Commit 356c9447 authored by Johann Koenig's avatar Johann Koenig Committed by Gerrit Code Review
Browse files

Merge changes I2e0cc72a,I63a7da78,I4aee2600,I71113505,I3dce6c70, ...

* changes:
  Only build multithreaded functions on mt builds.
  Don't build calc_psnr for high bit depth.
  Enable missing dual lpf test
  Remove unused VP10 functions.
  Mark VP10 functions as 'INLINE'
  Remove unused functions from test files
  Only build append_negative_gtest_filter when it is used.
  Add INLINE decoration to static test functions
parents 9a5e0d40 c7f58bcc
......@@ -40,30 +40,6 @@ static int round(double x) {
#endif
const int kNumCoeffs = 256;
const double PI = 3.1415926535898;
void reference2_16x16_idct_2d(double *input, double *output) {
double x;
for (int l = 0; l < 16; ++l) {
for (int k = 0; k < 16; ++k) {
double s = 0;
for (int i = 0; i < 16; ++i) {
for (int j = 0; j < 16; ++j) {
x = cos(PI * j * (l + 0.5) / 16.0) *
cos(PI * i * (k + 0.5) / 16.0) *
input[i * 16 + j] / 256;
if (i != 0)
x *= sqrt(2.0);
if (j != 0)
x *= sqrt(2.0);
s += x;
}
}
output[k*16+l] = s;
}
}
}
const double C1 = 0.995184726672197;
const double C2 = 0.98078528040323;
const double C3 = 0.956940335732209;
......
......@@ -67,43 +67,6 @@ void reference_dct_2d(int16_t input[64], double output[64]) {
output[i] *= 2;
}
void reference_idct_1d(double input[8], double output[8]) {
const double kPi = 3.141592653589793238462643383279502884;
const double kSqrt2 = 1.414213562373095048801688724209698;
for (int k = 0; k < 8; k++) {
output[k] = 0.0;
for (int n = 0; n < 8; n++) {
output[k] += input[n]*cos(kPi*(2*k+1)*n/16.0);
if (n == 0)
output[k] = output[k]/kSqrt2;
}
}
}
void reference_idct_2d(double input[64], int16_t output[64]) {
double out[64], out2[64];
// First transform rows
for (int i = 0; i < 8; ++i) {
double temp_in[8], temp_out[8];
for (int j = 0; j < 8; ++j)
temp_in[j] = input[j + i*8];
reference_idct_1d(temp_in, temp_out);
for (int j = 0; j < 8; ++j)
out[j + i*8] = temp_out[j];
}
// Then transform columns
for (int i = 0; i < 8; ++i) {
double temp_in[8], temp_out[8];
for (int j = 0; j < 8; ++j)
temp_in[j] = out[j*8 + i];
reference_idct_1d(temp_in, temp_out);
for (int j = 0; j < 8; ++j)
out2[j*8 + i] = temp_out[j];
}
for (int i = 0; i < 64; ++i)
output[i] = round(out2[i]/32);
}
TEST(VP9Idct8x8Test, AccuracyCheck) {
ACMRandom rnd(ACMRandom::DeterministicSeed());
const int count_test_block = 10000;
......
......@@ -590,7 +590,9 @@ INSTANTIATE_TEST_CASE_P(
make_tuple(&vpx_lpf_horizontal_16_sse2, &vpx_lpf_horizontal_16_c, 8, 1),
make_tuple(&vpx_lpf_horizontal_16_sse2, &vpx_lpf_horizontal_16_c, 8, 2),
make_tuple(&vpx_lpf_vertical_8_sse2, &vpx_lpf_vertical_8_c, 8, 1),
make_tuple(&wrapper_vertical_16_sse2, &wrapper_vertical_16_c, 8, 1)));
make_tuple(&wrapper_vertical_16_sse2, &wrapper_vertical_16_c, 8, 1),
make_tuple(&wrapper_vertical_16_dual_sse2,
&wrapper_vertical_16_dual_c, 8, 1)));
#endif // CONFIG_VP9_HIGHBITDEPTH
#endif
......
......@@ -26,6 +26,7 @@ extern void vpx_dsp_rtcd();
extern void vpx_scale_rtcd();
}
#if ARCH_X86 || ARCH_X86_64
static void append_negative_gtest_filter(const char *str) {
std::string filter = ::testing::FLAGS_gtest_filter;
// Negative patterns begin with one '-' followed by a ':' separated list.
