hashirama/uvg266
1
0
Fork 0
mirror of https://github.com/ultravideo/uvg266.git synced 2024-11-30 12:44:07 +00:00
Code Issues Actions 1 Projects Releases Wiki Activity

Rename pixel_t to kvz_pixel.

Browse source
This commit is contained in:
Arttu Ylä-Outinen 2015-06-30 11:43:48 +03:00
parent cecea44d37
commit f7f17a060c
36 changed files with 331 additions and 331 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

6
src/cu.h
View file

@ -123,9 +123,9 @@ int cu_array_free(cu_array_t *cua);
* - First pixel is the top-left pixel. * - First pixel is the top-left pixel.
*/ */
typedef struct { typedef struct {
pixel_t y[LCU_REF_PX_WIDTH + 1]; kvz_pixel y[LCU_REF_PX_WIDTH + 1];
pixel_t u[LCU_REF_PX_WIDTH / 2 + 1]; kvz_pixel u[LCU_REF_PX_WIDTH / 2 + 1];
pixel_t v[LCU_REF_PX_WIDTH / 2 + 1]; kvz_pixel v[LCU_REF_PX_WIDTH / 2 + 1];
} lcu_ref_px_t; } lcu_ref_px_t;
typedef struct { typedef struct {

18
src/encoderstate.c
View file

@ -441,9 +441,9 @@ static void encoder_state_worker_sao_reconstruct_lcu(void *opaque) {
int x; int x;
//TODO: copy only needed data //TODO: copy only needed data
pixel_t *new_y_data = MALLOC(pixel_t, frame->width * frame->height); kvz_pixel *new_y_data = MALLOC(kvz_pixel, frame->width * frame->height);
pixel_t *new_u_data = MALLOC(pixel_t, (frame->width * frame->height) >> 2); kvz_pixel *new_u_data = MALLOC(kvz_pixel, (frame->width * frame->height) >> 2);
pixel_t *new_v_data = MALLOC(pixel_t, (frame->width * frame->height) >> 2); kvz_pixel *new_v_data = MALLOC(kvz_pixel, (frame->width * frame->height) >> 2);
const int offset = frame->width * (data->y*LCU_WIDTH); const int offset = frame->width * (data->y*LCU_WIDTH);
const int offset_c = frame->width/2 * (data->y*LCU_WIDTH_C); const int offset_c = frame->width/2 * (data->y*LCU_WIDTH_C);
@ -453,15 +453,15 @@ static void encoder_state_worker_sao_reconstruct_lcu(void *opaque) {
num_pixels = frame->width * frame->height - offset; num_pixels = frame->width * frame->height - offset;
} }
memcpy(&new_y_data[offset], &frame->rec->y[offset], sizeof(pixel_t) * num_pixels); memcpy(&new_y_data[offset], &frame->rec->y[offset], sizeof(kvz_pixel) * num_pixels);
memcpy(&new_u_data[offset_c], &frame->rec->u[offset_c], sizeof(pixel_t) * num_pixels >> 2); memcpy(&new_u_data[offset_c], &frame->rec->u[offset_c], sizeof(kvz_pixel) * num_pixels >> 2);
memcpy(&new_v_data[offset_c], &frame->rec->v[offset_c], sizeof(pixel_t) * num_pixels >> 2); memcpy(&new_v_data[offset_c], &frame->rec->v[offset_c], sizeof(kvz_pixel) * num_pixels >> 2);
if (data->y>0) { if (data->y>0) {
//copy first row from buffer //copy first row from buffer
memcpy(&new_y_data[frame->width * (data->y*LCU_WIDTH-1)], &data->encoder_state->tile->hor_buf_before_sao->y[frame->width * (data->y-1)], frame->width * sizeof(pixel_t)); memcpy(&new_y_data[frame->width * (data->y*LCU_WIDTH-1)], &data->encoder_state->tile->hor_buf_before_sao->y[frame->width * (data->y-1)], frame->width * sizeof(kvz_pixel));
memcpy(&new_u_data[frame->width/2 * (data->y*LCU_WIDTH_C-1)], &data->encoder_state->tile->hor_buf_before_sao->u[frame->width/2 * (data->y-1)], frame->width/2 * sizeof(pixel_t)); memcpy(&new_u_data[frame->width/2 * (data->y*LCU_WIDTH_C-1)], &data->encoder_state->tile->hor_buf_before_sao->u[frame->width/2 * (data->y-1)], frame->width/2 * sizeof(kvz_pixel));
memcpy(&new_v_data[frame->width/2 * (data->y*LCU_WIDTH_C-1)], &data->encoder_state->tile->hor_buf_before_sao->v[frame->width/2 * (data->y-1)], frame->width/2 * sizeof(pixel_t)); memcpy(&new_v_data[frame->width/2 * (data->y*LCU_WIDTH_C-1)], &data->encoder_state->tile->hor_buf_before_sao->v[frame->width/2 * (data->y-1)], frame->width/2 * sizeof(kvz_pixel));
} }
for (x = 0; x < frame->width_in_lcu; x++) { for (x = 0; x < frame->width_in_lcu; x++) {

12
src/filter.c
View file

@ -83,7 +83,7 @@ const int8_t g_chroma_filter[8][4] =
/** /**
* \brief * \brief
*/ */
INLINE void filter_deblock_luma(const encoder_control_t * const encoder, pixel_t *src, int32_t offset, INLINE void filter_deblock_luma(const encoder_control_t * const encoder, kvz_pixel *src, int32_t offset,
int32_t tc, int8_t sw, int32_t tc, int8_t sw,
int8_t part_P_nofilter, int8_t part_Q_nofilter, int8_t part_P_nofilter, int8_t part_Q_nofilter,
int32_t thr_cut, int32_t thr_cut,
@ -143,7 +143,7 @@ INLINE void filter_deblock_luma(const encoder_control_t * const encoder, pixel_t
/** /**
* \brief * \brief
*/ */
INLINE void filter_deblock_chroma(const encoder_control_t * const encoder, pixel_t *src, int32_t offset, int32_t tc, INLINE void filter_deblock_chroma(const encoder_control_t * const encoder, kvz_pixel *src, int32_t offset, int32_t tc,
int8_t part_P_nofilter, int8_t part_Q_nofilter) int8_t part_P_nofilter, int8_t part_Q_nofilter)
{ {
int32_t delta; int32_t delta;
@ -188,8 +188,8 @@ void filter_deblock_edge_luma(encoder_state_t * const state,
int32_t beta_offset_div2 = encoder->beta_offset_div2; int32_t beta_offset_div2 = encoder->beta_offset_div2;
int32_t tc_offset_div2 = encoder->tc_offset_div2; int32_t tc_offset_div2 = encoder->tc_offset_div2;
// TODO: support 10+bits // TODO: support 10+bits
pixel_t *orig_src = &frame->rec->y[xpos + ypos*stride]; kvz_pixel *orig_src = &frame->rec->y[xpos + ypos*stride];
pixel_t *src = orig_src; kvz_pixel *src = orig_src;
int32_t step = 1; int32_t step = 1;
cu_info_t *cu_p = NULL; cu_info_t *cu_p = NULL;
int16_t x_cu = xpos>>MIN_SIZE,y_cu = ypos>>MIN_SIZE; int16_t x_cu = xpos>>MIN_SIZE,y_cu = ypos>>MIN_SIZE;
@ -381,8 +381,8 @@ void filter_deblock_edge_chroma(encoder_state_t * const state,
int32_t stride = frame->rec->stride >> 1; int32_t stride = frame->rec->stride >> 1;
int32_t tc_offset_div2 = encoder->tc_offset_div2; int32_t tc_offset_div2 = encoder->tc_offset_div2;
// TODO: support 10+bits // TODO: support 10+bits
pixel_t *src_u = &frame->rec->u[x + y*stride]; kvz_pixel *src_u = &frame->rec->u[x + y*stride];
pixel_t *src_v = &frame->rec->v[x + y*stride]; kvz_pixel *src_v = &frame->rec->v[x + y*stride];
// Init offset and step to EDGE_HOR // Init offset and step to EDGE_HOR
int32_t offset = stride; int32_t offset = stride;
int32_t step = 1; int32_t step = 1;

4
src/filter.h
View file

@ -43,11 +43,11 @@ void filter_deblock_edge_chroma(encoder_state_t *state,
int32_t xpos, int32_t ypos, int32_t xpos, int32_t ypos,
int8_t depth, int8_t dir); int8_t depth, int8_t dir);
void filter_deblock_lcu(encoder_state_t *state, int x_px, int y_px); void filter_deblock_lcu(encoder_state_t *state, int x_px, int y_px);
void filter_deblock_luma(const encoder_control_t * const encoder, pixel_t *src, int32_t offset, int32_t tc , int8_t sw, void filter_deblock_luma(const encoder_control_t * const encoder, kvz_pixel *src, int32_t offset, int32_t tc , int8_t sw,
int8_t part_p_nofilter, int8_t part_q_nofilter, int8_t part_p_nofilter, int8_t part_q_nofilter,
int32_t thr_cut, int32_t thr_cut,
int8_t filter_second_p, int8_t filter_second_q); int8_t filter_second_p, int8_t filter_second_q);
void filter_deblock_chroma(const encoder_control_t * const encoder, pixel_t *src, int32_t offset, int32_t tc, void filter_deblock_chroma(const encoder_control_t * const encoder, kvz_pixel *src, int32_t offset, int32_t tc,
int8_t part_p_nofilter, int8_t part_q_nofilter); int8_t part_p_nofilter, int8_t part_q_nofilter);
// SAO // SAO

28
src/image.c
View file

@ -52,7 +52,7 @@ kvz_picture *image_alloc(const int32_t width, const int32_t height)
unsigned int chroma_size = luma_size / 4; unsigned int chroma_size = luma_size / 4;
//Allocate memory //Allocate memory
im->fulldata = MALLOC(pixel_t, (luma_size + 2 * chroma_size)); im->fulldata = MALLOC(kvz_pixel, (luma_size + 2 * chroma_size));
if (!im->fulldata) { if (!im->fulldata) {
free(im); free(im);
return NULL; return NULL;
@ -156,9 +156,9 @@ yuv_t * yuv_t_alloc(int luma_size)
// Get buffers with separate mallocs in order to take advantage of // Get buffers with separate mallocs in order to take advantage of
// automatic buffer overrun checks. // automatic buffer overrun checks.
yuv_t *yuv = (yuv_t *)malloc(sizeof(*yuv)); yuv_t *yuv = (yuv_t *)malloc(sizeof(*yuv));
yuv->y = (pixel_t *)malloc(luma_size * sizeof(*yuv->y)); yuv->y = (kvz_pixel *)malloc(luma_size * sizeof(*yuv->y));
yuv->u = (pixel_t *)malloc(luma_size / 2 * sizeof(*yuv->u)); yuv->u = (kvz_pixel *)malloc(luma_size / 2 * sizeof(*yuv->u));
yuv->v = (pixel_t *)malloc(luma_size / 2 * sizeof(*yuv->v)); yuv->v = (kvz_pixel *)malloc(luma_size / 2 * sizeof(*yuv->v));
yuv->size = luma_size; yuv->size = luma_size;
return yuv; return yuv;
@ -183,10 +183,10 @@ void yuv_t_free(yuv_t * yuv)
* *
* \returns Sum of Absolute Differences * \returns Sum of Absolute Differences
*/ */
static unsigned cor_sad(const pixel_t *pic_data, const pixel_t *ref_data, static unsigned cor_sad(const kvz_pixel *pic_data, const kvz_pixel *ref_data,
int block_width, int block_height, unsigned pic_stride) int block_width, int block_height, unsigned pic_stride)
{ {
pixel_t ref = *ref_data; kvz_pixel ref = *ref_data;
int x, y; int x, y;
unsigned sad = 0; unsigned sad = 0;
@ -210,7 +210,7 @@ static unsigned cor_sad(const pixel_t *pic_data, const pixel_t *ref_data,
* *
* \returns Sum of Absolute Differences * \returns Sum of Absolute Differences
*/ */
static unsigned ver_sad(const pixel_t *pic_data, const pixel_t *ref_data, static unsigned ver_sad(const kvz_pixel *pic_data, const kvz_pixel *ref_data,
int block_width, int block_height, unsigned pic_stride) int block_width, int block_height, unsigned pic_stride)
{ {
int x, y; int x, y;
@ -236,7 +236,7 @@ static unsigned ver_sad(const pixel_t *pic_data, const pixel_t *ref_data,
* *
* \returns Sum of Absolute Differences * \returns Sum of Absolute Differences
*/ */
static unsigned hor_sad(const pixel_t *pic_data, const pixel_t *ref_data, static unsigned hor_sad(const kvz_pixel *pic_data, const kvz_pixel *ref_data,
int block_width, int block_height, unsigned pic_stride, unsigned ref_stride) int block_width, int block_height, unsigned pic_stride, unsigned ref_stride)
{ {
int x, y; int x, y;
@ -269,7 +269,7 @@ static unsigned image_interpolated_sad(const kvz_picture *pic, const kvz_picture
int pic_x, int pic_y, int ref_x, int ref_y, int pic_x, int pic_y, int ref_x, int ref_y,
int block_width, int block_height) int block_width, int block_height)
{ {
pixel_t *pic_data, *ref_data; kvz_pixel *pic_data, *ref_data;
int left, right, top, bottom; int left, right, top, bottom;
int result = 0; int result = 0;
@ -424,8 +424,8 @@ unsigned image_calc_sad(const kvz_picture *pic, const kvz_picture *ref, int pic_
{ {
// Reference block is completely inside the frame, so just calculate the // Reference block is completely inside the frame, so just calculate the
// SAD directly. This is the most common case, which is why it's first. // SAD directly. This is the most common case, which is why it's first.
const pixel_t *pic_data = &pic->y[pic_y * pic->stride + pic_x]; const kvz_pixel *pic_data = &pic->y[pic_y * pic->stride + pic_x];
const pixel_t *ref_data = &ref->y[ref_y * ref->stride + ref_x]; const kvz_pixel *ref_data = &ref->y[ref_y * ref->stride + ref_x];
return reg_sad(pic_data, ref_data, block_width, block_height, pic->stride, ref->stride); return reg_sad(pic_data, ref_data, block_width, block_height, pic->stride, ref->stride);
} else { } else {
// Call a routine that knows how to interpolate pixels outside the frame. // Call a routine that knows how to interpolate pixels outside the frame.
@ -434,7 +434,7 @@ unsigned image_calc_sad(const kvz_picture *pic, const kvz_picture *ref, int pic_
} }
unsigned pixels_calc_ssd(const pixel_t *const ref, const pixel_t *const rec, unsigned pixels_calc_ssd(const kvz_pixel *const ref, const kvz_pixel *const rec,
const int ref_stride, const int rec_stride, const int ref_stride, const int rec_stride,
const int width) const int width)
{ {
@ -467,7 +467,7 @@ unsigned pixels_calc_ssd(const pixel_t *const ref, const pixel_t *const rec,
* This should be inlined, but it's defined here for now to see if Visual * This should be inlined, but it's defined here for now to see if Visual
* Studios LTCG will inline it. * Studios LTCG will inline it.
*/ */
void pixels_blit(const pixel_t * const orig, pixel_t * const dst, void pixels_blit(const kvz_pixel * const orig, kvz_pixel * const dst,
const unsigned width, const unsigned height, const unsigned width, const unsigned height,
const unsigned orig_stride, const unsigned dst_stride) const unsigned orig_stride, const unsigned dst_stride)
{ {
@ -496,7 +496,7 @@ void pixels_blit(const pixel_t * const orig, pixel_t * const dst,
assert(orig != dst || orig_stride == dst_stride); assert(orig != dst || orig_stride == dst_stride);
for (y = 0; y < height; ++y) { for (y = 0; y < height; ++y) {
memcpy(&dst[y*dst_stride], &orig[y*orig_stride], width * sizeof(pixel_t)); memcpy(&dst[y*dst_stride], &orig[y*orig_stride], width * sizeof(kvz_pixel));
} }
} }

16
src/image.h
View file

@ -29,16 +29,16 @@
#include "kvazaar.h" #include "kvazaar.h"
typedef struct { typedef struct {
pixel_t y[LCU_LUMA_SIZE]; kvz_pixel y[LCU_LUMA_SIZE];
pixel_t u[LCU_CHROMA_SIZE]; kvz_pixel u[LCU_CHROMA_SIZE];
pixel_t v[LCU_CHROMA_SIZE]; kvz_pixel v[LCU_CHROMA_SIZE];
} lcu_yuv_t; } lcu_yuv_t;
typedef struct { typedef struct {
int size; int size;
pixel_t *y; kvz_pixel *y;
pixel_t *u; kvz_pixel *u;
pixel_t *v; kvz_pixel *v;
} yuv_t; } yuv_t;
@ -62,12 +62,12 @@ unsigned image_calc_sad(const kvz_picture *pic, const kvz_picture *ref, int pic_
int block_width, int block_height, int max_lcu_below); int block_width, int block_height, int max_lcu_below);
unsigned pixels_calc_ssd(const pixel_t *const ref, const pixel_t *const rec, unsigned pixels_calc_ssd(const kvz_pixel *const ref, const kvz_pixel *const rec,
const int ref_stride, const int rec_stride, const int ref_stride, const int rec_stride,
const int width); const int width);
void pixels_blit(const pixel_t* orig, pixel_t *dst, void pixels_blit(const kvz_pixel* orig, kvz_pixel *dst,
unsigned width, unsigned height, unsigned width, unsigned height,
unsigned orig_stride, unsigned dst_stride); unsigned orig_stride, unsigned dst_stride);

