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);
}
}
void ADD_SUFF(ByteGrayToIntArgbScaleConvert)(SCALE_PARAMS)
{
mlib_s32 dstScan = pDstInfo->scanStride;
mlib_s32 srcScan = pSrcInfo->scanStride;
mlib_d64 d0, d1, d2, d3, dd;
mlib_f32 ff, aa = vis_fones();
mlib_s32 i, j, x;
if (width < 16) {
for (j = 0; j < height; j++) {
mlib_u8 *src = srcBase;
mlib_s32 *dst = dstBase;
mlib_s32 tmpsxloc = sxloc;
PTR_ADD(src, (syloc >> shift) * srcScan);
for (i = 0; i < width; i++) {
x = src[tmpsxloc >> shift];
tmpsxloc += sxinc;
dst[i] = Gray2Argb(x);
}
PTR_ADD(dstBase, dstScan);
syloc += syinc;
}
return;
}
vis_alignaddr(NULL, 7);
for (j = 0; j < height; j++) {
mlib_u8 *src = srcBase;
mlib_s32 *dst = dstBase;
mlib_s32 *dst_end;
mlib_s32 tmpsxloc = sxloc;
PTR_ADD(src, (syloc >> shift) * srcScan);
dst_end = dst + width;
#pragma pipeloop(0)
for (; dst <= (dst_end - 4); dst += 4) {
LOAD_NEXT_U8(dd, src + ((tmpsxloc + 3*sxinc) >> shift));
LOAD_NEXT_U8(dd, src + ((tmpsxloc + 2*sxinc) >> shift));
LOAD_NEXT_U8(dd, src + ((tmpsxloc + sxinc) >> shift));
LOAD_NEXT_U8(dd, src + ((tmpsxloc ) >> shift));
tmpsxloc += 4*sxinc;
ff = vis_read_hi(dd);
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);
}
while (dst < dst_end) {
x = src[tmpsxloc >> shift];
tmpsxloc += sxinc;
*dst++ = Gray2Argb(x);
}
PTR_ADD(dstBase, dstScan);
syloc += syinc;
}
}
mlib_status
__mlib_VideoH263OverlappedMC_S16_U8(
mlib_s16 mc_block[64],
const mlib_u8 *ref_frame,
mlib_s32 mch,
mlib_s32 mcv,
mlib_s32 mah,
mlib_s32 mav,
mlib_s32 mbh,
mlib_s32 mbv,
mlib_s32 mlh,
mlib_s32 mlv,
mlib_s32 mrh,
mlib_s32 mrv,
mlib_s32 ref_stride)
{
mlib_d64 d0, d1, d2, d3, d4, d5, d6, d7, d8, d9;
mlib_d64 d10, d11, d12, d13, d14, d15;
mlib_d64 tmp1, tmp2, tmp3;
mlib_d64 dmask = vis_fexpand(vis_fones());
mlib_d64 denom = vis_fandnot(dmask, vis_fpadd16(dmask, dmask));
mlib_f32 reg_H0_00, reg_H0_01, reg_H0_10, reg_H0_20, reg_H0_21;
mlib_f32 reg_H1_00, reg_H1_10, reg_H1_11, reg_H1_20, reg_H2_00;
mlib_f32 reg_H2_01, reg_H2_10, reg_H2_11;
mlib_f32 frnd;
mlib_d64 *dp, *sd;
const mlib_u8 *sp1, *sp2, *sp3, *sp4, *sp5;
mlib_s32 ref_stride2 = ref_stride << 1, off;
sp1 = (ref_frame + mch + mcv * ref_stride);
sp2 = (ref_frame + mah + mav * ref_stride);
sp3 = (ref_frame + mlh + mlv * ref_stride);
