&wrapper_nc<vpx_highbd_lpf_vertical_4_c>, 8, 1),
make_tuple(&wrapper_nc<vpx_highbd_lpf_horizontal_8_sse2>,
&wrapper_nc<vpx_highbd_lpf_horizontal_8_c>, 8, 1),
- make_tuple(&vpx_highbd_lpf_horizontal_16_sse2,
- &vpx_highbd_lpf_horizontal_16_c, 8, 1),
- make_tuple(&vpx_highbd_lpf_horizontal_16_sse2,
- &vpx_highbd_lpf_horizontal_16_c, 8, 2),
+ make_tuple(&wrapper_nc<vpx_highbd_lpf_horizontal_edge_8_sse2>,
+ &wrapper_nc<vpx_highbd_lpf_horizontal_edge_8_c>, 8, 1),
+ make_tuple(&wrapper_nc<vpx_highbd_lpf_horizontal_edge_16_sse2>,
+ &wrapper_nc<vpx_highbd_lpf_horizontal_edge_16_c>, 8, 1),
make_tuple(&wrapper_nc<vpx_highbd_lpf_vertical_8_sse2>,
&wrapper_nc<vpx_highbd_lpf_vertical_8_c>, 8, 1),
make_tuple(&wrapper_nc<vpx_highbd_lpf_vertical_16_sse2>,
&wrapper_nc<vpx_highbd_lpf_vertical_4_c>, 10, 1),
make_tuple(&wrapper_nc<vpx_highbd_lpf_horizontal_8_sse2>,
&wrapper_nc<vpx_highbd_lpf_horizontal_8_c>, 10, 1),
- make_tuple(&vpx_highbd_lpf_horizontal_16_sse2,
- &vpx_highbd_lpf_horizontal_16_c, 10, 1),
- make_tuple(&vpx_highbd_lpf_horizontal_16_sse2,
- &vpx_highbd_lpf_horizontal_16_c, 10, 2),
+ make_tuple(&wrapper_nc<vpx_highbd_lpf_horizontal_edge_8_sse2>,
+ &wrapper_nc<vpx_highbd_lpf_horizontal_edge_8_c>, 10, 1),
+ make_tuple(&wrapper_nc<vpx_highbd_lpf_horizontal_edge_16_sse2>,
+ &wrapper_nc<vpx_highbd_lpf_horizontal_edge_16_c>, 10, 1),
make_tuple(&wrapper_nc<vpx_highbd_lpf_vertical_8_sse2>,
&wrapper_nc<vpx_highbd_lpf_vertical_8_c>, 10, 1),
make_tuple(&wrapper_nc<vpx_highbd_lpf_vertical_16_sse2>,
&wrapper_nc<vpx_highbd_lpf_vertical_4_c>, 12, 1),
make_tuple(&wrapper_nc<vpx_highbd_lpf_horizontal_8_sse2>,
&wrapper_nc<vpx_highbd_lpf_horizontal_8_c>, 12, 1),
- make_tuple(&vpx_highbd_lpf_horizontal_16_sse2,
- &vpx_highbd_lpf_horizontal_16_c, 12, 1),
- make_tuple(&vpx_highbd_lpf_horizontal_16_sse2,
- &vpx_highbd_lpf_horizontal_16_c, 12, 2),
+ make_tuple(&wrapper_nc<vpx_highbd_lpf_horizontal_edge_8_sse2>,
+ &wrapper_nc<vpx_highbd_lpf_horizontal_edge_8_c>, 12, 1),
+ make_tuple(&wrapper_nc<vpx_highbd_lpf_horizontal_edge_16_sse2>,
+ &wrapper_nc<vpx_highbd_lpf_horizontal_edge_16_c>, 12, 1),
make_tuple(&wrapper_nc<vpx_highbd_lpf_vertical_8_sse2>,
&wrapper_nc<vpx_highbd_lpf_vertical_8_c>, 12, 1),
make_tuple(&wrapper_nc<vpx_highbd_lpf_vertical_16_sse2>,
if (mask & 1) {
if (mask_16x16 & 1) {
if ((mask_16x16 & 3) == 3) {
- vpx_highbd_lpf_horizontal_16(s, pitch, lfi->mblim, lfi->lim,
- lfi->hev_thr, 2, bd);
+ vpx_highbd_lpf_horizontal_edge_16(s, pitch, lfi->mblim, lfi->lim,
+ lfi->hev_thr, bd);
count = 2;
} else {
- vpx_highbd_lpf_horizontal_16(s, pitch, lfi->mblim, lfi->lim,
- lfi->hev_thr, 1, bd);
+ vpx_highbd_lpf_horizontal_edge_8(s, pitch, lfi->mblim, lfi->lim,
+ lfi->hev_thr, bd);
}
