis_ld_d64_nf(sp + 2);		dd1 = vis_fpmerge(vis_read_lo(sd0), vis_read_hi(sd2));		dd0 = vis_fpmerge(vis_read_hi(sd0), vis_read_lo(sd1));		st0 = vis_bshuffle(dd0, dd1);		dd2 = vis_fpmerge(vis_read_hi(sd1), vis_read_lo(sd2));		dd3 = vis_faligndata(dd0, dd2);		st1 = vis_bshuffle(dd3, dd3);		st2 = vis_bshuffle(dd1, dd2);		vis_pst_16(st0, dp0, emask);		vis_pst_16(st1, dp1, emask);		vis_pst_16(st2, dp2, emask);	}

mlib_status__mlib_VideoAddBlock_U8_S16(	mlib_u8 *curr_block,	const mlib_s16 *mc_block,	mlib_s32 stride){	mlib_s32 y;	mlib_d64 *dp, *sp, s1hi, s1lo, s2hi, s2lo, dd;	mlib_f32 zeros = vis_fzeros();/* *   mlib_s32 mlib_imult = 0x100; *   mlib_f32 mult  = *(mlib_f32*) & mlib_imult; */	mlib_f32 mult = vis_to_float(0x100);	vis_write_gsr(7 << 3);	dp = (mlib_d64 *)curr_block;	sp = (mlib_d64 *)mc_block;#pragma pipeloop(0)	for (y = 0; y < 8; y++) {		dd = *dp;		s1hi = (*sp++);		s1lo = (*sp++);		s2hi = vis_fpmerge(zeros, vis_read_hi(dd));		s2lo = vis_fmul8x16al(vis_read_lo(dd), mult);		s1hi = vis_fpadd16(s1hi, s2hi);		s1lo = vis_fpadd16(s1lo, s2lo);		*dp = vis_fpack16_pair(s1hi, s1lo);		dp = (mlib_d64 *)((mlib_u8 *)dp + stride);	}	return (MLIB_SUCCESS);}

static mlib_statusmlib_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;//.........这里部分代码省略.........

mlib_status__mlib_VectorConvert_S16_U8_Mod(	mlib_s16 *z,	const mlib_u8 *x,	mlib_s32 n){	mlib_s32 i;	const mlib_u8 *src = x;	mlib_s16 *dst = z;	mlib_d64 *ddsrc, *ddst;	mlib_s32 len_64, even_length, rest_64, length = n;	mlib_f32 fzero = vis_fzeros();	mlib_d64 dd1, dd2, dd3, dd4;	mlib_f32 fm = vis_to_float(0x100);	if (length < 16) {		EXPAND(mlib_u8, mlib_s16);	}	while ((mlib_addr)dst & 7) {		(*dst++) = (*src++);		length--;	}	ddsrc = (mlib_d64 *)vis_alignaddr((void *)src, 0);	ddst = (mlib_d64 *)dst;	rest_64 = length & 7;	len_64 = length >> 3;	even_length = len_64 << 3;	dd2 = ddsrc[0];	if (!((mlib_addr)src & 7)) {/* * Both vectors are 64-bit aligned. We can process without * vis_faligndata * Peeling the 1 iteration. Then loop with step==2. */		if (i = (len_64 & 1)) {			dd1 = (*ddsrc++);			(*ddst++) = vis_fpmerge(fzero, vis_read_hi(dd1));			(*ddst++) = vis_fpmerge(fzero, vis_read_lo(dd1));		}#pragma pipeloop(1)#pragma unroll(1)		for (; i < len_64; i += 2) {			dd1 = (*ddsrc++);			dd2 = (*ddsrc++);			(*ddst++) = vis_fmul8x16al(vis_read_hi(dd1), fm);			(*ddst++) = vis_fpmerge(fzero, vis_read_lo(dd1));			(*ddst++) = vis_fmul8x16al(vis_read_hi(dd2), fm);			(*ddst++) = vis_fpmerge(fzero, vis_read_lo(dd2));		}	} else {/* * Source vector is not 64-bit aligned. Use vis_faligndata. * Peeling the 1 iteration. Then loop with step==2. */		i = 1;		if (len_64 & 1) {			dd1 = dd2;			dd2 = vis_ld_d64_nf(ddsrc + 1); i++;			dd3 = vis_faligndata(dd1, dd2);			(*ddst++) = vis_fpmerge(fzero, vis_read_hi(dd3));			(*ddst++) = vis_fpmerge(fzero, vis_read_lo(dd3));		}#pragma pipeloop(0)#pragma unroll(2)		for (; i <= len_64; i += 2) {			dd1 = dd2;			dd2 = vis_ld_d64_nf(ddsrc + i);			dd3 = vis_faligndata(dd1, dd2);			dd1 = dd2;			dd2 = vis_ld_d64_nf(ddsrc + i + 1);			dd4 = vis_faligndata(dd1, dd2);			(*ddst++) = vis_fmul8x16al(vis_read_hi(dd3), fm);			(*ddst++) = vis_fpmerge(fzero, vis_read_lo(dd3));			(*ddst++) = vis_fmul8x16al(vis_read_hi(dd4), fm);			(*ddst++) = vis_fpmerge(fzero, vis_read_lo(dd4));		}	}	for (i = 0; i < rest_64; i++)		dst[even_length + i] = src[even_length + i];	return (MLIB_SUCCESS);}

void ADD_SUFF(ByteGrayToIntArgbConvert)(BLIT_PARAMS){    mlib_s32 dstScan = pDstInfo->scanStride;    mlib_s32 srcScan = pSrcInfo->scanStride;    mlib_d64 d0, d1, d2, d3;    mlib_f32 ff, aa = vis_fones();    mlib_s32 i, j, x;    if (width < 8) {        for (j = 0; j < height; j++) {            mlib_u8  *src = srcBase;            mlib_s32 *dst = dstBase;            for (i = 0; i < width; i++) {                x = src[i];                dst[i] = Gray2Argb(x);            }            PTR_ADD(dstBase, dstScan);            PTR_ADD(srcBase, srcScan);        }        return;    }    if (srcScan == width && dstScan == 4*width) {        width *= height;        height = 1;    }    for (j = 0; j < height; j++) {        mlib_u8  *src = srcBase;        mlib_s32 *dst = dstBase;        mlib_s32 *dst_end;        dst_end = dst + width;        while (((mlib_s32)src & 3) && dst < dst_end) {            x = *src++;            *dst++ = Gray2Argb(x);        }#pragma pipeloop(0)        for (; dst <= (dst_end - 4); dst += 4) {            ff = *(mlib_f32*)src;            d0 = vis_fpmerge(aa, ff);            d1 = vis_fpmerge(ff, ff);            d2 = vis_fpmerge(vis_read_hi(d0), vis_read_hi(d1));            d3 = vis_fpmerge(vis_read_lo(d0), vis_read_lo(d1));            ((mlib_f32*)dst)[0] = vis_read_hi(d2);            ((mlib_f32*)dst)[1] = vis_read_lo(d2);            ((mlib_f32*)dst)[2] = vis_read_hi(d3);            ((mlib_f32*)dst)[3] = vis_read_lo(d3);            src += 4;        }        while (dst < dst_end) {            x = *src++;            *dst++ = Gray2Argb(x);        }        PTR_ADD(dstBase, dstScan);        PTR_ADD(srcBase, srcScan);    }}

