void
mlib_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_S32_S16_4_DstOff3_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;
/* pointer to start of destination */
mlib_s16 *dl;
/* aligned pointer to destination */
mlib_d64 *dp;
/* destination data */
mlib_d64 t0, t1, t2, t3;
/* destination data */
mlib_d64 acc;
/* loop variable */
mlib_s32 i;
const mlib_s16 *tab0 = &table[0][(mlib_u32)2147483648u];
const mlib_s16 *tab1 = &table[1][(mlib_u32)2147483648u];
const mlib_s16 *tab2 = &table[2][(mlib_u32)2147483648u];
const mlib_s16 *tab3 = &table[3][(mlib_u32)2147483648u];
sp = (void *)src;
dl = dst;
dp = (mlib_d64 *)dl;
vis_alignaddr((void *)0, 6);
s0 = (*sp++);
if (xsize >= 1) {
s1 = (*sp++);
#pragma pipeloop(0)
for (i = 0; i <= xsize - 2; i++) {
t3 = VIS_LD_U16_I(tab2, ((mlib_addr)2 * s1));
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(tab3, ((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 = s1;
s1 = (*sp++);
(*dp++) = acc;
}
t3 = VIS_LD_U16_I(tab2, ((mlib_addr)2 * s1));
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(tab3, ((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 = s1;
(*dp++) = acc;
}
dl = (mlib_s16 *)dp;
dl[0] = tab3[s0];
}
void
mlib_v_ImageLookUpSI_S32_S16_2_DstA8D1(
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;
/* 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;
dp = (mlib_d64 *)dl;
vis_alignaddr((void *)0, 6);
if (xsize >= 2) {
s0 = sp[0];
s1 = sp[1];
sp += 2;
#pragma pipeloop(0)
for (i = 0; i <= xsize - 4; i += 2, sp += 2) {
t3 = VIS_LD_U16_I(tab1, ((mlib_addr)2 * s1));
t2 = VIS_LD_U16_I(tab0, ((mlib_addr)2 * s1));
t1 = VIS_LD_U16_I(tab1, ((mlib_addr)2 * s0));
t0 = VIS_LD_U16_I(tab0, ((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 = sp[0];
s1 = sp[1];
(*dp++) = acc;
}
t3 = VIS_LD_U16_I(tab1, ((mlib_addr)2 * s1));
t2 = VIS_LD_U16_I(tab0, ((mlib_addr)2 * s1));
t1 = VIS_LD_U16_I(tab1, ((mlib_addr)2 * s0));
t0 = VIS_LD_U16_I(tab0, ((mlib_addr)2 * 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];
t1 = VIS_LD_U16_I(tab1, ((mlib_addr)2 * s0));
t0 = VIS_LD_U16_I(tab0, ((mlib_addr)2 * s0));
acc = vis_faligndata(t1, acc);
acc = vis_faligndata(t0, acc);
*(mlib_f32 *)dp = vis_read_hi(acc);
}
}
void mlib_v_ImageLookUpSI_S16_S16_4_DstOff2_D1(const mlib_s16 *src,
mlib_s16 *dst,
mlib_s32 xsize,
const mlib_s16 **table)
{
mlib_s16 *sp; /* pointer to source data */
mlib_s32 s0, s1; /* source data */
mlib_s16 *dl; /* pointer to start of destination */
mlib_d64 *dp; /* aligned pointer to destination */