58
src/inter.c
View file

@ -91,12 +91,12 @@ void inter_recon_lcu(const encoder_state_t * const state, const kvz_picture * co
// Chroma half-pel // Chroma half-pel
#define HALFPEL_CHROMA_WIDTH ((LCU_WIDTH>>1) + 8) #define HALFPEL_CHROMA_WIDTH ((LCU_WIDTH>>1) + 8)
int8_t chroma_halfpel = ((mv[0]>>2)&1) || ((mv[1]>>2)&1); //!< (luma integer mv) lsb is set -> chroma is half-pel int8_t chroma_halfpel = ((mv[0]>>2)&1) || ((mv[1]>>2)&1); //!< (luma integer mv) lsb is set -> chroma is half-pel
pixel_t halfpel_src_u[HALFPEL_CHROMA_WIDTH * HALFPEL_CHROMA_WIDTH]; //!< U source block for interpolation kvz_pixel halfpel_src_u[HALFPEL_CHROMA_WIDTH * HALFPEL_CHROMA_WIDTH]; //!< U source block for interpolation
pixel_t halfpel_src_v[HALFPEL_CHROMA_WIDTH * HALFPEL_CHROMA_WIDTH]; //!< V source block for interpolation kvz_pixel halfpel_src_v[HALFPEL_CHROMA_WIDTH * HALFPEL_CHROMA_WIDTH]; //!< V source block for interpolation
pixel_t *halfpel_src_off_u = &halfpel_src_u[HALFPEL_CHROMA_WIDTH * 4 + 4]; //!< halfpel_src_u with offset (4,4) kvz_pixel *halfpel_src_off_u = &halfpel_src_u[HALFPEL_CHROMA_WIDTH * 4 + 4]; //!< halfpel_src_u with offset (4,4)
pixel_t *halfpel_src_off_v = &halfpel_src_v[HALFPEL_CHROMA_WIDTH * 4 + 4]; //!< halfpel_src_v with offset (4,4) kvz_pixel *halfpel_src_off_v = &halfpel_src_v[HALFPEL_CHROMA_WIDTH * 4 + 4]; //!< halfpel_src_v with offset (4,4)
pixel_t halfpel_u[LCU_WIDTH * LCU_WIDTH]; //!< interpolated 2W x 2H block (u) kvz_pixel halfpel_u[LCU_WIDTH * LCU_WIDTH]; //!< interpolated 2W x 2H block (u)
pixel_t halfpel_v[LCU_WIDTH * LCU_WIDTH]; //!< interpolated 2W x 2H block (v) kvz_pixel halfpel_v[LCU_WIDTH * LCU_WIDTH]; //!< interpolated 2W x 2H block (v)
// Luma quarter-pel // Luma quarter-pel
int8_t fractional_mv = (mv[0]&1) || (mv[1]&1) || (mv[0]&2) || (mv[1]&2); // either of 2 lowest bits of mv set -> mv is fractional int8_t fractional_mv = (mv[0]&1) || (mv[1]&1) || (mv[0]&2) || (mv[1]&2); // either of 2 lowest bits of mv set -> mv is fractional
@ -114,19 +114,19 @@ void inter_recon_lcu(const encoder_state_t * const state, const kvz_picture * co
#define FILTER_SIZE_C 4 //Chroma filter size #define FILTER_SIZE_C 4 //Chroma filter size
// Fractional luma 1/4-pel // Fractional luma 1/4-pel
pixel_t qpel_src_y[(LCU_WIDTH+FILTER_SIZE_Y) * (LCU_WIDTH+FILTER_SIZE_Y)]; kvz_pixel qpel_src_y[(LCU_WIDTH+FILTER_SIZE_Y) * (LCU_WIDTH+FILTER_SIZE_Y)];
pixel_t* qpel_src_off_y = &qpel_src_y[(width+FILTER_SIZE_Y)*(FILTER_SIZE_Y>>1)+(FILTER_SIZE_Y>>1)]; kvz_pixel* qpel_src_off_y = &qpel_src_y[(width+FILTER_SIZE_Y)*(FILTER_SIZE_Y>>1)+(FILTER_SIZE_Y>>1)];
pixel_t qpel_dst_y[LCU_WIDTH*LCU_WIDTH*16]; kvz_pixel qpel_dst_y[LCU_WIDTH*LCU_WIDTH*16];
// Fractional chroma 1/8-pel // Fractional chroma 1/8-pel
int width_c = width>>1; int width_c = width>>1;
pixel_t octpel_src_u[((LCU_WIDTH>>1)+FILTER_SIZE_C) * ((LCU_WIDTH>>1)+FILTER_SIZE_C)]; kvz_pixel octpel_src_u[((LCU_WIDTH>>1)+FILTER_SIZE_C) * ((LCU_WIDTH>>1)+FILTER_SIZE_C)];
pixel_t* octpel_src_off_u = &octpel_src_u[(width_c+FILTER_SIZE_C)*(FILTER_SIZE_C>>1)+(FILTER_SIZE_C>>1)]; kvz_pixel* octpel_src_off_u = &octpel_src_u[(width_c+FILTER_SIZE_C)*(FILTER_SIZE_C>>1)+(FILTER_SIZE_C>>1)];
pixel_t octpel_dst_u[(LCU_WIDTH >> 1)*(LCU_WIDTH >> 1) * 64]; kvz_pixel octpel_dst_u[(LCU_WIDTH >> 1)*(LCU_WIDTH >> 1) * 64];
pixel_t octpel_src_v[((LCU_WIDTH >> 1) + FILTER_SIZE_C) * ((LCU_WIDTH >> 1) + FILTER_SIZE_C)]; kvz_pixel octpel_src_v[((LCU_WIDTH >> 1) + FILTER_SIZE_C) * ((LCU_WIDTH >> 1) + FILTER_SIZE_C)];
pixel_t* octpel_src_off_v = &octpel_src_v[(width_c + FILTER_SIZE_C)*(FILTER_SIZE_C >> 1) + (FILTER_SIZE_C >> 1)]; kvz_pixel* octpel_src_off_v = &octpel_src_v[(width_c + FILTER_SIZE_C)*(FILTER_SIZE_C >> 1) + (FILTER_SIZE_C >> 1)];
pixel_t octpel_dst_v[(LCU_WIDTH >> 1)*(LCU_WIDTH >> 1) * 64]; kvz_pixel octpel_dst_v[(LCU_WIDTH >> 1)*(LCU_WIDTH >> 1) * 64];
// Fractional luma // Fractional luma
extend_borders(xpos, ypos, mv[0]>>2, mv[1]>>2, state->tile->lcu_offset_x * LCU_WIDTH, state->tile->lcu_offset_y * LCU_WIDTH, extend_borders(xpos, ypos, mv[0]>>2, mv[1]>>2, state->tile->lcu_offset_x * LCU_WIDTH, state->tile->lcu_offset_y * LCU_WIDTH,
@ -156,7 +156,7 @@ void inter_recon_lcu(const encoder_state_t * const state, const kvz_picture * co
for(x = 0; x < width; ++x) { for(x = 0; x < width; ++x) {
int x_in_lcu = ((x+xpos) & ((LCU_WIDTH)-1)); int x_in_lcu = ((x+xpos) & ((LCU_WIDTH)-1));
int qpel_x = x*4+y_off_x; int qpel_x = x*4+y_off_x;
lcu->rec.y[y_in_lcu * LCU_WIDTH + x_in_lcu] = (pixel_t)qpel_dst_y[qpel_y*(width*4)+qpel_x]; lcu->rec.y[y_in_lcu * LCU_WIDTH + x_in_lcu] = (kvz_pixel)qpel_dst_y[qpel_y*(width*4)+qpel_x];
} }
} }
//Sample fractional pixels for chroma //Sample fractional pixels for chroma
@ -166,8 +166,8 @@ void inter_recon_lcu(const encoder_state_t * const state, const kvz_picture * co
for(x = 0; x < width_c; ++x) { for(x = 0; x < width_c; ++x) {
int x_in_lcu = ((x+(xpos>>1)) & ((LCU_WIDTH>>1)-1)); int x_in_lcu = ((x+(xpos>>1)) & ((LCU_WIDTH>>1)-1));
int qpel_x = x*8+c_off_x; int qpel_x = x*8+c_off_x;
lcu->rec.u[y_in_lcu * dst_width_c + x_in_lcu] = (pixel_t)octpel_dst_u[qpel_y*(width_c*8)+qpel_x]; lcu->rec.u[y_in_lcu * dst_width_c + x_in_lcu] = (kvz_pixel)octpel_dst_u[qpel_y*(width_c*8)+qpel_x];
lcu->rec.v[y_in_lcu * dst_width_c + x_in_lcu] = (pixel_t)octpel_dst_v[qpel_y*(width_c*8)+qpel_x]; lcu->rec.v[y_in_lcu * dst_width_c + x_in_lcu] = (kvz_pixel)octpel_dst_v[qpel_y*(width_c*8)+qpel_x];
} }
} }
} }
@ -219,8 +219,8 @@ void inter_recon_lcu(const encoder_state_t * const state, const kvz_picture * co
for (halfpel_x = abs_mv_x, x = xpos>>1; x < (xpos + width)>>1; halfpel_x += 2, x++) { for (halfpel_x = abs_mv_x, x = xpos>>1; x < (xpos + width)>>1; halfpel_x += 2, x++) {
int x_in_lcu = (x & ((LCU_WIDTH>>1)-1)); int x_in_lcu = (x & ((LCU_WIDTH>>1)-1));
int y_in_lcu = (y & ((LCU_WIDTH>>1)-1)); int y_in_lcu = (y & ((LCU_WIDTH>>1)-1));
lcu->rec.u[y_in_lcu*dst_width_c + x_in_lcu] = (pixel_t)halfpel_u[halfpel_y*LCU_WIDTH + halfpel_x]; lcu->rec.u[y_in_lcu*dst_width_c + x_in_lcu] = (kvz_pixel)halfpel_u[halfpel_y*LCU_WIDTH + halfpel_x];
lcu->rec.v[y_in_lcu*dst_width_c + x_in_lcu] = (pixel_t)halfpel_v[halfpel_y*LCU_WIDTH + halfpel_x]; lcu->rec.v[y_in_lcu*dst_width_c + x_in_lcu] = (kvz_pixel)halfpel_v[halfpel_y*LCU_WIDTH + halfpel_x];
} }
} }
} }
@ -339,17 +339,17 @@ void inter_recon_lcu(const encoder_state_t * const state, const kvz_picture * co
*/ */
void inter_recon_lcu_bipred(const encoder_state_t * const state, const kvz_picture * ref1, const kvz_picture * ref2, int32_t xpos, int32_t ypos, int32_t width, int16_t mv_param[2][2], lcu_t* lcu) { void inter_recon_lcu_bipred(const encoder_state_t * const state, const kvz_picture * ref1, const kvz_picture * ref2, int32_t xpos, int32_t ypos, int32_t width, int16_t mv_param[2][2], lcu_t* lcu) {
pixel_t temp_lcu_y[64 * 64]; kvz_pixel temp_lcu_y[64 * 64];
pixel_t temp_lcu_u[32 * 32]; kvz_pixel temp_lcu_u[32 * 32];
pixel_t temp_lcu_v[32 * 32]; kvz_pixel temp_lcu_v[32 * 32];
int temp_x, temp_y; int temp_x, temp_y;
// TODO: interpolated values require 14-bit accuracy for bi-prediction, current implementation of ipol filters round the value to 8bits // TODO: interpolated values require 14-bit accuracy for bi-prediction, current implementation of ipol filters round the value to 8bits
//Reconstruct both predictors //Reconstruct both predictors
inter_recon_lcu(state, ref1, xpos, ypos, width, mv_param[0], lcu); inter_recon_lcu(state, ref1, xpos, ypos, width, mv_param[0], lcu);
memcpy(temp_lcu_y, lcu->rec.y, sizeof(pixel_t) * 64 * 64); memcpy(temp_lcu_y, lcu->rec.y, sizeof(kvz_pixel) * 64 * 64);
memcpy(temp_lcu_u, lcu->rec.u, sizeof(pixel_t) * 32 * 32); memcpy(temp_lcu_u, lcu->rec.u, sizeof(kvz_pixel) * 32 * 32);
memcpy(temp_lcu_v, lcu->rec.v, sizeof(pixel_t) * 32 * 32); memcpy(temp_lcu_v, lcu->rec.v, sizeof(kvz_pixel) * 32 * 32);
inter_recon_lcu(state, ref2, xpos, ypos, width, mv_param[1], lcu); inter_recon_lcu(state, ref2, xpos, ypos, width, mv_param[1], lcu);
// After reconstruction, merge the predictors by taking an average of each pixel // After reconstruction, merge the predictors by taking an average of each pixel
@ -357,7 +357,7 @@ void inter_recon_lcu_bipred(const encoder_state_t * const state, const kvz_pictu
int y_in_lcu = ((ypos + temp_y) & ((LCU_WIDTH)-1)); int y_in_lcu = ((ypos + temp_y) & ((LCU_WIDTH)-1));
for (temp_x = 0; temp_x < width; ++temp_x) { for (temp_x = 0; temp_x < width; ++temp_x) {
int x_in_lcu = ((xpos + temp_x) & ((LCU_WIDTH)-1)); int x_in_lcu = ((xpos + temp_x) & ((LCU_WIDTH)-1));
lcu->rec.y[y_in_lcu * LCU_WIDTH + x_in_lcu] = (pixel_t)(((int)lcu->rec.y[y_in_lcu * LCU_WIDTH + x_in_lcu] + lcu->rec.y[y_in_lcu * LCU_WIDTH + x_in_lcu] = (kvz_pixel)(((int)lcu->rec.y[y_in_lcu * LCU_WIDTH + x_in_lcu] +
(int)temp_lcu_y[y_in_lcu * LCU_WIDTH + x_in_lcu] + 1) >> 1); (int)temp_lcu_y[y_in_lcu * LCU_WIDTH + x_in_lcu] + 1) >> 1);
} }
} }
@ -365,10 +365,10 @@ void inter_recon_lcu_bipred(const encoder_state_t * const state, const kvz_pictu
int y_in_lcu = (((ypos >> 1) + temp_y) & (LCU_WIDTH_C - 1)); int y_in_lcu = (((ypos >> 1) + temp_y) & (LCU_WIDTH_C - 1));
for (temp_x = 0; temp_x < width>>1; ++temp_x) { for (temp_x = 0; temp_x < width>>1; ++temp_x) {
int x_in_lcu = (((xpos >> 1) + temp_x) & (LCU_WIDTH_C - 1)); int x_in_lcu = (((xpos >> 1) + temp_x) & (LCU_WIDTH_C - 1));
lcu->rec.u[y_in_lcu * LCU_WIDTH_C + x_in_lcu] = (pixel_t)(((int)lcu->rec.u[y_in_lcu * LCU_WIDTH_C + x_in_lcu] + lcu->rec.u[y_in_lcu * LCU_WIDTH_C + x_in_lcu] = (kvz_pixel)(((int)lcu->rec.u[y_in_lcu * LCU_WIDTH_C + x_in_lcu] +
(int)temp_lcu_u[y_in_lcu * LCU_WIDTH_C + x_in_lcu] + 1) >> 1); (int)temp_lcu_u[y_in_lcu * LCU_WIDTH_C + x_in_lcu] + 1) >> 1);
lcu->rec.v[y_in_lcu * LCU_WIDTH_C + x_in_lcu] = (pixel_t)(((int)lcu->rec.v[y_in_lcu * LCU_WIDTH_C + x_in_lcu] + lcu->rec.v[y_in_lcu * LCU_WIDTH_C + x_in_lcu] = (kvz_pixel)(((int)lcu->rec.v[y_in_lcu * LCU_WIDTH_C + x_in_lcu] +
(int)temp_lcu_v[y_in_lcu * LCU_WIDTH_C + x_in_lcu] + 1) >> 1); (int)temp_lcu_v[y_in_lcu * LCU_WIDTH_C + x_in_lcu] + 1) >> 1);
} }
} }