voidmlib_v_VideoColorYUV2RGB444_all_align(	mlib_u8 *rgb,	const mlib_u8 *y,	const mlib_u8 *u,	const mlib_u8 *v,	mlib_s32 size){	mlib_u8 *dend;	mlib_f32 *sf0, *sf1, *sf2, *pfd, fzero = vis_fzeros();	mlib_s32 i, n, m, emask;	mlib_d64 *buff2, pbuff_arr2[BUFF_SIZE + 4];	mlib_d64 tmp_arr64[2];	mlib_d64 k01 = vis_to_double_dup(0x0000f375);	mlib_d64 k02 = vis_to_double_dup(0x3317e5fa);	mlib_d64 k11 = vis_to_double_dup(0xf3754097);	mlib_d64 k12 = vis_to_double_dup(0xe5fa0000);	mlib_d64 k21 = vis_to_double_dup(0x40970000);	mlib_d64 k22 = vis_to_double_dup(0x00003317);	mlib_d64 c_0 = vis_to_double_dup(0xe42010f4);	mlib_d64 c_1 = vis_to_double_dup(0x10f4dd60);	mlib_d64 c_2 = vis_to_double_dup(0xdd60e420);	mlib_d64 k_0 = vis_to_double_dup(0x25432543);	do {/* loop on buffer size */		if (size > 2 * BUFF_SIZE) {			n = 2 * BUFF_SIZE;		} else {			n = size;		}		m = n >> 2;		buff2 = pbuff_arr2;		sf0 = (mlib_f32 *)y;		sf1 = (mlib_f32 *)u;		sf2 = (mlib_f32 *)v;		dend = rgb + 3 * n - 1;		pfd = (mlib_f32 *)rgb;#pragma pipeloop(0)		for (i = 0; i < m; i++) {			mlib_d64 s00, s01, s02, s10, s11, s12, s20, s21, s22,				s_0;			mlib_f32 x0, x1, x2;			mlib_d64 d_0235, d_xx14, d_23xx, d_0145;			x0 = (*sf0++);			x1 = (*sf1++);			x2 = (*sf2++);			s_0 = vis_fmul8x16(x0, k_0);			s01 = vis_fmul8x16(x1, k01);			s11 = vis_fmul8x16(x1, k11);			s21 = vis_fmul8x16(x1, k21);			s02 = vis_fmul8x16(x2, k02);			s12 = vis_fmul8x16(x2, k12);			s22 = vis_fmul8x16(x2, k22);			s00 = vis_fpadd16(s_0, s01);			s10 = vis_fpadd16(s_0, s11);			s20 = vis_fpadd16(s_0, s21);			s02 = vis_fpadd16(s02, c_0);			s12 = vis_fpadd16(s12, c_1);			s22 = vis_fpadd16(s22, c_2);			s00 = vis_fpadd16(s00, s02);			s10 = vis_fpadd16(s10, s12);			s20 = vis_fpadd16(s20, s22);			d_0235 = vis_fpmerge(vis_fpack16(s00),				vis_fpack16(s10));			d_xx14 = vis_freg_pair(fzero, vis_fpack16(s20));/* * merge buff values to 3-channel array */			d_23xx = vis_faligndata(d_0235, d_0235);			d_0145 = vis_bshuffle(d_0235, d_xx14);			pfd[0] = vis_read_hi(d_0145);			pfd[1] = vis_read_hi(d_23xx);			pfd[2] = vis_read_lo(d_0145);			buff2 += 2;			pfd += 3;		}		if ((mlib_u8 *)pfd <= dend) {			mlib_d64 d_0235, d_xx14, d_23xx, d_0145;			mlib_f32 *tmp_arr32 = (mlib_f32 *)tmp_arr64;			mlib_d64 s00, s01, s02, s10, s11, s12, s20, s21, s22,				s_0;			mlib_f32 x0, x1, x2;			x0 = (*sf0++);//.........这里部分代码省略.........

mlib_status__mlib_VideoColorMerge3_S16(	mlib_s16 *colors,	const mlib_s16 *color1,	const mlib_s16 *color2,	const mlib_s16 *color3,	mlib_s32 n){	mlib_d64 *dp = (mlib_d64 *)colors;	mlib_d64 *sp0 = (mlib_d64 *)color1;	mlib_d64 *sp1 = (mlib_d64 *)color2;	mlib_d64 *sp2 = (mlib_d64 *)color3;	mlib_d64 sd0, sd1, sd2, sd3, sd4, sd5;	mlib_d64 dd0, dd1, dd2, dd3, dd4, dd5;	mlib_s32 i;#pragma pipeloop(1)	for (i = 0; i <= (n - 8); i += 8) {		sd0 = sp0[0];		sd1 = sp1[0];		sd2 = sp2[0];		sd3 = sp0[1];		sd4 = sp1[1];		sd5 = sp2[1];		vis_write_bmask(0x018923ab, 0);		dd0 = vis_bshuffle(sd0, sd1);		dd3 = vis_bshuffle(sd1, sd2);		dd2 = vis_bshuffle(sd3, sd4);		dd5 = vis_bshuffle(sd4, sd5);		vis_write_bmask(0x45cd67ef, 0);		dd1 = vis_bshuffle(sd0, sd1);		dd4 = vis_bshuffle(sd3, sd4);		vis_write_bmask(0x01238945, 0);		dp[0] = vis_bshuffle(dd0, sd2);		dp[3] = vis_bshuffle(dd2, sd5);		dp[1] = vis_freg_pair(vis_read_lo(dd3), vis_read_hi(dd1));		dp[4] = vis_freg_pair(vis_read_lo(dd5), vis_read_hi(dd4));		vis_write_bmask(0xcd4567ef, 0);		dp[2] = vis_bshuffle(dd1, sd2);		dp[5] = vis_bshuffle(dd4, sd5);		sp0 += 2;		sp1 += 2;		sp2 += 2;		dp += 6;	}	if (i <= (n - 4)) {		sd0 = sp0[0];		sd1 = sp1[0];		sd2 = sp2[0];		vis_write_bmask(0x018923ab, 0);		dd0 = vis_bshuffle(sd0, sd1);		dd3 = vis_bshuffle(sd1, sd2);		vis_write_bmask(0x45cd67ef, 0);		dd1 = vis_bshuffle(sd0, sd1);		vis_write_bmask(0x01238945, 0);		dp[0] = vis_bshuffle(dd0, sd2);		dp[1] = vis_freg_pair(vis_read_lo(dd3), vis_read_hi(dd1));		vis_write_bmask(0xcd4567ef, 0);		dp[2] = vis_bshuffle(dd1, sd2);		sp0++;		sp1++;		sp2++;		dp += 3;	}	for (; i < n; i++) {		colors[3 * i] = ((mlib_u16 *)color1)[i];		colors[3 * i + 1] = ((mlib_u16 *)color2)[i];		colors[3 * i + 2] = ((mlib_u16 *)color3)[i];	}	return (MLIB_SUCCESS);}