mlib_d64 t0, t1, t2, t3; /* destination data */
mlib_d64 acc; /* destination data */
mlib_s32 i; /* loop variable */
const mlib_s16 *tab0 = &table[0][32768];
const mlib_s16 *tab1 = &table[1][32768];
const mlib_s16 *tab2 = &table[2][32768];
const mlib_s16 *tab3 = &table[3][32768];
sp = (void *)src;
dl = dst;
dp = (mlib_d64 *) dl;
vis_alignaddr((void *) 0, 6);
s0 = (*sp++) << 1;
if (xsize >= 1) {
s1 = (*sp++) << 1;
#pragma pipeloop(0)
for(i = 0; i <= xsize - 2; i++) {
t3 = VIS_LD_U16_I(tab1, s1);
t2 = VIS_LD_U16_I(tab0, s1);
t1 = VIS_LD_U16_I(tab3, s0);
t0 = VIS_LD_U16_I(tab2, s0);
acc = vis_faligndata(t3, acc);
acc = vis_faligndata(t2, acc);
acc = vis_faligndata(t1, acc);
acc = vis_faligndata(t0, acc);
s0 = s1;
s1 = (*sp++) << 1;
*dp++ = acc;
}
t3 = VIS_LD_U16_I(tab1, s1);
t2 = VIS_LD_U16_I(tab0, s1);
t1 = VIS_LD_U16_I(tab3, s0);
t0 = VIS_LD_U16_I(tab2, s0);
acc = vis_faligndata(t3, acc);
acc = vis_faligndata(t2, acc);
acc = vis_faligndata(t1, acc);
acc = vis_faligndata(t0, acc);
s0 = s1;
*dp++ = acc;
}
dl = (mlib_s16*)dp;
s0 >>= 1;
dl[0] = tab2[s0];
dl[1] = tab3[s0];
}
void
mlib_v_ImageLookUpSI_S32_S16_3_D1(
const mlib_s32 *src,
mlib_s16 *dst,
mlib_s32 xsize,
const mlib_s16 **table)
{
/* pointer to source data */
mlib_s32 *sp;
/* pointer to start of destination */
mlib_s16 *dl;
/* aligned pointer to destination */
mlib_d64 *dp;
/* destination data */
mlib_d64 t0, t1, t2, t3;
/* destination data */
mlib_d64 acc0, acc1, acc2;
/* loop variable */
mlib_s32 i;
const mlib_s16 *tab0 = &table[0][(mlib_u32)2147483648u];
const mlib_s16 *tab1 = &table[1][(mlib_u32)2147483648u];
const mlib_s16 *tab2 = &table[2][(mlib_u32)2147483648u];
mlib_s32 s00, s01, s02, s03;
sp = (void *)src;
dl = dst;
dp = (mlib_d64 *)dl;
vis_alignaddr((void *)0, 6);
i = 0;
if (xsize >= 4) {
s00 = sp[0];
s01 = sp[1];
s02 = sp[2];
s03 = sp[3];
sp += 4;
#pragma pipeloop(0)
for (i = 0; i <= xsize - 8; i += 4, sp += 4) {
t3 = VIS_LD_U16_I(tab0, ((mlib_addr)2 * s01));
t2 = VIS_LD_U16_I(tab2, ((mlib_addr)2 * s00));
t1 = VIS_LD_U16_I(tab1, ((mlib_addr)2 * s00));
t0 = VIS_LD_U16_I(tab0, ((mlib_addr)2 * s00));
acc0 = vis_faligndata(t3, acc0);
acc0 = vis_faligndata(t2, acc0);
acc0 = vis_faligndata(t1, acc0);
acc0 = vis_faligndata(t0, acc0);
t3 = VIS_LD_U16_I(tab1, ((mlib_addr)2 * s02));
t2 = VIS_LD_U16_I(tab0, ((mlib_addr)2 * s02));
t1 = VIS_LD_U16_I(tab2, ((mlib_addr)2 * s01));
t0 = VIS_LD_U16_I(tab1, ((mlib_addr)2 * s01));
acc1 = vis_faligndata(t3, acc1);
acc1 = vis_faligndata(t2, acc1);
acc1 = vis_faligndata(t1, acc1);
acc1 = vis_faligndata(t0, acc1);
