void
mlib_ImageLineXor8000(
const mlib_u8 *src,
mlib_u8 *dst,
mlib_s32 size)
{
mlib_u8 *dend;
mlib_d64 *dptr;
mlib_d64 *sptr;
mlib_d64 s0, s1;
mlib_d64 mask8000 = vis_to_double_dup(0x80008000);
mlib_s32 j;
mlib_s32 emask;
/* prepare the destination addresses */
dptr = (mlib_d64 *)((mlib_addr)dst & (~7));
j = (mlib_addr)dptr - (mlib_addr)dst;
dend = (mlib_u8 *)dst + size - 1;
/* prepare the source address */
sptr = (mlib_d64 *)VIS_ALIGNADDR(src, j);
/* generate edge mask for the start point */
emask = vis_edge8(dst, dend);
s1 = vis_ld_d64_nf(sptr);
if (emask != 0xff) {
s0 = s1;
s1 = vis_ld_d64_nf(sptr + 1);
s0 = vis_fxor(vis_faligndata(s0, s1), mask8000);
vis_pst_8(s0, dptr++, emask);
sptr++;
j += 8;
}
#pragma pipeloop(0)
for (; j <= (size - 16); j += 8) {
s0 = s1;
s1 = sptr[1];
(*dptr++) = vis_fxor(vis_faligndata(s0, s1), mask8000);
sptr++;
}
if (j <= (size - 8)) {
s0 = s1;
s1 = vis_ld_d64_nf(sptr + 1);
(*dptr++) = vis_fxor(vis_faligndata(s0, s1), mask8000);
sptr++;
j += 8;
}
if (j < size) {
s0 = vis_fxor(vis_faligndata(s1, vis_ld_d64_nf(sptr + 1)),
mask8000);
emask = vis_edge8(dptr, dend);
vis_pst_8(s0, dptr, emask);
}
}
mlib_status
mlib_ImageChannelMerge2_S16(
mlib_s16 *dst_s16_0,
const mlib_s16 *src_s16_0,
const mlib_s16 *src_s16_1,
mlib_s32 height,
mlib_s32 width,
mlib_s32 dst_stride,
mlib_s32 src0_stride,
mlib_s32 src1_stride)
{
mlib_s32 i, j, k, n = width * 2;
mlib_u32 *dp, s0, s1;
mlib_s16 *f_ptr, *s_ptr;
mlib_d64 *sd0_ptr, *sd1_ptr;
mlib_d64 sd0, sd1, sd2, sd3, s0h, s1h, s0l, s1l;
mlib_u32 bm0 = 0x018923ab;
mlib_u32 bm1 = 0x45cd67ef;
mlib_d64 dd0, dd1, dd2, dd3;
for (j = 0; j < height; j++) {
i = 0;
if ((mlib_addr)dst_s16_0 & 3) {
dst_s16_0[0] = src_s16_0[0];
f_ptr = (mlib_s16 *)(src_s16_1);
s_ptr = (mlib_s16 *)(src_s16_0 + 1);
i++;
} else {
f_ptr = (mlib_s16 *)src_s16_0;
s_ptr = (mlib_s16 *)src_s16_1;
}
if (((mlib_addr)(dst_s16_0 + i) & 7) && (i < (n - 1))) {
dst_s16_0[i + 0] = (*f_ptr++);
dst_s16_0[i + 1] = (*s_ptr++);
i += 2;
}
dp = (mlib_u32 *)(dst_s16_0 + i);
if ((((mlib_addr)f_ptr & 7) == 0) &&
(((mlib_addr)s_ptr & 7) == 0)) {
#pragma pipeloop(0)
for (; i < (n - 15); i += 16) {
sd0 = ((mlib_d64 *)f_ptr)[0];
sd1 = ((mlib_d64 *)s_ptr)[0];
sd2 = ((mlib_d64 *)f_ptr)[1];
sd3 = ((mlib_d64 *)s_ptr)[1];
vis_write_bmask(bm0, 0);
dd0 = vis_bshuffle(sd0, sd1);
dd2 = vis_bshuffle(sd2, sd3);
vis_write_bmask(bm1, 0);
dd1 = vis_bshuffle(sd0, sd1);
dd3 = vis_bshuffle(sd2, sd3);
((mlib_d64 *)dp)[0] = dd0;