mlib_status__mlib_VideoInterpX_U8_U8(	mlib_u8 *curr_block,	const mlib_u8 *ref_block,	mlib_s32 width,	mlib_s32 height,	mlib_s32 frame_stride,	mlib_s32 field_stride){	mlib_d64 s0, s1, s2, s3, s4, s5, s6, s7, d0, d1, d2, d3, *sd, *dd;	mlib_d64 dzero = vis_fzero();	const mlib_f32 fm1 = vis_to_float(0x100);	mlib_f32 fzero = vis_read_hi(dzero);	mlib_d64 rounder = vis_fone();	mlib_s32 y;	vis_write_gsr((6 << 3) + ((mlib_u32)ref_block & 7));	dd = (mlib_d64 *)curr_block;	sd = (mlib_d64 *)((mlib_addr)ref_block & ~7);	if (width == 8) {		y = height >> 2;		if (((mlib_s32)(ref_block + 1) & 7)) {			do {				s0 = sd[0];				s1 = sd[1];				sd = (mlib_d64 *)((mlib_u8 *)sd + field_stride);				d0 = vis_faligndata(s0, s1);				s2 = sd[0];				s3 = sd[1];				sd = (mlib_d64 *)((mlib_u8 *)sd + field_stride);				d1 = vis_faligndata(s2, s3);				s4 = sd[0];				s5 = sd[1];				sd = (mlib_d64 *)((mlib_u8 *)sd + field_stride);				d2 = vis_faligndata(s4, s5);				s6 = sd[0];				s7 = sd[1];				sd = (mlib_d64 *)((mlib_u8 *)sd + field_stride);				d3 = vis_faligndata(s6, s7);				vis_alignaddr((void *)(ref_block + 1), 0);				s0 = vis_faligndata(s0, s1);				s1 = vis_faligndata(s2, s3);				s2 = vis_faligndata(s4, s5);				s3 = vis_faligndata(s6, s7);				MLIB_V_VIDEOINTERP(d0, d0, s0);				MLIB_V_VIDEOINTERP(d1, d1, s1);				MLIB_V_VIDEOINTERP(d2, d2, s2);				MLIB_V_VIDEOINTERP4(d3, d3, s3);				*dd = d0;				dd = (mlib_d64 *)((mlib_u8 *)dd + field_stride);				*dd = d1;				dd = (mlib_d64 *)((mlib_u8 *)dd + field_stride);				*dd = d2;				dd = (mlib_d64 *)((mlib_u8 *)dd + field_stride);				*dd = d3;				dd = (mlib_d64 *)((mlib_u8 *)dd + field_stride);				vis_alignaddr((void *)ref_block, 0);			} while (--y);		} else {			do {				s0 = sd[0];				s1 = sd[1];				sd = (mlib_d64 *)((mlib_u8 *)sd + field_stride);				d0 = vis_faligndata(s0, s1);				s2 = sd[0];				s3 = sd[1];				sd = (mlib_d64 *)((mlib_u8 *)sd + field_stride);				d1 = vis_faligndata(s2, s3);				s4 = sd[0];				s5 = sd[1];				sd = (mlib_d64 *)((mlib_u8 *)sd + field_stride);				d2 = vis_faligndata(s4, s5);				s6 = sd[0];				s7 = sd[1];				sd = (mlib_d64 *)((mlib_u8 *)sd + field_stride);				d3 = vis_faligndata(s6, s7);				MLIB_V_VIDEOINTERP4(d0, d0, s1);				MLIB_V_VIDEOINTERP4(d1, d1, s3);				MLIB_V_VIDEOINTERP4(d2, d2, s5);				MLIB_V_VIDEOINTERP4(d3, d3, s7);				*dd = d0;				dd = (mlib_d64 *)((mlib_u8 *)dd + field_stride);				*dd = d1;				dd = (mlib_d64 *)((mlib_u8 *)dd + field_stride);				*dd = d2;				dd = (mlib_d64 *)((mlib_u8 *)dd + field_stride);				*dd = d3;				dd = (mlib_d64 *)((mlib_u8 *)dd + field_stride);			} while (--y);		}	} else {

mlib_status__mlib_VideoColorARGB2JFIFYCC422(	mlib_u8 *y,	mlib_u8 *cb,	mlib_u8 *cr,	const mlib_u8 *argb,	mlib_s32 n){	mlib_d64 *sp = (mlib_d64 *)argb, *py = (mlib_d64 *)y;	mlib_f32 *pcb = (mlib_f32 *)cb, *pcr = (mlib_f32 *)cr;	mlib_u8 *yend = y + n, *cbend = cb + (n >> 1);	mlib_d64 sd01, sd23, sd45, sd67, sd04, sd26, sd15, sd37;	mlib_d64 dh0, dh1, dl0, dl1, z0, z1;	mlib_s32 i;	mlib_f32 k11 = vis_to_float((mlib_s32)(K11 * 8192));	mlib_f32 k12 = vis_to_float((mlib_s32)(K12 * 8192));	mlib_f32 k13 = vis_to_float((mlib_s32)(K13 * 8192));	mlib_f32 k21 = vis_to_float((mlib_s32)(K21 * 4096));	mlib_f32 k22 = vis_to_float((mlib_s32)(K22 * 4096));	mlib_f32 k23 = vis_to_float((mlib_s32)(K23 * 4096));	mlib_f32 k31 = vis_to_float((mlib_s32)(K31 * 4096));	mlib_f32 k32 = vis_to_float((mlib_s32)(K32 * 4096));	mlib_f32 k33 = vis_to_float((mlib_s32)(K33 * 4096));	mlib_d64 off128 = vis_to_double_dup(0x10101010);	mlib_d64 off0 = vis_to_double_dup(0x00100010);	if (n <= 0)		return (MLIB_FAILURE);	vis_write_gsr(2 << 3);	n = n >> 3;#pragma pipeloop(0)	for (i = 0; i < n; i++) {		sd01 = (*sp++);		sd23 = (*sp++);		sd45 = (*sp++);		sd67 = (*sp++);		CHANNELSEPARATE_U8_422(sd01, sd23, sd45, sd67, dh0, dh1, dl0,			dl1);		CHANNELWEIGHT_U8_2p(vis_read_lo(dh0), vis_read_hi(dh1),			vis_read_lo(dh1), vis_read_lo(dl0), vis_read_hi(dl1),			vis_read_lo(dl1), k11, k12, k13, off0, z0, z1);		z1 = vis_fpadd16(z1, off0);		py[0] = vis_fpmerge(vis_fpack16(z0), vis_fpack16(z1));		CHANNELWEIGHT_U8_2p(vis_read_lo(dh0), vis_read_hi(dh1),			vis_read_lo(dh1), vis_read_lo(dl0), vis_read_hi(dl1),			vis_read_lo(dl1), k21, k22, k23, off128, z0, z1);		pcb[0] = vis_fpack16(vis_fpadd16(z0, z1));		CHANNELWEIGHT_U8_2p(vis_read_lo(dh0), vis_read_hi(dh1),			vis_read_lo(dh1), vis_read_lo(dl0), vis_read_hi(dl1),			vis_read_lo(dl1), k31, k32, k33, off128, z0, z1);		pcr[0] = vis_fpack16(vis_fpadd16(z0, z1));		py++;		pcb++;		pcr++;	}	if ((mlib_u8 *)pcb < cbend) {		mlib_d64 yd;		mlib_f32 cbf, crf;		mlib_s32 ymask, cmask;		sd01 = (*sp++);		sd23 = vis_ld_d64_nf(sp); sp++;		sd45 = vis_ld_d64_nf(sp); sp++;		sd67 = vis_ld_d64_nf(sp);		CHANNELSEPARATE_U8_422(sd01, sd23, sd45, sd67, dh0, dh1, dl0,			dl1);		CHANNELWEIGHT_U8_2p(vis_read_lo(dh0), vis_read_hi(dh1),			vis_read_lo(dh1), vis_read_lo(dl0), vis_read_hi(dl1),			vis_read_lo(dl1), k11, k12, k13, off0, z0, z1);		z1 = vis_fpadd16(z1, off0);		yd = vis_fpmerge(vis_fpack16(z0), vis_fpack16(z1));		CHANNELWEIGHT_U8_2p(vis_read_lo(dh0), vis_read_hi(dh1),			vis_read_lo(dh1), vis_read_lo(dl0), vis_read_hi(dl1),			vis_read_lo(dl1), k21, k22, k23, off128, z0, z1);		cbf = vis_fpack16(vis_fpadd16(z0, z1));		CHANNELWEIGHT_U8_2p(vis_read_lo(dh0), vis_read_hi(dh1),			vis_read_lo(dh1), vis_read_lo(dl0), vis_read_hi(dl1),			vis_read_lo(dl1), k31, k32, k33, off128, z0, z1);		crf = vis_fpack16(vis_fpadd16(z0, z1));		ymask = vis_edge8(py, yend - 1);		vis_pst_8(yd, py, ymask);		cmask = vis_edge8(pcb, cbend - 1);		if (cmask & 0xf0) {			vis_pst_8(vis_freg_pair(cbf, vis_fzeros()), pcb, cmask);			vis_pst_8(vis_freg_pair(crf, vis_fzeros()), pcr, cmask);		} else {			vis_pst_8(vis_freg_pair(vis_fzeros(), cbf), pcb - 1,				cmask);//.........这里部分代码省略.........