80
src/intra.c
View file

@ -87,7 +87,7 @@ void intra_set_block_mode(videoframe_t *frame,uint32_t x_cu, uint32_t y_cu, uint
* \param width block width * \param width block width
* \returns DC prediction * \returns DC prediction
*/ */
pixel_t intra_get_dc_pred(const pixel_t *pic, uint16_t picwidth, uint8_t width) kvz_pixel intra_get_dc_pred(const kvz_pixel *pic, uint16_t picwidth, uint8_t width)
{ {
int32_t i, sum = 0; int32_t i, sum = 0;
@ -100,7 +100,7 @@ pixel_t intra_get_dc_pred(const pixel_t *pic, uint16_t picwidth, uint8_t width)
} }
// return the average // return the average
return (pixel_t)((sum + width) / (width + width)); return (kvz_pixel)((sum + width) / (width + width));
} }
/** /**
@ -175,11 +175,11 @@ int8_t intra_get_dir_luma_predictor(const uint32_t x, const uint32_t y, int8_t*
* \param preds output buffer for 3 predictions * \param preds output buffer for 3 predictions
* \returns (predictions are found)?1:0 * \returns (predictions are found)?1:0
*/ */
void intra_filter(pixel_t *ref, int32_t stride,int32_t width, int8_t mode) void intra_filter(kvz_pixel *ref, int32_t stride,int32_t width, int8_t mode)
{ {
#define FWIDTH (LCU_WIDTH*2+1) #define FWIDTH (LCU_WIDTH*2+1)
pixel_t filtered[FWIDTH * FWIDTH]; //!< temporary buffer for filtered samples kvz_pixel filtered[FWIDTH * FWIDTH]; //!< temporary buffer for filtered samples
pixel_t *filteredShift = &filtered[FWIDTH+1]; //!< pointer to temporary buffer with offset (1,1) kvz_pixel *filteredShift = &filtered[FWIDTH+1]; //!< pointer to temporary buffer with offset (1,1)
int x,y; int x,y;
if (!mode) { if (!mode) {
@ -222,9 +222,9 @@ void intra_filter(pixel_t *ref, int32_t stride,int32_t width, int8_t mode)
* \param recstride Stride for rec pixel arrays. * \param recstride Stride for rec pixel arrays.
* \param dst * \param dst
*/ */
void intra_get_pred(const encoder_control_t * const encoder, const pixel_t *rec, const pixel_t *rec_filtered, int recstride, pixel_t *dst, int width, int mode, int is_chroma) void intra_get_pred(const encoder_control_t * const encoder, const kvz_pixel *rec, const kvz_pixel *rec_filtered, int recstride, kvz_pixel *dst, int width, int mode, int is_chroma)
{ {
const pixel_t *ref_pixels = rec; const kvz_pixel *ref_pixels = rec;
if (is_chroma || mode == 1 || width == 4) { if (is_chroma || mode == 1 || width == 4) {
// For chroma, DC and 4x4 blocks, always use unfiltered reference. // For chroma, DC and 4x4 blocks, always use unfiltered reference.
} else if (mode == 0) { } else if (mode == 0) {
@ -244,7 +244,7 @@ void intra_get_pred(const encoder_control_t * const encoder, const pixel_t *rec,
intra_get_planar_pred(ref_pixels, recstride, width, dst, width); intra_get_planar_pred(ref_pixels, recstride, width, dst, width);
} else if (mode == 1) { } else if (mode == 1) {
int i; int i;
pixel_t val = intra_get_dc_pred(ref_pixels, recstride, width); kvz_pixel val = intra_get_dc_pred(ref_pixels, recstride, width);
for (i = 0; i < width * width; i++) { for (i = 0; i < width * width; i++) {
dst[i] = val; dst[i] = val;
} }
@ -271,11 +271,11 @@ void intra_get_pred(const encoder_control_t * const encoder, const pixel_t *rec,
* \param chroma chroma-block flag * \param chroma chroma-block flag
*/ */
void intra_recon(const encoder_control_t * const encoder, pixel_t* rec, int32_t recstride, uint32_t width, pixel_t* dst, int32_t dststride, int8_t mode, int8_t chroma) void intra_recon(const encoder_control_t * const encoder, kvz_pixel* rec, int32_t recstride, uint32_t width, kvz_pixel* dst, int32_t dststride, int8_t mode, int8_t chroma)
{ {
pixel_t pred[LCU_WIDTH * LCU_WIDTH]; kvz_pixel pred[LCU_WIDTH * LCU_WIDTH];
pixel_t rec_filtered_temp[(LCU_WIDTH * 2 + 8) * (LCU_WIDTH * 2 + 8) + 1]; kvz_pixel rec_filtered_temp[(LCU_WIDTH * 2 + 8) * (LCU_WIDTH * 2 + 8) + 1];
pixel_t *recf = &rec_filtered_temp[recstride + 1]; kvz_pixel *recf = &rec_filtered_temp[recstride + 1];
// Generate filtered reference pixels. // Generate filtered reference pixels.
{ {
@ -304,7 +304,7 @@ void intra_recon(const encoder_control_t * const encoder, pixel_t* rec, int32_t
* edge pixels filled with the reconstructed pixels. * edge pixels filled with the reconstructed pixels.
*/ */
void intra_build_reference_border(const encoder_control_t * const encoder, int32_t x_luma, int32_t y_luma, int16_t out_width, void intra_build_reference_border(const encoder_control_t * const encoder, int32_t x_luma, int32_t y_luma, int16_t out_width,
pixel_t *dst, int32_t dst_stride, int8_t chroma, kvz_pixel *dst, int32_t dst_stride, int8_t chroma,
int32_t pic_width, int32_t pic_height, int32_t pic_width, int32_t pic_height,
lcu_t *lcu) lcu_t *lcu)
{ {
@ -362,7 +362,7 @@ void intra_build_reference_border(const encoder_control_t * const encoder, int32
{ 64, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4 } { 64, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4 }
}; };
const pixel_t dc_val = 1 << (encoder->bitdepth - 1); const kvz_pixel dc_val = 1 << (encoder->bitdepth - 1);
const int is_chroma = chroma ? 1 : 0; const int is_chroma = chroma ? 1 : 0;
// input picture pointer // input picture pointer
@ -377,12 +377,12 @@ void intra_build_reference_border(const encoder_control_t * const encoder, int32
int x_local = (x_luma&0x3f)>>is_chroma, y_local = (y_luma&0x3f)>>is_chroma; int x_local = (x_luma&0x3f)>>is_chroma, y_local = (y_luma&0x3f)>>is_chroma;
pixel_t *left_ref = !chroma ? &lcu->left_ref.y[1] : (chroma == 1) ? &lcu->left_ref.u[1] : &lcu->left_ref.v[1]; kvz_pixel *left_ref = !chroma ? &lcu->left_ref.y[1] : (chroma == 1) ? &lcu->left_ref.u[1] : &lcu->left_ref.v[1];
pixel_t *top_ref = !chroma ? &lcu->top_ref.y[1] : (chroma == 1) ? &lcu->top_ref.u[1] : &lcu->top_ref.v[1]; kvz_pixel *top_ref = !chroma ? &lcu->top_ref.y[1] : (chroma == 1) ? &lcu->top_ref.u[1] : &lcu->top_ref.v[1];
pixel_t *rec_ref = !chroma ? lcu->rec.y : (chroma == 1) ? lcu->rec.u : lcu->rec.v; kvz_pixel *rec_ref = !chroma ? lcu->rec.y : (chroma == 1) ? lcu->rec.u : lcu->rec.v;
pixel_t *left_border = &left_ref[y_local]; kvz_pixel *left_border = &left_ref[y_local];
pixel_t *top_border = &top_ref[x_local]; kvz_pixel *top_border = &top_ref[x_local];
uint32_t left_stride = 1; uint32_t left_stride = 1;
if(x_local) { if(x_local) {
@ -399,7 +399,7 @@ void intra_build_reference_border(const encoder_control_t * const encoder, int32
// Get the number of reference pixels based on the PU coordinate within the LCU. // Get the number of reference pixels based on the PU coordinate within the LCU.
int num_ref_pixels = num_ref_pixels_left[y_in_lcu / 4][x_in_lcu / 4] >> is_chroma; int num_ref_pixels = num_ref_pixels_left[y_in_lcu / 4][x_in_lcu / 4] >> is_chroma;
int i; int i;
pixel_t nearest_pixel; kvz_pixel nearest_pixel;
// Max pixel we can copy from src is yy + outwidth - 1 because the dst // Max pixel we can copy from src is yy + outwidth - 1 because the dst
// extends one pixel to the left. // extends one pixel to the left.
@ -421,7 +421,7 @@ void intra_build_reference_border(const encoder_control_t * const encoder, int32
} }
} else { } else {
// If we are on the left edge, extend the first pixel of the top row. // If we are on the left edge, extend the first pixel of the top row.
pixel_t nearest_pixel = y > 0 ? top_border[0] : dc_val; kvz_pixel nearest_pixel = y > 0 ? top_border[0] : dc_val;
int i; int i;
for (i = 1; i < out_width - 1; i++) { for (i = 1; i < out_width - 1; i++) {
dst[i * dst_stride] = nearest_pixel; dst[i * dst_stride] = nearest_pixel;
@ -433,7 +433,7 @@ void intra_build_reference_border(const encoder_control_t * const encoder, int32
// Get the number of reference pixels based on the PU coordinate within the LCU. // Get the number of reference pixels based on the PU coordinate within the LCU.
int num_ref_pixels = num_ref_pixels_top[y_in_lcu / 4][x_in_lcu / 4] >> is_chroma; int num_ref_pixels = num_ref_pixels_top[y_in_lcu / 4][x_in_lcu / 4] >> is_chroma;
int i; int i;
pixel_t nearest_pixel; kvz_pixel nearest_pixel;
// Max pixel we can copy from src is yy + outwidth - 1 because the dst // Max pixel we can copy from src is yy + outwidth - 1 because the dst
// extends one pixel to the left. // extends one pixel to the left.
@ -452,7 +452,7 @@ void intra_build_reference_border(const encoder_control_t * const encoder, int32
} }
} else { } else {
// Extend nearest pixel. // Extend nearest pixel.
pixel_t nearest_pixel = x > 0 ? left_border[0] : dc_val; kvz_pixel nearest_pixel = x > 0 ? left_border[0] : dc_val;
int i; int i;
for(i = 1; i < out_width; i++) for(i = 1; i < out_width; i++)
{ {
@ -484,7 +484,7 @@ const int32_t inv_ang_table[9] = {0, 4096, 1638, 910, 630, 482, 390, 315, 256};
* \brief this functions constructs the angular intra prediction from border samples * \brief this functions constructs the angular intra prediction from border samples
* *
*/ */
void intra_get_angular_pred(const encoder_control_t * const encoder, const pixel_t* src, int32_t src_stride, pixel_t* dst, int32_t dst_stride, int32_t width, int32_t dir_mode, int8_t filter) void intra_get_angular_pred(const encoder_control_t * const encoder, const kvz_pixel* src, int32_t src_stride, kvz_pixel* dst, int32_t dst_stride, int32_t width, int32_t dir_mode, int8_t filter)
{ {
int32_t k,l; int32_t k,l;
int32_t blk_size = width; int32_t blk_size = width;
@ -499,10 +499,10 @@ void intra_get_angular_pred(const encoder_control_t * const encoder, const pixel
int32_t inv_angle = inv_ang_table[abs_ang]; int32_t inv_angle = inv_ang_table[abs_ang];
// Do angular predictions // Do angular predictions
pixel_t *ref_main; kvz_pixel *ref_main;
pixel_t *ref_side; kvz_pixel *ref_side;
pixel_t ref_above[2 * LCU_WIDTH + 1]; kvz_pixel ref_above[2 * LCU_WIDTH + 1];
pixel_t ref_left[2 * LCU_WIDTH + 1]; kvz_pixel ref_left[2 * LCU_WIDTH + 1];
// Tell clang-analyzer that everything is ok. // Tell clang-analyzer that everything is ok.
assert(width == 4 || width == 8 || width == 16 || width == 32); assert(width == 4 || width == 8 || width == 16 || width == 32);
@ -564,7 +564,7 @@ void intra_get_angular_pred(const encoder_control_t * const encoder, const pixel
// Do linear filtering // Do linear filtering
for (l = 0; l < blk_size; l++) { for (l = 0; l < blk_size; l++) {
ref_main_index = l + delta_int + 1; ref_main_index = l + delta_int + 1;
dst[k * dst_stride + l] = (pixel_t) ( (minus_delta_fract * ref_main[ref_main_index] dst[k * dst_stride + l] = (kvz_pixel) ( (minus_delta_fract * ref_main[ref_main_index]
+ delta_fract * ref_main[ref_main_index + 1] + 16) >> 5); + delta_fract * ref_main[ref_main_index + 1] + 16) >> 5);
} }
} else { } else {
@ -578,7 +578,7 @@ void intra_get_angular_pred(const encoder_control_t * const encoder, const pixel
// Flip the block if this is the horizontal mode // Flip the block if this is the horizontal mode
if (!mode_ver) { if (!mode_ver) {
pixel_t tmp; kvz_pixel tmp;
for (k=0;k<blk_size-1;k++) { for (k=0;k<blk_size-1;k++) {
for (l=k+1;l<blk_size;l++) { for (l=k+1;l<blk_size;l++) {
tmp = dst[k * dst_stride + l]; tmp = dst[k * dst_stride + l];
@ -592,7 +592,7 @@ void intra_get_angular_pred(const encoder_control_t * const encoder, const pixel
void intra_dc_pred_filtering(const pixel_t *src, int32_t src_stride, pixel_t *dst, int32_t dst_stride, int32_t width, int32_t height ) void intra_dc_pred_filtering(const kvz_pixel *src, int32_t src_stride, kvz_pixel *dst, int32_t dst_stride, int32_t width, int32_t height )
{ {
int32_t x, y, dst_stride2, src_stride2; int32_t x, y, dst_stride2, src_stride2;
@ -619,7 +619,7 @@ void intra_dc_pred_filtering(const pixel_t *src, int32_t src_stride, pixel_t *ds
This function derives the prediction samples for planar mode (intra coding). This function derives the prediction samples for planar mode (intra coding).
*/ */
void intra_get_planar_pred(const pixel_t* src, int32_t srcstride, uint32_t width, pixel_t* dst, int32_t dststride) void intra_get_planar_pred(const kvz_pixel* src, int32_t srcstride, uint32_t width, kvz_pixel* dst, int32_t dststride)
{ {
int32_t k, l, bottom_left, top_right; int32_t k, l, bottom_left, top_right;
int32_t hor_pred; int32_t hor_pred;
@ -651,7 +651,7 @@ void intra_get_planar_pred(const pixel_t* src, int32_t srcstride, uint32_t width
for (l = 0; l < (int32_t)blk_size; l++) { for (l = 0; l < (int32_t)blk_size; l++) {
hor_pred += right_column[k]; hor_pred += right_column[k];
top_row[l] += bottom_row[l]; top_row[l] += bottom_row[l];
dst[k * dststride + l] = (pixel_t)((hor_pred + top_row[l]) >> shift_2d); dst[k * dststride + l] = (kvz_pixel)((hor_pred + top_row[l]) >> shift_2d);
} }
} }
} }
@ -689,10 +689,10 @@ void intra_recon_lcu_luma(encoder_state_t * const state, int x, int y, int depth
const uint32_t pic_height = state->tile->frame->height; const uint32_t pic_height = state->tile->frame->height;
// Pointers to reconstruction arrays // Pointers to reconstruction arrays
pixel_t *recbase_y = &lcu->rec.y[lcu_px.x + lcu_px.y * LCU_WIDTH]; kvz_pixel *recbase_y = &lcu->rec.y[lcu_px.x + lcu_px.y * LCU_WIDTH];
pixel_t rec[(LCU_WIDTH*2+8)*(LCU_WIDTH*2+8)]; kvz_pixel rec[(LCU_WIDTH*2+8)*(LCU_WIDTH*2+8)];
pixel_t *rec_shift = &rec[width * 2 + 8 + 1]; kvz_pixel *rec_shift = &rec[width * 2 + 8 + 1];
int32_t rec_stride = LCU_WIDTH; int32_t rec_stride = LCU_WIDTH;
@ -744,16 +744,16 @@ void intra_recon_lcu_chroma(encoder_state_t * const state, int x, int y, int dep
const uint32_t pic_height = state->tile->frame->height; const uint32_t pic_height = state->tile->frame->height;
// Pointers to reconstruction arrays // Pointers to reconstruction arrays
pixel_t *recbase_u = &lcu->rec.u[lcu_px.x/2 + (lcu_px.y * LCU_WIDTH)/4]; kvz_pixel *recbase_u = &lcu->rec.u[lcu_px.x/2 + (lcu_px.y * LCU_WIDTH)/4];
pixel_t *recbase_v = &lcu->rec.v[lcu_px.x/2 + (lcu_px.y * LCU_WIDTH)/4]; kvz_pixel *recbase_v = &lcu->rec.v[lcu_px.x/2 + (lcu_px.y * LCU_WIDTH)/4];
pixel_t rec[(LCU_WIDTH*2+8)*(LCU_WIDTH*2+8)]; kvz_pixel rec[(LCU_WIDTH*2+8)*(LCU_WIDTH*2+8)];
int32_t rec_stride = LCU_WIDTH; int32_t rec_stride = LCU_WIDTH;
// Reconstruct chroma. // Reconstruct chroma.
if (!(x & 4 || y & 4)) { if (!(x & 4 || y & 4)) {
pixel_t *rec_shift_c = &rec[width_c * 2 + 8 + 1]; kvz_pixel *rec_shift_c = &rec[width_c * 2 + 8 + 1];
intra_build_reference_border(encoder, x, y,(int16_t)width_c * 2 + 8, rec, (int16_t)width_c * 2 + 8, 1, intra_build_reference_border(encoder, x, y,(int16_t)width_c * 2 + 8, rec, (int16_t)width_c * 2 + 8, 1,
pic_width/2, pic_height/2, lcu); pic_width/2, pic_height/2, lcu);
intra_recon(encoder, intra_recon(encoder,

16
src/intra.h
View file

@ -35,19 +35,19 @@
int8_t intra_get_dir_luma_predictor(uint32_t x, uint32_t y, int8_t* preds, int8_t intra_get_dir_luma_predictor(uint32_t x, uint32_t y, int8_t* preds,
const cu_info_t* cur_cu, const cu_info_t* left_cu, const cu_info_t* above_cu); const cu_info_t* cur_cu, const cu_info_t* left_cu, const cu_info_t* above_cu);
void intra_dc_pred_filtering(const pixel_t* src, int32_t src_stride, pixel_t* dst, int32_t dst_stride, int32_t width, int32_t height ); void intra_dc_pred_filtering(const kvz_pixel* src, int32_t src_stride, kvz_pixel* dst, int32_t dst_stride, int32_t width, int32_t height );
void intra_build_reference_border(const encoder_control_t *encoder, int32_t x_luma, int32_t y_luma, int16_t out_width, pixel_t *dst, int32_t dst_stride, int8_t chroma, int32_t pic_width, int32_t pic_height, lcu_t *lcu); void intra_build_reference_border(const encoder_control_t *encoder, int32_t x_luma, int32_t y_luma, int16_t out_width, kvz_pixel *dst, int32_t dst_stride, int8_t chroma, int32_t pic_width, int32_t pic_height, lcu_t *lcu);
void intra_filter(pixel_t* ref, int32_t stride, int32_t width, int8_t mode); void intra_filter(kvz_pixel* ref, int32_t stride, int32_t width, int8_t mode);
/* Predictions */ /* Predictions */
void intra_get_pred(const encoder_control_t * const encoder, const pixel_t *rec, const pixel_t *rec_filtered, int recstride, pixel_t *dst, int width, int mode, int is_chroma); void intra_get_pred(const encoder_control_t * const encoder, const kvz_pixel *rec, const kvz_pixel *rec_filtered, int recstride, kvz_pixel *dst, int width, int mode, int is_chroma);
pixel_t intra_get_dc_pred(const pixel_t* pic, uint16_t pic_width, uint8_t width); kvz_pixel intra_get_dc_pred(const kvz_pixel* pic, uint16_t pic_width, uint8_t width);
void intra_get_planar_pred(const pixel_t* src,int32_t srcstride, uint32_t width, pixel_t* dst, int32_t dststride); void intra_get_planar_pred(const kvz_pixel* src,int32_t srcstride, uint32_t width, kvz_pixel* dst, int32_t dststride);
void intra_get_angular_pred(const encoder_control_t *encoder, const pixel_t* src, int32_t src_stride, pixel_t* dst, int32_t dst_stride, int32_t width, int32_t dir_mode, int8_t filter); void intra_get_angular_pred(const encoder_control_t *encoder, const kvz_pixel* src, int32_t src_stride, kvz_pixel* dst, int32_t dst_stride, int32_t width, int32_t dir_mode, int8_t filter);
void intra_recon(const encoder_control_t *encoder, pixel_t* rec, int32_t rec_stride, uint32_t width, pixel_t* dst, int32_t dst_stride, int8_t mode, int8_t chroma); void intra_recon(const encoder_control_t *encoder, kvz_pixel* rec, int32_t rec_stride, uint32_t width, kvz_pixel* dst, int32_t dst_stride, int8_t mode, int8_t chroma);
void intra_recon_lcu_luma(encoder_state_t *state, int x, int y, int depth, int8_t intra_mode, cu_info_t *cur_cu, lcu_t *lcu); void intra_recon_lcu_luma(encoder_state_t *state, int x, int y, int depth, int8_t intra_mode, cu_info_t *cur_cu, lcu_t *lcu);
void intra_recon_lcu_chroma(encoder_state_t *state, int x, int y, int depth, int8_t intra_mode, cu_info_t *cur_cu, lcu_t *lcu); void intra_recon_lcu_chroma(encoder_state_t *state, int x, int y, int depth, int8_t intra_mode, cu_info_t *cur_cu, lcu_t *lcu);

14
src/kvazaar.h
View file

@ -41,9 +41,9 @@ extern "C" {
#define KVZ_BIT_DEPTH 8 #define KVZ_BIT_DEPTH 8
#if KVZ_BIT_DEPTH == 8 #if KVZ_BIT_DEPTH == 8
typedef uint8_t pixel_t; typedef uint8_t kvz_pixel;
#else #else
typedef uint16_t pixel_t; typedef uint16_t kvz_pixel;
#endif #endif
/** /**
@ -140,12 +140,12 @@ typedef struct bitstream_chunk_t kvz_payload;
* \brief Struct which contains all picture data * \brief Struct which contains all picture data
*/ */
typedef struct kvz_picture { typedef struct kvz_picture {
pixel_t *fulldata; //!< \brief Allocated buffer (only used in the base_image) kvz_pixel *fulldata; //!< \brief Allocated buffer (only used in the base_image)
pixel_t *y; //!< \brief Pointer to luma pixel array. kvz_pixel *y; //!< \brief Pointer to luma pixel array.
pixel_t *u; //!< \brief Pointer to chroma U pixel array. kvz_pixel *u; //!< \brief Pointer to chroma U pixel array.
pixel_t *v; //!< \brief Pointer to chroma V pixel array. kvz_pixel *v; //!< \brief Pointer to chroma V pixel array.
pixel_t *data[3]; //!< \brief Alternate access method to same data. kvz_pixel *data[3]; //!< \brief Alternate access method to same data.
int32_t width; //!< \brief Luma pixel array width. int32_t width; //!< \brief Luma pixel array width.
int32_t height; //!< \brief Luma pixel array height. int32_t height; //!< \brief Luma pixel array height.