((mlib_d64 *)dp)[1] = dd1;
((mlib_d64 *)dp)[2] = dd2;
((mlib_d64 *)dp)[3] = dd3;
f_ptr += 8;
s_ptr += 8;
dp += 8;
}
} else if (((mlib_addr)f_ptr & 7) == ((mlib_addr)s_ptr & 7)) {
mlib_d64 s0h, s1h, s0l, s1l;
sd0_ptr = VIS_ALIGNADDR(f_ptr, 0);
sd1_ptr = VIS_ALIGNADDR(s_ptr, 0);
s0h = (*sd0_ptr++);
s1h = (*sd1_ptr++);
#pragma pipeloop(0)
for (; i < (n - 7); i += 8) {
s0l = (*sd0_ptr++);
s1l = (*sd1_ptr++);
sd0 = vis_faligndata(s0h, s0l);
sd1 = vis_faligndata(s1h, s1l);
vis_write_bmask(bm0, 0);
dd0 = vis_bshuffle(sd0, sd1);
vis_write_bmask(bm1, 0);
dd1 = vis_bshuffle(sd0, sd1);
((mlib_d64 *)dp)[0] = dd0;
((mlib_d64 *)dp)[1] = dd1;
s0h = s0l;
s1h = s1l;
f_ptr += 4;
s_ptr += 4;
dp += 4;
}
} else {
sd0_ptr = VIS_ALIGNADDR(f_ptr, 0);
sd1_ptr = VIS_ALIGNADDR(s_ptr, 0);
s0h = vis_ld_d64_nf(sd0_ptr); sd0_ptr++;
s1h = vis_ld_d64_nf(sd1_ptr); sd1_ptr++;
#pragma pipeloop(0)
for (k = 0; i < (n - 7); i += 8, k++) {
VIS_ALIGNADDR(f_ptr, 0);
s0l = vis_ld_d64_nf(sd0_ptr); sd0_ptr++;
sd0 = vis_faligndata(s0h, s0l);
VIS_ALIGNADDR(s_ptr, 0);
s1l = vis_ld_d64_nf(sd1_ptr); sd1_ptr++;
sd1 = vis_faligndata(s1h, s1l);
vis_write_bmask(bm0, 0);
dd0 = vis_bshuffle(sd0, sd1);
vis_write_bmask(bm1, 0);
dd1 = vis_bshuffle(sd0, sd1);
((mlib_d64 *)dp)[0] = dd0;
((mlib_d64 *)dp)[1] = dd1;
s0h = s0l;
s1h = s1l;
dp += 4;
}
f_ptr += (k << 2);
s_ptr += (k << 2);
}
for (; i < (n - 1); i += 2) {
s0 = (mlib_u16)((*f_ptr++));
s1 = (mlib_u16)((*s_ptr++));
(*dp++) = (s0 << 16) + s1;
}
if (i < n)
dst_s16_0[i] = *f_ptr;
src_s16_0 += src0_stride;
src_s16_1 += src1_stride;
dst_s16_0 += dst_stride;
}
return (MLIB_SUCCESS);
}
mlib_status
mlib_ImageChannelMerge3_S16(
mlib_s16 *dst_s16_0,
const mlib_s16 *src_s16_0,
const mlib_s16 *src_s16_1,
const mlib_s16 *src_s16_2,
mlib_s32 height,
mlib_s32 width,
mlib_s32 dst_stride,
mlib_s32 src0_stride,
mlib_s32 src1_stride,
mlib_s32 src2_stride)
{
mlib_s32 i, j, k, n = width * 3;
mlib_f32 *dp;
mlib_u32 bm0 = 0x0189ff23;
mlib_u32 bm2 = 0xabff45cd;
mlib_u32 bm4 = 0xff67efff;
mlib_u32 bm1 = 0x01238967;
mlib_u32 bm3 = 0x01ab4567;
mlib_u32 bm5 = 0xcd2345ef;
mlib_d64 sd0, sd1, sd2;
mlib_d64 dd0, xx0, dd1, xx1;
mlib_d64 dd2, xx2;
vis_write_gsr(8 << 3);
for (j = 0; j < height; j++) {
i = 0;
k = 0;
for (; (i < (n - 2)) && ((mlib_addr)(dst_s16_0 + i) & 7);
i += 3, k++) {
dst_s16_0[i + 0] = src_s16_0[k];
dst_s16_0[i + 1] = src_s16_1[k];
dst_s16_0[i + 2] = src_s16_2[k];
}
dp = (mlib_f32 *)(dst_s16_0 + i);
if (((mlib_addr)(src_s16_0 + k) & 7) ||