t3 = VIS_LD_U16_I(tab2, ((mlib_addr)2 * s03));
t2 = VIS_LD_U16_I(tab1, ((mlib_addr)2 * s03));
t1 = VIS_LD_U16_I(tab0, ((mlib_addr)2 * s03));
t0 = VIS_LD_U16_I(tab2, ((mlib_addr)2 * s02));
acc2 = vis_faligndata(t3, acc2);
acc2 = vis_faligndata(t2, acc2);
acc2 = vis_faligndata(t1, acc2);
acc2 = vis_faligndata(t0, acc2);
s00 = sp[0];
s01 = sp[1];
s02 = sp[2];
s03 = sp[3];
(*dp++) = acc0;
(*dp++) = acc1;
(*dp++) = acc2;
}
t3 = VIS_LD_U16_I(tab0, ((mlib_addr)2 * s01));
t2 = VIS_LD_U16_I(tab2, ((mlib_addr)2 * s00));
t1 = VIS_LD_U16_I(tab1, ((mlib_addr)2 * s00));
t0 = VIS_LD_U16_I(tab0, ((mlib_addr)2 * s00));
acc0 = vis_faligndata(t3, acc0);
acc0 = vis_faligndata(t2, acc0);
acc0 = vis_faligndata(t1, acc0);
acc0 = vis_faligndata(t0, acc0);
t3 = VIS_LD_U16_I(tab1, ((mlib_addr)2 * s02));
t2 = VIS_LD_U16_I(tab0, ((mlib_addr)2 * s02));
t1 = VIS_LD_U16_I(tab2, ((mlib_addr)2 * s01));
t0 = VIS_LD_U16_I(tab1, ((mlib_addr)2 * s01));
acc1 = vis_faligndata(t3, acc1);
acc1 = vis_faligndata(t2, acc1);
acc1 = vis_faligndata(t1, acc1);
acc1 = vis_faligndata(t0, acc1);
t3 = VIS_LD_U16_I(tab2, ((mlib_addr)2 * s03));
t2 = VIS_LD_U16_I(tab1, ((mlib_addr)2 * s03));
t1 = VIS_LD_U16_I(tab0, ((mlib_addr)2 * s03));
t0 = VIS_LD_U16_I(tab2, ((mlib_addr)2 * s02));
acc2 = vis_faligndata(t3, acc2);
acc2 = vis_faligndata(t2, acc2);
acc2 = vis_faligndata(t1, acc2);
acc2 = vis_faligndata(t0, acc2);
(*dp++) = acc0;
(*dp++) = acc1;
(*dp++) = acc2;
i += 4;
}
dl = (mlib_s16 *)dp;
#pragma pipeloop(0)
for (; i < xsize; i++) {
s00 = sp[0];
dl[0] = tab0[s00];
dl[1] = tab1[s00];
dl[2] = tab2[s00];
dl += 3;
sp++;
}
}
void mlib_v_ImageLookUp_U16_U16_124_D1(const mlib_u16 *src,
mlib_u16 *dst,
mlib_s32 xsize,
const mlib_u16 *table0,
const mlib_u16 *table1,
const mlib_u16 *table2,
const mlib_u16 *table3)
{
mlib_u16 *sp; /* pointer to source data */
mlib_s32 s0, s1, s2, s3; /* source data */
mlib_u16 *dl; /* pointer to start of destination */
mlib_u16 *dend; /* pointer to end of destination */
mlib_d64 *dp; /* aligned pointer to destination */
mlib_d64 t0, t1, t2; /* destination data */
mlib_d64 t3, acc0; /* destination data */
mlib_s32 emask; /* edge mask */
mlib_s32 i, num; /* loop variable */
dl = dst;
sp = (void *)src;
dp = (mlib_d64 *) dl;
dend = dl + xsize - 1;
vis_alignaddr((void *) 0, 6);
i = 0;
if (xsize >= 4) {
s0 = sp[0];
s1 = sp[1];
s2 = sp[2];
s3 = sp[3];
sp += 4;
#pragma pipeloop(0)
for(i = 0; i <= xsize - 8; i+=4, sp += 4) {
t3 = VIS_LD_U16_I(table3, 2*s3);
t2 = VIS_LD_U16_I(table2, 2*s2);
t1 = VIS_LD_U16_I(table1, 2*s1);