mlib_status__mlib_VectorConvert_S32_S8_Mod(	mlib_s32 *z,	const mlib_s8 *x,	mlib_s32 n){	mlib_s8 *psrc = (mlib_s8 *)x;	mlib_s32 *pdst = (mlib_s32 *)z;	mlib_f32 fone = vis_to_float(0x10001);	mlib_d64 *dpsrc, dsrc0, dsrc1, dsrc, dst0, dst1, dst2, dst3, done =		vis_to_double_dup(0x1000100);	mlib_s32 i = 0;	if (n <= 0)		return (MLIB_FAILURE);	if ((mlib_addr)pdst & 7) {		(*pdst++) = (*psrc++);		i = 1;	}	dpsrc = (mlib_d64 *)vis_alignaddr(psrc, 0);	dsrc = vis_ld_d64_nf(dpsrc);	vis_write_bmask(0x00012223, 0);	if ((mlib_addr)psrc & 7) {		dsrc1 = vis_ld_d64_nf(dpsrc + 1);		dsrc = vis_faligndata(dsrc, dsrc1);#pragma pipeloop(1)#pragma unroll(1)		for (; i <= (n - 8); i += 8) {			dst1 = vis_fpmerge(vis_read_hi(dsrc),				vis_read_hi(dsrc));			dst1 = vis_fmul8sux16(dst1, done);			dst0 = vis_bshuffle(dst1, dst1);			dst1 = vis_fmuld8ulx16(fone, vis_read_lo(dst1));			dst3 = vis_fpmerge(vis_read_lo(dsrc),				vis_read_lo(dsrc));			dst3 = vis_fmul8sux16(dst3, done);			dst2 = vis_fmuld8ulx16(fone, vis_read_hi(dst3));			dst3 = vis_fmuld8ulx16(fone, vis_read_lo(dst3));			dsrc0 = dsrc1;			dsrc1 = vis_ld_d64_nf(dpsrc + 2);			dsrc = vis_faligndata(dsrc0, dsrc1);			((mlib_d64 *)pdst)[0] = dst0;			((mlib_d64 *)pdst)[1] = dst1;			((mlib_d64 *)pdst)[2] = dst2;			((mlib_d64 *)pdst)[3] = dst3;			pdst += 8;			psrc += 8;			dpsrc++;		}	} else {#pragma pipeloop(1)#pragma unroll(1)		for (; i <= (n - 8); i += 8) {			dst1 = vis_fpmerge(vis_read_hi(dsrc),				vis_read_hi(dsrc));			dst1 = vis_fmul8sux16(dst1, done);			dst0 = vis_bshuffle(dst1, dst1);			dst1 = vis_fmuld8ulx16(fone, vis_read_lo(dst1));			dst3 = vis_fpmerge(vis_read_lo(dsrc),				vis_read_lo(dsrc));			dst3 = vis_fmul8sux16(dst3, done);			dst2 = vis_bshuffle(dst3, dst3);			dst3 = vis_fmuld8ulx16(fone, vis_read_lo(dst3));			dsrc = vis_ld_d64_nf(dpsrc + 1);			((mlib_d64 *)pdst)[0] = dst0;			((mlib_d64 *)pdst)[1] = dst1;			((mlib_d64 *)pdst)[2] = dst2;			((mlib_d64 *)pdst)[3] = dst3;			pdst += 8;			psrc += 8;			dpsrc++;		}	}	for (; i < n; i++)		(*pdst++) = (*psrc++);	return (MLIB_SUCCESS);}

mlib_status__mlib_VideoColorCMYK2JFIFYCCK444(	mlib_u8 *y,	mlib_u8 *cb,	mlib_u8 *cr,	mlib_u8 *k,	const mlib_u8 *cmyk,	mlib_s32 n){	mlib_d64 buff_arr[(SIZE / 2) + 2];	mlib_f32 *py, *pcb, *pcr, *pk;	mlib_d64 *buff;	mlib_d64 sdh, sdl, dr, dg, db, dd;	mlib_s32 i, m, size, num;	mlib_f32 k11 = vis_to_float((mlib_s32)(K11 * 8192));	mlib_f32 k12 = vis_to_float((mlib_s32)(K12 * 8192));	mlib_f32 k13 = vis_to_float((mlib_s32)(K13 * 8192));	mlib_f32 k21 = vis_to_float((mlib_s32)(K21 * 8192));	mlib_f32 k22 = vis_to_float((mlib_s32)(K22 * 8192));	mlib_f32 k23 = vis_to_float((mlib_s32)(K23 * 8192));	mlib_f32 k31 = vis_to_float((mlib_s32)(K31 * 8192));	mlib_f32 k32 = vis_to_float((mlib_s32)(K32 * 8192));	mlib_f32 k33 = vis_to_float((mlib_s32)(K33 * 8192));	mlib_d64 off128 = vis_to_double_dup(0x10101010);	mlib_d64 off255 = vis_to_double_dup(0x1ff01ff0);	vis_write_gsr(2 << 3);/* * 4-pixel loop */	for (size = 0; size < n; size += num) {		num = n - size;		if (num > SIZE)			num = SIZE;		m = (num + 3) / 4;		mlib_channel_separate((mlib_d64 *)cmyk + size / 2, buff_arr, m);		m = (num / 4) & ~1;		py = (mlib_f32 *)y + size / 4;		pcb = (mlib_f32 *)cb + size / 4;		pcr = (mlib_f32 *)cr + size / 4;		pk = (mlib_f32 *)k + size / 4;		buff = buff_arr;#pragma pipeloop(0)		for (i = 0; i < m; i++) {			sdh = buff[0];			sdl = buff[1];			CHANNELWEIGHT_U8(vis_read_hi(sdh), vis_read_lo(sdh),				vis_read_hi(sdl), k11, k12, k13, off255, py[0]);			CHANNELWEIGHT_U8(vis_read_hi(sdh), vis_read_lo(sdh),				vis_read_hi(sdl), k21, k22, k23, off128,				pcb[0]);			CHANNELWEIGHT_U8(vis_read_hi(sdh), vis_read_lo(sdh),				vis_read_hi(sdl), k31, k32, k33, off128,				pcr[0]);			py++;			pcb++;			pcr++;			(*pk++) = vis_read_lo(sdl);			buff += 2;		}	}	if (n & 7) {		mlib_s32 emask = (0xFF00 >> (n & 7)) & 0xFF;		mlib_d64 rbuff[4];		mlib_f32 *prbuff = (mlib_f32 *)rbuff;		sdh = (*buff++);		sdl = (*buff++);		CHANNELWEIGHT_U8(vis_read_hi(sdh), vis_read_lo(sdh),			vis_read_hi(sdl), k11, k12, k13, off255, prbuff[0]);		CHANNELWEIGHT_U8(vis_read_hi(sdh), vis_read_lo(sdh),			vis_read_hi(sdl), k21, k22, k23, off128, prbuff[2]);		CHANNELWEIGHT_U8(vis_read_hi(sdh), vis_read_lo(sdh),			vis_read_hi(sdl), k31, k32, k33, off128, prbuff[4]);		prbuff[6] = vis_read_lo(sdl);		sdh = (*buff++);		sdl = (*buff++);		CHANNELWEIGHT_U8(vis_read_hi(sdh), vis_read_lo(sdh),			vis_read_hi(sdl), k11, k12, k13, off255, prbuff[1]);		CHANNELWEIGHT_U8(vis_read_hi(sdh), vis_read_lo(sdh),			vis_read_hi(sdl), k21, k22, k23, off128, prbuff[3]);		CHANNELWEIGHT_U8(vis_read_hi(sdh), vis_read_lo(sdh),			vis_read_hi(sdl), k31, k32, k33, off128, prbuff[5]);		prbuff[7] = vis_read_lo(sdl);		vis_pst_8(rbuff[0], py, emask);		vis_pst_8(rbuff[1], pcb, emask);		vis_pst_8(rbuff[2], pcr, emask);		vis_pst_8(rbuff[3], pk, emask);	}