......@@ -33,6 +34,7 @@ static void append_negative_gtest_filter(const char *str) {
filter += str;
::testing::FLAGS_gtest_filter = filter;
}
#endif // ARCH_X86 || ARCH_X86_64
int main(int argc, char **argv) {
::testing::InitGoogleTest(&argc, argv);
......@@ -55,7 +57,7 @@ int main(int argc, char **argv) {
append_negative_gtest_filter(":AVX.*:AVX/*");
if (!(simd_caps & HAS_AVX2))
append_negative_gtest_filter(":AVX2.*:AVX2/*");
#endif
#endif // ARCH_X86 || ARCH_X86_64
#if !CONFIG_SHARED
// Shared library builds don't support whitebox tests
......
......@@ -19,8 +19,7 @@
// Macros
#define GET_PARAM(k) std::tr1::get< k >(GetParam())
static double compute_psnr(const vpx_image_t *img1,
const vpx_image_t *img2) {
static INLINE double compute_psnr(const vpx_image_t *img1, const vpx_image_t *img2) {
assert((img1->fmt == img2->fmt) &&
(img1->d_w == img2->d_w) &&
(img1->d_h == img2->d_h));
......
......@@ -48,7 +48,7 @@ static std::string GetDataPath() {
#undef TO_STRING
#undef STRINGIFY
static FILE *OpenTestDataFile(const std::string& file_name) {
static INLINE FILE *OpenTestDataFile(const std::string& file_name) {
const std::string path_to_source = GetDataPath() + "/" + file_name;
return fopen(path_to_source.c_str(), "rb");
}
......
......@@ -34,14 +34,14 @@ static INLINE void inter_predictor(const uint8_t *src, int src_stride,
}
#if CONFIG_VP9_HIGHBITDEPTH
static void high_inter_predictor(const uint8_t *src, int src_stride,
uint8_t *dst, int dst_stride,
const int subpel_x,
const int subpel_y,
const struct scale_factors *sf,
int w, int h, int ref,
const InterpKernel *kernel,
int xs, int ys, int bd) {
static INLINE void high_inter_predictor(const uint8_t *src, int src_stride,
uint8_t *dst, int dst_stride,
const int subpel_x,
const int subpel_y,
const struct scale_factors *sf,
int w, int h, int ref,
const InterpKernel *kernel,
int xs, int ys, int bd) {
sf->highbd_predict[subpel_x != 0][subpel_y != 0][ref](
src, src_stride, dst, dst_stride,
kernel[subpel_x], xs, kernel[subpel_y], ys, w, h, bd);
......@@ -77,8 +77,9 @@ static MV mi_mv_pred_q2(const MODE_INFO *mi, int idx, int block0, int block1) {
}
// TODO(jkoleszar): yet another mv clamping function :-(
static MV clamp_mv_to_umv_border_sb(const MACROBLOCKD *xd, const MV *src_mv,
int bw, int bh, int ss_x, int ss_y) {
static INLINE MV clamp_mv_to_umv_border_sb(const MACROBLOCKD *xd,
const MV *src_mv,
int bw, int bh, int ss_x, int ss_y) {
// If the MV points so far into the UMV border that no visible pixels
// are used for reconstruction, the subpel part of the MV can be
// discarded and the MV limited to 16 pixels with equivalent results.
......@@ -102,8 +103,8 @@ static MV clamp_mv_to_umv_border_sb(const MACROBLOCKD *xd, const MV *src_mv,
return clamped_mv;
}
static MV average_split_mvs(const struct macroblockd_plane *pd,
const MODE_INFO *mi, int ref, int block) {
static INLINE MV average_split_mvs(const struct macroblockd_plane *pd,
const MODE_INFO *mi, int ref, int block) {
const int ss_idx = ((pd->subsampling_x > 0) << 1) | (pd->subsampling_y > 0);
MV res = {0, 0};
switch (ss_idx) {
......