} else if (mask_8x8 & 1) {
if ((mask_8x8 & 3) == 3) {
if (mask & 1) {
if (mask_16x16 & 1) {
if ((mask_16x16 & 3) == 3) {
- vpx_highbd_lpf_horizontal_16(s, pitch, lfi->mblim, lfi->lim,
- lfi->hev_thr, 2, bd);
+ vpx_highbd_lpf_horizontal_edge_16(s, pitch, lfi->mblim, lfi->lim,
+ lfi->hev_thr, bd);
count = 2;
} else {
- vpx_highbd_lpf_horizontal_16(s, pitch, lfi->mblim, lfi->lim,
- lfi->hev_thr, 1, bd);
+ vpx_highbd_lpf_horizontal_edge_8(s, pitch, lfi->mblim, lfi->lim,
+ lfi->hev_thr, bd);
}
} else if (mask_8x8 & 1) {
if ((mask_8x8 & 3) == 3) {
}
}
-void vpx_highbd_lpf_horizontal_16_c(uint16_t *s, int p, const uint8_t *blimit,
- const uint8_t *limit, const uint8_t *thresh,
- int count, int bd) {
+static void highbd_mb_lpf_horizontal_edge_w(uint16_t *s, int p,
+ const uint8_t *blimit,
+ const uint8_t *limit,
+ const uint8_t *thresh,
+ int count, int bd) {
int i;
// loop filter designed to work using chars so that we can make maximum use
}
}
+void vpx_highbd_lpf_horizontal_edge_8_c(uint16_t *s, int p,
+ const uint8_t *blimit,
+ const uint8_t *limit,
+ const uint8_t *thresh, int bd) {
+ highbd_mb_lpf_horizontal_edge_w(s, p, blimit, limit, thresh, 1, bd);
+}
+
+void vpx_highbd_lpf_horizontal_edge_16_c(uint16_t *s, int p,
+ const uint8_t *blimit,
+ const uint8_t *limit,
+ const uint8_t *thresh, int bd) {
+ highbd_mb_lpf_horizontal_edge_w(s, p, blimit, limit, thresh, 2, bd);
+}
+
static void highbd_mb_lpf_vertical_edge_w(uint16_t *s, int p,
const uint8_t *blimit,
const uint8_t *limit,
add_proto qw/void vpx_highbd_lpf_vertical_4_dual/, "uint16_t *s, int pitch, const uint8_t *blimit0, const uint8_t *limit0, const uint8_t *thresh0, const uint8_t *blimit1, const uint8_t *limit1, const uint8_t *thresh1, int bd";
specialize qw/vpx_highbd_lpf_vertical_4_dual sse2/;
- add_proto qw/void vpx_highbd_lpf_horizontal_16/, "uint16_t *s, int pitch, const uint8_t *blimit, const uint8_t *limit, const uint8_t *thresh, int count, int bd";
- specialize qw/vpx_highbd_lpf_horizontal_16 sse2/;
+ add_proto qw/void vpx_highbd_lpf_horizontal_edge_8/, "uint16_t *s, int pitch, const uint8_t *blimit, const uint8_t *limit, const uint8_t *thresh, int bd";
+ specialize qw/vpx_highbd_lpf_horizontal_edge_8 sse2/;
+
+ add_proto qw/void vpx_highbd_lpf_horizontal_edge_16/, "uint16_t *s, int pitch, const uint8_t *blimit, const uint8_t *limit, const uint8_t *thresh, int bd";
+ specialize qw/vpx_highbd_lpf_horizontal_edge_16 sse2/;
add_proto qw/void vpx_highbd_lpf_horizontal_8/, "uint16_t *s, int pitch, const uint8_t *blimit, const uint8_t *limit, const uint8_t *thresh, int bd";
specialize qw/vpx_highbd_lpf_horizontal_8 sse2/;
// TODO(debargha, peter): Break up large functions into smaller ones
// in this file.