((mlib_addr)(src_s16_1 + k) & 7) ||
((mlib_addr)(src_s16_2 + k) & 7)) {
mlib_d64 s0h, s0l, s1h, s1l, s2h, s2l;
mlib_d64 *sp0;
mlib_d64 *sp1;
mlib_d64 *sp2;
sp0 = VIS_ALIGNADDR((src_s16_0 + k), 0);
s0h = vis_ld_d64_nf(sp0); sp0++;
sp1 = VIS_ALIGNADDR((src_s16_1 + k), 0);
s1h = vis_ld_d64_nf(sp1); sp1++;
sp2 = VIS_ALIGNADDR((src_s16_2 + k), 0);
s2h = vis_ld_d64_nf(sp2); sp2++;
#pragma pipeloop(0)
for (; i < (n - 11); i += 12, k += 4) {
s0l = vis_ld_d64_nf(sp0); sp0++;
s1l = vis_ld_d64_nf(sp1); sp1++;
s2l = vis_ld_d64_nf(sp2); sp2++;
VIS_ALIGNADDR((src_s16_0 + k), 0);
sd0 = vis_faligndata(s0h, s0l);
VIS_ALIGNADDR((src_s16_1 + k), 0);
sd1 = vis_faligndata(s1h, s1l);
VIS_ALIGNADDR((src_s16_2 + k), 0);
sd2 = vis_faligndata(s2h, s2l);
vis_write_bmask(bm0, 0);
xx0 = vis_bshuffle(sd0, sd1);
vis_write_bmask(bm1, 0);
dd0 = vis_bshuffle(xx0, sd2);
vis_write_bmask(bm2, 0);
xx1 = vis_bshuffle(sd0, sd1);
vis_write_bmask(bm3, 0);
dd1 = vis_bshuffle(xx1, sd2);
vis_write_bmask(bm4, 0);
xx2 = vis_bshuffle(sd0, sd1);
vis_write_bmask(bm5, 0);
dd2 = vis_bshuffle(xx2, sd2);
((mlib_d64 *)dp)[0] = dd0;
((mlib_d64 *)dp)[1] = dd1;
((mlib_d64 *)dp)[2] = dd2;
dp += 6;
s0h = s0l;
s1h = s1l;
s2h = s2l;
}
} else {
#pragma pipeloop(0)
for (; i < (n - 11); i += 12, k += 4) {
sd0 = *((mlib_d64 *)(src_s16_0 + k));
sd1 = *((mlib_d64 *)(src_s16_1 + k));
sd2 = *((mlib_d64 *)(src_s16_2 + k));
vis_write_bmask(bm0, 0);
xx0 = vis_bshuffle(sd0, sd1);
vis_write_bmask(bm1, 0);
dd0 = vis_bshuffle(xx0, sd2);
vis_write_bmask(bm2, 0);
xx1 = vis_bshuffle(sd0, sd1);
vis_write_bmask(bm3, 0);
dd1 = vis_bshuffle(xx1, sd2);
vis_write_bmask(bm4, 0);
xx2 = vis_bshuffle(sd0, sd1);
vis_write_bmask(bm5, 0);
dd2 = vis_bshuffle(xx2, sd2);
((mlib_d64 *)dp)[0] = dd0;
((mlib_d64 *)dp)[1] = dd1;
((mlib_d64 *)dp)[2] = dd2;
dp += 6;
}
}
for (; i < (n - 2); i += 3, k++) {
dst_s16_0[i + 0] = src_s16_0[k];
dst_s16_0[i + 1] = src_s16_1[k];
dst_s16_0[i + 2] = src_s16_2[k];
}
dst_s16_0 += dst_stride;
src_s16_0 += src0_stride;
src_s16_1 += src1_stride;
src_s16_2 += src2_stride;
}
return (MLIB_SUCCESS);
}
mlib_status
mlib_ImageChannelMerge4_S16(
mlib_s16 *dst_s16_0,
const mlib_s16 *src_s16_0,
const mlib_s16 *src_s16_1,
const mlib_s16 *src_s16_2,
const mlib_s16 *src_s16_3,
mlib_s32 height,
mlib_s32 width,
mlib_s32 dst_stride,
mlib_s32 src0_stride,
mlib_s32 src1_stride,
mlib_s32 src2_stride,
mlib_s32 src3_stride)
{
mlib_s32 i, j, n = width << 2;