2
src/rdo.c
View file

@ -170,7 +170,7 @@ int intra_rdo_cost_compare(uint32_t *rdo_costs,int8_t rdo_modes_to_check, uint32
** Only for luma ** Only for luma
*/ */
uint32_t rdo_cost_intra(encoder_state_t * const state, pixel_t *pred, pixel_t *orig_block, int width, int8_t mode, int tr_depth) uint32_t rdo_cost_intra(encoder_state_t * const state, kvz_pixel *pred, kvz_pixel *orig_block, int width, int8_t mode, int tr_depth)
{ {
const encoder_control_t * const encoder = state->encoder_control; const encoder_control_t * const encoder = state->encoder_control;
coeff_t pre_quant_coeff[LCU_WIDTH*LCU_WIDTH>>2]; coeff_t pre_quant_coeff[LCU_WIDTH*LCU_WIDTH>>2];

2
src/rdo.h
View file

@ -39,7 +39,7 @@ int intra_rdo_cost_compare(uint32_t *rdo_costs,int8_t rdo_modes_to_check, uint32
void rdoq(encoder_state_t *state, coeff_t *coef, coeff_t *dest_coeff, int32_t width, void rdoq(encoder_state_t *state, coeff_t *coef, coeff_t *dest_coeff, int32_t width,
int32_t height, int8_t type, int8_t scan_mode, int8_t block_type, int8_t tr_depth); int32_t height, int8_t type, int8_t scan_mode, int8_t block_type, int8_t tr_depth);
uint32_t rdo_cost_intra(encoder_state_t *state, pixel_t* pred, pixel_t* orig_block, int width, int8_t mode, int tr_depth); uint32_t rdo_cost_intra(encoder_state_t *state, kvz_pixel* pred, kvz_pixel* orig_block, int width, int8_t mode, int tr_depth);
int32_t get_coeff_cost(const encoder_state_t *state, coeff_t *coeff, int32_t width, int32_t type, int8_t scan_mode); int32_t get_coeff_cost(const encoder_state_t *state, coeff_t *coeff, int32_t width, int32_t type, int8_t scan_mode);

88
src/sao.c
View file

@ -44,7 +44,7 @@ static const vector2d_t g_sao_edge_offsets[SAO_NUM_EO][2] = {
// Mapping of edge_idx values to eo-classes. // Mapping of edge_idx values to eo-classes.
static int sao_calc_eo_cat(pixel_t a, pixel_t b, pixel_t c) static int sao_calc_eo_cat(kvz_pixel a, kvz_pixel b, kvz_pixel c)
{ {
// Mapping relationships between a, b and c to eo_idx. // Mapping relationships between a, b and c to eo_idx.
static const int sao_eo_idx_to_eo_category[] = { 1, 2, 0, 3, 4 }; static const int sao_eo_idx_to_eo_category[] = { 1, 2, 0, 3, 4 };
@ -55,7 +55,7 @@ static int sao_calc_eo_cat(pixel_t a, pixel_t b, pixel_t c)
} }
int sao_band_ddistortion(const encoder_state_t * const state, const pixel_t *orig_data, const pixel_t *rec_data, int sao_band_ddistortion(const encoder_state_t * const state, const kvz_pixel *orig_data, const kvz_pixel *rec_data,
int block_width, int block_height, int block_width, int block_height,
int band_pos, int sao_bands[4]) int band_pos, int sao_bands[4])
{ {
@ -82,7 +82,7 @@ int sao_band_ddistortion(const encoder_state_t * const state, const pixel_t *ori
} }
int sao_edge_ddistortion(const pixel_t *orig_data, const pixel_t *rec_data, int sao_edge_ddistortion(const kvz_pixel *orig_data, const kvz_pixel *rec_data,
int block_width, int block_height, int block_width, int block_height,
int eo_class, int offsets[NUM_SAO_EDGE_CATEGORIES]) int eo_class, int offsets[NUM_SAO_EDGE_CATEGORIES])
{ {
@ -93,10 +93,10 @@ int sao_edge_ddistortion(const pixel_t *orig_data, const pixel_t *rec_data,
for (y = 1; y < block_height - 1; ++y) { for (y = 1; y < block_height - 1; ++y) {
for (x = 1; x < block_width - 1; ++x) { for (x = 1; x < block_width - 1; ++x) {
const pixel_t *c_data = &rec_data[y * block_width + x]; const kvz_pixel *c_data = &rec_data[y * block_width + x];
pixel_t a = c_data[a_ofs.y * block_width + a_ofs.x]; kvz_pixel a = c_data[a_ofs.y * block_width + a_ofs.x];
pixel_t c = c_data[0]; kvz_pixel c = c_data[0];
pixel_t b = c_data[b_ofs.y * block_width + b_ofs.x]; kvz_pixel b = c_data[b_ofs.y * block_width + b_ofs.x];
int offset = offsets[sao_calc_eo_cat(a, b, c)]; int offset = offsets[sao_calc_eo_cat(a, b, c)];
@ -326,7 +326,7 @@ static int calc_sao_band_offsets(int sao_bands[2][32], int offsets[4],
* \param rec_data Reconstructed pixel data. 64x64 for luma, 32x32 for chroma. * \param rec_data Reconstructed pixel data. 64x64 for luma, 32x32 for chroma.
* \param sao_bands an array of bands for original and reconstructed block * \param sao_bands an array of bands for original and reconstructed block
*/ */
static void calc_sao_bands(const encoder_state_t * const state, const pixel_t *orig_data, const pixel_t *rec_data, static void calc_sao_bands(const encoder_state_t * const state, const kvz_pixel *orig_data, const kvz_pixel *rec_data,
int block_width, int block_height, int block_width, int block_height,
int sao_bands[2][32]) int sao_bands[2][32])
{ {
@ -349,7 +349,7 @@ static void calc_sao_bands(const encoder_state_t * const state, const pixel_t *o
* \param dir_offsets * \param dir_offsets
* \param is_chroma 0 for luma, 1 for chroma. Indicates * \param is_chroma 0 for luma, 1 for chroma. Indicates
*/ */
static void calc_sao_edge_dir(const pixel_t *orig_data, const pixel_t *rec_data, static void calc_sao_edge_dir(const kvz_pixel *orig_data, const kvz_pixel *rec_data,
int eo_class, int block_width, int block_height, int eo_class, int block_width, int block_height,
int cat_sum_cnt[2][NUM_SAO_EDGE_CATEGORIES]) int cat_sum_cnt[2][NUM_SAO_EDGE_CATEGORIES])
{ {
@ -362,10 +362,10 @@ static void calc_sao_edge_dir(const pixel_t *orig_data, const pixel_t *rec_data,
// their neighbours. // their neighbours.
for (y = 1; y < block_height - 1; ++y) { for (y = 1; y < block_height - 1; ++y) {
for (x = 1; x < block_width - 1; ++x) { for (x = 1; x < block_width - 1; ++x) {
const pixel_t *c_data = &rec_data[y * block_width + x]; const kvz_pixel *c_data = &rec_data[y * block_width + x];
pixel_t a = c_data[a_ofs.y * block_width + a_ofs.x]; kvz_pixel a = c_data[a_ofs.y * block_width + a_ofs.x];
pixel_t c = c_data[0]; kvz_pixel c = c_data[0];
pixel_t b = c_data[b_ofs.y * block_width + b_ofs.x]; kvz_pixel b = c_data[b_ofs.y * block_width + b_ofs.x];
int eo_cat = sao_calc_eo_cat(a, b, c); int eo_cat = sao_calc_eo_cat(a, b, c);
@ -376,7 +376,7 @@ static void calc_sao_edge_dir(const pixel_t *orig_data, const pixel_t *rec_data,
} }
static void sao_reconstruct_color(const encoder_control_t * const encoder, static void sao_reconstruct_color(const encoder_control_t * const encoder,
const pixel_t *rec_data, pixel_t *new_rec_data, const kvz_pixel *rec_data, kvz_pixel *new_rec_data,
const sao_info_t *sao, const sao_info_t *sao,
int stride, int new_stride, int stride, int new_stride,
int block_width, int block_height, int block_width, int block_height,
@ -401,15 +401,15 @@ static void sao_reconstruct_color(const encoder_control_t * const encoder,
for (x = 0; x < block_width; ++x) { for (x = 0; x < block_width; ++x) {
vector2d_t a_ofs = g_sao_edge_offsets[sao->eo_class][0]; vector2d_t a_ofs = g_sao_edge_offsets[sao->eo_class][0];
vector2d_t b_ofs = g_sao_edge_offsets[sao->eo_class][1]; vector2d_t b_ofs = g_sao_edge_offsets[sao->eo_class][1];
const pixel_t *c_data = &rec_data[y * stride + x]; const kvz_pixel *c_data = &rec_data[y * stride + x];
pixel_t *new_data = &new_rec_data[y * new_stride + x]; kvz_pixel *new_data = &new_rec_data[y * new_stride + x];
pixel_t a = c_data[a_ofs.y * stride + a_ofs.x]; kvz_pixel a = c_data[a_ofs.y * stride + a_ofs.x];
pixel_t c = c_data[0]; kvz_pixel c = c_data[0];
pixel_t b = c_data[b_ofs.y * stride + b_ofs.x]; kvz_pixel b = c_data[b_ofs.y * stride + b_ofs.x];
int eo_cat = sao_calc_eo_cat(a, b, c); int eo_cat = sao_calc_eo_cat(a, b, c);
new_data[0] = (pixel_t)CLIP(0, (1 << KVZ_BIT_DEPTH) - 1, c_data[0] + sao->offsets[eo_cat + offset_v]); new_data[0] = (kvz_pixel)CLIP(0, (1 << KVZ_BIT_DEPTH) - 1, c_data[0] + sao->offsets[eo_cat + offset_v]);
} }
} }
} }
@ -527,7 +527,7 @@ static void sao_calc_edge_block_dims(const videoframe_t * const frame, color_t c
rec->x = (rec->x == 0 ? 0 : -1); rec->x = (rec->x == 0 ? 0 : -1);
} }
void sao_reconstruct(const encoder_control_t * const encoder, videoframe_t * frame, const pixel_t *old_rec, void sao_reconstruct(const encoder_control_t * const encoder, videoframe_t * frame, const kvz_pixel *old_rec,
unsigned x_ctb, unsigned y_ctb, unsigned x_ctb, unsigned y_ctb,
const sao_info_t *sao, color_t color_i) const sao_info_t *sao, color_t color_i)
{ {
@ -536,12 +536,12 @@ void sao_reconstruct(const encoder_control_t * const encoder, videoframe_t * fra
const int lcu_stride = LCU_WIDTH >> is_chroma; const int lcu_stride = LCU_WIDTH >> is_chroma;
const int buf_stride = lcu_stride + 2; const int buf_stride = lcu_stride + 2;
pixel_t *recdata = frame->rec->data[color_i]; kvz_pixel *recdata = frame->rec->data[color_i];
pixel_t buf_rec[(LCU_WIDTH + 2) * (LCU_WIDTH + 2)]; kvz_pixel buf_rec[(LCU_WIDTH + 2) * (LCU_WIDTH + 2)];
pixel_t new_rec[LCU_WIDTH * LCU_WIDTH]; kvz_pixel new_rec[LCU_WIDTH * LCU_WIDTH];
// Calling CU_TO_PIXEL with depth 1 is the same as using block size of 32. // Calling CU_TO_PIXEL with depth 1 is the same as using block size of 32.
pixel_t *lcu_rec = &recdata[CU_TO_PIXEL(x_ctb, y_ctb, is_chroma, frame->rec->stride>>is_chroma)]; kvz_pixel *lcu_rec = &recdata[CU_TO_PIXEL(x_ctb, y_ctb, is_chroma, frame->rec->stride>>is_chroma)];
const pixel_t *old_lcu_rec = &old_rec[CU_TO_PIXEL(x_ctb, y_ctb, is_chroma, pic_stride)]; const kvz_pixel *old_lcu_rec = &old_rec[CU_TO_PIXEL(x_ctb, y_ctb, is_chroma, pic_stride)];
vector2d_t ofs; vector2d_t ofs;
vector2d_t tl = { 1, 1 }; vector2d_t tl = { 1, 1 };
@ -593,7 +593,7 @@ void sao_reconstruct(const encoder_control_t * const encoder, videoframe_t * fra
static void sao_search_edge_sao(const encoder_state_t * const state, static void sao_search_edge_sao(const encoder_state_t * const state,
const pixel_t * data[], const pixel_t * recdata[], const kvz_pixel * data[], const kvz_pixel * recdata[],
int block_width, int block_height, int block_width, int block_height,
unsigned buf_cnt, unsigned buf_cnt,
sao_info_t *sao_out, sao_info_t *sao_top, sao_info_t *sao_out, sao_info_t *sao_top,
@ -671,7 +671,7 @@ static void sao_search_edge_sao(const encoder_state_t * const state,
} }
static void sao_search_band_sao(const encoder_state_t * const state, const pixel_t * data[], const pixel_t * recdata[], static void sao_search_band_sao(const encoder_state_t * const state, const kvz_pixel * data[], const kvz_pixel * recdata[],
int block_width, int block_height, int block_width, int block_height,
unsigned buf_cnt, unsigned buf_cnt,
sao_info_t *sao_out, sao_info_t *sao_top, sao_info_t *sao_out, sao_info_t *sao_top,
@ -719,7 +719,7 @@ static void sao_search_band_sao(const encoder_state_t * const state, const pixel
* \param buf_cnt Number of pointers data and recdata have. * \param buf_cnt Number of pointers data and recdata have.
* \param sao_out Output parameter for the best sao parameters. * \param sao_out Output parameter for the best sao parameters.
*/ */
static void sao_search_best_mode(const encoder_state_t * const state, const pixel_t * data[], const pixel_t * recdata[], static void sao_search_best_mode(const encoder_state_t * const state, const kvz_pixel * data[], const kvz_pixel * recdata[],
int block_width, int block_height, int block_width, int block_height,
unsigned buf_cnt, unsigned buf_cnt,
sao_info_t *sao_out, sao_info_t *sao_top, sao_info_t *sao_out, sao_info_t *sao_top,
@ -831,10 +831,10 @@ void sao_search_chroma(const encoder_state_t * const state, const videoframe_t *
{ {
int block_width = (LCU_WIDTH / 2); int block_width = (LCU_WIDTH / 2);
int block_height = (LCU_WIDTH / 2); int block_height = (LCU_WIDTH / 2);
const pixel_t *orig_list[2]; const kvz_pixel *orig_list[2];
const pixel_t *rec_list[2]; const kvz_pixel *rec_list[2];
pixel_t orig[2][LCU_CHROMA_SIZE]; kvz_pixel orig[2][LCU_CHROMA_SIZE];
pixel_t rec[2][LCU_CHROMA_SIZE]; kvz_pixel rec[2][LCU_CHROMA_SIZE];
color_t color_i; color_t color_i;
// Check for right and bottom boundaries. // Check for right and bottom boundaries.
@ -849,8 +849,8 @@ void sao_search_chroma(const encoder_state_t * const state, const videoframe_t *
// Copy data to temporary buffers and init orig and rec lists to point to those buffers. // Copy data to temporary buffers and init orig and rec lists to point to those buffers.
for (color_i = COLOR_U; color_i <= COLOR_V; ++color_i) { for (color_i = COLOR_U; color_i <= COLOR_V; ++color_i) {
pixel_t *data = &frame->source->data[color_i][CU_TO_PIXEL(x_ctb, y_ctb, 1, frame->source->stride / 2)]; kvz_pixel *data = &frame->source->data[color_i][CU_TO_PIXEL(x_ctb, y_ctb, 1, frame->source->stride / 2)];
pixel_t *recdata = &frame->rec->data[color_i][CU_TO_PIXEL(x_ctb, y_ctb, 1, frame->rec->stride / 2)]; kvz_pixel *recdata = &frame->rec->data[color_i][CU_TO_PIXEL(x_ctb, y_ctb, 1, frame->rec->stride / 2)];
pixels_blit(data, orig[color_i - 1], block_width, block_height, pixels_blit(data, orig[color_i - 1], block_width, block_height,
frame->source->stride / 2, block_width); frame->source->stride / 2, block_width);
pixels_blit(recdata, rec[color_i - 1], block_width, block_height, pixels_blit(recdata, rec[color_i - 1], block_width, block_height,
@ -865,12 +865,12 @@ void sao_search_chroma(const encoder_state_t * const state, const videoframe_t *
void sao_search_luma(const encoder_state_t * const state, const videoframe_t *frame, unsigned x_ctb, unsigned y_ctb, sao_info_t *sao, sao_info_t *sao_top, sao_info_t *sao_left, int32_t merge_cost[3]) void sao_search_luma(const encoder_state_t * const state, const videoframe_t *frame, unsigned x_ctb, unsigned y_ctb, sao_info_t *sao, sao_info_t *sao_top, sao_info_t *sao_left, int32_t merge_cost[3])
{ {
pixel_t orig[LCU_LUMA_SIZE]; kvz_pixel orig[LCU_LUMA_SIZE];
pixel_t rec[LCU_LUMA_SIZE]; kvz_pixel rec[LCU_LUMA_SIZE];
const pixel_t * orig_list[1] = { NULL }; const kvz_pixel * orig_list[1] = { NULL };
const pixel_t * rec_list[1] = { NULL }; const kvz_pixel * rec_list[1] = { NULL };
pixel_t *data = &frame->source->y[CU_TO_PIXEL(x_ctb, y_ctb, 0, frame->source->stride)]; kvz_pixel *data = &frame->source->y[CU_TO_PIXEL(x_ctb, y_ctb, 0, frame->source->stride)];
pixel_t *recdata = &frame->rec->y[CU_TO_PIXEL(x_ctb, y_ctb, 0, frame->rec->stride)]; kvz_pixel *recdata = &frame->rec->y[CU_TO_PIXEL(x_ctb, y_ctb, 0, frame->rec->stride)];
int block_width = LCU_WIDTH; int block_width = LCU_WIDTH;
int block_height = LCU_WIDTH; int block_height = LCU_WIDTH;
@ -901,9 +901,9 @@ void sao_reconstruct_frame(encoder_state_t * const state)
// These are needed because SAO needs the pre-SAO pixels form left and // These are needed because SAO needs the pre-SAO pixels form left and
// top LCUs. Single pixel wide buffers, like what search_lcu takes, would // top LCUs. Single pixel wide buffers, like what search_lcu takes, would
// be enough though. // be enough though.
pixel_t *new_y_data = MALLOC(pixel_t, frame->rec->width * frame->rec->height); kvz_pixel *new_y_data = MALLOC(kvz_pixel, frame->rec->width * frame->rec->height);
pixel_t *new_u_data = MALLOC(pixel_t, (frame->rec->width * frame->rec->height) >> 2); kvz_pixel *new_u_data = MALLOC(kvz_pixel, (frame->rec->width * frame->rec->height) >> 2);
pixel_t *new_v_data = MALLOC(pixel_t, (frame->rec->width * frame->rec->height) >> 2); kvz_pixel *new_v_data = MALLOC(kvz_pixel, (frame->rec->width * frame->rec->height) >> 2);
pixels_blit(frame->rec->y, new_y_data, frame->rec->width, frame->rec->height, frame->rec->stride, frame->rec->width); pixels_blit(frame->rec->y, new_y_data, frame->rec->width, frame->rec->height, frame->rec->stride, frame->rec->width);
pixels_blit(frame->rec->u, new_u_data, frame->rec->width/2, frame->rec->height/2, frame->rec->stride/2, frame->rec->width/2); pixels_blit(frame->rec->u, new_u_data, frame->rec->width/2, frame->rec->height/2, frame->rec->stride/2, frame->rec->width/2);