......@@ -238,19 +238,6 @@ static void set_offsets(VP10_COMP *cpi, const TileInfo *const tile,
xd->tile = *tile;
}
static void duplicate_mode_info_in_sb(VP10_COMMON *cm, MACROBLOCKD *xd,
int mi_row, int mi_col,
BLOCK_SIZE bsize) {
const int block_width = num_8x8_blocks_wide_lookup[bsize];
const int block_height = num_8x8_blocks_high_lookup[bsize];
int i, j;
for (j = 0; j < block_height; ++j)
for (i = 0; i < block_width; ++i) {
if (mi_row + j < cm->mi_rows && mi_col + i < cm->mi_cols)
xd->mi[j * xd->mi_stride + i] = xd->mi[0];
}
}
static void set_block_size(VP10_COMP * const cpi,
MACROBLOCK *const x,
MACROBLOCKD *const xd,
......@@ -1106,36 +1093,6 @@ void vp10_setup_src_planes(MACROBLOCK *x, const YV12_BUFFER_CONFIG *src,
x->e_mbd.plane[i].subsampling_y);
}
static void set_mode_info_seg_skip(MACROBLOCK *x, TX_MODE tx_mode,
RD_COST *rd_cost, BLOCK_SIZE bsize) {
MACROBLOCKD *const xd = &x->e_mbd;
MB_MODE_INFO *const mbmi = &xd->mi[0]->mbmi;
INTERP_FILTER filter_ref;
if (xd->up_available)
filter_ref = xd->mi[-xd->mi_stride]->mbmi.interp_filter;
else if (xd->left_available)
filter_ref = xd->mi[-1]->mbmi.interp_filter;
else
filter_ref = EIGHTTAP;
mbmi->sb_type = bsize;
mbmi->mode = ZEROMV;
mbmi->tx_size = VPXMIN(max_txsize_lookup[bsize],
tx_mode_to_biggest_tx_size[tx_mode]);
mbmi->skip = 1;
mbmi->uv_mode = DC_PRED;
mbmi->ref_frame[0] = LAST_FRAME;
mbmi->ref_frame[1] = NONE;
mbmi->mv[0].as_int = 0;
mbmi->interp_filter = filter_ref;
xd->mi[0]->bmi[0].as_mv[0].as_int = 0;
x->skip = 1;
vp10_rd_cost_init(rd_cost);
}
static int set_segment_rdmult(VP10_COMP *const cpi,
MACROBLOCK *const x,
int8_t segment_id) {
......@@ -2628,83 +2585,6 @@ static TX_MODE select_tx_mode(const VP10_COMP *cpi, MACROBLOCKD *const xd) {
return cpi->common.tx_mode;
}
static void fill_mode_info_sb(VP10_COMMON *cm, MACROBLOCK *x,
int mi_row, int mi_col,
BLOCK_SIZE bsize,
PC_TREE *pc_tree) {
MACROBLOCKD *xd = &x->e_mbd;
int bsl = b_width_log2_lookup[bsize], hbs = (1 << bsl) / 4;
PARTITION_TYPE partition = pc_tree->partitioning;
BLOCK_SIZE subsize = get_subsize(bsize, partition);
assert(bsize >= BLOCK_8X8);
if (mi_row >= cm->mi_rows || mi_col >= cm->mi_cols)
return;
switch (partition) {
case PARTITION_NONE:
set_mode_info_offsets(cm, x, xd, mi_row, mi_col);
*(xd->mi[0]) = pc_tree->none.mic;
*(x->mbmi_ext) = pc_tree->none.mbmi_ext;
duplicate_mode_info_in_sb(cm, xd, mi_row, mi_col, bsize);
break;
case PARTITION_VERT:
set_mode_info_offsets(cm, x, xd, mi_row, mi_col);
*(xd->mi[0]) = pc_tree->vertical[0].mic;
*(x->mbmi_ext) = pc_tree->vertical[0].mbmi_ext;
duplicate_mode_info_in_sb(cm, xd, mi_row, mi_col, subsize);
if (mi_col + hbs < cm->mi_cols) {
set_mode_info_offsets(cm, x, xd, mi_row, mi_col + hbs);
*(xd->mi[0]) = pc_tree->vertical[1].mic;
*(x->mbmi_ext) = pc_tree->vertical[1].mbmi_ext;
duplicate_mode_info_in_sb(cm, xd, mi_row, mi_col + hbs, subsize);
}
break;
case PARTITION_HORZ:
set_mode_info_offsets(cm, x, xd, mi_row, mi_col);
*(xd->mi[0]) = pc_tree->horizontal[0].mic;
*(x->mbmi_ext) = pc_tree->horizontal[0].mbmi_ext;