void mlib_v_ImageLookUpSI_U16_U8_4_DstOff3_D1(const mlib_u16 *src,                                              mlib_u8        *dst,                                              mlib_s32       xsize,                                              const mlib_u8  **table){  mlib_u16 *sp;                        /* pointer to source data */  mlib_s32 s0, s1, s2;                 /* source data */  mlib_u8 *dl;                         /* pointer to start of destination */  mlib_d64 *dp;                        /* aligned pointer to destination */  mlib_d64 t0, t1, t2;                 /* destination data */  mlib_d64 t3, t4, t5;                 /* destination data */  mlib_d64 t6, t7, acc;                /* destination data */  mlib_s32 i;                          /* loop variable */  const mlib_u8 *tab0 = &table[0][0];  const mlib_u8 *tab1 = &table[1][0];  const mlib_u8 *tab2 = &table[2][0];  const mlib_u8 *tab3 = &table[3][0];  sp = (void *)src;  dl = dst;  dp = (mlib_d64 *) dl;  vis_alignaddr((void *)0, 7);  s0 = *sp++;  if (xsize >= 2) {    s1 = sp[0];    s2 = sp[1];    sp += 2;#pragma pipeloop(0)    for (i = 0; i <= xsize - 4; i += 2, sp += 2) {      t7 = VIS_LD_U8_I(tab2, s2);      t6 = VIS_LD_U8_I(tab1, s2);      t5 = VIS_LD_U8_I(tab0, s2);      t4 = VIS_LD_U8_I(tab3, s1);      t3 = VIS_LD_U8_I(tab2, s1);      t2 = VIS_LD_U8_I(tab1, s1);      t1 = VIS_LD_U8_I(tab0, s1);      t0 = VIS_LD_U8_I(tab3, s0);      acc = vis_faligndata(t7, acc);      acc = vis_faligndata(t6, acc);      acc = vis_faligndata(t5, acc);      acc = vis_faligndata(t4, acc);      acc = vis_faligndata(t3, acc);      acc = vis_faligndata(t2, acc);      acc = vis_faligndata(t1, acc);      acc = vis_faligndata(t0, acc);      s0 = s2;      s1 = sp[0];      s2 = sp[1];      *dp++ = acc;    }    t7 = VIS_LD_U8_I(tab2, s2);    t6 = VIS_LD_U8_I(tab1, s2);    t5 = VIS_LD_U8_I(tab0, s2);    t4 = VIS_LD_U8_I(tab3, s1);    t3 = VIS_LD_U8_I(tab2, s1);    t2 = VIS_LD_U8_I(tab1, s1);    t1 = VIS_LD_U8_I(tab0, s1);    t0 = VIS_LD_U8_I(tab3, s0);    acc = vis_faligndata(t7, acc);    acc = vis_faligndata(t6, acc);    acc = vis_faligndata(t5, acc);    acc = vis_faligndata(t4, acc);    acc = vis_faligndata(t3, acc);    acc = vis_faligndata(t2, acc);    acc = vis_faligndata(t1, acc);    acc = vis_faligndata(t0, acc);    s0 = s2;    *dp++ = acc;  }  dl = (mlib_u8 *) dp;  if ((xsize & 1) != 0) {    s1 = sp[0];    t7 = VIS_LD_U8_I(tab2, s1);    t6 = VIS_LD_U8_I(tab1, s1);    t5 = VIS_LD_U8_I(tab0, s1);    t4 = VIS_LD_U8_I(tab3, s0);    acc = vis_faligndata(t7, acc);    acc = vis_faligndata(t6, acc);    acc = vis_faligndata(t5, acc);    acc = vis_faligndata(t4, acc);    *(mlib_f32 *) dl = vis_read_hi(acc);    dl += 4;    s0 = s1;  }  dl[0] = tab3[s0];}

static mlib_statusmlib_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;//.........这里部分代码省略.........

mlib_status__mlib_VectorConvert_S32_U8_Mod(	mlib_s32 *z,	const mlib_u8 *x,	mlib_s32 n){	mlib_u8 *psrc = (mlib_u8 *)x;	mlib_s32 *pdst = (mlib_s32 *)z;	mlib_f32 fzero = vis_fzero(), fone1 = vis_to_float(0x100), fone2 =		vis_to_float(0x10001);	mlib_d64 *dpsrc, dsrc0, dsrc1, dsrc, dst0, dst1, dst2, dst3;	mlib_s32 i = 0, off;	if (n <= 0)		return (MLIB_FAILURE);	if ((mlib_addr)pdst & 7) {		(*pdst++) = (*psrc++);		i = 1;	}	dpsrc = (mlib_d64 *)vis_alignaddr(psrc, 0);	dsrc = dpsrc[0];	off = (mlib_addr)psrc & 7;	if (off) {		dsrc1 = dsrc;		vis_alignaddr((void *)0, 7);		vis_write_bmask(0x11111111 * off, 0x40516273);#pragma pipeloop(0)#pragma unroll(2)		for (; i <= (n - 8); i += 8) {			dsrc0 = dsrc1;			dsrc1 = vis_ld_d64_nf(dpsrc + 1);			dsrc = vis_bshuffle(dsrc0, dsrc1);			dst0 = vis_fmuld8ulx16(vis_read_hi(dsrc), fone2);			dst1 = vis_fmuld8ulx16(vis_read_lo(dsrc), fone2);			dsrc = vis_faligndata(dsrc, dsrc);			dst2 = vis_fmuld8ulx16(vis_read_hi(dsrc), fone2);			dst3 = vis_fmuld8ulx16(vis_read_lo(dsrc), fone2);			((mlib_d64 *)pdst)[0] = dst0;			((mlib_d64 *)pdst)[1] = dst1;			((mlib_d64 *)pdst)[2] = dst2;			((mlib_d64 *)pdst)[3] = dst3;			pdst += 8;			psrc += 8;			dpsrc++;		}	} else {#pragma pipeloop(1)#pragma unroll(1)		for (; i <= (n - 8); i += 8) {			dst1 = vis_fmul8x16al(vis_read_hi(dsrc), fone1);			dst0 = vis_fpmerge(fzero, vis_read_hi(dst1));			dst1 = vis_fpmerge(fzero, vis_read_lo(dst1));			dst3 = vis_fpmerge(vis_read_lo(dsrc),				vis_read_lo(dsrc));			dst2 = vis_fmuld8ulx16(vis_read_hi(dst3), fone2);			dst3 = vis_fmuld8ulx16(vis_read_lo(dst3), fone2);			dsrc = vis_ld_d64_nf(dpsrc + 1);			((mlib_d64 *)pdst)[0] = dst0;			((mlib_d64 *)pdst)[1] = dst1;			((mlib_d64 *)pdst)[2] = dst2;			((mlib_d64 *)pdst)[3] = dst3;			pdst += 8;			psrc += 8;			dpsrc++;		}	}	for (; i < n; i++)		(*pdst++) = (*psrc++);	return (MLIB_SUCCESS);}