sp4 = (ref_frame + mrh + 8 + mrv * ref_stride);
sp5 = (ref_frame + mbh + (mbv + 8) * ref_stride);
dp = (mlib_d64 *)mc_block;
reg_H0_00 = vis_to_float(0x40505050);
reg_H0_01 = vis_to_float(0x50505040);
reg_H0_10 = vis_to_float(0x50505050);
reg_H0_20 = vis_to_float(0x50506060);
reg_H0_21 = vis_to_float(0x60605050);
frnd = vis_to_float(0x20202020);
/*
* central
*/
sd = (mlib_d64 *)vis_alignaddr((void *)sp1, 0);
tmp1 = sd[0];
tmp2 = sd[1];
tmp3 = sd[2];
tmp1 = vis_faligndata(tmp1, tmp2);
tmp2 = vis_faligndata(tmp2, tmp3);
ACCSET(d0, tmp1, reg_H0_00);
ACCSET(d1, tmp2, reg_H0_01);
sd = (mlib_d64 *)((mlib_u8 *)sd + ref_stride2);
tmp1 = sd[0];
tmp2 = sd[1];
tmp3 = sd[2];
tmp1 = vis_faligndata(tmp1, tmp2);
tmp2 = vis_faligndata(tmp2, tmp3);
ACCSET(d2, tmp1, reg_H0_10);
ACCSET(d3, tmp2, reg_H0_10);
sd = (mlib_d64 *)((mlib_u8 *)sd + ref_stride2);
tmp1 = sd[0];
tmp2 = sd[1];
tmp3 = sd[2];
tmp1 = vis_faligndata(tmp1, tmp2);
tmp2 = vis_faligndata(tmp2, tmp3);
ACCSET(d4, tmp1, reg_H0_20);
ACCSET(d5, tmp2, reg_H0_21);
sd = (mlib_d64 *)((mlib_u8 *)sd + ref_stride2);
tmp1 = sd[0];
tmp2 = sd[1];
tmp3 = sd[2];
tmp1 = vis_faligndata(tmp1, tmp2);
tmp2 = vis_faligndata(tmp2, tmp3);
ACCSET(d6, tmp1, reg_H0_20);
ACCSET(d7, tmp2, reg_H0_21);
sd = (mlib_d64 *)((mlib_u8 *)sd + ref_stride2);
tmp1 = sd[0];
tmp2 = sd[1];
tmp3 = sd[2];
tmp1 = vis_faligndata(tmp1, tmp2);
tmp2 = vis_faligndata(tmp2, tmp3);
ACCSET(d8, tmp1, reg_H0_20);
ACCSET(d9, tmp2, reg_H0_21);
sd = (mlib_d64 *)((mlib_u8 *)sd + ref_stride2);
tmp1 = sd[0];
tmp2 = sd[1];
tmp3 = sd[2];
tmp1 = vis_faligndata(tmp1, tmp2);
tmp2 = vis_faligndata(tmp2, tmp3);
ACCSET(d10, tmp1, reg_H0_20);
ACCSET(d11, tmp2, reg_H0_21);
sd = (mlib_d64 *)((mlib_u8 *)sd + ref_stride2);
tmp1 = sd[0];
tmp2 = sd[1];
tmp3 = sd[2];
tmp1 = vis_faligndata(tmp1, tmp2);
tmp2 = vis_faligndata(tmp2, tmp3);
ACCSET(d12, tmp1, reg_H0_10);
ACCSET(d13, tmp2, reg_H0_10);
sd = (mlib_d64 *)((mlib_u8 *)sd + ref_stride2);
tmp1 = sd[0];
tmp2 = sd[1];
tmp3 = vis_ld_d64_nf(sd + 2);
tmp1 = vis_faligndata(tmp1, tmp2);
tmp2 = vis_faligndata(tmp2, tmp3);
ACCSET(d14, tmp1, reg_H0_00);
ACCSET(d15, tmp2, reg_H0_01);
/*
* left
*/
reg_H2_00 = vis_to_float(0x20101010);
reg_H2_01 = vis_to_float(0x10101020);
reg_H2_10 = vis_to_float(0x20201010);