2
src/sao.h
View file

@ -58,7 +58,7 @@ typedef struct sao_info_t {
void init_sao_info(sao_info_t *sao); void init_sao_info(sao_info_t *sao);
void sao_search_chroma(const encoder_state_t * state, const videoframe_t *frame, unsigned x_ctb, unsigned y_ctb, sao_info_t *sao, sao_info_t *sao_top, sao_info_t *sao_left, int32_t merge_cost[3]); void sao_search_chroma(const encoder_state_t * state, const videoframe_t *frame, unsigned x_ctb, unsigned y_ctb, sao_info_t *sao, sao_info_t *sao_top, sao_info_t *sao_left, int32_t merge_cost[3]);
void sao_search_luma(const encoder_state_t * state, const videoframe_t *frame, unsigned x_ctb, unsigned y_ctb, sao_info_t *sao, sao_info_t *sao_top, sao_info_t *sao_left, int32_t merge_cost[3]); void sao_search_luma(const encoder_state_t * state, const videoframe_t *frame, unsigned x_ctb, unsigned y_ctb, sao_info_t *sao, sao_info_t *sao_top, sao_info_t *sao_left, int32_t merge_cost[3]);
void sao_reconstruct(const encoder_control_t * encoder, videoframe_t *frame, const pixel_t *old_rec, void sao_reconstruct(const encoder_control_t * encoder, videoframe_t *frame, const kvz_pixel *old_rec,
unsigned x_ctb, unsigned y_ctb, unsigned x_ctb, unsigned y_ctb,
const sao_info_t *sao, color_t color_i); const sao_info_t *sao, color_t color_i);
void sao_reconstruct_frame(encoder_state_t *state); void sao_reconstruct_frame(encoder_state_t *state);

82
src/search.c
View file

@ -860,13 +860,13 @@ static unsigned search_frac(const encoder_state_t * const state,
//create buffer for block + extra for filter //create buffer for block + extra for filter
int src_stride = block_width+FILTER_SIZE+1; int src_stride = block_width+FILTER_SIZE+1;
pixel_t src[(LCU_WIDTH+FILTER_SIZE+1) * (LCU_WIDTH+FILTER_SIZE+1)]; kvz_pixel src[(LCU_WIDTH+FILTER_SIZE+1) * (LCU_WIDTH+FILTER_SIZE+1)];
pixel_t* src_off = &src[HALF_FILTER+HALF_FILTER*(block_width+FILTER_SIZE+1)]; kvz_pixel* src_off = &src[HALF_FILTER+HALF_FILTER*(block_width+FILTER_SIZE+1)];
//destination buffer for interpolation //destination buffer for interpolation
int dst_stride = (block_width+1)*4; int dst_stride = (block_width+1)*4;
pixel_t dst[(LCU_WIDTH+1) * (LCU_WIDTH+1) * 16]; kvz_pixel dst[(LCU_WIDTH+1) * (LCU_WIDTH+1) * 16];
pixel_t* dst_off = &dst[dst_stride*4+4]; kvz_pixel* dst_off = &dst[dst_stride*4+4];
extend_borders(orig->x, orig->y, mv.x-1, mv.y-1, extend_borders(orig->x, orig->y, mv.x-1, mv.y-1,
state->tile->lcu_offset_x * LCU_WIDTH, state->tile->lcu_offset_x * LCU_WIDTH,
@ -885,8 +885,8 @@ static unsigned search_frac(const encoder_state_t * const state,
for (i = 0; i < 9; ++i) { for (i = 0; i < 9; ++i) {
const vector2d_t *pattern = &square[i]; const vector2d_t *pattern = &square[i];
pixel_t tmp_filtered[LCU_WIDTH*LCU_WIDTH]; kvz_pixel tmp_filtered[LCU_WIDTH*LCU_WIDTH];
pixel_t tmp_pic[LCU_WIDTH*LCU_WIDTH]; kvz_pixel tmp_pic[LCU_WIDTH*LCU_WIDTH];
int y,x; int y,x;
for(y = 0; y < block_width; ++y) { for(y = 0; y < block_width; ++y) {
@ -925,8 +925,8 @@ static unsigned search_frac(const encoder_state_t * const state,
for (i = 0; i < 9; ++i) { for (i = 0; i < 9; ++i) {
const vector2d_t *pattern = &square[i]; const vector2d_t *pattern = &square[i];
pixel_t tmp_filtered[LCU_WIDTH*LCU_WIDTH]; kvz_pixel tmp_filtered[LCU_WIDTH*LCU_WIDTH];
pixel_t tmp_pic[LCU_WIDTH*LCU_WIDTH]; kvz_pixel tmp_pic[LCU_WIDTH*LCU_WIDTH];
int y,x; int y,x;
for(y = 0; y < block_width; ++y) { for(y = 0; y < block_width; ++y) {
@ -1123,8 +1123,8 @@ static int search_cu_inter(const encoder_state_t * const state, int x, int y, in
uint32_t cost = 0; uint32_t cost = 0;
int8_t cu_mv_cand = 0; int8_t cu_mv_cand = 0;
int16_t mv[2][2]; int16_t mv[2][2];
pixel_t tmp_block[64 * 64]; kvz_pixel tmp_block[64 * 64];
pixel_t tmp_pic[64 * 64]; kvz_pixel tmp_pic[64 * 64];
// Force L0 and L1 references // Force L0 and L1 references
if (state->global->refmap[merge_cand[i].ref[0]].list == 2 || state->global->refmap[merge_cand[j].ref[1]].list == 1) continue; if (state->global->refmap[merge_cand[i].ref[0]].list == 2 || state->global->refmap[merge_cand[j].ref[1]].list == 1) continue;
@ -1636,9 +1636,9 @@ static double search_intra_trdepth(encoder_state_t * const state,
const bool reconstruct_chroma = !(x_px & 4 || y_px & 4); const bool reconstruct_chroma = !(x_px & 4 || y_px & 4);
struct { struct {
pixel_t y[TR_MAX_WIDTH*TR_MAX_WIDTH]; kvz_pixel y[TR_MAX_WIDTH*TR_MAX_WIDTH];
pixel_t u[TR_MAX_WIDTH*TR_MAX_WIDTH]; kvz_pixel u[TR_MAX_WIDTH*TR_MAX_WIDTH];
pixel_t v[TR_MAX_WIDTH*TR_MAX_WIDTH]; kvz_pixel v[TR_MAX_WIDTH*TR_MAX_WIDTH];
} nosplit_pixels; } nosplit_pixels;
cu_cbf_t nosplit_cbf; cu_cbf_t nosplit_cbf;
@ -1885,7 +1885,7 @@ static INLINE uint8_t select_best_mode_index(const int8_t *modes, const double *
* coefficients of the residual. * coefficients of the residual.
*/ */
static double get_cost(encoder_state_t * const state, static double get_cost(encoder_state_t * const state,
pixel_t *pred, pixel_t *orig_block, kvz_pixel *pred, kvz_pixel *orig_block,
cost_pixel_nxn_func *satd_func, cost_pixel_nxn_func *satd_func,
cost_pixel_nxn_func *sad_func, cost_pixel_nxn_func *sad_func,
int width) int width)
@ -1914,8 +1914,8 @@ static double get_cost(encoder_state_t * const state,
static void search_intra_chroma_rough(encoder_state_t * const state, static void search_intra_chroma_rough(encoder_state_t * const state,
int x_px, int y_px, int depth, int x_px, int y_px, int depth,
const pixel_t *orig_u, const pixel_t *orig_v, int16_t origstride, const kvz_pixel *orig_u, const kvz_pixel *orig_v, int16_t origstride,
const pixel_t *rec_u, const pixel_t *rec_v, int16_t recstride, const kvz_pixel *rec_u, const kvz_pixel *rec_v, int16_t recstride,
int8_t luma_mode, int8_t luma_mode,
int8_t modes[5], double costs[5]) int8_t modes[5], double costs[5])
{ {
@ -1931,11 +1931,11 @@ static void search_intra_chroma_rough(encoder_state_t * const state,
cost_pixel_nxn_func *const satd_func = pixels_get_satd_func(width); cost_pixel_nxn_func *const satd_func = pixels_get_satd_func(width);
//cost_pixel_nxn_func *const sad_func = pixels_get_sad_func(width); //cost_pixel_nxn_func *const sad_func = pixels_get_sad_func(width);
pixel_t _pred[LCU_WIDTH * LCU_WIDTH + 1 + SIMD_ALIGNMENT]; kvz_pixel _pred[LCU_WIDTH * LCU_WIDTH + 1 + SIMD_ALIGNMENT];
pixel_t *pred = ALIGNED_POINTER(_pred, SIMD_ALIGNMENT); kvz_pixel *pred = ALIGNED_POINTER(_pred, SIMD_ALIGNMENT);
pixel_t _orig_block[LCU_WIDTH * LCU_WIDTH + 1 + SIMD_ALIGNMENT]; kvz_pixel _orig_block[LCU_WIDTH * LCU_WIDTH + 1 + SIMD_ALIGNMENT];
pixel_t *orig_block = ALIGNED_POINTER(_orig_block, SIMD_ALIGNMENT); kvz_pixel *orig_block = ALIGNED_POINTER(_orig_block, SIMD_ALIGNMENT);
pixels_blit(orig_u, orig_block, width, width, origstride, width); pixels_blit(orig_u, orig_block, width, width, origstride, width);
for (int i = 0; i < 5; ++i) { for (int i = 0; i < 5; ++i) {
@ -1986,8 +1986,8 @@ static void search_intra_chroma_rough(encoder_state_t * const state,
* \return Number of prediction modes in param modes. * \return Number of prediction modes in param modes.
*/ */
static int8_t search_intra_rough(encoder_state_t * const state, static int8_t search_intra_rough(encoder_state_t * const state,
pixel_t *orig, int32_t origstride, kvz_pixel *orig, int32_t origstride,
pixel_t *rec, int16_t recstride, kvz_pixel *rec, int16_t recstride,
int width, int8_t *intra_preds, int width, int8_t *intra_preds,
int8_t modes[35], double costs[35]) int8_t modes[35], double costs[35])
{ {
@ -1995,15 +1995,15 @@ static int8_t search_intra_rough(encoder_state_t * const state,
cost_pixel_nxn_func *sad_func = pixels_get_sad_func(width); cost_pixel_nxn_func *sad_func = pixels_get_sad_func(width);
// Temporary block arrays // Temporary block arrays
pixel_t _pred[LCU_WIDTH * LCU_WIDTH + 1 + SIMD_ALIGNMENT]; kvz_pixel _pred[LCU_WIDTH * LCU_WIDTH + 1 + SIMD_ALIGNMENT];
pixel_t *pred = ALIGNED_POINTER(_pred, SIMD_ALIGNMENT); kvz_pixel *pred = ALIGNED_POINTER(_pred, SIMD_ALIGNMENT);
pixel_t _orig_block[LCU_WIDTH * LCU_WIDTH + 1 + SIMD_ALIGNMENT]; kvz_pixel _orig_block[LCU_WIDTH * LCU_WIDTH + 1 + SIMD_ALIGNMENT];
pixel_t *orig_block = ALIGNED_POINTER(_orig_block, SIMD_ALIGNMENT); kvz_pixel *orig_block = ALIGNED_POINTER(_orig_block, SIMD_ALIGNMENT);
pixel_t rec_filtered_temp[(LCU_WIDTH * 2 + 8) * (LCU_WIDTH * 2 + 8) + 1]; kvz_pixel rec_filtered_temp[(LCU_WIDTH * 2 + 8) * (LCU_WIDTH * 2 + 8) + 1];
pixel_t *recf = &rec_filtered_temp[recstride + 1]; kvz_pixel *recf = &rec_filtered_temp[recstride + 1];
assert(width == 4 || width == 8 || width == 16 || width == 32); assert(width == 4 || width == 8 || width == 16 || width == 32);
@ -2150,8 +2150,8 @@ static int8_t search_intra_rough(encoder_state_t * const state,
*/ */
static int8_t search_intra_rdo(encoder_state_t * const state, static int8_t search_intra_rdo(encoder_state_t * const state,
int x_px, int y_px, int depth, int x_px, int y_px, int depth,
pixel_t *orig, int32_t origstride, kvz_pixel *orig, int32_t origstride,
pixel_t *rec, int16_t recstride, kvz_pixel *rec, int16_t recstride,
int8_t *intra_preds, int8_t *intra_preds,
int modes_to_check, int modes_to_check,
int8_t modes[35], double costs[35], int8_t modes[35], double costs[35],
@ -2160,13 +2160,13 @@ static int8_t search_intra_rdo(encoder_state_t * const state,
const int tr_depth = CLIP(1, MAX_PU_DEPTH, depth + state->encoder_control->tr_depth_intra); const int tr_depth = CLIP(1, MAX_PU_DEPTH, depth + state->encoder_control->tr_depth_intra);
const int width = LCU_WIDTH >> depth; const int width = LCU_WIDTH >> depth;
pixel_t pred[LCU_WIDTH * LCU_WIDTH + 1]; kvz_pixel pred[LCU_WIDTH * LCU_WIDTH + 1];
pixel_t orig_block[LCU_WIDTH * LCU_WIDTH + 1]; kvz_pixel orig_block[LCU_WIDTH * LCU_WIDTH + 1];
int rdo_mode; int rdo_mode;
int pred_mode; int pred_mode;
pixel_t rec_filtered_temp[(LCU_WIDTH * 2 + 8) * (LCU_WIDTH * 2 + 8) + 1]; kvz_pixel rec_filtered_temp[(LCU_WIDTH * 2 + 8) * (LCU_WIDTH * 2 + 8) + 1];
pixel_t *recf = &rec_filtered_temp[recstride + 1]; kvz_pixel *recf = &rec_filtered_temp[recstride + 1];
// Generate filtered reference pixels. // Generate filtered reference pixels.
{ {
@ -2273,8 +2273,8 @@ static double search_cu_intra(encoder_state_t * const state,
cu_info_t *cur_cu = &lcu->cu[cu_index]; cu_info_t *cur_cu = &lcu->cu[cu_index];
pixel_t rec_buffer[(LCU_WIDTH * 2 + 1) * (LCU_WIDTH * 2 + 1)]; kvz_pixel rec_buffer[(LCU_WIDTH * 2 + 1) * (LCU_WIDTH * 2 + 1)];
pixel_t *cu_in_rec_buffer = &rec_buffer[cu_width * 2 + 8 + 1]; kvz_pixel *cu_in_rec_buffer = &rec_buffer[cu_width * 2 + 8 + 1];
int8_t candidate_modes[3]; int8_t candidate_modes[3];
@ -2304,7 +2304,7 @@ static double search_cu_intra(encoder_state_t * const state,
double costs[35]; double costs[35];
// Find best intra mode for 2Nx2N. // Find best intra mode for 2Nx2N.
pixel_t *ref_pixels = &lcu->ref.y[lcu_px.x + lcu_px.y * LCU_WIDTH]; kvz_pixel *ref_pixels = &lcu->ref.y[lcu_px.x + lcu_px.y * LCU_WIDTH];
unsigned pu_index = PU_INDEX(x_px >> 2, y_px >> 2); unsigned pu_index = PU_INDEX(x_px >> 2, y_px >> 2);
int8_t number_of_modes; int8_t number_of_modes;
@ -2499,8 +2499,8 @@ static double search_cu(encoder_state_t * const state, int x, int y, int depth,
} }
if (num_modes != 1 && num_modes != 5) { if (num_modes != 1 && num_modes != 5) {
pixel_t rec_u[(LCU_WIDTH_C * 2 + 8) * (LCU_WIDTH_C * 2 + 8)]; kvz_pixel rec_u[(LCU_WIDTH_C * 2 + 8) * (LCU_WIDTH_C * 2 + 8)];
pixel_t rec_v[(LCU_WIDTH_C * 2 + 8) * (LCU_WIDTH_C * 2 + 8)]; kvz_pixel rec_v[(LCU_WIDTH_C * 2 + 8) * (LCU_WIDTH_C * 2 + 8)];
const int16_t width_c = MAX(LCU_WIDTH_C >> depth, TR_MIN_WIDTH); const int16_t width_c = MAX(LCU_WIDTH_C >> depth, TR_MIN_WIDTH);
const int16_t rec_stride = width_c * 2 + 8; const int16_t rec_stride = width_c * 2 + 8;
@ -2518,8 +2518,8 @@ static double search_cu(encoder_state_t * const state, int x, int y, int depth,
lcu); lcu);
vector2d_t lcu_cpx = { lcu_px.x / 2, lcu_px.y / 2 }; vector2d_t lcu_cpx = { lcu_px.x / 2, lcu_px.y / 2 };
pixel_t *ref_u = &lcu->ref.u[lcu_cpx.x + lcu_cpx.y * LCU_WIDTH_C]; kvz_pixel *ref_u = &lcu->ref.u[lcu_cpx.x + lcu_cpx.y * LCU_WIDTH_C];
pixel_t *ref_v = &lcu->ref.v[lcu_cpx.x + lcu_cpx.y * LCU_WIDTH_C]; kvz_pixel *ref_v = &lcu->ref.v[lcu_cpx.x + lcu_cpx.y * LCU_WIDTH_C];
search_intra_chroma_rough(state, x, y, depth, search_intra_chroma_rough(state, x, y, depth,
ref_u, ref_v, LCU_WIDTH_C, ref_u, ref_v, LCU_WIDTH_C,