mlib_status__mlib_VideoColorARGB2RGB(	mlib_u8 *rgb,	const mlib_u8 *argb,	mlib_s32 n){/* 8-byte aligned start points in dst */	mlib_f32 *dp = (mlib_f32 *)rgb;/* 8-byte aligned start point in src */	mlib_d64 *sp = (mlib_d64 *)argb;/* 8-byte source data */	mlib_d64 sd0, sd1, sd2, sd3;/* intermediate variables */	mlib_d64 sda, sdb, sdc, sdd;/* intermediate variables */	mlib_d64 sde, sdf, sdg, sdh;/* dst data */	mlib_d64 dd0, dd1, dd2;	mlib_s32 i;	if (n <= 0)		return (MLIB_FAILURE);#pragma pipeloop(1)	for (i = 0; i <= (n - 16); i += 16) {/* --r0g0b0--r1g1b1 */		sd0 = sp[0];/* --r2g2b2--r3g3b3 */		sd1 = sp[1];/* --r4g4b4--r5g5b5 */		sd2 = sp[2];/* --r6g6b6--r7g7b7 */		sd3 = sp[3];		EXTRACT_U8_43R;		dp[0] = vis_read_hi(dd0);		dp[3] = vis_read_lo(dd0);		dp[1] = vis_read_hi(dd1);		dp[4] = vis_read_lo(dd1);		dp[2] = vis_read_hi(dd2);		dp[5] = vis_read_lo(dd2);		SPLIT_S32(((mlib_u32 *)dp)[6], ((mlib_u32 *)dp)[7],			((mlib_u32 *)dp)[8], ((mlib_u32 *)sp)[8],			((mlib_u32 *)sp)[9], ((mlib_u32 *)sp)[10],			((mlib_u32 *)sp)[11]);		SPLIT_S32(((mlib_u32 *)dp)[9], ((mlib_u32 *)dp)[10],			((mlib_u32 *)dp)[11], ((mlib_u32 *)sp)[12],			((mlib_u32 *)sp)[13], ((mlib_u32 *)sp)[14],			((mlib_u32 *)sp)[15]);		sp += 8;		dp += 12;	}#pragma unroll(1)	for (; i < n; i++) {		SPLIT_S32_3_U8(((mlib_u32 *)argb)[i], rgb[3 * i],			rgb[3 * i + 1], rgb[3 * i + 2]);	}	return (MLIB_SUCCESS);}

mlib_status__mlib_VectorConvert_S16_S8_Mod(	mlib_s16 *z,	const mlib_s8 *x,	mlib_s32 n){	mlib_s32 i;	const mlib_s8 *src = x;	mlib_s16 *dst = z;	mlib_d64 *ddsrc, *ddst;	mlib_d64 four_16_ones = vis_to_double_dup(0x01000100);	mlib_f32 fzero = vis_fzeros();	mlib_s32 len_64, even_length, rest_64, length = n, off;	mlib_d64 dd0, dd1, dd2, dd4, dd5, dd6, dd7;	if (length < 16) {		EXPAND(mlib_s8, mlib_s16);	}	while ((mlib_addr)dst & 7) {		(*dst++) = (*src++);		length--;	}	ddsrc = (mlib_d64 *)vis_alignaddr((void *)src, 0);	ddst = (mlib_d64 *)dst;	rest_64 = length & 7;	len_64 = length >> 3;	even_length = len_64 << 3;	dd2 = ddsrc[0];	off = (mlib_addr)src & 7;	if (!off) {/* * Both vectors are 64-bit aligned. *//* * Peeling of 1 iteration. */		if (i = (len_64 & 1)) {			dd1 = (*ddsrc++);			(*ddst++) =				vis_fmul8sux16(vis_fpmerge(vis_read_hi(dd1),				fzero), four_16_ones);			(*ddst++) =				vis_fmul8sux16(vis_fpmerge(vis_read_lo(dd1),				fzero), four_16_ones);		}#pragma pipeloop(0)#pragma unroll(4)		for (; i < len_64; i += 2) {			dd1 = (*ddsrc++);			dd2 = (*ddsrc++);			(*ddst++) =				vis_fmul8sux16(vis_fpmerge(vis_read_hi(dd1),				fzero), four_16_ones);			(*ddst++) =				vis_fmul8sux16(vis_fpmerge(vis_read_lo(dd1),				fzero), four_16_ones);			(*ddst++) =				vis_fmul8sux16(vis_fpmerge(vis_read_hi(dd2),				fzero), four_16_ones);			(*ddst++) =				vis_fmul8sux16(vis_fpmerge(vis_read_lo(dd2),				fzero), four_16_ones);		}	} else {/* * Source vector is not 64-bit aligned. * Peeling of 1 iteration. Then loop with step==2. */		vis_alignaddr((void *)0, 1);		vis_write_bmask(0x11111111 * off, 0x04152637);		i = 1;		if (len_64 & 1) {			dd1 = dd2;			dd2 = vis_ld_d64_nf(ddsrc + 1); i++;			dd4 = vis_bshuffle(dd1, dd2);			dd5 = vis_faligndata(dd4, dd4);			(*ddst++) = vis_fmul8sux16(dd4, four_16_ones);			(*ddst++) = vis_fmul8sux16(dd5, four_16_ones);		}#pragma pipeloop(0)#pragma unroll(4)		for (; i <= len_64; i += 2) {			dd0 = dd2;			dd1 = vis_ld_d64_nf(ddsrc + i);			dd2 = vis_ld_d64_nf(ddsrc + i + 1);			dd4 = vis_bshuffle(dd0, dd1);			dd6 = vis_bshuffle(dd1, dd2);			dd5 = vis_faligndata(dd4, dd4);			dd7 = vis_faligndata(dd6, dd6);			(*ddst++) = vis_fmul8sux16(dd4, four_16_ones);			(*ddst++) = vis_fmul8sux16(dd5, four_16_ones);//.........这里部分代码省略.........