duplicate_mode_info_in_sb(cm, xd, mi_row, mi_col, subsize);
if (mi_row + hbs < cm->mi_rows) {
set_mode_info_offsets(cm, x, xd, mi_row + hbs, mi_col);
*(xd->mi[0]) = pc_tree->horizontal[1].mic;
*(x->mbmi_ext) = pc_tree->horizontal[1].mbmi_ext;
duplicate_mode_info_in_sb(cm, xd, mi_row + hbs, mi_col, subsize);
}
break;
case PARTITION_SPLIT: {
fill_mode_info_sb(cm, x, mi_row, mi_col, subsize, pc_tree->split[0]);
fill_mode_info_sb(cm, x, mi_row, mi_col + hbs, subsize,
pc_tree->split[1]);
fill_mode_info_sb(cm, x, mi_row + hbs, mi_col, subsize,
pc_tree->split[2]);
fill_mode_info_sb(cm, x, mi_row + hbs, mi_col + hbs, subsize,
pc_tree->split[3]);
break;
}
default:
break;
}
}
// Reset the prediction pixel ready flag recursively.
static void pred_pixel_ready_reset(PC_TREE *pc_tree, BLOCK_SIZE bsize) {
pc_tree->none.pred_pixel_ready = 0;
pc_tree->horizontal[0].pred_pixel_ready = 0;
pc_tree->horizontal[1].pred_pixel_ready = 0;
pc_tree->vertical[0].pred_pixel_ready = 0;
pc_tree->vertical[1].pred_pixel_ready = 0;
if (bsize > BLOCK_8X8) {
BLOCK_SIZE subsize = get_subsize(bsize, PARTITION_SPLIT);
int i;
for (i = 0; i < 4; ++i)
pred_pixel_ready_reset(pc_tree->split[i], subsize);
}
}
static int get_skip_encode_frame(const VP10_COMMON *cm, ThreadData *const td) {
unsigned int intra_count = 0, inter_count = 0;
int j;
......
......@@ -2145,42 +2145,6 @@ typedef struct {
uint32_t samples[4]; // total/y/u/v
} PSNR_STATS;
static void calc_psnr(const YV12_BUFFER_CONFIG *a, const YV12_BUFFER_CONFIG *b,
PSNR_STATS *psnr) {
static const double peak = 255.0;
const int widths[3] = {
a->y_crop_width, a->uv_crop_width, a->uv_crop_width};
const int heights[3] = {
a->y_crop_height, a->uv_crop_height, a->uv_crop_height};
const uint8_t *a_planes[3] = {a->y_buffer, a->u_buffer, a->v_buffer};
const int a_strides[3] = {a->y_stride, a->uv_stride, a->uv_stride};
const uint8_t *b_planes[3] = {b->y_buffer, b->u_buffer, b->v_buffer};
const int b_strides[3] = {b->y_stride, b->uv_stride, b->uv_stride};
int i;
uint64_t total_sse = 0;
uint32_t total_samples = 0;
for (i = 0; i < 3; ++i) {
const int w = widths[i];
const int h = heights[i];
const uint32_t samples = w * h;
const uint64_t sse = get_sse(a_planes[i], a_strides[i],
b_planes[i], b_strides[i],
w, h);
psnr->sse[1 + i] = sse;
psnr->samples[1 + i] = samples;
psnr->psnr[1 + i] = vpx_sse_to_psnr(samples, peak, (double)sse);
total_sse += sse;
total_samples += samples;
}
psnr->sse[0] = total_sse;
psnr->samples[0] = total_samples;
psnr->psnr[0] = vpx_sse_to_psnr((double)total_samples, peak,
(double)total_sse);
}
#if CONFIG_VP9_HIGHBITDEPTH
static void calc_highbd_psnr(const YV12_BUFFER_CONFIG *a,
const YV12_BUFFER_CONFIG *b,
......@@ -2233,6 +2197,44 @@ static void calc_highbd_psnr(const YV12_BUFFER_CONFIG *a,
psnr->psnr[0] = vpx_sse_to_psnr((double)total_samples, peak,
(double)total_sse);
}
#else // !CONFIG_VP9_HIGHBITDEPTH
static void calc_psnr(const YV12_BUFFER_CONFIG *a, const YV12_BUFFER_CONFIG *b,
PSNR_STATS *psnr) {
static const double peak = 255.0;
