#include "libyuv/scale.h"
#include <assert.h>
#include <string.h>
#include "libyuv/cpu_id.h"
#include "libyuv/planar_functions.h"
#include "libyuv/row.h"
#include "libyuv/scale_argb.h"
#include "libyuv/scale_row.h"
#ifdef __cplusplus
namespace libyuv {
extern "C" {
#endif
static __inline int Abs(int v) { … }
static void ScaleARGBDown2(int src_width,
int src_height,
int dst_width,
int dst_height,
int src_stride,
int dst_stride,
const uint8_t* src_argb,
uint8_t* dst_argb,
int x,
int dx,
int y,
int dy,
enum FilterMode filtering) { … }
static int ScaleARGBDown4Box(int src_width,
int src_height,
int dst_width,
int dst_height,
int src_stride,
int dst_stride,
const uint8_t* src_argb,
uint8_t* dst_argb,
int x,
int dx,
int y,
int dy) { … }
static void ScaleARGBDownEven(int src_width,
int src_height,
int dst_width,
int dst_height,
int src_stride,
int dst_stride,
const uint8_t* src_argb,
uint8_t* dst_argb,
int x,
int dx,
int y,
int dy,
enum FilterMode filtering) { … }
static int ScaleARGBBilinearDown(int src_width,
int src_height,
int dst_width,
int dst_height,
int src_stride,
int dst_stride,
const uint8_t* src_argb,
uint8_t* dst_argb,
int x,
int dx,
int y,
int dy,
enum FilterMode filtering) { … }
static int ScaleARGBBilinearUp(int src_width,
int src_height,
int dst_width,
int dst_height,
int src_stride,
int dst_stride,
const uint8_t* src_argb,
uint8_t* dst_argb,
int x,
int dx,
int y,
int dy,
enum FilterMode filtering) { … }
#ifdef YUVSCALEUP
static int ScaleYUVToARGBBilinearUp(int src_width,
int src_height,
int dst_width,
int dst_height,
int src_stride_y,
int src_stride_u,
int src_stride_v,
int dst_stride_argb,
const uint8_t* src_y,
const uint8_t* src_u,
const uint8_t* src_v,
uint8_t* dst_argb,
int x,
int dx,
int y,
int dy,
enum FilterMode filtering) {
int j;
void (*I422ToARGBRow)(const uint8_t* y_buf, const uint8_t* u_buf,
const uint8_t* v_buf, uint8_t* rgb_buf, int width) =
I422ToARGBRow_C;
#if defined(HAS_I422TOARGBROW_SSSE3)
if (TestCpuFlag(kCpuHasSSSE3)) {
I422ToARGBRow = I422ToARGBRow_Any_SSSE3;
if (IS_ALIGNED(src_width, 8)) {
I422ToARGBRow = I422ToARGBRow_SSSE3;
}
}
#endif
#if defined(HAS_I422TOARGBROW_AVX2)
if (TestCpuFlag(kCpuHasAVX2)) {
I422ToARGBRow = I422ToARGBRow_Any_AVX2;
if (IS_ALIGNED(src_width, 16)) {
I422ToARGBRow = I422ToARGBRow_AVX2;
}
}
#endif
#if defined(HAS_I422TOARGBROW_AVX512BW)
if (TestCpuFlag(kCpuHasAVX512BW | kCpuHasAVX512VL) ==
(kCpuHasAVX512BW | kCpuHasAVX512VL)) {
I422ToARGBRow = I422ToARGBRow_Any_AVX512BW;
if (IS_ALIGNED(src_width, 32)) {
I422ToARGBRow = I422ToARGBRow_AVX512BW;
}
}
#endif
#if defined(HAS_I422TOARGBROW_NEON)
if (TestCpuFlag(kCpuHasNEON)) {
I422ToARGBRow = I422ToARGBRow_Any_NEON;
if (IS_ALIGNED(src_width, 8)) {
I422ToARGBRow = I422ToARGBRow_NEON;
}
}
#endif
#if defined(HAS_I422TOARGBROW_SVE2)
if (TestCpuFlag(kCpuHasSVE2)) {
I422ToARGBRow = I422ToARGBRow_SVE2;
}
#endif
#if defined(HAS_I422TOARGBROW_MSA)
if (TestCpuFlag(kCpuHasMSA)) {
I422ToARGBRow = I422ToARGBRow_Any_MSA;
if (IS_ALIGNED(src_width, 8)) {
I422ToARGBRow = I422ToARGBRow_MSA;
}
}
#endif
#if defined(HAS_I422TOARGBROW_LSX)
if (TestCpuFlag(kCpuHasLSX)) {
I422ToARGBRow = I422ToARGBRow_Any_LSX;
if (IS_ALIGNED(src_width, 16)) {
I422ToARGBRow = I422ToARGBRow_LSX;
}
}
#endif
#if defined(HAS_I422TOARGBROW_LASX)
if (TestCpuFlag(kCpuHasLASX)) {
I422ToARGBRow = I422ToARGBRow_Any_LASX;
