Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Change rounding method #28

Merged
merged 1 commit into from
Oct 5, 2023
Merged

Change rounding method #28

Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
109 changes: 47 additions & 62 deletions lib/color.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -87,28 +87,13 @@ HWY_ATTR void rgb2ycbcr(uint8_t *HWY_RESTRICT in, std::vector<uint8_t *> &out, i
*/
HWY_ATTR void subsample_core(std::vector<uint8_t *> &in, std::vector<int16_t *> &out, int width,
int YCCtype) {
// int nc = (YCCtype == YCC::GRAY) ? 1 : 3;
int scale_x = YCC_HV[YCCtype][0] >> 4;
int scale_y = YCC_HV[YCCtype][0] & 0xF;

int shift = 0;
int bound = 1;
while ((scale_x * scale_y) > bound) {
bound += bound;
shift++;
}
int half = 0;
if (shift) {
half = 1 << (shift - 1);
}
size_t pos = 0;
size_t pos_Chroma = 0;
// HWY_CAPPED(uint8_t, 16) u8;
// HWY_CAPPED(int16_t, 8) s16;

hn::FixedTag<uint8_t, 16> u8;
hn::FixedTag<int16_t, 8> s16;
const auto c128 = Set(s16, 128);
const auto vhalf = Set(s16, half);
const auto c128 = Set(s16, 128);

switch (YCCtype) {
case YCC::GRAY:
for (int i = 0; i < LINES; i += DCTSIZE) {
Expand Down Expand Up @@ -292,14 +277,14 @@ HWY_ATTR void subsample_core(std::vector<uint8_t *> &in, std::vector<int16_t *>
auto cb71 = Sub(PromoteUpperTo(s16, v7_1), c128);

// clang-format off
Store(hn::ShiftRight<1>(Add(Padd(s16, cb00, cb01), vhalf)), s16, out[1] + pos_Chroma + 8 * 0);
Store(hn::ShiftRight<1>(Add(Padd(s16, cb10, cb11), vhalf)), s16, out[1] + pos_Chroma + 8 * 1);
Store(hn::ShiftRight<1>(Add(Padd(s16, cb20, cb21), vhalf)), s16, out[1] + pos_Chroma + 8 * 2);
Store(hn::ShiftRight<1>(Add(Padd(s16, cb30, cb31), vhalf)), s16, out[1] + pos_Chroma + 8 * 3);
Store(hn::ShiftRight<1>(Add(Padd(s16, cb40, cb41), vhalf)), s16, out[1] + pos_Chroma + 8 * 4);
Store(hn::ShiftRight<1>(Add(Padd(s16, cb50, cb51), vhalf)), s16, out[1] + pos_Chroma + 8 * 5);
Store(hn::ShiftRight<1>(Add(Padd(s16, cb60, cb61), vhalf)), s16, out[1] + pos_Chroma + 8 * 6);
Store(hn::ShiftRight<1>(Add(Padd(s16, cb70, cb71), vhalf)), s16, out[1] + pos_Chroma + 8 * 7);
Store(hn::ShiftRight<1>(Padd(s16, cb00, cb01)), s16, out[1] + pos_Chroma + 8 * 0);
Store(hn::ShiftRight<1>(Padd(s16, cb10, cb11)), s16, out[1] + pos_Chroma + 8 * 1);
Store(hn::ShiftRight<1>(Padd(s16, cb20, cb21)), s16, out[1] + pos_Chroma + 8 * 2);
Store(hn::ShiftRight<1>(Padd(s16, cb30, cb31)), s16, out[1] + pos_Chroma + 8 * 3);
Store(hn::ShiftRight<1>(Padd(s16, cb40, cb41)), s16, out[1] + pos_Chroma + 8 * 4);
Store(hn::ShiftRight<1>(Padd(s16, cb50, cb51)), s16, out[1] + pos_Chroma + 8 * 5);
Store(hn::ShiftRight<1>(Padd(s16, cb60, cb61)), s16, out[1] + pos_Chroma + 8 * 6);
Store(hn::ShiftRight<1>(Padd(s16, cb70, cb71)), s16, out[1] + pos_Chroma + 8 * 7);
// clang-format on

auto v0_2 = Load(u8, sp2 + 0 * width);
Expand Down Expand Up @@ -329,14 +314,14 @@ HWY_ATTR void subsample_core(std::vector<uint8_t *> &in, std::vector<int16_t *>
cb71 = Sub(PromoteUpperTo(s16, v7_2), c128);