mlib_status__mlib_VideoColorJFIFYCC2RGB444(    mlib_u8 *rgb,    const mlib_u8 *y,    const mlib_u8 *cb,    const mlib_u8 *cr,    mlib_s32 size){    mlib_u8 *dend;    mlib_f32 *sf0, *sf1, *sf2, *pfd;    mlib_f32 fzero = vis_fzeros();    mlib_s32 i, n, m, emask;    mlib_d64 tmp_arr64[2];    mlib_d64 k01 = vis_to_double_dup(0x0000f4fd);    mlib_d64 k02 = vis_to_double_dup(0x2cdde926);    mlib_d64 k11 = vis_to_double_dup(0xf4fd38b4);    mlib_d64 k12 = vis_to_double_dup(0xe9260000);    mlib_d64 k21 = vis_to_double_dup(0x38b40000);    mlib_d64 k22 = vis_to_double_dup(0x00002cdd);    mlib_d64 c_0 = vis_to_double_dup(0xe9a110ff);    mlib_d64 c_1 = vis_to_double_dup(0x10ffe3b6);    mlib_d64 c_2 = vis_to_double_dup(0xe3b6e9a1);    mlib_d64 k_0 = vis_to_double_dup(0x20002000);    if (size <= 0)        return (MLIB_FAILURE);    vis_write_gsr((2 << 3) + 2);    vis_write_bmask(0x0489AB37, 0);    do {        /* loop on buffer size */        if (size > 2 * BUFF_SIZE) {            n = 2 * BUFF_SIZE;        } else {            n = size;        }        m = (n - 1) >> 2;        sf0 = (mlib_f32 *)y;        sf1 = (mlib_f32 *)cb;        sf2 = (mlib_f32 *)cr;        dend = rgb + 3 * n - 1;        pfd = (mlib_f32 *)rgb;#pragma pipeloop(0)#pragma unroll(4)        for (i = 0; i < m; i++) {            mlib_d64 s00, s01, s02, s10, s11, s12, s20, s21, s22,                     s_0;            mlib_d64 d_0235, d_0145;            mlib_f32 x0, x1, x2;            x0 = (*sf0++);            x1 = (*sf1++);            x2 = (*sf2++);            s_0 = vis_fmul8x16(x0, k_0);            s01 = vis_fmul8x16(x1, k01);            s11 = vis_fmul8x16(x1, k11);            s21 = vis_fmul8x16(x1, k21);            s02 = vis_fmul8x16(x2, k02);            s12 = vis_fmul8x16(x2, k12);            s22 = vis_fmul8x16(x2, k22);            s00 = vis_fpadd16(s_0, s01);            s10 = vis_fpadd16(s_0, s11);            s20 = vis_fpadd16(s_0, s21);            s02 = vis_fpadd16(s02, c_0);            s12 = vis_fpadd16(s12, c_1);            s22 = vis_fpadd16(s22, c_2);            s00 = vis_fpadd16(s00, s02);            s10 = vis_fpadd16(s10, s12);            s20 = vis_fpadd16(s20, s22);            d_0235 = vis_fpack16_pair(s00, s10);            s20 = vis_freg_pair(vis_fpack16(s20), fzero);            d_0145 = vis_bshuffle(d_0235, s20);            d_0235 = vis_fpack32(d_0235, d_0235);            d_0235 = vis_fpmerge(vis_read_hi(d_0235),                                 vis_read_lo(d_0235));            pfd[0] = vis_read_hi(d_0145);            pfd[1] = vis_read_hi(d_0235);            pfd[2] = vis_read_lo(d_0145);            pfd += 3;        }        /*         * last pixels         */        if ((mlib_u8 *)pfd <= dend) {            mlib_d64 s00, s01, s02, s10, s11, s12, s20, s21, s22,                     s_0;//.........这里部分代码省略.........

mlib_status__mlib_VectorConvert_S32_S16_Mod(	mlib_s32 *z,	const mlib_s16 *x,	mlib_s32 n){	mlib_s32 i;	const mlib_s16 *src = x;	mlib_s32 *dst = z;	mlib_d64 *ddsrc, *ddst;	mlib_s32 len_64, even_length, rest_64, length = n;	mlib_d64 dd1, dd2, dd3, dd4;	mlib_f32 two_16_ones = vis_to_float(0x10001);	if (length < 16) {		EXPAND(mlib_s16, mlib_s32);	}	while ((mlib_addr)dst & 7) {		(*dst++) = (*src++);		length--;	}	ddsrc = (mlib_d64 *)vis_alignaddr((void *)src, 0);	ddst = (mlib_d64 *)dst;	rest_64 = length & 3;	len_64 = length >> 2;	even_length = len_64 << 2;	dd2 = ddsrc[0];	if (!((mlib_addr)(src) & 7)) {/* * Source vector is 64-bit aligned. We can process it without * vis_faligndata. * Peeling of 1 iteration. */		if (i = (len_64 & 1)) {			dd3 = (*ddsrc++);/* * Now obtaining of the 4*32 - signed objects */			(*ddst++) =				vis_fmuld8ulx16(two_16_ones, vis_read_hi(dd3));			(*ddst++) =				vis_fmuld8ulx16(two_16_ones, vis_read_lo(dd3));		}#pragma pipeloop(1)#pragma unroll(1)		for (; i < len_64; i += 2) {			dd3 = (*ddsrc++);			dd4 = (*ddsrc++);/* * Now obtaining of the 4*32 - signed objects */			(*ddst++) =				vis_fmuld8ulx16(two_16_ones, vis_read_hi(dd3));			(*ddst++) =				vis_fmuld8ulx16(two_16_ones, vis_read_lo(dd3));/* * Now obtaining of the 4*32 - signed objects */			(*ddst++) =				vis_fmuld8ulx16(two_16_ones, vis_read_hi(dd4));			(*ddst++) =				vis_fmuld8ulx16(two_16_ones, vis_read_lo(dd4));		}	} else {/* * Source vector is not 64-bit aligned. Use vis_faligndata. * Peeling of 1 iteration. */		i = 1;		if ((len_64 & 1)) {			i++;			dd1 = dd2;			dd2 = vis_ld_d64_nf(ddsrc + 1);			dd3 = vis_faligndata(dd1, dd2);/* * Now obtaining of the 4*32 - signed objects */			(*ddst++) =				vis_fmuld8ulx16(two_16_ones, vis_read_hi(dd3));			(*ddst++) =				vis_fmuld8ulx16(two_16_ones, vis_read_lo(dd3));		}/* * Now loop with step == 2. */#pragma pipeloop(1)#pragma unroll(1)		for (; i <= len_64; i += 2) {			dd1 = dd2;//.........这里部分代码省略.........