if (IS_ALIGNED(src_width, 32)) {
I422ToARGBRow = I422ToARGBRow_LASX;
}
}
#endif
#if defined(HAS_I422TOARGBROW_RVV)
if (TestCpuFlag(kCpuHasRVV)) {
I422ToARGBRow = I422ToARGBRow_RVV;
}
#endif
void (*InterpolateRow)(uint8_t* dst_argb, const uint8_t* src_argb,
ptrdiff_t src_stride, int dst_width,
int source_y_fraction) = InterpolateRow_C;
#if defined(HAS_INTERPOLATEROW_SSSE3)
if (TestCpuFlag(kCpuHasSSSE3)) {
InterpolateRow = InterpolateRow_Any_SSSE3;
if (IS_ALIGNED(dst_width, 4)) {
InterpolateRow = InterpolateRow_SSSE3;
}
}
#endif
#if defined(HAS_INTERPOLATEROW_AVX2)
if (TestCpuFlag(kCpuHasAVX2)) {
InterpolateRow = InterpolateRow_Any_AVX2;
if (IS_ALIGNED(dst_width, 8)) {
InterpolateRow = InterpolateRow_AVX2;
}
}
#endif
#if defined(HAS_INTERPOLATEROW_NEON)
if (TestCpuFlag(kCpuHasNEON)) {
InterpolateRow = InterpolateRow_Any_NEON;
if (IS_ALIGNED(dst_width, 4)) {
InterpolateRow = InterpolateRow_NEON;
}
}
#endif
#if defined(HAS_INTERPOLATEROW_MSA)
if (TestCpuFlag(kCpuHasMSA)) {
InterpolateRow = InterpolateRow_Any_MSA;
if (IS_ALIGNED(dst_width, 8)) {
InterpolateRow = InterpolateRow_MSA;
}
}
#endif
#if defined(HAS_INTERPOLATEROW_LSX)
if (TestCpuFlag(kCpuHasLSX)) {
InterpolateRow = InterpolateRow_Any_LSX;
if (IS_ALIGNED(dst_width, 8)) {
InterpolateRow = InterpolateRow_LSX;
}
}
#endif
#if defined(HAS_INTERPOLATEROW_RVV)
if (TestCpuFlag(kCpuHasRVV)) {
InterpolateRow = InterpolateRow_RVV;
}
#endif
void (*ScaleARGBFilterCols)(uint8_t* dst_argb, const uint8_t* src_argb,
int dst_width, int x, int dx) =
filtering ? ScaleARGBFilterCols_C : ScaleARGBCols_C;
if (src_width >= 32768) {
ScaleARGBFilterCols =
filtering ? ScaleARGBFilterCols64_C : ScaleARGBCols64_C;
}
#if defined(HAS_SCALEARGBFILTERCOLS_SSSE3)
if (filtering && TestCpuFlag(kCpuHasSSSE3) && src_width < 32768) {
ScaleARGBFilterCols = ScaleARGBFilterCols_SSSE3;
}
#endif
#if defined(HAS_SCALEARGBFILTERCOLS_NEON)
if (filtering && TestCpuFlag(kCpuHasNEON)) {
ScaleARGBFilterCols = ScaleARGBFilterCols_Any_NEON;
if (IS_ALIGNED(dst_width, 4)) {
ScaleARGBFilterCols = ScaleARGBFilterCols_NEON;
}
}
#endif
#if defined(HAS_SCALEARGBFILTERCOLS_MSA)
if (filtering && TestCpuFlag(kCpuHasMSA)) {
ScaleARGBFilterCols = ScaleARGBFilterCols_Any_MSA;
if (IS_ALIGNED(dst_width, 8)) {
ScaleARGBFilterCols = ScaleARGBFilterCols_MSA;
}
}
#endif
#if defined(HAS_SCALEARGBFILTERCOLS_LSX)
if (filtering && TestCpuFlag(kCpuHasLSX)) {
ScaleARGBFilterCols = ScaleARGBFilterCols_Any_LSX;
if (IS_ALIGNED(dst_width, 8)) {
ScaleARGBFilterCols = ScaleARGBFilterCols_LSX;
}
}
#endif
#if defined(HAS_SCALEARGBCOLS_SSE2)
if (!filtering && TestCpuFlag(kCpuHasSSE2) && src_width < 32768) {
ScaleARGBFilterCols = ScaleARGBCols_SSE2;
}
#endif
#if defined(HAS_SCALEARGBCOLS_NEON)
if (!filtering && TestCpuFlag(kCpuHasNEON)) {
ScaleARGBFilterCols = ScaleARGBCols_Any_NEON;
if (IS_ALIGNED(dst_width, 8)) {
ScaleARGBFilterCols = ScaleARGBCols_NEON;
}
}
#endif
#if defined(HAS_SCALEARGBCOLS_MSA)
if (!filtering && TestCpuFlag(kCpuHasMSA)) {
ScaleARGBFilterCols = ScaleARGBCols_Any_MSA;
if (IS_ALIGNED(dst_width, 4)) {
ScaleARGBFilterCols = ScaleARGBCols_MSA;
}
}
#endif
#if defined(HAS_SCALEARGBCOLS_LSX)
if (!filtering && TestCpuFlag(kCpuHasLSX)) {
ScaleARGBFilterCols = ScaleARGBCols_Any_LSX;
if (IS_ALIGNED(dst_width, 4)) {
ScaleARGBFilterCols = ScaleARGBCols_LSX;
}
}
#endif
if (!filtering && src_width * 2 == dst_width && x < 0x8000) {