22
src/strategies/avx2/ipol-avx2.c
View file

@ -80,7 +80,7 @@ __m128i eight_tap_filter_x4_and_flip_16bit(__m128i data0, __m128i data1, __m128i
return a; return a;
} }
void eight_tap_filter_and_flip_avx2(int8_t filter[4][8], pixel_t *src, int16_t src_stride, int16_t* __restrict dst) void eight_tap_filter_and_flip_avx2(int8_t filter[4][8], kvz_pixel *src, int16_t src_stride, int16_t* __restrict dst)
{ {
//Load 2 rows per xmm register //Load 2 rows per xmm register
@ -104,7 +104,7 @@ void eight_tap_filter_and_flip_avx2(int8_t filter[4][8], pixel_t *src, int16_t s
eight_tap_filter_x8_and_flip(rows01, rows23, rows45, rows67, (__m128i*)(&filter[3]), (__m128i*)(dst + 3 * dst_stride)); eight_tap_filter_x8_and_flip(rows01, rows23, rows45, rows67, (__m128i*)(&filter[3]), (__m128i*)(dst + 3 * dst_stride));
} }
static INLINE void eight_tap_filter_and_flip_16bit_avx2(int8_t filter[4][8], int16_t *src, int16_t src_stride, int offset, int combined_shift, pixel_t* __restrict dst, int16_t dst_stride) static INLINE void eight_tap_filter_and_flip_16bit_avx2(int8_t filter[4][8], int16_t *src, int16_t src_stride, int offset, int combined_shift, kvz_pixel* __restrict dst, int16_t dst_stride)
{ {
//Load a row per xmm register //Load a row per xmm register
@ -157,7 +157,7 @@ static INLINE void eight_tap_filter_and_flip_16bit_avx2(int8_t filter[4][8], int
} }
int16_t eight_tap_filter_hor_avx2(int8_t *filter, pixel_t *data) int16_t eight_tap_filter_hor_avx2(int8_t *filter, kvz_pixel *data)
{ {
union { union {
__m128i vector; __m128i vector;
@ -185,7 +185,7 @@ int32_t eight_tap_filter_hor_16bit_avx2(int8_t *filter, int16_t *data)
return temp; return temp;
} }
int16_t eight_tap_filter_ver_avx2(int8_t *filter, pixel_t *data, int16_t stride) int16_t eight_tap_filter_ver_avx2(int8_t *filter, kvz_pixel *data, int16_t stride)
{ {
int16_t temp = 0; int16_t temp = 0;
for (int i = 0; i < 8; ++i) for (int i = 0; i < 8; ++i)
@ -207,7 +207,7 @@ int32_t eight_tap_filter_ver_16bit_avx2(int8_t *filter, int16_t *data, int16_t s
return temp; return temp;
} }
int16_t four_tap_filter_hor_avx2(int8_t *filter, pixel_t *data) int16_t four_tap_filter_hor_avx2(int8_t *filter, kvz_pixel *data)
{ {
int16_t temp = 0; int16_t temp = 0;
for (int i = 0; i < 4; ++i) for (int i = 0; i < 4; ++i)
@ -229,7 +229,7 @@ int32_t four_tap_filter_hor_16bit_avx2(int8_t *filter, int16_t *data)
return temp; return temp;
} }
int16_t four_tap_filter_ver_avx2(int8_t *filter, pixel_t *data, int16_t stride) int16_t four_tap_filter_ver_avx2(int8_t *filter, kvz_pixel *data, int16_t stride)
{ {
int16_t temp = 0; int16_t temp = 0;
for (int i = 0; i < 4; ++i) for (int i = 0; i < 4; ++i)
@ -251,7 +251,7 @@ int32_t four_tap_filter_ver_16bit_avx2(int8_t *filter, int16_t *data, int16_t st
return temp; return temp;
} }
void filter_inter_quarterpel_luma_avx2(const encoder_control_t * const encoder, pixel_t *src, int16_t src_stride, int width, int height, pixel_t *dst, int16_t dst_stride, int8_t hor_flag, int8_t ver_flag) void filter_inter_quarterpel_luma_avx2(const encoder_control_t * const encoder, kvz_pixel *src, int16_t src_stride, int width, int height, kvz_pixel *dst, int16_t dst_stride, int8_t hor_flag, int8_t ver_flag)
{ {
int32_t x, y; int32_t x, y;
@ -312,7 +312,7 @@ void filter_inter_quarterpel_luma_avx2(const encoder_control_t * const encoder,
* \param dst_stride stride of destination image * \param dst_stride stride of destination image
* *
*/ */
void filter_inter_halfpel_chroma_avx2(const encoder_control_t * const encoder, pixel_t *src, int16_t src_stride, int width, int height, pixel_t *dst, int16_t dst_stride, int8_t hor_flag, int8_t ver_flag) void filter_inter_halfpel_chroma_avx2(const encoder_control_t * const encoder, kvz_pixel *src, int16_t src_stride, int width, int height, kvz_pixel *dst, int16_t dst_stride, int8_t hor_flag, int8_t ver_flag)
{ {
/* ____________ /* ____________
* | B0,0|ae0,0| * | B0,0|ae0,0|
@ -373,7 +373,7 @@ void filter_inter_halfpel_chroma_avx2(const encoder_control_t * const encoder, p
} }
} }
void filter_inter_octpel_chroma_avx2(const encoder_control_t * const encoder, pixel_t *src, int16_t src_stride, int width, int height, pixel_t *dst, int16_t dst_stride, int8_t hor_flag, int8_t ver_flag) void filter_inter_octpel_chroma_avx2(const encoder_control_t * const encoder, kvz_pixel *src, int16_t src_stride, int width, int height, kvz_pixel *dst, int16_t dst_stride, int8_t hor_flag, int8_t ver_flag)
{ {
int32_t x, y; int32_t x, y;
@ -481,8 +481,8 @@ void filter_inter_octpel_chroma_avx2(const encoder_control_t * const encoder, pi
} }
} }
void extend_borders_avx2(int xpos, int ypos, int mv_x, int mv_y, int off_x, int off_y, pixel_t *ref, int ref_width, int ref_height, void extend_borders_avx2(int xpos, int ypos, int mv_x, int mv_y, int off_x, int off_y, kvz_pixel *ref, int ref_width, int ref_height,
int filterSize, int width, int height, pixel_t *dst) { int filterSize, int width, int height, kvz_pixel *dst) {
int16_t mv[2] = { mv_x, mv_y }; int16_t mv[2] = { mv_x, mv_y };
int halfFilterSize = filterSize >> 1; int halfFilterSize = filterSize >> 1;

6
src/strategies/avx2/picture-avx2.c
View file

@ -98,7 +98,7 @@ static INLINE uint32_t m256i_horizontal_sum(const __m256i sum)
} }
static unsigned sad_8bit_8x8_avx2(const pixel_t *buf1, const pixel_t *buf2) static unsigned sad_8bit_8x8_avx2(const kvz_pixel *buf1, const kvz_pixel *buf2)
{ {
const __m256i *const a = (const __m256i *)buf1; const __m256i *const a = (const __m256i *)buf1;
const __m256i *const b = (const __m256i *)buf2; const __m256i *const b = (const __m256i *)buf2;
@ -108,7 +108,7 @@ static unsigned sad_8bit_8x8_avx2(const pixel_t *buf1, const pixel_t *buf2)
} }
static unsigned sad_8bit_16x16_avx2(const pixel_t *buf1, const pixel_t *buf2) static unsigned sad_8bit_16x16_avx2(const kvz_pixel *buf1, const kvz_pixel *buf2)
{ {
const __m256i *const a = (const __m256i *)buf1; const __m256i *const a = (const __m256i *)buf1;
const __m256i *const b = (const __m256i *)buf2; const __m256i *const b = (const __m256i *)buf2;
@ -118,7 +118,7 @@ static unsigned sad_8bit_16x16_avx2(const pixel_t *buf1, const pixel_t *buf2)
} }
static unsigned sad_8bit_32x32_avx2(const pixel_t *buf1, const pixel_t *buf2) static unsigned sad_8bit_32x32_avx2(const kvz_pixel *buf1, const kvz_pixel *buf2)
{ {
const __m256i *const a = (const __m256i *)buf1; const __m256i *const a = (const __m256i *)buf1;
const __m256i *const b = (const __m256i *)buf2; const __m256i *const b = (const __m256i *)buf2;

18
src/strategies/generic/ipol-generic.c
View file

@ -32,7 +32,7 @@
extern int8_t g_luma_filter[4][8]; extern int8_t g_luma_filter[4][8];
extern int8_t g_chroma_filter[8][4]; extern int8_t g_chroma_filter[8][4];
int16_t eight_tap_filter_hor_generic(int8_t *filter, pixel_t *data) int16_t eight_tap_filter_hor_generic(int8_t *filter, kvz_pixel *data)
{ {
int16_t temp = 0; int16_t temp = 0;
for (int i = 0; i < 8; ++i) for (int i = 0; i < 8; ++i)
@ -54,7 +54,7 @@ int32_t eight_tap_filter_hor_16bit_generic(int8_t *filter, int16_t *data)
return temp; return temp;
} }
int16_t eight_tap_filter_ver_generic(int8_t *filter, pixel_t *data, int16_t stride) int16_t eight_tap_filter_ver_generic(int8_t *filter, kvz_pixel *data, int16_t stride)
{ {
int16_t temp = 0; int16_t temp = 0;
for (int i = 0; i < 8; ++i) for (int i = 0; i < 8; ++i)
@ -76,7 +76,7 @@ int32_t eight_tap_filter_ver_16bit_generic(int8_t *filter, int16_t *data, int16_
return temp; return temp;
} }
int16_t four_tap_filter_hor_generic(int8_t *filter, pixel_t *data) int16_t four_tap_filter_hor_generic(int8_t *filter, kvz_pixel *data)
{ {
int16_t temp = 0; int16_t temp = 0;
for (int i = 0; i < 4; ++i) for (int i = 0; i < 4; ++i)
@ -98,7 +98,7 @@ int32_t four_tap_filter_hor_16bit_generic(int8_t *filter, int16_t *data)
return temp; return temp;
} }
int16_t four_tap_filter_ver_generic(int8_t *filter, pixel_t *data, int16_t stride) int16_t four_tap_filter_ver_generic(int8_t *filter, kvz_pixel *data, int16_t stride)
{ {
int16_t temp = 0; int16_t temp = 0;
for (int i = 0; i < 4; ++i) for (int i = 0; i < 4; ++i)
@ -120,7 +120,7 @@ int32_t four_tap_filter_ver_16bit_generic(int8_t *filter, int16_t *data, int16_t
return temp; return temp;
} }
void filter_inter_quarterpel_luma_generic(const encoder_control_t * const encoder, pixel_t *src, int16_t src_stride, int width, int height, pixel_t *dst, int16_t dst_stride, int8_t hor_flag, int8_t ver_flag) void filter_inter_quarterpel_luma_generic(const encoder_control_t * const encoder, kvz_pixel *src, int16_t src_stride, int width, int height, kvz_pixel *dst, int16_t dst_stride, int8_t hor_flag, int8_t ver_flag)
{ {
//TODO: horizontal and vertical only filtering //TODO: horizontal and vertical only filtering
int32_t x, y; int32_t x, y;
@ -180,7 +180,7 @@ void filter_inter_quarterpel_luma_generic(const encoder_control_t * const encode
* \param dst_stride stride of destination image * \param dst_stride stride of destination image
* *
*/ */
void filter_inter_halfpel_chroma_generic(const encoder_control_t * const encoder, pixel_t *src, int16_t src_stride, int width, int height, pixel_t *dst, int16_t dst_stride, int8_t hor_flag, int8_t ver_flag) void filter_inter_halfpel_chroma_generic(const encoder_control_t * const encoder, kvz_pixel *src, int16_t src_stride, int width, int height, kvz_pixel *dst, int16_t dst_stride, int8_t hor_flag, int8_t ver_flag)
{ {
/* ____________ /* ____________
* | B0,0|ae0,0| * | B0,0|ae0,0|
@ -241,7 +241,7 @@ void filter_inter_halfpel_chroma_generic(const encoder_control_t * const encoder
} }
} }
void filter_inter_octpel_chroma_generic(const encoder_control_t * const encoder, pixel_t *src, int16_t src_stride, int width, int height, pixel_t *dst, int16_t dst_stride, int8_t hor_flag, int8_t ver_flag) void filter_inter_octpel_chroma_generic(const encoder_control_t * const encoder, kvz_pixel *src, int16_t src_stride, int width, int height, kvz_pixel *dst, int16_t dst_stride, int8_t hor_flag, int8_t ver_flag)
{ {
int32_t x, y; int32_t x, y;
@ -349,8 +349,8 @@ void filter_inter_octpel_chroma_generic(const encoder_control_t * const encoder,
} }
} }
void extend_borders_generic(int xpos, int ypos, int mv_x, int mv_y, int off_x, int off_y, pixel_t *ref, int ref_width, int ref_height, void extend_borders_generic(int xpos, int ypos, int mv_x, int mv_y, int off_x, int off_y, kvz_pixel *ref, int ref_width, int ref_height,
int filterSize, int width, int height, pixel_t *dst) { int filterSize, int width, int height, kvz_pixel *dst) {
int16_t mv[2] = { mv_x, mv_y }; int16_t mv[2] = { mv_x, mv_y };
int halfFilterSize = filterSize >> 1; int halfFilterSize = filterSize >> 1;

6
src/strategies/generic/nal-generic.c
View file

@ -29,7 +29,7 @@
#include "nal.h" #include "nal.h"
static void array_checksum_generic(const pixel_t* data, static void array_checksum_generic(const kvz_pixel* data,
const int height, const int width, const int height, const int width,
const int stride, const int stride,
unsigned char checksum_out[SEI_HASH_MAX_LENGTH]) { unsigned char checksum_out[SEI_HASH_MAX_LENGTH]) {
@ -52,7 +52,7 @@ static void array_checksum_generic(const pixel_t* data,
checksum_out[3] = (checksum) & 0xff; checksum_out[3] = (checksum) & 0xff;
} }
static void array_checksum_generic4(const pixel_t* data, static void array_checksum_generic4(const kvz_pixel* data,
const int height, const int width, const int height, const int width,
const int stride, const int stride,
unsigned char checksum_out[SEI_HASH_MAX_LENGTH]) { unsigned char checksum_out[SEI_HASH_MAX_LENGTH]) {
@ -95,7 +95,7 @@ static void array_checksum_generic4(const pixel_t* data,
checksum_out[3] = (checksum) & 0xff; checksum_out[3] = (checksum) & 0xff;
} }
static void array_checksum_generic8(const pixel_t* data, static void array_checksum_generic8(const kvz_pixel* data,
const int height, const int width, const int height, const int width,
const int stride, const int stride,
unsigned char checksum_out[SEI_HASH_MAX_LENGTH]) { unsigned char checksum_out[SEI_HASH_MAX_LENGTH]) {