voidmlib_v_ImageLookUpSI_U16_S16_2_D1(    const mlib_u16 *src,    mlib_s16 *dst,    mlib_s32 xsize,    const mlib_s16 **table){/* pointer to source data */	mlib_u16 *sp;/* source data */	mlib_s32 s0, s1, s2;/* pointer to start of destination */	mlib_s16 *dl;/* aligned pointer to destination */	mlib_d64 *dp;/* destination data */	mlib_d64 t0, t1, t2;/* destination data */	mlib_d64 t3, acc0;/* loop variable */	mlib_s32 i;	const mlib_s16 *tab0 = &table[0][0];	const mlib_s16 *tab1 = &table[1][0];/* destination data */	mlib_d64 acc1;	sp = (void *)src;	dl = dst;	vis_alignaddr((void *)0, 6);	s0 = (*sp++);	(*dl++) = tab0[s0];	dp = (mlib_d64 *)dl;	xsize--;	s0 <<= 1;	if (xsize >= 2) {		s1 = (sp[0] << 1);		s2 = (sp[1] << 1);		sp += 2;		vis_write_bmask(0x012389ab, 0);#pragma pipeloop(0)		for (i = 0; i <= xsize - 4; i += 2, sp += 2) {			t3 = VIS_LD_U16_I(tab0, s2);			t2 = VIS_LD_U16_I(tab1, s1);			t1 = VIS_LD_U16_I(tab0, s1);			t0 = VIS_LD_U16_I(tab1, s0);			acc1 = vis_faligndata(t3, acc1);			acc1 = vis_faligndata(t2, acc1);			acc0 = vis_faligndata(t1, acc0);			acc0 = vis_faligndata(t0, acc0);			s0 = s2;			s1 = (sp[0] << 1);			s2 = (sp[1] << 1);			(*dp++) = vis_bshuffle(acc0, acc1);		}		t3 = VIS_LD_U16_I(tab0, s2);		t2 = VIS_LD_U16_I(tab1, s1);		t1 = VIS_LD_U16_I(tab0, s1);		t0 = VIS_LD_U16_I(tab1, s0);		acc1 = vis_faligndata(t3, acc1);		acc1 = vis_faligndata(t2, acc1);		acc0 = vis_faligndata(t1, acc0);		acc0 = vis_faligndata(t0, acc0);		s0 = s2;		(*dp++) = vis_bshuffle(acc0, acc1);	}	dl = (mlib_s16 *)dp;	if ((xsize & 1) != 0) {		s1 = (sp[0] << 1);		t1 = VIS_LD_U16_I(tab0, s1);		t0 = VIS_LD_U16_I(tab1, s0);		acc0 = vis_faligndata(t1, acc0);		acc0 = vis_faligndata(t0, acc0);		*(mlib_f32 *)dp = vis_read_hi(acc0);		s0 = s1;		dl += 2;	}	s0 >>= 1;	*dl = tab1[s0];}

void mlib_v_ImageLookUpSI_S16_U16_2_DstA8D1(const mlib_s16 *src,                                            mlib_u16       *dst,                                            mlib_s32       xsize,                                            const mlib_u16 **table){  mlib_s16 *sp;              /* pointer to source data */  mlib_s32 s0, s1;           /* source data */  mlib_u16 *dl;              /* pointer to start of destination */  mlib_d64 *dp;              /* aligned pointer to destination */  mlib_d64 t0, t1, t2;       /* destination data */  mlib_d64 t3, acc;          /* destination data */  mlib_s32 i;                /* loop variable */  const mlib_u16 *tab0 = &table[0][32768];  const mlib_u16 *tab1 = &table[1][32768];  sp   = (void *)src;  dl   = dst;  dp   = (mlib_d64 *) dl;  vis_alignaddr((void *) 0, 6);  if (xsize >= 2) {    s0 = (sp[0] << 1);    s1 = (sp[1] << 1);    sp += 2;#pragma pipeloop(0)    for(i = 0; i <= xsize - 4; i+=2, sp+=2) {      t3 = VIS_LD_U16_I(tab1, s1);      t2 = VIS_LD_U16_I(tab0, s1);      t1 = VIS_LD_U16_I(tab1, s0);      t0 = VIS_LD_U16_I(tab0, s0);      acc = vis_faligndata(t3, acc);      acc = vis_faligndata(t2, acc);      acc = vis_faligndata(t1, acc);      acc = vis_faligndata(t0, acc);      s0 = (sp[0] << 1);      s1 = (sp[1] << 1);      *dp++ = acc;    }    t3 = VIS_LD_U16_I(tab1, s1);    t2 = VIS_LD_U16_I(tab0, s1);    t1 = VIS_LD_U16_I(tab1, s0);    t0 = VIS_LD_U16_I(tab0, s0);    acc = vis_faligndata(t3, acc);    acc = vis_faligndata(t2, acc);    acc = vis_faligndata(t1, acc);    acc = vis_faligndata(t0, acc);    *dp++ = acc;  }  if ((xsize & 1) != 0) {    s0 = (sp[0] << 1);    t1 = VIS_LD_U16_I(tab1, s0);    t0 = VIS_LD_U16_I(tab0, s0);    acc = vis_faligndata(t1, acc);    acc = vis_faligndata(t0, acc);    *(mlib_f32*)dp = vis_read_hi(acc);  }}

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);}

voidmlib_v_ImageLookUpSI_S32_S16_2_D1(    const mlib_s32 *src,    mlib_s16 *dst,    mlib_s32 xsize,    const mlib_s16 **table){    /* pointer to source data */    mlib_s32 *sp;    /* source data */    mlib_s32 s0, s1, s2;    /* pointer to start of destination */    mlib_s16 *dl;    /* aligned pointer to destination */    mlib_d64 *dp;    /* destination data */    mlib_d64 t0, t1, t2;    /* destination data */    mlib_d64 t3, acc;    /* loop variable */    mlib_s32 i;    const mlib_s16 *tab0 = &table[0][(mlib_u32)2147483648u];    const mlib_s16 *tab1 = &table[1][(mlib_u32)2147483648u];    sp = (void *)src;    dl = dst;    vis_alignaddr((void *)0, 6);    s0 = (*sp++);    (*dl++) = tab0[s0];    dp = (mlib_d64 *)dl;    xsize--;    if (xsize >= 2) {        s1 = sp[0];        s2 = sp[1];        sp += 2;#pragma pipeloop(0)        for (i = 0; i <= xsize - 4; i += 2, sp += 2) {            t3 = VIS_LD_U16_I(tab0, ((mlib_addr)2 * s2));            t2 = VIS_LD_U16_I(tab1, ((mlib_addr)2 * s1));            t1 = VIS_LD_U16_I(tab0, ((mlib_addr)2 * s1));            t0 = VIS_LD_U16_I(tab1, ((mlib_addr)2 * s0));            acc = vis_faligndata(t3, acc);            acc = vis_faligndata(t2, acc);            acc = vis_faligndata(t1, acc);            acc = vis_faligndata(t0, acc);            s0 = s2;            s1 = sp[0];            s2 = sp[1];            (*dp++) = acc;        }        t3 = VIS_LD_U16_I(tab0, ((mlib_addr)2 * s2));        t2 = VIS_LD_U16_I(tab1, ((mlib_addr)2 * s1));        t1 = VIS_LD_U16_I(tab0, ((mlib_addr)2 * s1));        t0 = VIS_LD_U16_I(tab1, ((mlib_addr)2 * s0));        acc = vis_faligndata(t3, acc);        acc = vis_faligndata(t2, acc);        acc = vis_faligndata(t1, acc);        acc = vis_faligndata(t0, acc);        s0 = s2;        (*dp++) = acc;    }    dl = (mlib_s16 *)dp;    if ((xsize & 1) != 0) {        s1 = sp[0];        t1 = VIS_LD_U16_I(tab0, ((mlib_addr)2 * s1));        t0 = VIS_LD_U16_I(tab1, ((mlib_addr)2 * s0));        acc = vis_faligndata(t1, acc);        acc = vis_faligndata(t0, acc);        *(mlib_f32 *)dp = vis_read_hi(acc);        s0 = s1;        dl += 2;    }    *dl = tab1[s0];}

//.........这里部分代码省略.........	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);