const int widths[3] = {
a->y_crop_width, a->uv_crop_width, a->uv_crop_width};
const int heights[3] = {
a->y_crop_height, a->uv_crop_height, a->uv_crop_height};
const uint8_t *a_planes[3] = {a->y_buffer, a->u_buffer, a->v_buffer};
const int a_strides[3] = {a->y_stride, a->uv_stride, a->uv_stride};
const uint8_t *b_planes[3] = {b->y_buffer, b->u_buffer, b->v_buffer};
const int b_strides[3] = {b->y_stride, b->uv_stride, b->uv_stride};
int i;
uint64_t total_sse = 0;
uint32_t total_samples = 0;
for (i = 0; i < 3; ++i) {
const int w = widths[i];
const int h = heights[i];
const uint32_t samples = w * h;
const uint64_t sse = get_sse(a_planes[i], a_strides[i],
b_planes[i], b_strides[i],
w, h);
psnr->sse[1 + i] = sse;
psnr->samples[1 + i] = samples;
psnr->psnr[1 + i] = vpx_sse_to_psnr(samples, peak, (double)sse);
total_sse += sse;
total_samples += samples;
}
psnr->sse[0] = total_sse;
psnr->samples[0] = total_samples;
psnr->psnr[0] = vpx_sse_to_psnr((double)total_samples, peak,
(double)total_sse);
}
#endif // CONFIG_VP9_HIGHBITDEPTH
static void generate_psnr_packet(VP10_COMP *cpi) {
......
......@@ -52,18 +52,6 @@ static int enc_worker_hook(EncWorkerData *const thread_data, void *unused) {
return 0;
}
static int get_max_tile_cols(VP10_COMP *cpi) {
const int aligned_width = ALIGN_POWER_OF_TWO(cpi->oxcf.width, MI_SIZE_LOG2);
int mi_cols = aligned_width >> MI_SIZE_LOG2;
int min_log2_tile_cols, max_log2_tile_cols;
int log2_tile_cols;
vp10_get_tile_n_bits(mi_cols, &min_log2_tile_cols, &max_log2_tile_cols);
log2_tile_cols = clamp(cpi->oxcf.tile_columns,
min_log2_tile_cols, max_log2_tile_cols);
return (1 << log2_tile_cols);
}
void vp10_encode_tiles_mt(VP10_COMP *cpi) {
VP10_COMMON *const cm = &cpi->common;
const int tile_cols = 1 << cm->log2_tile_cols;
......
......@@ -407,6 +407,7 @@ static void pack_tokens_into_partitions(VP8_COMP *cpi, unsigned char *cx_data,
}
#if CONFIG_MULTITHREAD
static void pack_mb_row_tokens(VP8_COMP *cpi, vp8_writer *w)
{
int mb_row;
......@@ -421,6 +422,7 @@ static void pack_mb_row_tokens(VP8_COMP *cpi, vp8_writer *w)
}
}
#endif // CONFIG_MULTITHREAD
static void write_mv_ref
(
......@@ -1675,7 +1677,7 @@ void vp8_pack_bitstream(VP8_COMP *cpi, unsigned char *dest, unsigned char * dest
if (cpi->b_multi_threaded)
pack_mb_row_tokens(cpi, &cpi->bc[1]);
else
#endif
#endif // CONFIG_MULTITHREAD
vp8_pack_tokens(&cpi->bc[1], cpi->tok, cpi->tok_count);
vp8_stop_encode(&cpi->bc[1]);
......
......@@ -700,6 +700,7 @@ static void init_encode_frame_mb_context(VP8_COMP *cpi)
vp8_zero(x->count_mb_ref_frame_usage);
}
#if CONFIG_MULTITHREAD
static void sum_coef_counts(MACROBLOCK *x, MACROBLOCK *x_thread)
{
int i = 0;
......@@ -729,6 +730,7 @@ static void sum_coef_counts(MACROBLOCK *x, MACROBLOCK *x_thread)
}
while (++i < BLOCK_TYPES);
}
#endif // CONFIG_MULTITHREAD
void vp8_encode_frame(VP8_COMP *cpi)
{
......@@ -927,7 +929,7 @@ void vp8_encode_frame(VP8_COMP *cpi)
}
else
#endif
#endif // CONFIG_MULTITHREAD
{
/* for each macroblock row in image */
......