reg_H2_11 = vis_to_float(0x10102020);
off = (mlib_addr)sp3 & 7;
sd = (mlib_d64 *)((mlib_u8 *)sp3 - off);
vis_write_bmask(0x11111111 * off + 0x01234567, 0);
tmp1 = sd[0];
tmp2 = sd[1];
tmp1 = vis_bshuffle(tmp1, tmp2);
ACCADD(d0, tmp1, reg_H2_00);
sd = (mlib_d64 *)((mlib_u8 *)sd + ref_stride2);
tmp1 = sd[0];
tmp2 = sd[1];
tmp1 = vis_bshuffle(tmp1, tmp2);
ACCADD(d2, tmp1, reg_H2_10);
sd = (mlib_d64 *)((mlib_u8 *)sd + ref_stride2);
tmp1 = sd[0];
tmp2 = sd[1];
tmp1 = vis_bshuffle(tmp1, tmp2);
ACCADD(d4, tmp1, reg_H2_10);
sd = (mlib_d64 *)((mlib_u8 *)sd + ref_stride2);
tmp1 = sd[0];
tmp2 = sd[1];
tmp1 = vis_bshuffle(tmp1, tmp2);
ACCADD(d6, tmp1, reg_H2_10);
sd = (mlib_d64 *)((mlib_u8 *)sd + ref_stride2);
tmp1 = sd[0];
tmp2 = sd[1];
tmp1 = vis_bshuffle(tmp1, tmp2);
ACCADD(d8, tmp1, reg_H2_10);
sd = (mlib_d64 *)((mlib_u8 *)sd + ref_stride2);
tmp1 = sd[0];
tmp2 = sd[1];
tmp1 = vis_bshuffle(tmp1, tmp2);
ACCADD(d10, tmp1, reg_H2_10);
sd = (mlib_d64 *)((mlib_u8 *)sd + ref_stride2);
tmp1 = sd[0];
tmp2 = sd[1];
tmp1 = vis_bshuffle(tmp1, tmp2);
ACCADD(d12, tmp1, reg_H2_10);
sd = (mlib_d64 *)((mlib_u8 *)sd + ref_stride2);
tmp1 = sd[0];
tmp2 = vis_ld_d64_nf(sd + 1);
tmp1 = vis_bshuffle(tmp1, tmp2);
ACCADD(d14, tmp1, reg_H2_00);
/*
* right
*/
sd = (mlib_d64 *)vis_alignaddr((void *)sp4, 0);
tmp1 = sd[0];
tmp2 = sd[1];
tmp1 = vis_faligndata(tmp1, tmp2);
ACCADD(d1, tmp1, reg_H2_01);
sd = (mlib_d64 *)((mlib_u8 *)sd + ref_stride2);
tmp1 = sd[0];
tmp2 = sd[1];
tmp1 = vis_faligndata(tmp1, tmp2);
ACCADD(d3, tmp1, reg_H2_11);
sd = (mlib_d64 *)((mlib_u8 *)sd + ref_stride2);
tmp1 = sd[0];
tmp2 = sd[1];
tmp1 = vis_faligndata(tmp1, tmp2);
ACCADD(d5, tmp1, reg_H2_11);
sd = (mlib_d64 *)((mlib_u8 *)sd + ref_stride2);
tmp1 = sd[0];
tmp2 = sd[1];
tmp1 = vis_faligndata(tmp1, tmp2);
ACCADD(d7, tmp1, reg_H2_11);
sd = (mlib_d64 *)((mlib_u8 *)sd + ref_stride2);
tmp1 = sd[0];
tmp2 = sd[1];
tmp1 = vis_faligndata(tmp1, tmp2);
ACCADD(d9, tmp1, reg_H2_11);
sd = (mlib_d64 *)((mlib_u8 *)sd + ref_stride2);
tmp1 = sd[0];
tmp2 = sd[1];
tmp1 = vis_faligndata(tmp1, tmp2);
ACCADD(d11, tmp1, reg_H2_11);
sd = (mlib_d64 *)((mlib_u8 *)sd + ref_stride2);
tmp1 = sd[0];
tmp2 = sd[1];
tmp1 = vis_faligndata(tmp1, tmp2);
ACCADD(d13, tmp1, reg_H2_11);
sd = (mlib_d64 *)((mlib_u8 *)sd + ref_stride2);