mlib_s16 *fi_ptr, *se_ptr, *th_ptr, *fo_ptr;
mlib_d64 *dp;
for (j = 0; j < height; j++) {
i = 0;
if ((mlib_addr)(dst_s16_0 + i) & 7) {
dst_s16_0[i++] = src_s16_0[0];
if ((mlib_addr)(dst_s16_0 + i) & 7) {
dst_s16_0[i++] = src_s16_1[0];
if ((mlib_addr)(dst_s16_0 + i) & 7) {
dst_s16_0[i++] = src_s16_2[0];
}
}
}
if (i == 0) {
fi_ptr = (mlib_s16 *)src_s16_0;
se_ptr = (mlib_s16 *)src_s16_1;
th_ptr = (mlib_s16 *)src_s16_2;
fo_ptr = (mlib_s16 *)src_s16_3;
} else if (i == 1) {
fi_ptr = (mlib_s16 *)src_s16_1;
se_ptr = (mlib_s16 *)src_s16_2;
th_ptr = (mlib_s16 *)src_s16_3;
fo_ptr = (mlib_s16 *)(src_s16_0 + 1);
} else if (i == 2) {
fi_ptr = (mlib_s16 *)src_s16_2;
se_ptr = (mlib_s16 *)src_s16_3;
th_ptr = (mlib_s16 *)(src_s16_0 + 1);
fo_ptr = (mlib_s16 *)(src_s16_1 + 1);
} else if (i == 3) {
fi_ptr = (mlib_s16 *)src_s16_3;
se_ptr = (mlib_s16 *)(src_s16_0 + 1);
th_ptr = (mlib_s16 *)(src_s16_1 + 1);
fo_ptr = (mlib_s16 *)(src_s16_2 + 1);
}
dp = (mlib_d64 *)(dst_s16_0 + i);
if ((n - i) > 16) {
if (((mlib_addr)fi_ptr & 7) ||
((mlib_addr)se_ptr & 7) ||
((mlib_addr)th_ptr & 7) ||
((mlib_addr)fo_ptr & 7)) {
mlib_d64 sd0, sd1, sd2, sd3;
mlib_d64 dd0, dd1, dd2, dd3, dr02, dr13;
mlib_d64 s0h, s0l, s1h, s1l, s2h, s2l, s3h, s3l;
mlib_d64 *sp0;
mlib_d64 *sp1;
mlib_d64 *sp2;
mlib_d64 *sp3;
sp0 = VIS_ALIGNADDR(fi_ptr, 0);
s0h = (*sp0++);
sp1 = VIS_ALIGNADDR(se_ptr, 0);
s1h = (*sp1++);
sp2 = VIS_ALIGNADDR(th_ptr, 0);
s2h = (*sp2++);
sp3 = VIS_ALIGNADDR(fo_ptr, 0);
s3h = (*sp3++);
#pragma pipeloop(0)
for (; i < (n - 15); i += 16) {
s0l = vis_ld_d64_nf(sp0); sp0++;
s1l = vis_ld_d64_nf(sp1); sp1++;
s2l = vis_ld_d64_nf(sp2); sp2++;
s3l = vis_ld_d64_nf(sp3); sp3++;
VIS_ALIGNADDR(fi_ptr, 0);
sd0 = vis_faligndata(s0h, s0l);
VIS_ALIGNADDR(se_ptr, 0);
sd1 = vis_faligndata(s1h, s1l);
VIS_ALIGNADDR(th_ptr, 0);
sd2 = vis_faligndata(s2h, s2l);
VIS_ALIGNADDR(fo_ptr, 0);
sd3 = vis_faligndata(s3h, s3l);
dr02 =
vis_fpmerge(vis_read_hi(sd0),
vis_read_hi(sd2));
dr13 =
vis_fpmerge(vis_read_hi(sd1),
vis_read_hi(sd3));
dd0 =
vis_fpmerge(vis_read_hi(dr02),
vis_read_hi(dr13));
dp[0] =
vis_fpmerge(vis_read_hi(dd0),
vis_read_lo(dd0));
dd1 =
vis_fpmerge(vis_read_lo(dr02),
vis_read_lo(dr13));
dp[1] =
vis_fpmerge(vis_read_hi(dd1),
vis_read_lo(dd1));
dr02 =
vis_fpmerge(vis_read_lo(sd0),
vis_read_lo(sd2));
dr13 =
vis_fpmerge(vis_read_lo(sd1),
vis_read_lo(sd3));
dd2 =
vis_fpmerge(vis_read_hi(dr02),
vis_read_hi(dr13));
dp[2] =
vis_fpmerge(vis_read_hi(dd2),
vis_read_lo(dd2));
dd3 =