t0 = VIS_LD_U16_I(table0, 2*s0);
acc0 = vis_faligndata(t3, acc0);
acc0 = vis_faligndata(t2, acc0);
acc0 = vis_faligndata(t1, acc0);
acc0 = vis_faligndata(t0, acc0);
s0 = sp[0];
s1 = sp[1];
s2 = sp[2];
s3 = sp[3];
*dp++ = acc0;
}
t3 = VIS_LD_U16_I(table3, 2*s3);
t2 = VIS_LD_U16_I(table2, 2*s2);
t1 = VIS_LD_U16_I(table1, 2*s1);
t0 = VIS_LD_U16_I(table0, 2*s0);
acc0 = vis_faligndata(t3, acc0);
acc0 = vis_faligndata(t2, acc0);
acc0 = vis_faligndata(t1, acc0);
acc0 = vis_faligndata(t0, acc0);
*dp++ = acc0;
}
if ((mlib_addr) dp <= (mlib_addr) dend) {
num = (mlib_u16*) dend - (mlib_u16*) dp;
sp += num;
num ++;
if (num == 1) {
s0 = (mlib_s32) *sp;
sp --;
t0 = VIS_LD_U16_I(table0, 2*s0);
acc0 = vis_faligndata(t0, acc0);
} else if (num == 2) {
s0 = (mlib_s32) *sp;
sp --;
t0 = VIS_LD_U16_I(table1, 2*s0);
acc0 = vis_faligndata(t0, acc0);
s0 = (mlib_s32) *sp;
sp --;
t0 = VIS_LD_U16_I(table0, 2*s0);
acc0 = vis_faligndata(t0, acc0);
} else if (num == 3) {
s0 = (mlib_s32) *sp;
sp --;
t0 = VIS_LD_U16_I(table2, 2*s0);
acc0 = vis_faligndata(t0, acc0);
s0 = (mlib_s32) *sp;
sp --;
t0 = VIS_LD_U16_I(table1, 2*s0);
acc0 = vis_faligndata(t0, acc0);
s0 = (mlib_s32) *sp;
sp --;
t0 = VIS_LD_U16_I(table0, 2*s0);
acc0 = vis_faligndata(t0, acc0);
}
emask = vis_edge16(dp, dend);
vis_pst_16(acc0, dp, emask);
}
}
void
mlib_v_ImageLookUp_S16_S16_3_D1(
const mlib_s16 *src,
mlib_s16 *dst,
mlib_s32 xsize,
const mlib_s16 *table0,
const mlib_s16 *table1,
const mlib_s16 *table2)
{
/* pointer to source data */
mlib_s16 *sp;
/* source data */
mlib_s32 s0, s1, s2, s3;
/* pointer to start of destination */
mlib_s16 *dl;
/* pointer to end of destination */
mlib_s16 *dend;
/* aligned pointer to destination */
mlib_d64 *dp;
/* destination data */
mlib_d64 t0, t1, t2, t3;
/* destination data */
mlib_d64 acc0, acc1;
/* edge mask */
mlib_s32 emask;
/* loop variable */
mlib_s32 i, num;
const mlib_s16 *table;
dl = dst;
sp = (void *)src;
dp = (mlib_d64 *)dl;
dend = dl + xsize - 1;
vis_alignaddr((void *)0, 6);
i = 0;
if (xsize >= 4) {
s0 = sp[0] << 1;
s1 = sp[1] << 1;
s2 = sp[2] << 1;
s3 = sp[3] << 1;
sp += 4;
vis_write_bmask(0x012389ab, 0);
#pragma pipeloop(0)
for (i = 0; i <= xsize - 8; i += 4, sp += 4) {
t3 = VIS_LD_U16_I(table0, s3);
t2 = VIS_LD_U16_I(table2, s2);
t1 = VIS_LD_U16_I(table1, s1);
t0 = VIS_LD_U16_I(table0, s0);
acc1 = vis_faligndata(t3, acc1);
acc1 = vis_faligndata(t2, acc1);
acc0 = vis_faligndata(t1, acc0);
acc0 = vis_faligndata(t0, acc0);
s0 = sp[0] << 1;
s1 = sp[1] << 1;
s2 = sp[2] << 1;
s3 = sp[3] << 1;