tmp1 = sd[0];
tmp2 = vis_ld_d64_nf(sd + 1);
tmp1 = vis_faligndata(tmp1, tmp2);
ACCADD(d15, tmp1, reg_H2_01);
/*
* above
*/
reg_H1_10 = vis_to_float(0x10102020);
reg_H1_11 = vis_to_float(0x20201010);
reg_H1_20 = vis_to_float(0x10101010);
off = (mlib_addr)sp2 & 7;
sd = (mlib_d64 *)((mlib_u8 *)sp2 - off);
vis_write_bmask(0x11111111 * off + 0x01234567, 0);
reg_H1_00 = vis_to_float(0x20202020);
tmp1 = sd[0];
tmp2 = sd[1];
tmp3 = sd[2];
tmp1 = vis_bshuffle(tmp1, tmp2);
tmp2 = vis_bshuffle(tmp2, tmp3);
ACCPUT(dp[0], d0, tmp1, reg_H1_00);
ACCPUT(dp[1], d1, tmp2, reg_H1_00);
sd = (mlib_d64 *)((mlib_u8 *)sd + ref_stride2);
tmp1 = sd[0];
tmp2 = sd[1];
tmp3 = sd[2];
tmp1 = vis_bshuffle(tmp1, tmp2);
tmp2 = vis_bshuffle(tmp2, tmp3);
ACCPUT(dp[2], d2, tmp1, reg_H1_10);
ACCPUT(dp[3], d3, tmp2, reg_H1_11);
sd = (mlib_d64 *)((mlib_u8 *)sd + ref_stride2);
tmp1 = sd[0];
tmp2 = sd[1];
tmp3 = sd[2];
tmp1 = vis_bshuffle(tmp1, tmp2);
tmp2 = vis_bshuffle(tmp2, tmp3);
ACCPUT(dp[4], d4, tmp1, reg_H1_20);
ACCPUT(dp[5], d5, tmp2, reg_H1_20);
sd = (mlib_d64 *)((mlib_u8 *)sd + ref_stride2);
tmp1 = sd[0];
tmp2 = sd[1];
tmp3 = vis_ld_d64_nf(sd + 2);
tmp1 = vis_bshuffle(tmp1, tmp2);
tmp2 = vis_bshuffle(tmp2, tmp3);
ACCPUT(dp[6], d6, tmp1, reg_H1_20);
ACCPUT(dp[7], d7, tmp2, reg_H1_20);
/*
* below
*/
sd = (mlib_d64 *)vis_alignaddr((void *)sp5, 0);
tmp1 = sd[0];
tmp2 = sd[1];
tmp3 = sd[2];
tmp1 = vis_faligndata(tmp1, tmp2);
tmp2 = vis_faligndata(tmp2, tmp3);
ACCPUT(dp[8], d8, tmp1, reg_H1_20);
ACCPUT(dp[9], d9, tmp2, reg_H1_20);
sd = (mlib_d64 *)((mlib_u8 *)sd + ref_stride2);
tmp1 = sd[0];
tmp2 = sd[1];
tmp3 = sd[2];
tmp1 = vis_faligndata(tmp1, tmp2);
tmp2 = vis_faligndata(tmp2, tmp3);
ACCPUT(dp[10], d10, tmp1, reg_H1_20);
ACCPUT(dp[11], d11, tmp2, reg_H1_20);
sd = (mlib_d64 *)((mlib_u8 *)sd + ref_stride2);
tmp1 = sd[0];
tmp2 = sd[1];
tmp3 = sd[2];
tmp1 = vis_faligndata(tmp1, tmp2);
tmp2 = vis_faligndata(tmp2, tmp3);
ACCPUT(dp[12], d12, tmp1, reg_H1_10);
ACCPUT(dp[13], d13, tmp2, reg_H1_11);
sd = (mlib_d64 *)((mlib_u8 *)sd + ref_stride2);
tmp1 = sd[0];
tmp2 = sd[1];
tmp3 = vis_ld_d64_nf(sd + 2);
tmp1 = vis_faligndata(tmp1, tmp2);
tmp2 = vis_faligndata(tmp2, tmp3);
ACCPUT(dp[14], d14, tmp1, reg_H1_00);
ACCPUT(dp[15], d15, tmp2, reg_H1_00);
return (MLIB_SUCCESS);
}