vis_fpmerge(vis_read_lo(dr02),
vis_read_lo(dr13));
dp[3] =
vis_fpmerge(vis_read_hi(dd3),
vis_read_lo(dd3));
dp += 4;
s0h = s0l;
s1h = s1l;
s2h = s2l;
s3h = s3l;
fi_ptr += 4;
se_ptr += 4;
th_ptr += 4;
fo_ptr += 4;
}
} else {
mlib_d64 sd0, sd1, sd2, sd3;
mlib_d64 dd0, dd1, dd2, dd3, dr02, dr13;
#pragma pipeloop(0)
for (; i < (n - 15); i += 16) {
sd0 = ((mlib_d64 *)fi_ptr)[0];
sd1 = ((mlib_d64 *)se_ptr)[0];
sd2 = ((mlib_d64 *)th_ptr)[0];
sd3 = ((mlib_d64 *)fo_ptr)[0];
dr02 =
vis_fpmerge(vis_read_hi(sd0),
vis_read_hi(sd2));
dr13 =
vis_fpmerge(vis_read_hi(sd1),
vis_read_hi(sd3));
dd0 =
vis_fpmerge(vis_read_hi(dr02),
vis_read_hi(dr13));
dp[0] =
vis_fpmerge(vis_read_hi(dd0),
vis_read_lo(dd0));
dd1 =
vis_fpmerge(vis_read_lo(dr02),
vis_read_lo(dr13));
dp[1] =
vis_fpmerge(vis_read_hi(dd1),
vis_read_lo(dd1));
dr02 =
vis_fpmerge(vis_read_lo(sd0),
vis_read_lo(sd2));
dr13 =
vis_fpmerge(vis_read_lo(sd1),
vis_read_lo(sd3));
dd2 =
vis_fpmerge(vis_read_hi(dr02),
vis_read_hi(dr13));
dp[2] =
vis_fpmerge(vis_read_hi(dd2),
vis_read_lo(dd2));
dd3 =
vis_fpmerge(vis_read_lo(dr02),
vis_read_lo(dr13));
dp[3] =
vis_fpmerge(vis_read_hi(dd3),
vis_read_lo(dd3));
dp += 4;
fi_ptr += 4;
se_ptr += 4;
th_ptr += 4;
fo_ptr += 4;
}
}
}
#pragma pipeloop(0)
for (; i < (n - 3); i += 4) {
dst_s16_0[i + 0] = (*fi_ptr++);
dst_s16_0[i + 1] = (*se_ptr++);
dst_s16_0[i + 2] = (*th_ptr++);
dst_s16_0[i + 3] = (*fo_ptr++);
}
if (i < (n - 2)) {
dst_s16_0[i + 0] = *fi_ptr;
dst_s16_0[i + 1] = *se_ptr;
dst_s16_0[i + 2] = *th_ptr;
} else if (i < (n - 1)) {
dst_s16_0[i + 0] = *fi_ptr;
dst_s16_0[i + 1] = *se_ptr;
} else if (i < n) {
dst_s16_0[i + 0] = *fi_ptr;
}
dst_s16_0 += dst_stride;
src_s16_0 += src0_stride;
src_s16_1 += src1_stride;
src_s16_2 += src2_stride;
src_s16_3 += src3_stride;
}
return (MLIB_SUCCESS);
}
mask = mask0 >> offset;
src = da[0];
da[0] = (src & (~mask)) | (sa[0] & mask);
da++;
sa++;
size = size - 8 + offset;
b_size = size >> 3; /* size in bytes */
/* prepare the destination addresses */
dp = (mlib_d64 *) ((mlib_addr) da & (~7));
j = (mlib_addr) dp - (mlib_addr) da;
dend = da + b_size - 1;
/* prepare the source address */
sp = (mlib_d64 *) VIS_ALIGNADDR(sa, j);
/* generate edge mask for the start point */
emask = vis_edge8(da, dend);
s1 = vis_ld_d64_nf(sp);
if (emask != 0xff) {
s0 = s1;
s1 = vis_ld_d64_nf(sp+1);
s0 = vis_faligndata(s0, s1);
vis_pst_8(s0, dp++, emask);
sp++;
j += 8;
}
#pragma pipeloop(0)
for (; j <= (b_size - 8); j += 8) {