(*dp++) = vis_bshuffle(acc0, acc1);
table = table0;
table0 = table1;
table1 = table2;
table2 = table;
}
t3 = VIS_LD_U16_I(table0, s3);
t2 = VIS_LD_U16_I(table2, s2);
t1 = VIS_LD_U16_I(table1, s1);
t0 = VIS_LD_U16_I(table0, s0);
acc1 = vis_faligndata(t3, acc1);
acc1 = vis_faligndata(t2, acc1);
acc0 = vis_faligndata(t1, acc0);
acc0 = vis_faligndata(t0, acc0);
(*dp++) = vis_bshuffle(acc0, acc1);
table = table0;
table0 = table1;
table1 = table2;
table2 = table;
i += 4;
}
if ((mlib_addr)dp <= (mlib_addr)dend) {
num = (mlib_s16 *)dend - (mlib_s16 *)dp;
sp += num;
num++;
if (num == 1) {
s0 = (mlib_s32)*sp;
sp--;
t0 = VIS_LD_U16_I(table0, s0 << 1);
acc0 = vis_faligndata(t0, acc0);
} else if (num == 2) {
s0 = (mlib_s32)*sp;
sp--;
t0 = VIS_LD_U16_I(table1, s0 << 1);
acc0 = vis_faligndata(t0, acc0);
s0 = (mlib_s32)*sp;
sp--;
t0 = VIS_LD_U16_I(table0, s0 << 1);
acc0 = vis_faligndata(t0, acc0);
} else if (num == 3) {
s0 = (mlib_s32)*sp;
sp--;
t0 = VIS_LD_U16_I(table2, s0 << 1);
acc0 = vis_faligndata(t0, acc0);
s0 = (mlib_s32)*sp;
sp--;
t0 = VIS_LD_U16_I(table1, s0 << 1);
acc0 = vis_faligndata(t0, acc0);
s0 = (mlib_s32)*sp;
sp--;
t0 = VIS_LD_U16_I(table0, s0 << 1);
acc0 = vis_faligndata(t0, acc0);
}
emask = vis_edge16(dp, dend);
vis_pst_16(acc0, dp, emask);
}
}
void
mlib_v_ImageLookUpSI_S32_S16_2_DstA8D1(
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;
/* 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 = (void *)&(((mlib_u8 **)table)[0][HALF_U64]);
const mlib_s16 *tab1 = (void *)&(((mlib_u8 **)table)[1][HALF_U64]);
/* destination data */
mlib_d64 acc1;
sp = (void *)src;
dl = dst;
dp = (mlib_d64 *)dl;
vis_alignaddr((void *)0, 6);
if (xsize >= 2) {
s0 = sp[0];
s1 = sp[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(tab1, ((mlib_addr)2 * s1));
t2 = VIS_LD_U16_I(tab0, ((mlib_addr)2 * s1));
t1 = VIS_LD_U16_I(tab1, ((mlib_addr)2 * s0));
t0 = VIS_LD_U16_I(tab0, ((mlib_addr)2 * s0));
acc1 = vis_faligndata(t3, acc1);
acc1 = vis_faligndata(t2, acc1);
acc0 = vis_faligndata(t1, acc0);
acc0 = vis_faligndata(t0, acc0);
s0 = sp[0];
s1 = sp[1];
(*dp++) = vis_bshuffle(acc0, acc1);
}
t3 = VIS_LD_U16_I(tab1, ((mlib_addr)2 * s1));
t2 = VIS_LD_U16_I(tab0, ((mlib_addr)2 * s1));
t1 = VIS_LD_U16_I(tab1, ((mlib_addr)2 * s0));
t0 = VIS_LD_U16_I(tab0, ((mlib_addr)2 * s0));
acc1 = vis_faligndata(t3, acc1);
acc1 = vis_faligndata(t2, acc1);
acc0 = vis_faligndata(t1, acc0);
acc0 = vis_faligndata(t0, acc0);
(*dp++) = vis_bshuffle(acc0, acc1);
}
if ((xsize & 1) != 0) {
s0 = sp[0];