// clang-format off
Store(hn::ShiftRight<1>(Add(Padd(s16, cb00, cb01), vhalf)), s16, out[2] + pos_Chroma + 8 * 0);
Store(hn::ShiftRight<1>(Add(Padd(s16, cb10, cb11), vhalf)), s16, out[2] + pos_Chroma + 8 * 1);
Store(hn::ShiftRight<1>(Add(Padd(s16, cb20, cb21), vhalf)), s16, out[2] + pos_Chroma + 8 * 2);
Store(hn::ShiftRight<1>(Add(Padd(s16, cb30, cb31), vhalf)), s16, out[2] + pos_Chroma + 8 * 3);
Store(hn::ShiftRight<1>(Add(Padd(s16, cb40, cb41), vhalf)), s16, out[2] + pos_Chroma + 8 * 4);
Store(hn::ShiftRight<1>(Add(Padd(s16, cb50, cb51), vhalf)), s16, out[2] + pos_Chroma + 8 * 5);
Store(hn::ShiftRight<1>(Add(Padd(s16, cb60, cb61), vhalf)), s16, out[2] + pos_Chroma + 8 * 6);
Store(hn::ShiftRight<1>(Add(Padd(s16, cb70, cb71), vhalf)), s16, out[2] + pos_Chroma + 8 * 7);
Store(hn::ShiftRight<1>(Padd(s16, cb00, cb01)), s16, out[2] + pos_Chroma + 8 * 0);
Store(hn::ShiftRight<1>(Padd(s16, cb10, cb11)), s16, out[2] + pos_Chroma + 8 * 1);
Store(hn::ShiftRight<1>(Padd(s16, cb20, cb21)), s16, out[2] + pos_Chroma + 8 * 2);
Store(hn::ShiftRight<1>(Padd(s16, cb30, cb31)), s16, out[2] + pos_Chroma + 8 * 3);
Store(hn::ShiftRight<1>(Padd(s16, cb40, cb41)), s16, out[2] + pos_Chroma + 8 * 4);
Store(hn::ShiftRight<1>(Padd(s16, cb50, cb51)), s16, out[2] + pos_Chroma + 8 * 5);
Store(hn::ShiftRight<1>(Padd(s16, cb60, cb61)), s16, out[2] + pos_Chroma + 8 * 6);
Store(hn::ShiftRight<1>(Padd(s16, cb70, cb71)), s16, out[2] + pos_Chroma + 8 * 7);
// clang-format on
pos += 128;
pos_Chroma += 64;
Expand Down Expand Up @@ -404,14 +389,14 @@ HWY_ATTR void subsample_core(std::vector<uint8_t *> &in, std::vector<int16_t *>
auto cb70 = Sub(PromoteLowerTo(s16, v7_1), c128);
auto cb71 = Sub(PromoteUpperTo(s16, v7_1), c128);

Store(hn::ShiftRight<1>(Add(Add(cb00, cb10), vhalf)), s16, out[1] + pos_Chroma + 8 * 0);
Store(hn::ShiftRight<1>(Add(Add(cb20, cb30), vhalf)), s16, out[1] + pos_Chroma + 8 * 1);
Store(hn::ShiftRight<1>(Add(Add(cb40, cb50), vhalf)), s16, out[1] + pos_Chroma + 8 * 2);
Store(hn::ShiftRight<1>(Add(Add(cb60, cb70), vhalf)), s16, out[1] + pos_Chroma + 8 * 3);
Store(hn::ShiftRight<1>(Add(Add(cb01, cb11), vhalf)), s16, out[1] + pos_Chroma + 8 * 8);
Store(hn::ShiftRight<1>(Add(Add(cb21, cb31), vhalf)), s16, out[1] + pos_Chroma + 8 * 9);
Store(hn::ShiftRight<1>(Add(Add(cb41, cb51), vhalf)), s16, out[1] + pos_Chroma + 8 * 10);
Store(hn::ShiftRight<1>(Add(Add(cb61, cb71), vhalf)), s16, out[1] + pos_Chroma + 8 * 11);
Store(hn::ShiftRight<1>(Add(cb00, cb10)), s16, out[1] + pos_Chroma + 8 * 0);
Store(hn::ShiftRight<1>(Add(cb20, cb30)), s16, out[1] + pos_Chroma + 8 * 1);
Store(hn::ShiftRight<1>(Add(cb40, cb50)), s16, out[1] + pos_Chroma + 8 * 2);
Store(hn::ShiftRight<1>(Add(cb60, cb70)), s16, out[1] + pos_Chroma + 8 * 3);
Store(hn::ShiftRight<1>(Add(cb01, cb11)), s16, out[1] + pos_Chroma + 8 * 8);
Store(hn::ShiftRight<1>(Add(cb21, cb31)), s16, out[1] + pos_Chroma + 8 * 9);
Store(hn::ShiftRight<1>(Add(cb41, cb51)), s16, out[1] + pos_Chroma + 8 * 10);
Store(hn::ShiftRight<1>(Add(cb61, cb71)), s16, out[1] + pos_Chroma + 8 * 11);

auto v0_2 = Load(u8, sp2 + 0 * width);
auto v1_2 = Load(u8, sp2 + 1 * width);
Expand Down Expand Up @@ -439,14 +424,14 @@ HWY_ATTR void subsample_core(std::vector<uint8_t *> &in, std::vector<int16_t *>
cb70 = Sub(PromoteLowerTo(s16, v7_2), c128);
cb71 = Sub(PromoteUpperTo(s16, v7_2), c128);

Store(hn::ShiftRight<1>(Add(Add(cb00, cb10), vhalf)), s16, out[2] + pos_Chroma + 8 * 0);
Store(hn::ShiftRight<1>(Add(Add(cb20, cb30), vhalf)), s16, out[2] + pos_Chroma + 8 * 1);
Store(hn::ShiftRight<1>(Add(Add(cb40, cb50), vhalf)), s16, out[2] + pos_Chroma + 8 * 2);
Store(hn::ShiftRight<1>(Add(Add(cb60, cb70), vhalf)), s16, out[2] + pos_Chroma + 8 * 3);
Store(hn::ShiftRight<1>(Add(Add(cb01, cb11), vhalf)), s16, out[2] + pos_Chroma + 8 * 8);
Store(hn::ShiftRight<1>(Add(Add(cb21, cb31), vhalf)), s16, out[2] + pos_Chroma + 8 * 9);
Store(hn::ShiftRight<1>(Add(Add(cb41, cb51), vhalf)), s16, out[2] + pos_Chroma + 8 * 10);
Store(hn::ShiftRight<1>(Add(Add(cb61, cb71), vhalf)), s16, out[2] + pos_Chroma + 8 * 11);
Store(hn::ShiftRight<1>(Add(cb00, cb10)), s16, out[2] + pos_Chroma + 8 * 0);
Store(hn::ShiftRight<1>(Add(cb20, cb30)), s16, out[2] + pos_Chroma + 8 * 1);
Store(hn::ShiftRight<1>(Add(cb40, cb50)), s16, out[2] + pos_Chroma + 8 * 2);
Store(hn::ShiftRight<1>(Add(cb60, cb70)), s16, out[2] + pos_Chroma + 8 * 3);
Store(hn::ShiftRight<1>(Add(cb01, cb11)), s16, out[2] + pos_Chroma + 8 * 8);
Store(hn::ShiftRight<1>(Add(cb21, cb31)), s16, out[2] + pos_Chroma + 8 * 9);
Store(hn::ShiftRight<1>(Add(cb41, cb51)), s16, out[2] + pos_Chroma + 8 * 10);
Store(hn::ShiftRight<1>(Add(cb61, cb71)), s16, out[2] + pos_Chroma + 8 * 11);
}
}
break;
Expand Down Expand Up @@ -511,10 +496,10 @@ HWY_ATTR void subsample_core(std::vector<uint8_t *> &in, std::vector<int16_t *>
auto cb71 = Sub(PromoteUpperTo(s16, v7_1), c128);

// clang-format off
Store(hn::ShiftRight<2>(Add(Padd(s16, Add(cb00, cb10), Add(cb01, cb11)), vhalf)), s16, out[1] + pos_Chroma + 8 * 0);
Store(hn::ShiftRight<2>(Add(Padd(s16, Add(cb20, cb30), Add(cb21, cb31)), vhalf)), s16, out[1] + pos_Chroma + 8 * 1);
Store(hn::ShiftRight<2>(Add(Padd(s16, Add(cb40, cb50), Add(cb41, cb51)), vhalf)), s16, out[1] + pos_Chroma + 8 * 2);
Store(hn::ShiftRight<2>(Add(Padd(s16, Add(cb60, cb70), Add(cb61, cb71)), vhalf)), s16, out[1] + pos_Chroma + 8 * 3);
Store(hn::ShiftRight<2>(Padd(s16, Add(cb00, cb10), Add(cb01, cb11))), s16, out[1] + pos_Chroma + 8 * 0);
Store(hn::ShiftRight<2>(Padd(s16, Add(cb20, cb30), Add(cb21, cb31))), s16, out[1] + pos_Chroma + 8 * 1);
Store(hn::ShiftRight<2>(Padd(s16, Add(cb40, cb50), Add(cb41, cb51))), s16, out[1] + pos_Chroma + 8 * 2);
Store(hn::ShiftRight<2>(Padd(s16, Add(cb60, cb70), Add(cb61, cb71))), s16, out[1] + pos_Chroma + 8 * 3);
// clang-format on