......@@ -2248,42 +2248,6 @@ typedef struct {
uint32_t samples[4]; // total/y/u/v
} PSNR_STATS;
static void calc_psnr(const YV12_BUFFER_CONFIG *a, const YV12_BUFFER_CONFIG *b,
PSNR_STATS *psnr) {
static const double peak = 255.0;
const int widths[3] = {
a->y_crop_width, a->uv_crop_width, a->uv_crop_width};
const int heights[3] = {
a->y_crop_height, a->uv_crop_height, a->uv_crop_height};
const uint8_t *a_planes[3] = {a->y_buffer, a->u_buffer, a->v_buffer};
const int a_strides[3] = {a->y_stride, a->uv_stride, a->uv_stride};
const uint8_t *b_planes[3] = {b->y_buffer, b->u_buffer, b->v_buffer};
const int b_strides[3] = {b->y_stride, b->uv_stride, b->uv_stride};
int i;
uint64_t total_sse = 0;
uint32_t total_samples = 0;
for (i = 0; i < 3; ++i) {
const int w = widths[i];
const int h = heights[i];
const uint32_t samples = w * h;
const uint64_t sse = get_sse(a_planes[i], a_strides[i],
b_planes[i], b_strides[i],
w, h);
psnr->sse[1 + i] = sse;
psnr->samples[1 + i] = samples;
psnr->psnr[1 + i] = vpx_sse_to_psnr(samples, peak, (double)sse);
total_sse += sse;
total_samples += samples;
}
psnr->sse[0] = total_sse;
psnr->samples[0] = total_samples;
psnr->psnr[0] = vpx_sse_to_psnr((double)total_samples, peak,
(double)total_sse);
}
#if CONFIG_VP9_HIGHBITDEPTH
static void calc_highbd_psnr(const YV12_BUFFER_CONFIG *a,
const YV12_BUFFER_CONFIG *b,
......@@ -2336,6 +2300,44 @@ static void calc_highbd_psnr(const YV12_BUFFER_CONFIG *a,
psnr->psnr[0] = vpx_sse_to_psnr((double)total_samples, peak,
(double)total_sse);
}
#else // !CONFIG_VP9_HIGHBITDEPTH
static void calc_psnr(const YV12_BUFFER_CONFIG *a, const YV12_BUFFER_CONFIG *b,
PSNR_STATS *psnr) {
static const double peak = 255.0;
const int widths[3] = {
a->y_crop_width, a->uv_crop_width, a->uv_crop_width};
const int heights[3] = {
a->y_crop_height, a->uv_crop_height, a->uv_crop_height};
const uint8_t *a_planes[3] = {a->y_buffer, a->u_buffer, a->v_buffer};
const int a_strides[3] = {a->y_stride, a->uv_stride, a->uv_stride};
const uint8_t *b_planes[3] = {b->y_buffer, b->u_buffer, b->v_buffer};
const int b_strides[3] = {b->y_stride, b->uv_stride, b->uv_stride};
int i;
uint64_t total_sse = 0;
uint32_t total_samples = 0;
for (i = 0; i < 3; ++i) {
const int w = widths[i];
const int h = heights[i];
const uint32_t samples = w * h;
const uint64_t sse = get_sse(a_planes[i], a_strides[i],
b_planes[i], b_strides[i],
w, h);
psnr->sse[1 + i] = sse;
psnr->samples[1 + i] = samples;
psnr->psnr[1 + i] = vpx_sse_to_psnr(samples, peak, (double)sse);
total_sse += sse;
total_samples += samples;
}
psnr->sse[0] = total_sse;
psnr->samples[0] = total_samples;
psnr->psnr[0] = vpx_sse_to_psnr((double)total_samples, peak,
(double)total_sse);
}
#endif // CONFIG_VP9_HIGHBITDEPTH
static void generate_psnr_packet(VP9_COMP *cpi) {
......
Markdown is supported
0% or .
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment