int VP8GetCostLuma16(VP8EncIterator* const it, const VP8ModeScore* const rd) {
VP8Residual res;
VP8Encoder* const enc = it->enc_;
int x, y;
int R = 0;
VP8IteratorNzToBytes(it); // re-import the non-zero context
// DC
VP8InitResidual(0, 1, enc, &res);
VP8SetResidualCoeffs(rd->y_dc_levels, &res);
R += VP8GetResidualCost(it->top_nz_[8] + it->left_nz_[8], &res);
// AC
VP8InitResidual(1, 0, enc, &res);
for (y = 0; y < 4; ++y) {
for (x = 0; x < 4; ++x) {
const int ctx = it->top_nz_[x] + it->left_nz_[y];
VP8SetResidualCoeffs(rd->y_ac_levels[x + y * 4], &res);
R += VP8GetResidualCost(ctx, &res);
it->top_nz_[x] = it->left_nz_[y] = (res.last >= 0);
}
}
return R;
}
static void CodeResiduals(VP8BitWriter* const bw, VP8EncIterator* const it,
const VP8ModeScore* const rd) {
int x, y, ch;
VP8Residual res;
uint64_t pos1, pos2, pos3;
const int i16 = (it->mb_->type_ == 1);
const int segment = it->mb_->segment_;
VP8Encoder* const enc = it->enc_;
VP8IteratorNzToBytes(it);
pos1 = VP8BitWriterPos(bw);
if (i16) {
VP8InitResidual(0, 1, enc, &res);
VP8SetResidualCoeffs(rd->y_dc_levels, &res);
it->top_nz_[8] = it->left_nz_[8] =
PutCoeffs(bw, it->top_nz_[8] + it->left_nz_[8], &res);
VP8InitResidual(1, 0, enc, &res);
} else {
VP8InitResidual(0, 3, enc, &res);
}
// luma-AC
for (y = 0; y < 4; ++y) {
for (x = 0; x < 4; ++x) {
const int ctx = it->top_nz_[x] + it->left_nz_[y];
VP8SetResidualCoeffs(rd->y_ac_levels[x + y * 4], &res);
it->top_nz_[x] = it->left_nz_[y] = PutCoeffs(bw, ctx, &res);
}
}
pos2 = VP8BitWriterPos(bw);
// U/V
VP8InitResidual(0, 2, enc, &res);
for (ch = 0; ch <= 2; ch += 2) {
for (y = 0; y < 2; ++y) {
for (x = 0; x < 2; ++x) {
const int ctx = it->top_nz_[4 + ch + x] + it->left_nz_[4 + ch + y];
VP8SetResidualCoeffs(rd->uv_levels[ch * 2 + x + y * 2], &res);
it->top_nz_[4 + ch + x] = it->left_nz_[4 + ch + y] =
PutCoeffs(bw, ctx, &res);
}
}
}
pos3 = VP8BitWriterPos(bw);
it->luma_bits_ = pos2 - pos1;
it->uv_bits_ = pos3 - pos2;
it->bit_count_[segment][i16] += it->luma_bits_;
it->bit_count_[segment][2] += it->uv_bits_;
VP8IteratorBytesToNz(it);
}
static int RecordTokens(VP8EncIterator* const it, const VP8ModeScore* const rd,
VP8TBuffer* const tokens) {
int x, y, ch;
VP8Residual res;
VP8Encoder* const enc = it->enc_;
VP8IteratorNzToBytes(it);
if (it->mb_->type_ == 1) { // i16x16
const int ctx = it->top_nz_[8] + it->left_nz_[8];
VP8InitResidual(0, 1, enc, &res);
VP8SetResidualCoeffs(rd->y_dc_levels, &res);
it->top_nz_[8] = it->left_nz_[8] =
VP8RecordCoeffTokens(ctx, 1,
res.first, res.last, res.coeffs, tokens);
VP8RecordCoeffs(ctx, &res);
VP8InitResidual(1, 0, enc, &res);
} else {
VP8InitResidual(0, 3, enc, &res);
}
// luma-AC
for (y = 0; y < 4; ++y) {
for (x = 0; x < 4; ++x) {
const int ctx = it->top_nz_[x] + it->left_nz_[y];
VP8SetResidualCoeffs(rd->y_ac_levels[x + y * 4], &res);
it->top_nz_[x] = it->left_nz_[y] =
VP8RecordCoeffTokens(ctx, res.coeff_type,
res.first, res.last, res.coeffs, tokens);
VP8RecordCoeffs(ctx, &res);
}
}
// U/V
VP8InitResidual(0, 2, enc, &res);
for (ch = 0; ch <= 2; ch += 2) {
for (y = 0; y < 2; ++y) {
for (x = 0; x < 2; ++x) {
const int ctx = it->top_nz_[4 + ch + x] + it->left_nz_[4 + ch + y];
VP8SetResidualCoeffs(rd->uv_levels[ch * 2 + x + y * 2], &res);
it->top_nz_[4 + ch + x] = it->left_nz_[4 + ch + y] =
VP8RecordCoeffTokens(ctx, 2,
res.first, res.last, res.coeffs, tokens);
VP8RecordCoeffs(ctx, &res);
}
}
}
VP8IteratorBytesToNz(it);
return !tokens->error_;
}
// Same as CodeResiduals, but doesn't actually write anything.
// Instead, it just records the event distribution.
static void RecordResiduals(VP8EncIterator* const it,
const VP8ModeScore* const rd) {
int x, y, ch;
VP8Residual res;
VP8Encoder* const enc = it->enc_;
VP8IteratorNzToBytes(it);
if (it->mb_->type_ == 1) { // i16x16
VP8InitResidual(0, 1, enc, &res);
VP8SetResidualCoeffs(rd->y_dc_levels, &res);
it->top_nz_[8] = it->left_nz_[8] =
VP8RecordCoeffs(it->top_nz_[8] + it->left_nz_[8], &res);
VP8InitResidual(1, 0, enc, &res);
} else {
VP8InitResidual(0, 3, enc, &res);
}
// luma-AC
for (y = 0; y < 4; ++y) {
for (x = 0; x < 4; ++x) {
const int ctx = it->top_nz_[x] + it->left_nz_[y];
VP8SetResidualCoeffs(rd->y_ac_levels[x + y * 4], &res);
it->top_nz_[x] = it->left_nz_[y] = VP8RecordCoeffs(ctx, &res);
}
}
// U/V
VP8InitResidual(0, 2, enc, &res);
for (ch = 0; ch <= 2; ch += 2) {
for (y = 0; y < 2; ++y) {
for (x = 0; x < 2; ++x) {
const int ctx = it->top_nz_[4 + ch + x] + it->left_nz_[4 + ch + y];
VP8SetResidualCoeffs(rd->uv_levels[ch * 2 + x + y * 2], &res);
it->top_nz_[4 + ch + x] = it->left_nz_[4 + ch + y] =
VP8RecordCoeffs(ctx, &res);
}
}
}
VP8IteratorBytesToNz(it);
}
int VP8GetCostLuma4(VP8EncIterator* const it, const int16_t levels[16]) {
const int x = (it->i4_ & 3), y = (it->i4_ >> 2);
VP8Residual res;
VP8Encoder* const enc = it->enc_;
int R = 0;
int ctx;
VP8InitResidual(0, 3, enc, &res);
ctx = it->top_nz_[x] + it->left_nz_[y];
VP8SetResidualCoeffs(levels, &res);
R += VP8GetResidualCost(ctx, &res);
return R;
}
int VP8GetCostUV(VP8EncIterator* const it, const VP8ModeScore* const rd) {
VP8Residual res;
VP8Encoder* const enc = it->enc_;
int ch, x, y;
int R = 0;
VP8IteratorNzToBytes(it); // re-import the non-zero context
VP8InitResidual(0, 2, enc, &res);
for (ch = 0; ch <= 2; ch += 2) {
for (y = 0; y < 2; ++y) {
for (x = 0; x < 2; ++x) {
const int ctx = it->top_nz_[4 + ch + x] + it->left_nz_[4 + ch + y];
VP8SetResidualCoeffs(rd->uv_levels[ch * 2 + x + y * 2], &res);
R += VP8GetResidualCost(ctx, &res);
it->top_nz_[4 + ch + x] = it->left_nz_[4 + ch + y] = (res.last >= 0);
}
}
}
return R;
}