30
src/strategies/generic/picture-generic.c
View file

@ -28,7 +28,7 @@
// Function to clip int16_t to pixel. (0-255 or 0-1023) // Function to clip int16_t to pixel. (0-255 or 0-1023)
// Assumes PIXEL_MAX to be 2^n-1 // Assumes PIXEL_MAX to be 2^n-1
pixel_t fast_clip_16bit_to_pixel(int16_t value) kvz_pixel fast_clip_16bit_to_pixel(int16_t value)
{ {
// Ensure that compiler generates arithmetic shift from ">>" // Ensure that compiler generates arithmetic shift from ">>"
#if defined(_MSC_VER) || defined(__GNUC__) || defined(__clang__) #if defined(_MSC_VER) || defined(__GNUC__) || defined(__clang__)
@ -50,7 +50,7 @@ pixel_t fast_clip_16bit_to_pixel(int16_t value)
// Function to clip int32_t to pixel. (0-255 or 0-1023) // Function to clip int32_t to pixel. (0-255 or 0-1023)
// Assumes PIXEL_MAX to be 2^n-1 // Assumes PIXEL_MAX to be 2^n-1
pixel_t fast_clip_32bit_to_pixel(int32_t value) kvz_pixel fast_clip_32bit_to_pixel(int32_t value)
{ {
// Ensure that compiler generates arithmetic shift from ">>" // Ensure that compiler generates arithmetic shift from ">>"
#if defined(_MSC_VER) || defined(__GNUC__) || defined(__clang__) #if defined(_MSC_VER) || defined(__GNUC__) || defined(__clang__)
@ -84,7 +84,7 @@ pixel_t fast_clip_32bit_to_pixel(int32_t value)
* *
* \returns Sum of Absolute Differences * \returns Sum of Absolute Differences
*/ */
static unsigned reg_sad_generic(const pixel_t * const data1, const pixel_t * const data2, static unsigned reg_sad_generic(const kvz_pixel * const data1, const kvz_pixel * const data2,
const int width, const int height, const unsigned stride1, const unsigned stride2) const int width, const int height, const unsigned stride1, const unsigned stride2)
{ {
int y, x; int y, x;
@ -104,7 +104,7 @@ static unsigned reg_sad_generic(const pixel_t * const data1, const pixel_t * con
* \brief Calculate SATD between two 4x4 blocks inside bigger arrays. * \brief Calculate SATD between two 4x4 blocks inside bigger arrays.
* From HM 13.0 * From HM 13.0
*/ */
static unsigned satd_8bit_4x4_generic(const pixel_t *piOrg, const pixel_t *piCur) static unsigned satd_8bit_4x4_generic(const kvz_pixel *piOrg, const kvz_pixel *piCur)
{ {
int32_t k, satd = 0, diff[16], m[16], d[16]; int32_t k, satd = 0, diff[16], m[16], d[16];
for (k = 0; k < 16; ++k) { for (k = 0; k < 16; ++k) {
@ -191,8 +191,8 @@ static unsigned satd_8bit_4x4_generic(const pixel_t *piOrg, const pixel_t *piCur
/** /**
* \brief Calculate SATD between two 8x8 blocks inside bigger arrays. * \brief Calculate SATD between two 8x8 blocks inside bigger arrays.
*/ */
unsigned satd_16bit_8x8_general(const pixel_t * piOrg, const int32_t iStrideOrg, unsigned satd_16bit_8x8_general(const kvz_pixel * piOrg, const int32_t iStrideOrg,
const pixel_t * piCur, const int32_t iStrideCur) const kvz_pixel * piCur, const int32_t iStrideCur)
{ {
int32_t k, i, j, jj, sad = 0; int32_t k, i, j, jj, sad = 0;
int32_t diff[64], m1[8][8], m2[8][8], m3[8][8]; int32_t diff[64], m1[8][8], m2[8][8], m3[8][8];
@ -307,10 +307,10 @@ cost_pixel_nxn_func satd_8bit_32x32_generic;
cost_pixel_nxn_func satd_8bit_64x64_generic; cost_pixel_nxn_func satd_8bit_64x64_generic;
// These macros define sadt_16bit_NxN for N = 8, 16, 32, 64 // These macros define sadt_16bit_NxN for N = 8, 16, 32, 64
SATD_NXN(8, pixel_t, 8bit) SATD_NXN(8, kvz_pixel, 8bit)
SATD_NXN(16, pixel_t, 8bit) SATD_NXN(16, kvz_pixel, 8bit)
SATD_NXN(32, pixel_t, 8bit) SATD_NXN(32, kvz_pixel, 8bit)
SATD_NXN(64, pixel_t, 8bit) SATD_NXN(64, kvz_pixel, 8bit)
// Function macro for defining SAD calculating functions // Function macro for defining SAD calculating functions
// for fixed size blocks. // for fixed size blocks.
@ -336,11 +336,11 @@ static cost_pixel_nxn_func sad_8bit_64x64_generic;
// These macros define sad_16bit_nxn functions for n = 4, 8, 16, 32, 64 // These macros define sad_16bit_nxn functions for n = 4, 8, 16, 32, 64
// with function signatures of cost_16bit_nxn_func. // with function signatures of cost_16bit_nxn_func.
// They are used through get_pixel_sad_func. // They are used through get_pixel_sad_func.
SAD_NXN(4, pixel_t, 8bit) SAD_NXN(4, kvz_pixel, 8bit)
SAD_NXN(8, pixel_t, 8bit) SAD_NXN(8, kvz_pixel, 8bit)
SAD_NXN(16, pixel_t, 8bit) SAD_NXN(16, kvz_pixel, 8bit)
SAD_NXN(32, pixel_t, 8bit) SAD_NXN(32, kvz_pixel, 8bit)
SAD_NXN(64, pixel_t, 8bit) SAD_NXN(64, kvz_pixel, 8bit)
int strategy_register_picture_generic(void* opaque) int strategy_register_picture_generic(void* opaque)

4
src/strategies/generic/picture-generic.h
View file

@ -24,10 +24,10 @@ int strategy_register_picture_generic(void* opaque);
// Function to clip int16_t to pixel. (0-255 or 0-1023) // Function to clip int16_t to pixel. (0-255 or 0-1023)
// Assumes PIXEL_MAX to be 2^n-1 // Assumes PIXEL_MAX to be 2^n-1
pixel_t fast_clip_16bit_to_pixel(int16_t value); kvz_pixel fast_clip_16bit_to_pixel(int16_t value);
// Function to clip int32_t to pixel. (0-255 or 0-1023) // Function to clip int32_t to pixel. (0-255 or 0-1023)
// Assumes PIXEL_MAX to be 2^n-1 // Assumes PIXEL_MAX to be 2^n-1
pixel_t fast_clip_32bit_to_pixel(int32_t value); kvz_pixel fast_clip_32bit_to_pixel(int32_t value);
#endif //STRATEGIES_PICTURE_GENERIC_H_ #endif //STRATEGIES_PICTURE_GENERIC_H_

4
src/strategies/sse2/picture-sse2.c
View file

@ -31,7 +31,7 @@
# include <stdlib.h> # include <stdlib.h>
static unsigned reg_sad_sse2(const pixel_t * const data1, const pixel_t * const data2, static unsigned reg_sad_sse2(const kvz_pixel * const data1, const kvz_pixel * const data2,
const int width, const int height, const unsigned stride1, const unsigned stride2) const int width, const int height, const unsigned stride1, const unsigned stride2)
{ {
int y, x; int y, x;
@ -56,7 +56,7 @@ static unsigned reg_sad_sse2(const pixel_t * const data1, const pixel_t * const
return sad; return sad;
} }
static unsigned sad_8bit_4x4_sse2(const pixel_t *buf1, const pixel_t *buf2) static unsigned sad_8bit_4x4_sse2(const kvz_pixel *buf1, const kvz_pixel *buf2)
{ {
const __m128i *const mbuf1 = (const __m128i *)buf1; const __m128i *const mbuf1 = (const __m128i *)buf1;
const __m128i *const mbuf2 = (const __m128i *)buf2; const __m128i *const mbuf2 = (const __m128i *)buf2;

2
src/strategies/sse41/picture-sse41.c
View file

@ -31,7 +31,7 @@
# include <stdlib.h> # include <stdlib.h>
static unsigned reg_sad_sse41(const pixel_t * const data1, const pixel_t * const data2, static unsigned reg_sad_sse41(const kvz_pixel * const data1, const kvz_pixel * const data2,
const int width, const int height, const unsigned stride1, const unsigned stride2) const int width, const int height, const unsigned stride1, const unsigned stride2)
{ {
int y, x; int y, x;

6
src/strategies/strategies-ipol.h
View file

@ -25,11 +25,11 @@
#include "encoder.h" #include "encoder.h"
typedef unsigned(ipol_func)(const encoder_control_t * encoder, pixel_t *src, int16_t src_stride, int width, int height, pixel_t *dst, typedef unsigned(ipol_func)(const encoder_control_t * encoder, kvz_pixel *src, int16_t src_stride, int width, int height, kvz_pixel *dst,
int16_t dst_stride, int8_t hor_flag, int8_t ver_flag); int16_t dst_stride, int8_t hor_flag, int8_t ver_flag);
typedef unsigned(epol_func)(int xpos, int ypos, int mv_x, int mv_y, int off_x, int off_y, pixel_t *ref, int ref_width, int ref_height, typedef unsigned(epol_func)(int xpos, int ypos, int mv_x, int mv_y, int off_x, int off_y, kvz_pixel *ref, int ref_width, int ref_height,
int filterSize, int width, int height, pixel_t *dst); int filterSize, int width, int height, kvz_pixel *dst);
// Declare function pointers. // Declare function pointers.

2
src/strategies/strategies-nal.c
View file

@ -22,7 +22,7 @@
#include "generic/nal-generic.h" #include "generic/nal-generic.h"
void (*array_checksum)(const pixel_t* data, void (*array_checksum)(const kvz_pixel* data,
const int height, const int width, const int height, const int width,
const int stride, const int stride,
unsigned char checksum_out[SEI_HASH_MAX_LENGTH]); unsigned char checksum_out[SEI_HASH_MAX_LENGTH]);

2
src/strategies/strategies-nal.h
View file

@ -30,7 +30,7 @@
* \param width Width of the picture. * \param width Width of the picture.
* \param stride Width of one row in the pixel array. * \param stride Width of one row in the pixel array.
*/ */
typedef void (*array_checksum_func)(const pixel_t* data, typedef void (*array_checksum_func)(const kvz_pixel* data,
const int height, const int width, const int height, const int width,
const int stride, const int stride,
unsigned char checksum_out[SEI_HASH_MAX_LENGTH]); unsigned char checksum_out[SEI_HASH_MAX_LENGTH]);

4
src/strategies/strategies-picture.h
View file

@ -23,10 +23,10 @@
#include "../image.h" #include "../image.h"
typedef unsigned(reg_sad_func)(const pixel_t *const data1, const pixel_t *const data2, typedef unsigned(reg_sad_func)(const kvz_pixel *const data1, const kvz_pixel *const data2,
const int width, const int height, const int width, const int height,
const unsigned stride1, const unsigned stride2); const unsigned stride1, const unsigned stride2);
typedef unsigned (cost_pixel_nxn_func)(const pixel_t *block1, const pixel_t *block2); typedef unsigned (cost_pixel_nxn_func)(const kvz_pixel *block1, const kvz_pixel *block2);
// Declare function pointers. // Declare function pointers.

12
src/strategies/x86_asm/picture-x86-asm-sad.h
View file

@ -24,13 +24,13 @@
\brief assembly functions header for sad \brief assembly functions header for sad
*/ */
unsigned kvz_sad_4x4_avx(const pixel_t*, const pixel_t*); unsigned kvz_sad_4x4_avx(const kvz_pixel*, const kvz_pixel*);
unsigned kvz_sad_8x8_avx(const pixel_t*, const pixel_t*); unsigned kvz_sad_8x8_avx(const kvz_pixel*, const kvz_pixel*);
unsigned kvz_sad_16x16_avx(const pixel_t*, const pixel_t*); unsigned kvz_sad_16x16_avx(const kvz_pixel*, const kvz_pixel*);
unsigned kvz_sad_4x4_stride_avx(const pixel_t *data1, const pixel_t *data2, unsigned stride); unsigned kvz_sad_4x4_stride_avx(const kvz_pixel *data1, const kvz_pixel *data2, unsigned stride);
unsigned kvz_sad_8x8_stride_avx(const pixel_t *data1, const pixel_t *data2, unsigned stride); unsigned kvz_sad_8x8_stride_avx(const kvz_pixel *data1, const kvz_pixel *data2, unsigned stride);
unsigned kvz_sad_16x16_stride_avx(const pixel_t *data1, const pixel_t *data2, unsigned stride); unsigned kvz_sad_16x16_stride_avx(const kvz_pixel *data1, const kvz_pixel *data2, unsigned stride);
#endif #endif

10
src/strategies/x86_asm/picture-x86-asm-satd.h
View file

@ -24,10 +24,10 @@
/*! \file picture-x86-asm-satd.h /*! \file picture-x86-asm-satd.h
\brief assembly functions header for satd \brief assembly functions header for satd
*/ */
unsigned kvz_satd_4x4_avx(const pixel_t *org, const pixel_t *cur); unsigned kvz_satd_4x4_avx(const kvz_pixel *org, const kvz_pixel *cur);
unsigned kvz_satd_8x8_avx(const pixel_t *org, const pixel_t *cur); unsigned kvz_satd_8x8_avx(const kvz_pixel *org, const kvz_pixel *cur);
unsigned kvz_satd_16x16_avx(const pixel_t *org, const pixel_t *cur); unsigned kvz_satd_16x16_avx(const kvz_pixel *org, const kvz_pixel *cur);
unsigned kvz_satd_32x32_avx(const pixel_t *org, const pixel_t *cur); unsigned kvz_satd_32x32_avx(const kvz_pixel *org, const kvz_pixel *cur);
unsigned kvz_satd_64x64_avx(const pixel_t *org, const pixel_t *cur); unsigned kvz_satd_64x64_avx(const kvz_pixel *org, const kvz_pixel *cur);
#endif #endif

12
src/strategies/x86_asm/picture-x86-asm.c
View file

@ -29,7 +29,7 @@
#include "picture-x86-asm-sad.h" #include "picture-x86-asm-sad.h"
#include "picture-x86-asm-satd.h" #include "picture-x86-asm-satd.h"
static unsigned kvz_sad_32x32_avx(const pixel_t *data1, const pixel_t *data2) static unsigned kvz_sad_32x32_avx(const kvz_pixel *data1, const kvz_pixel *data2)
{ {
unsigned sad = 0; unsigned sad = 0;
sad += kvz_sad_16x16_avx(data1, data2); sad += kvz_sad_16x16_avx(data1, data2);
@ -39,7 +39,7 @@ static unsigned kvz_sad_32x32_avx(const pixel_t *data1, const pixel_t *data2)
return sad; return sad;
} }
static unsigned kvz_sad_32x32_stride_avx(const pixel_t *data1, const pixel_t *data2, unsigned stride) static unsigned kvz_sad_32x32_stride_avx(const kvz_pixel *data1, const kvz_pixel *data2, unsigned stride)
{ {
unsigned sad = 0; unsigned sad = 0;
sad += kvz_sad_16x16_stride_avx(data1, data2, stride); sad += kvz_sad_16x16_stride_avx(data1, data2, stride);
@ -49,7 +49,7 @@ static unsigned kvz_sad_32x32_stride_avx(const pixel_t *data1, const pixel_t *da
return sad; return sad;
} }
static unsigned kvz_sad_64x64_avx(const pixel_t *data1, const pixel_t *data2) static unsigned kvz_sad_64x64_avx(const kvz_pixel *data1, const kvz_pixel *data2)
{ {
unsigned sad = 0; unsigned sad = 0;
sad += kvz_sad_32x32_avx(data1, data2); sad += kvz_sad_32x32_avx(data1, data2);
@ -59,7 +59,7 @@ static unsigned kvz_sad_64x64_avx(const pixel_t *data1, const pixel_t *data2)
return sad; return sad;
} }
static unsigned kvz_sad_64x64_stride_avx(const pixel_t *data1, const pixel_t *data2, unsigned stride) static unsigned kvz_sad_64x64_stride_avx(const kvz_pixel *data1, const kvz_pixel *data2, unsigned stride)
{ {
unsigned sad = 0; unsigned sad = 0;
sad += kvz_sad_32x32_stride_avx(data1, data2, stride); sad += kvz_sad_32x32_stride_avx(data1, data2, stride);
@ -69,7 +69,7 @@ static unsigned kvz_sad_64x64_stride_avx(const pixel_t *data1, const pixel_t *da
return sad; return sad;
} }
static unsigned kvz_sad_other_avx(const pixel_t * const data1, const pixel_t * const data2, static unsigned kvz_sad_other_avx(const kvz_pixel * const data1, const kvz_pixel * const data2,
const int width, const int height, const unsigned stride1, const unsigned stride2) const int width, const int height, const unsigned stride1, const unsigned stride2)
{ {
int y, x; int y, x;
@ -84,7 +84,7 @@ static unsigned kvz_sad_other_avx(const pixel_t * const data1, const pixel_t * c
return sad; return sad;
} }
static unsigned reg_sad_x86_asm(const pixel_t * const data1, const pixel_t * const data2, static unsigned reg_sad_x86_asm(const kvz_pixel * const data1, const kvz_pixel * const data2,
const int width, const int height, const unsigned stride1, const unsigned stride2) const int width, const int height, const unsigned stride1, const unsigned stride2)
{ {
if (width == 4 && height == 4) { if (width == 4 && height == 4) {

2
src/strategyselector.h
View file

@ -157,7 +157,7 @@ static const strategy_to_select_t strategies_to_select[] = {
{ NULL, NULL }, { NULL, NULL },
}; };
unsigned satd_8bit_8x8_generic(const pixel_t * const block1, const pixel_t * const block2); unsigned satd_8bit_8x8_generic(const kvz_pixel * const block1, const kvz_pixel * const block2);
#endif //STRATEGYSELECTOR_H_ #endif //STRATEGYSELECTOR_H_