t1 = VIS_LD_U16_I(tab1, ((mlib_addr)2 * s0));
t0 = VIS_LD_U16_I(tab0, ((mlib_addr)2 * s0));
acc0 = vis_faligndata(t1, acc0);
acc0 = vis_faligndata(t0, acc0);
*(mlib_f32 *)dp = vis_read_hi(acc0);
}
}
void
mlib_v_ImageLookUpSI_S32_S16_4_DstOff3_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;
/* pointer to start of destination */
mlib_s16 *dl;
/* aligned pointer to destination */
mlib_d64 *dp;
/* destination data */
mlib_d64 t0, t1, t2, t3;
/* destination data */
mlib_d64 acc0;
/* loop variable */
mlib_s32 i;
const mlib_s16 *tab0 = (void *)&(((mlib_u8 **)table)[0][HALF_U64]);
const mlib_s16 *tab1 = (void *)&(((mlib_u8 **)table)[1][HALF_U64]);
const mlib_s16 *tab2 = (void *)&(((mlib_u8 **)table)[2][HALF_U64]);
const mlib_s16 *tab3 = (void *)&(((mlib_u8 **)table)[3][HALF_U64]);
/* destination data */
mlib_d64 acc1;
sp = (void *)src;
dl = dst;
dp = (mlib_d64 *)dl;
vis_alignaddr((void *)0, 6);
s0 = (*sp++);
if (xsize >= 1) {
s1 = (*sp++);
vis_write_bmask(0x012389ab, 0);
#pragma pipeloop(0)
for (i = 0; i <= xsize - 2; i++) {
t3 = VIS_LD_U16_I(tab2, ((mlib_addr)2 * s1));
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(tab3, ((mlib_addr)2 * s0));
acc1 = vis_faligndata(t3, acc1);
acc1 = vis_faligndata(t2, acc1);
acc0 = vis_faligndata(t1, acc0);
acc0 = vis_faligndata(t0, acc0);
s0 = s1;
s1 = (*sp++);
(*dp++) = vis_bshuffle(acc0, acc1);
}
t3 = VIS_LD_U16_I(tab2, ((mlib_addr)2 * s1));
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(tab3, ((mlib_addr)2 * s0));
acc1 = vis_faligndata(t3, acc1);
acc1 = vis_faligndata(t2, acc1);
acc0 = vis_faligndata(t1, acc0);
acc0 = vis_faligndata(t0, acc0);
s0 = s1;
(*dp++) = vis_bshuffle(acc0, acc1);
}
dl = (mlib_s16 *)dp;
dl[0] = tab3[s0];
}
void
mlib_v_ImageLookUpSI_S32_S16_3_D1(
const mlib_s32 *src,
mlib_s16 *dst,
mlib_s32 xsize,
const mlib_s16 **table)
{
/* pointer to source data */
mlib_s32 *sp;
/* pointer to start of destination */
mlib_s16 *dl;
/* aligned pointer to destination */
mlib_d64 *dp;
/* destination data */
mlib_d64 t0, t1, t2, t3;
/* destination data */
mlib_d64 acc00, acc01, acc02;
/* destination data */
mlib_d64 acc10, acc11, acc12;
/* loop variable */
mlib_s32 i;
const mlib_s16 *tab0 = (void *)&(((mlib_u8 **)table)[0][HALF_U64]);
const mlib_s16 *tab1 = (void *)&(((mlib_u8 **)table)[1][HALF_U64]);
const mlib_s16 *tab2 = (void *)&(((mlib_u8 **)table)[2][HALF_U64]);
mlib_s32 s00, s01, s02, s03;
sp = (void *)src;
dl = dst;
dp = (mlib_d64 *)dl;
vis_alignaddr((void *)0, 6);
i = 0;
if (xsize >= 4) {
s00 = sp[0];
s01 = sp[1];
s02 = sp[2];
s03 = sp[3];
sp += 4;
vis_write_bmask(0x012389ab, 0);
#pragma pipeloop(0)
for (i = 0; i <= xsize - 8; i += 4, sp += 4) {
t3 = VIS_LD_U16_I(tab0, ((mlib_addr)2 * s01));
t2 = VIS_LD_U16_I(tab2, ((mlib_addr)2 * s00));
t1 = VIS_LD_U16_I(tab1, ((mlib_addr)2 * s00));
t0 = VIS_LD_U16_I(tab0, ((mlib_addr)2 * s00));
acc10 = vis_faligndata(t3, acc10);
acc10 = vis_faligndata(t2, acc10);
acc00 = vis_faligndata(t1, acc00);
acc00 = vis_faligndata(t0, acc00);
t3 = VIS_LD_U16_I(tab1, ((mlib_addr)2 * s02));
t2 = VIS_LD_U16_I(tab0, ((mlib_addr)2 * s02));
t1 = VIS_LD_U16_I(tab2, ((mlib_addr)2 * s01));
t0 = VIS_LD_U16_I(tab1, ((mlib_addr)2 * s01));
acc11 = vis_faligndata(t3, acc11);
acc11 = vis_faligndata(t2, acc11);
acc01 = vis_faligndata(t1, acc01);
acc01 = vis_faligndata(t0, acc01);
t3 = VIS_LD_U16_I(tab2, ((mlib_addr)2 * s03));
t2 = VIS_LD_U16_I(tab1, ((mlib_addr)2 * s03));
t1 = VIS_LD_U16_I(tab0, ((mlib_addr)2 * s03));
t0 = VIS_LD_U16_I(tab2, ((mlib_addr)2 * s02));
acc12 = vis_faligndata(t3, acc12);
acc12 = vis_faligndata(t2, acc12);
acc02 = vis_faligndata(t1, acc02);
acc02 = vis_faligndata(t0, acc02);
s00 = sp[0];
s01 = sp[1];
s02 = sp[2];
s03 = sp[3];
(*dp++) = vis_bshuffle(acc00, acc10);
(*dp++) = vis_bshuffle(acc01, acc11);
(*dp++) = vis_bshuffle(acc02, acc12);
}
t3 = VIS_LD_U16_I(tab0, ((mlib_addr)2 * s01));
t2 = VIS_LD_U16_I(tab2, ((mlib_addr)2 * s00));
t1 = VIS_LD_U16_I(tab1, ((mlib_addr)2 * s00));
t0 = VIS_LD_U16_I(tab0, ((mlib_addr)2 * s00));
acc10 = vis_faligndata(t3, acc10);
acc10 = vis_faligndata(t2, acc10);
acc00 = vis_faligndata(t1, acc00);
acc00 = vis_faligndata(t0, acc00);
t3 = VIS_LD_U16_I(tab1, ((mlib_addr)2 * s02));
t2 = VIS_LD_U16_I(tab0, ((mlib_addr)2 * s02));
t1 = VIS_LD_U16_I(tab2, ((mlib_addr)2 * s01));
t0 = VIS_LD_U16_I(tab1, ((mlib_addr)2 * s01));
acc11 = vis_faligndata(t3, acc11);
acc11 = vis_faligndata(t2, acc11);
acc01 = vis_faligndata(t1, acc01);
acc01 = vis_faligndata(t0, acc01);
t3 = VIS_LD_U16_I(tab2, ((mlib_addr)2 * s03));
t2 = VIS_LD_U16_I(tab1, ((mlib_addr)2 * s03));
t1 = VIS_LD_U16_I(tab0, ((mlib_addr)2 * s03));
t0 = VIS_LD_U16_I(tab2, ((mlib_addr)2 * s02));
acc12 = vis_faligndata(t3, acc12);
acc12 = vis_faligndata(t2, acc12);
acc02 = vis_faligndata(t1, acc02);
acc02 = vis_faligndata(t0, acc02);
(*dp++) = vis_bshuffle(acc00, acc10);
(*dp++) = vis_bshuffle(acc01, acc11);
(*dp++) = vis_bshuffle(acc02, acc12);
i += 4;
}
dl = (mlib_s16 *)dp;
#pragma pipeloop(0)
for (; i < xsize; i++) {
s00 = sp[0];
dl[0] = tab0[s00];
dl[1] = tab1[s00];
dl[2] = tab2[s00];
dl += 3;
sp++;
}
}