ScaleARGBFilterCols = ScaleARGBColsUp2_C;
#if defined(HAS_SCALEARGBCOLSUP2_SSE2)
if (TestCpuFlag(kCpuHasSSE2) && IS_ALIGNED(dst_width, 8)) {
ScaleARGBFilterCols = ScaleARGBColsUp2_SSE2;
}
#endif
}
const int max_y = (src_height - 1) << 16;
if (y > max_y) {
y = max_y;
}
const int kYShift = 1;
int yi = y >> 16;
int uv_yi = yi >> kYShift;
const uint8_t* src_row_y = src_y + yi * (intptr_t)src_stride_y;
const uint8_t* src_row_u = src_u + uv_yi * (intptr_t)src_stride_u;
const uint8_t* src_row_v = src_v + uv_yi * (intptr_t)src_stride_v;
const int row_size = (dst_width * 4 + 31) & ~31;
align_buffer_64(row, row_size * 2 + src_width * 4);
uint8_t* argb_row = row + row_size * 2;
uint8_t* rowptr = row;
int rowstride = row_size;
int lasty = yi;
if (!row)
return 1;
ScaleARGBFilterCols(rowptr, src_row_y, dst_width, x, dx);
if (src_height > 1) {
src_row_y += src_stride_y;
if (yi & 1) {
src_row_u += src_stride_u;
src_row_v += src_stride_v;
}
}
ScaleARGBFilterCols(rowptr + rowstride, src_row_y, dst_width, x, dx);
if (src_height > 2) {
src_row_y += src_stride_y;
if (!(yi & 1)) {
src_row_u += src_stride_u;
src_row_v += src_stride_v;
}
}
for (j = 0; j < dst_height; ++j) {
yi = y >> 16;
if (yi != lasty) {
if (y > max_y) {
y = max_y;
yi = y >> 16;
uv_yi = yi >> kYShift;
src_row_y = src_y + yi * (intptr_t)src_stride_y;
src_row_u = src_u + uv_yi * (intptr_t)src_stride_u;
src_row_v = src_v + uv_yi * (intptr_t)src_stride_v;
}
if (yi != lasty) {
I422ToARGBRow(src_row_y, src_row_u, src_row_v, argb_row, src_width);
ScaleARGBFilterCols(rowptr, argb_row, dst_width, x, dx);
rowptr += rowstride;
rowstride = -rowstride;
lasty = yi;
src_row_y += src_stride_y;
if (yi & 1) {
src_row_u += src_stride_u;
src_row_v += src_stride_v;
}
}
}
if (filtering == kFilterLinear) {
InterpolateRow(dst_argb, rowptr, 0, dst_width * 4, 0);
} else {
int yf = (y >> 8) & 255;
InterpolateRow(dst_argb, rowptr, rowstride, dst_width * 4, yf);
}
dst_argb += dst_stride_argb;
y += dy;
}
free_aligned_buffer_64(row);
return 0;
}
#endif
static void ScaleARGBSimple(int src_width,
int src_height,
int dst_width,
int dst_height,
int src_stride,
int dst_stride,
const uint8_t* src_argb,
uint8_t* dst_argb,
int x,
int dx,
int y,
int dy) { … }
static int ScaleARGB(const uint8_t* src,
int src_stride,
int src_width,
int src_height,
uint8_t* dst,
int dst_stride,
int dst_width,
int dst_height,
int clip_x,
int clip_y,
int clip_width,
int clip_height,
enum FilterMode filtering) { … }
LIBYUV_API
int ARGBScaleClip(const uint8_t* src_argb,
int src_stride_argb,
int src_width,
int src_height,
uint8_t* dst_argb,
int dst_stride_argb,
int dst_width,
int dst_height,
int clip_x,
int clip_y,
int clip_width,
int clip_height,
enum FilterMode filtering) { … }
LIBYUV_API
int ARGBScale(const uint8_t* src_argb,
int src_stride_argb,
int src_width,
int src_height,
uint8_t* dst_argb,
int dst_stride_argb,
int dst_width,
int dst_height,
enum FilterMode filtering) { … }
LIBYUV_API
int YUVToARGBScaleClip(const uint8_t* src_y,
int src_stride_y,
const uint8_t* src_u,
int src_stride_u,
const uint8_t* src_v,
int src_stride_v,
uint32_t src_fourcc,
int src_width,
int src_height,
uint8_t* dst_argb,
int dst_stride_argb,
uint32_t dst_fourcc,
int dst_width,
int dst_height,
int clip_x,
int clip_y,
int clip_width,
int clip_height,
enum FilterMode filtering) { … }
#ifdef __cplusplus
}
}
#endif