22
src/transform.c
View file

@ -326,8 +326,8 @@ int quantize_residual(encoder_state_t *const state,
const cu_info_t *const cur_cu, const int width, const color_t color, const cu_info_t *const cur_cu, const int width, const color_t color,
const coeff_scan_order_t scan_order, const int use_trskip, const coeff_scan_order_t scan_order, const int use_trskip,
const int in_stride, const int out_stride, const int in_stride, const int out_stride,
const pixel_t *const ref_in, const pixel_t *const pred_in, const kvz_pixel *const ref_in, const kvz_pixel *const pred_in,
pixel_t *rec_out, coeff_t *coeff_out) kvz_pixel *rec_out, coeff_t *coeff_out)
{ {
// Temporary arrays to pass data to and from quant and transform functions. // Temporary arrays to pass data to and from quant and transform functions.
int16_t residual[TR_MAX_WIDTH * TR_MAX_WIDTH]; int16_t residual[TR_MAX_WIDTH * TR_MAX_WIDTH];
@ -440,11 +440,11 @@ int quantize_residual_trskip(
const cu_info_t *const cur_cu, const int width, const color_t color, const cu_info_t *const cur_cu, const int width, const color_t color,
const coeff_scan_order_t scan_order, int8_t *trskip_out, const coeff_scan_order_t scan_order, int8_t *trskip_out,
const int in_stride, const int out_stride, const int in_stride, const int out_stride,
const pixel_t *const ref_in, const pixel_t *const pred_in, const kvz_pixel *const ref_in, const kvz_pixel *const pred_in,
pixel_t *rec_out, coeff_t *coeff_out) kvz_pixel *rec_out, coeff_t *coeff_out)
{ {
struct { struct {
pixel_t rec[4*4]; kvz_pixel rec[4*4];
coeff_t coeff[4*4]; coeff_t coeff[4*4];
uint32_t cost; uint32_t cost;
int has_coeffs; int has_coeffs;
@ -543,9 +543,9 @@ void quantize_lcu_luma_residual(encoder_state_t * const state, int32_t x, int32_
const int luma_offset = lcu_px.x + lcu_px.y * LCU_WIDTH; const int luma_offset = lcu_px.x + lcu_px.y * LCU_WIDTH;
// Pointers to current location in arrays with prediction. // Pointers to current location in arrays with prediction.
pixel_t *recbase_y = &lcu->rec.y[luma_offset]; kvz_pixel *recbase_y = &lcu->rec.y[luma_offset];
// Pointers to current location in arrays with reference. // Pointers to current location in arrays with reference.
const pixel_t *base_y = &lcu->ref.y[luma_offset]; const kvz_pixel *base_y = &lcu->ref.y[luma_offset];
// Pointers to current location in arrays with kvantized coefficients. // Pointers to current location in arrays with kvantized coefficients.
coeff_t *orig_coeff_y = &lcu->coeff.y[luma_offset]; coeff_t *orig_coeff_y = &lcu->coeff.y[luma_offset];
@ -633,10 +633,10 @@ void quantize_lcu_chroma_residual(encoder_state_t * const state, int32_t x, int3
cbf_clear(&cur_cu->cbf.v, depth); cbf_clear(&cur_cu->cbf.v, depth);
const int chroma_offset = lcu_px.x / 2 + lcu_px.y / 2 * LCU_WIDTH_C; const int chroma_offset = lcu_px.x / 2 + lcu_px.y / 2 * LCU_WIDTH_C;
pixel_t *recbase_u = &lcu->rec.u[chroma_offset]; kvz_pixel *recbase_u = &lcu->rec.u[chroma_offset];
pixel_t *recbase_v = &lcu->rec.v[chroma_offset]; kvz_pixel *recbase_v = &lcu->rec.v[chroma_offset];
const pixel_t *base_u = &lcu->ref.u[chroma_offset]; const kvz_pixel *base_u = &lcu->ref.u[chroma_offset];
const pixel_t *base_v = &lcu->ref.v[chroma_offset]; const kvz_pixel *base_v = &lcu->ref.v[chroma_offset];
coeff_t *orig_coeff_u = &lcu->coeff.u[chroma_offset]; coeff_t *orig_coeff_u = &lcu->coeff.u[chroma_offset];
coeff_t *orig_coeff_v = &lcu->coeff.v[chroma_offset]; coeff_t *orig_coeff_v = &lcu->coeff.v[chroma_offset];
coeff_scan_order_t scan_idx_chroma; coeff_scan_order_t scan_idx_chroma;

14
src/yuv_io.c
View file

@ -28,10 +28,10 @@
#include "yuv_io.h" #include "yuv_io.h"
static void fill_after_frame(unsigned height, unsigned array_width, static void fill_after_frame(unsigned height, unsigned array_width,
unsigned array_height, pixel_t *data) unsigned array_height, kvz_pixel *data)
{ {
pixel_t* p = data + height * array_width; kvz_pixel* p = data + height * array_width;
pixel_t* end = data + array_width * array_height; kvz_pixel* end = data + array_width * array_height;
while (p < end) { while (p < end) {
// Fill the line by copying the line above. // Fill the line by copying the line above.
@ -43,11 +43,11 @@ static void fill_after_frame(unsigned height, unsigned array_width,
static int read_and_fill_frame_data(FILE *file, static int read_and_fill_frame_data(FILE *file,
unsigned width, unsigned height, unsigned width, unsigned height,
unsigned array_width, pixel_t *data) unsigned array_width, kvz_pixel *data)
{ {
pixel_t* p = data; kvz_pixel* p = data;
pixel_t* end = data + array_width * height; kvz_pixel* end = data + array_width * height;
pixel_t fill_char; kvz_pixel fill_char;
unsigned i; unsigned i;
while (p < end) { while (p < end) {

18
tests/intra_sad_tests.c
View file

@ -35,7 +35,7 @@
////////////////////////////////////////////////////////////////////////// //////////////////////////////////////////////////////////////////////////
// GLOBALS // GLOBALS
pixel_t * bufs[NUM_TESTS][6][2]; kvz_pixel * bufs[NUM_TESTS][6][2];
static struct { static struct {
int log_width; // for selecting dim from bufs int log_width; // for selecting dim from bufs
@ -45,7 +45,7 @@ static struct {
////////////////////////////////////////////////////////////////////////// //////////////////////////////////////////////////////////////////////////
// SETUP, TEARDOWN AND HELPER FUNCTIONS // SETUP, TEARDOWN AND HELPER FUNCTIONS
static void init_gradient(int x_px, int y_px, int width, int slope, pixel_t *buf) static void init_gradient(int x_px, int y_px, int width, int slope, kvz_pixel *buf)
{ {
for (int y = 0; y < width; ++y) { for (int y = 0; y < width; ++y) {
for (int x = 0; x < width; ++x) { for (int x = 0; x < width; ++x) {
@ -68,10 +68,10 @@ static void setup_tests()
for (int w = LCU_MIN_LOG_W; w <= LCU_MAX_LOG_W; ++w) { for (int w = LCU_MIN_LOG_W; w <= LCU_MAX_LOG_W; ++w) {
unsigned size = 1 << (w * 2); unsigned size = 1 << (w * 2);
bufs[test][w][0] = malloc(size * sizeof(pixel_t) + SIMD_ALIGNMENT); bufs[test][w][0] = malloc(size * sizeof(kvz_pixel) + SIMD_ALIGNMENT);
bufs[test][w][0] = ALIGNED_POINTER(bufs[test][w][0], SIMD_ALIGNMENT); bufs[test][w][0] = ALIGNED_POINTER(bufs[test][w][0], SIMD_ALIGNMENT);
bufs[test][w][1] = malloc(size * sizeof(pixel_t) + SIMD_ALIGNMENT); bufs[test][w][1] = malloc(size * sizeof(kvz_pixel) + SIMD_ALIGNMENT);
bufs[test][w][1] = ALIGNED_POINTER(bufs[test][w][1], SIMD_ALIGNMENT); bufs[test][w][1] = ALIGNED_POINTER(bufs[test][w][1], SIMD_ALIGNMENT);
} }
} }
@ -104,7 +104,7 @@ static void tear_down_tests()
} }
static unsigned test_calc_sad(const pixel_t * buf1, const pixel_t * buf2, int dim) static unsigned test_calc_sad(const kvz_pixel * buf1, const kvz_pixel * buf2, int dim)
{ {
unsigned result = 0; unsigned result = 0;
for (int i = 0; i < dim * dim; ++i) { for (int i = 0; i < dim * dim; ++i) {
@ -125,8 +125,8 @@ TEST test_black_and_white(void)
const int test = 0; const int test = 0;
const int width = 1 << test_env.log_width; const int width = 1 << test_env.log_width;
pixel_t * buf1 = bufs[test][test_env.log_width][0]; kvz_pixel * buf1 = bufs[test][test_env.log_width][0];
pixel_t * buf2 = bufs[test][test_env.log_width][1]; kvz_pixel * buf2 = bufs[test][test_env.log_width][1];
unsigned result1 = test_env.tested_func(buf1, buf2); unsigned result1 = test_env.tested_func(buf1, buf2);
unsigned result2 = test_env.tested_func(buf2, buf1); unsigned result2 = test_env.tested_func(buf2, buf1);
@ -149,8 +149,8 @@ TEST test_gradient(void)
const int test = 1; const int test = 1;
const int width = 1 << test_env.log_width; const int width = 1 << test_env.log_width;
pixel_t * buf1 = bufs[test][test_env.log_width][0]; kvz_pixel * buf1 = bufs[test][test_env.log_width][0];
pixel_t * buf2 = bufs[test][test_env.log_width][1]; kvz_pixel * buf2 = bufs[test][test_env.log_width][1];
unsigned result = test_calc_sad(buf1, buf2, width); unsigned result = test_calc_sad(buf1, buf2, width);
unsigned result1 = test_env.tested_func(buf1, buf2); unsigned result1 = test_env.tested_func(buf1, buf2);

18
tests/satd_tests.c
View file

@ -33,7 +33,7 @@
////////////////////////////////////////////////////////////////////////// //////////////////////////////////////////////////////////////////////////
// GLOBALS // GLOBALS
pixel_t * satd_bufs[NUM_TESTS][7][2]; kvz_pixel * satd_bufs[NUM_TESTS][7][2];
static struct { static struct {
int log_width; // for selecting dim from satd_bufs int log_width; // for selecting dim from satd_bufs
@ -53,8 +53,8 @@ static void setup_tests()
for (int w = LCU_MIN_LOG_W; w <= LCU_MAX_LOG_W; ++w) { for (int w = LCU_MIN_LOG_W; w <= LCU_MAX_LOG_W; ++w) {
unsigned size = 1 << (w * 2); unsigned size = 1 << (w * 2);
satd_bufs[test][w][0] = malloc(size * sizeof(pixel_t)); satd_bufs[test][w][0] = malloc(size * sizeof(kvz_pixel));
satd_bufs[test][w][1] = malloc(size * sizeof(pixel_t)); satd_bufs[test][w][1] = malloc(size * sizeof(kvz_pixel));
} }
} }
@ -110,8 +110,8 @@ TEST satd_test_black_and_white(void)
const int test = 0; const int test = 0;
pixel_t * buf1 = satd_bufs[test][satd_test_env.log_width][0]; kvz_pixel * buf1 = satd_bufs[test][satd_test_env.log_width][0];
pixel_t * buf2 = satd_bufs[test][satd_test_env.log_width][1]; kvz_pixel * buf2 = satd_bufs[test][satd_test_env.log_width][1];
unsigned result1 = satd_test_env.tested_func(buf1, buf2); unsigned result1 = satd_test_env.tested_func(buf1, buf2);
unsigned result2 = satd_test_env.tested_func(buf2, buf1); unsigned result2 = satd_test_env.tested_func(buf2, buf1);
@ -128,8 +128,8 @@ TEST satd_test_checkers(void)
const int test = 1; const int test = 1;
pixel_t * buf1 = satd_bufs[test][satd_test_env.log_width][0]; kvz_pixel * buf1 = satd_bufs[test][satd_test_env.log_width][0];
pixel_t * buf2 = satd_bufs[test][satd_test_env.log_width][1]; kvz_pixel * buf2 = satd_bufs[test][satd_test_env.log_width][1];
unsigned result1 = satd_test_env.tested_func(buf1, buf2); unsigned result1 = satd_test_env.tested_func(buf1, buf2);
unsigned result2 = satd_test_env.tested_func(buf2, buf1); unsigned result2 = satd_test_env.tested_func(buf2, buf1);
@ -147,8 +147,8 @@ TEST satd_test_gradient(void)
const int test = 2; const int test = 2;
pixel_t * buf1 = satd_bufs[test][satd_test_env.log_width][0]; kvz_pixel * buf1 = satd_bufs[test][satd_test_env.log_width][0];
pixel_t * buf2 = satd_bufs[test][satd_test_env.log_width][1]; kvz_pixel * buf2 = satd_bufs[test][satd_test_env.log_width][1];
unsigned result1 = satd_test_env.tested_func(buf1, buf2); unsigned result1 = satd_test_env.tested_func(buf1, buf2);
unsigned result2 = satd_test_env.tested_func(buf2, buf1); unsigned result2 = satd_test_env.tested_func(buf2, buf1);

20
tests/speed_tests.c
View file

@ -40,8 +40,8 @@
////////////////////////////////////////////////////////////////////////// //////////////////////////////////////////////////////////////////////////
// GLOBALS // GLOBALS
pixel_t * bufs[NUM_TESTS]; // SIMD aligned pointers. kvz_pixel * bufs[NUM_TESTS]; // SIMD aligned pointers.
pixel_t * actual_bufs[NUM_TESTS]; // pointers returned by malloc. kvz_pixel * actual_bufs[NUM_TESTS]; // pointers returned by malloc.
static struct test_env_t { static struct test_env_t {
int log_width; // for selecting dim from bufs int log_width; // for selecting dim from bufs
@ -53,7 +53,7 @@ static struct test_env_t {
////////////////////////////////////////////////////////////////////////// //////////////////////////////////////////////////////////////////////////
// SETUP, TEARDOWN AND HELPER FUNCTIONS // SETUP, TEARDOWN AND HELPER FUNCTIONS
static void init_gradient(int x_px, int y_px, int width, int slope, pixel_t *buf) static void init_gradient(int x_px, int y_px, int width, int slope, kvz_pixel *buf)
{ {
for (int y = 0; y < width; ++y) { for (int y = 0; y < width; ++y) {
for (int x = 0; x < width; ++x) { for (int x = 0; x < width; ++x) {
@ -71,7 +71,7 @@ static void setup_tests()
for (int test = 0; test < NUM_TESTS; ++test) { for (int test = 0; test < NUM_TESTS; ++test) {
unsigned size = NUM_CHUNKS * 64 * 64; unsigned size = NUM_CHUNKS * 64 * 64;
actual_bufs[test] = malloc(size * sizeof(pixel_t) + SIMD_ALIGNMENT); actual_bufs[test] = malloc(size * sizeof(kvz_pixel) + SIMD_ALIGNMENT);
bufs[test] = ALIGNED_POINTER(actual_bufs[test], SIMD_ALIGNMENT); bufs[test] = ALIGNED_POINTER(actual_bufs[test], SIMD_ALIGNMENT);
} }
@ -112,9 +112,9 @@ TEST test_intra_speed(const int width)
uint64_t sum = 0; uint64_t sum = 0;
for (int offset = 0; offset < NUM_CHUNKS * 64 * 64; offset += NUM_CHUNKS * size) { for (int offset = 0; offset < NUM_CHUNKS * 64 * 64; offset += NUM_CHUNKS * size) {
// Compare the first chunk against the 35 other chunks to simulate real usage. // Compare the first chunk against the 35 other chunks to simulate real usage.
pixel_t * buf1 = &bufs[test][offset]; kvz_pixel * buf1 = &bufs[test][offset];
for (int chunk = 1; chunk < NUM_CHUNKS; ++chunk) { for (int chunk = 1; chunk < NUM_CHUNKS; ++chunk) {
pixel_t * buf2 = &bufs[test][chunk * size + offset]; kvz_pixel * buf2 = &bufs[test][chunk * size + offset];
cost_pixel_nxn_func *tested_func = test_env.tested_func; cost_pixel_nxn_func *tested_func = test_env.tested_func;
sum += tested_func(buf1, buf2); sum += tested_func(buf1, buf2);
@ -151,9 +151,9 @@ TEST test_inter_speed(const int width)
for (int offset = 0; offset < NUM_CHUNKS * 64 * 64; offset += NUM_CHUNKS * size) { for (int offset = 0; offset < NUM_CHUNKS * 64 * 64; offset += NUM_CHUNKS * size) {
// Treat 4 consecutive chunks as one chunk with double width and height, // Treat 4 consecutive chunks as one chunk with double width and height,
// and do a 8x8 grid search against the first chunk to simulate real usage. // and do a 8x8 grid search against the first chunk to simulate real usage.
pixel_t * buf1 = &bufs[test][offset]; kvz_pixel * buf1 = &bufs[test][offset];
for (int chunk = 0; chunk < NUM_CHUNKS; chunk += 4) { for (int chunk = 0; chunk < NUM_CHUNKS; chunk += 4) {
pixel_t * buf2 = &bufs[test][chunk * size + offset]; kvz_pixel * buf2 = &bufs[test][chunk * size + offset];
for (int y = 0; y < 8; ++y) { for (int y = 0; y < 8; ++y) {
for (int x = 0; x < 8; ++x) { for (int x = 0; x < 8; ++x) {
const int stride1 = 2 * 64; const int stride1 = 2 * 64;
@ -203,8 +203,8 @@ TEST dct_speed(const int width)
for (int offset = 0; offset < NUM_CHUNKS * 64 * 64; offset += NUM_CHUNKS * size) { for (int offset = 0; offset < NUM_CHUNKS * 64 * 64; offset += NUM_CHUNKS * size) {
// Compare the first chunk against the 35 other chunks to simulate real usage. // Compare the first chunk against the 35 other chunks to simulate real usage.
for (int chunk = 0; chunk < NUM_CHUNKS; ++chunk) { for (int chunk = 0; chunk < NUM_CHUNKS; ++chunk) {
pixel_t * buf1 = &bufs[test][offset]; kvz_pixel * buf1 = &bufs[test][offset];
pixel_t * buf2 = &bufs[test][chunk * size + offset]; kvz_pixel * buf2 = &bufs[test][chunk * size + offset];
for (int p = 0; p < size; ++p) { for (int p = 0; p < size; ++p) {
tmp_residual[p] = (int16_t)(buf1[p] - buf2[p]); tmp_residual[p] = (int16_t)(buf1[p] - buf2[p]);
} }