auto v0_2 = Load(u8, sp2 + 0 * width);
Expand Down Expand Up @@ -544,10 +529,10 @@ HWY_ATTR void subsample_core(std::vector<uint8_t *> &in, std::vector<int16_t *>
cb71 = Sub(PromoteUpperTo(s16, v7_2), c128);

// clang-format off
Store(hn::ShiftRight<2>(Add(Padd(s16, Add(cb00, cb10), Add(cb01, cb11)), vhalf)), s16, out[2] + pos_Chroma + 8 * 0);
Store(hn::ShiftRight<2>(Add(Padd(s16, Add(cb20, cb30), Add(cb21, cb31)), vhalf)), s16, out[2] + pos_Chroma + 8 * 1);
Store(hn::ShiftRight<2>(Add(Padd(s16, Add(cb40, cb50), Add(cb41, cb51)), vhalf)), s16, out[2] + pos_Chroma + 8 * 2);
Store(hn::ShiftRight<2>(Add(Padd(s16, Add(cb60, cb70), Add(cb61, cb71)), vhalf)), s16, out[2] + pos_Chroma + 8 * 3);
Store(hn::ShiftRight<2>(Padd(s16, Add(cb00, cb10), Add(cb01, cb11))), s16, out[2] + pos_Chroma + 8 * 0);
Store(hn::ShiftRight<2>(Padd(s16, Add(cb20, cb30), Add(cb21, cb31))), s16, out[2] + pos_Chroma + 8 * 1);
Store(hn::ShiftRight<2>(Padd(s16, Add(cb40, cb50), Add(cb41, cb51))), s16, out[2] + pos_Chroma + 8 * 2);
Store(hn::ShiftRight<2>(Padd(s16, Add(cb60, cb70), Add(cb61, cb71))), s16, out[2] + pos_Chroma + 8 * 3);
// clang-format on

pos += 128;
Expand Down Expand Up @@ -585,7 +570,7 @@ HWY_ATTR void subsample_core(std::vector<uint8_t *> &in, std::vector<int16_t *>
auto t0 = Padd(s16, cb00, cb01);
auto t1 = Padd(s16, cb10, cb11);
auto tb00 = Padd(s16, t0, t1);
Store(hn::ShiftRight<2>(Add(Padd(s16, t0, t1), vhalf)), s16, out[1] + pos_Chroma + p);
Store(hn::ShiftRight<2>(Padd(s16, t0, t1)), s16, out[1] + pos_Chroma + p);

// Cr
cb00 = Sub(PromoteLowerTo(s16, v0_2), c128);
Expand All @@ -596,7 +581,7 @@ HWY_ATTR void subsample_core(std::vector<uint8_t *> &in, std::vector<int16_t *>
t0 = Padd(s16, cb00, cb01);
t1 = Padd(s16, cb10, cb11);
tb00 = Padd(s16, t0, t1);
Store(hn::ShiftRight<2>(Add(Padd(s16, t0, t1), vhalf)), s16, out[2] + pos_Chroma + p);
Store(hn::ShiftRight<2>(Padd(s16, t0, t1)), s16, out[2] + pos_Chroma + p);
p += 8;
}
pos += 256;
Expand Down Expand Up @@ -649,8 +634,8 @@ HWY_ATTR void subsample_core(std::vector<uint8_t *> &in, std::vector<int16_t *>
cb = Padd(s16, tb0, tb1);
cr = Padd(s16, tr0, tr1);
} else {
cb = hn::ShiftRight<3>(Add(Add(cb, Padd(s16, tb0, tb1)), vhalf));
cr = hn::ShiftRight<3>(Add(Add(cr, Padd(s16, tr0, tr1)), vhalf));
cb = hn::ShiftRight<3>(Add(cb, Padd(s16, tb0, tb1)));
cr = hn::ShiftRight<3>(Add(cr, Padd(s16, tr0, tr1)));
Store(cb, s16, out[1] + pos_Chroma + pc);
Store(cr, s16, out[2] + pos_Chroma + pc);
pc += 8;
Expand Down