-static void highbd_mb_lpf_horizontal_edge_w_sse2_8(uint16_t *s,
- int p,
- const uint8_t *_blimit,
- const uint8_t *_limit,
- const uint8_t *_thresh,
- int bd) {
+void vpx_highbd_lpf_horizontal_edge_8_sse2(uint16_t *s, int p,
+ const uint8_t *_blimit,
+ const uint8_t *_limit,
+ const uint8_t *_thresh, int bd) {
const __m128i zero = _mm_set1_epi16(0);
const __m128i one = _mm_set1_epi16(1);
__m128i blimit, limit, thresh;
_mm_store_si128((__m128i *)(s - 0 * p), q0);
}
-static void highbd_mb_lpf_horizontal_edge_w_sse2_16(uint16_t *s,
- int p,
- const uint8_t *_blimit,
- const uint8_t *_limit,
- const uint8_t *_thresh,
- int bd) {
- highbd_mb_lpf_horizontal_edge_w_sse2_8(s, p, _blimit, _limit, _thresh, bd);
- highbd_mb_lpf_horizontal_edge_w_sse2_8(s + 8, p, _blimit, _limit, _thresh,
- bd);
-}
-
-// TODO(yunqingwang): remove count and call these 2 functions(8 or 16) directly.
-void vpx_highbd_lpf_horizontal_16_sse2(uint16_t *s, int p,
- const uint8_t *_blimit,
- const uint8_t *_limit,
- const uint8_t *_thresh,
- int count, int bd) {
- if (count == 1)
- highbd_mb_lpf_horizontal_edge_w_sse2_8(s, p, _blimit, _limit, _thresh, bd);
- else
- highbd_mb_lpf_horizontal_edge_w_sse2_16(s, p, _blimit, _limit, _thresh, bd);
+void vpx_highbd_lpf_horizontal_edge_16_sse2(uint16_t *s, int p,
+ const uint8_t *_blimit,
+ const uint8_t *_limit,
+ const uint8_t *_thresh, int bd) {
+ vpx_highbd_lpf_horizontal_edge_8_sse2(s, p, _blimit, _limit, _thresh, bd);
+ vpx_highbd_lpf_horizontal_edge_8_sse2(s + 8, p, _blimit, _limit, _thresh, bd);
}
void vpx_highbd_lpf_horizontal_8_sse2(uint16_t *s, int p,
highbd_transpose(src, p, dst, 8, 2);
// Loop filtering
- highbd_mb_lpf_horizontal_edge_w_sse2_8(t_dst + 8 * 8, 8, blimit, limit,
- thresh, bd);
+ vpx_highbd_lpf_horizontal_edge_8_sse2(t_dst + 8 * 8, 8, blimit, limit,
+ thresh, bd);
src[0] = t_dst;
src[1] = t_dst + 8 * 8;
dst[0] = s - 8;
highbd_transpose8x16(s, s + 8 * p, p, t_dst + 8 * 16, 16);
// Loop filtering
- highbd_mb_lpf_horizontal_edge_w_sse2_16(t_dst + 8 * 16, 16, blimit, limit,
- thresh, bd);
+ vpx_highbd_lpf_horizontal_edge_16_sse2(t_dst + 8 * 16, 16, blimit, limit,
+ thresh, bd);
// Transpose back
highbd_transpose8x16(t_dst, t_dst + 8 * 16, 16, s - 8, p);