potok/opencv
1
0
Fork 0
mirror of https://github.com/opencv/opencv.git synced 2026-07-21 19:33:03 +04:00
Code Packages Releases Wiki Activity

core: add RVV convertScale data type conversions

Browse Source
This commit is contained in:
Ziyuan Li
2026-05-29 13:24:03 +08:00
parent 3bb68212da
commit 18dad63e2e
1 changed files with 148 additions and 0 deletions
Download Patch File Download Diff File Expand all files Collapse all files
hal/riscv-rvv/src/core/convert_scale.cpp
+148
View File
@@ -62,6 +62,141 @@ inline int convertScale_8U32F(const uchar* src, size_t src_step, uchar* dst, siz
return CV_HAL_ERROR_OK;
}
inline int convertScale_8U16U(const uchar* src, size_t src_step, uchar* dst, size_t dst_step, int width, int height, double alpha, double beta)
{
if (alpha == 1.0 && beta == 0.0)
{
for (int i = 0; i < height; i++)
{
const uchar* src_row = src + i * src_step;
ushort* dst_row = reinterpret_cast<ushort*>(dst + i * dst_step);
int vl;
for (int j = 0; j < width; j += vl)
{
vl = __riscv_vsetvl_e8m2(width - j);
auto vec_src = __riscv_vle8_v_u8m2(src_row + j, vl);
auto vec_dst = __riscv_vzext_vf2(vec_src, vl);
__riscv_vse16_v_u16m4(dst_row + j, vec_dst, vl);
}
}
return CV_HAL_ERROR_OK;
}
int vlmax = __riscv_vsetvlmax_e32m8();
auto vec_b = __riscv_vfmv_v_f_f32m8(beta, vlmax);
float a = alpha;
for (int i = 0; i < height; i++)
{
const uchar* src_row = src + i * src_step;
ushort* dst_row = reinterpret_cast<ushort*>(dst + i * dst_step);
int vl;
for (int j = 0; j < width; j += vl)
{
vl = __riscv_vsetvl_e8m2(width - j);
auto vec_src = __riscv_vle8_v_u8m2(src_row + j, vl);
auto vec_src_u16 = __riscv_vzext_vf2(vec_src, vl);
auto vec_src_f32 = __riscv_vfwcvt_f(vec_src_u16, vl);
auto vec_fma = __riscv_vfmadd(vec_src_f32, a, vec_b, vl);
auto vec_dst = __riscv_vfncvt_xu(vec_fma, vl);
__riscv_vse16_v_u16m4(dst_row + j, vec_dst, vl);
}
}
return CV_HAL_ERROR_OK;
}
inline int convertScale_8U16S(const uchar* src, size_t src_step, uchar* dst, size_t dst_step, int width, int height, double alpha, double beta)
{
if (alpha == 1.0 && beta == 0.0)
{
for (int i = 0; i < height; i++)
{
const uchar* src_row = src + i * src_step;
short* dst_row = reinterpret_cast<short*>(dst + i * dst_step);
int vl;
for (int j = 0; j < width; j += vl)
{
vl = __riscv_vsetvl_e8m2(width - j);
auto vec_src = __riscv_vle8_v_u8m2(src_row + j, vl);
auto vec_dst = __riscv_vreinterpret_v_u16m4_i16m4(__riscv_vzext_vf2(vec_src, vl));
__riscv_vse16_v_i16m4(dst_row + j, vec_dst, vl);
}
}
return CV_HAL_ERROR_OK;
}
int vlmax = __riscv_vsetvlmax_e32m8();
auto vec_b = __riscv_vfmv_v_f_f32m8(beta, vlmax);
float a = alpha;
for (int i = 0; i < height; i++)
{
const uchar* src_row = src + i * src_step;
short* dst_row = reinterpret_cast<short*>(dst + i * dst_step);
int vl;
for (int j = 0; j < width; j += vl)
{
vl = __riscv_vsetvl_e8m2(width - j);
auto vec_src = __riscv_vle8_v_u8m2(src_row + j, vl);
auto vec_src_u16 = __riscv_vzext_vf2(vec_src, vl);
auto vec_src_f32 = __riscv_vfwcvt_f(vec_src_u16, vl);
auto vec_fma = __riscv_vfmadd(vec_src_f32, a, vec_b, vl);
auto vec_dst = __riscv_vfncvt_x(vec_fma, vl);
__riscv_vse16_v_i16m4(dst_row + j, vec_dst, vl);
}
}
return CV_HAL_ERROR_OK;
}
inline int convertScale_16U8U(const uchar* src, size_t src_step, uchar* dst, size_t dst_step, int width, int height, double alpha, double beta)
{
if (alpha == 1.0 && beta == 0.0)
{
for (int i = 0; i < height; i++)
{
const ushort* src_row = reinterpret_cast<const ushort*>(src + i * src_step);
uchar* dst_row = dst + i * dst_step;
int vl;
for (int j = 0; j < width; j += vl)
{
vl = __riscv_vsetvl_e16m4(width - j);
auto vec_src = __riscv_vle16_v_u16m4(src_row + j, vl);
auto vec_dst = __riscv_vnclipu(vec_src, 0, __RISCV_VXRM_RNU, vl);
__riscv_vse8_v_u8m2(dst_row + j, vec_dst, vl);
}
}
return CV_HAL_ERROR_OK;
}
int vlmax = __riscv_vsetvlmax_e32m8();
auto vec_b = __riscv_vfmv_v_f_f32m8(beta, vlmax);
float a = alpha;
for (int i = 0; i < height; i++)
{
const ushort* src_row = reinterpret_cast<const ushort*>(src + i * src_step);
uchar* dst_row = dst + i * dst_step;
int vl;
for (int j = 0; j < width; j += vl)
{
vl = __riscv_vsetvl_e16m4(width - j);
auto vec_src = __riscv_vle16_v_u16m4(src_row + j, vl);
auto vec_src_f32 = __riscv_vfwcvt_f(vec_src, vl);
auto vec_fma = __riscv_vfmadd(vec_src_f32, a, vec_b, vl);
auto vec_dst_u16 = __riscv_vfncvt_xu(vec_fma, vl);
auto vec_dst = __riscv_vnclipu(vec_dst_u16, 0, __RISCV_VXRM_RNU, vl);
__riscv_vse8_v_u8m2(dst_row + j, vec_dst, vl);
}
}
return CV_HAL_ERROR_OK;
}
inline int convertScale_16U32F(const uchar* src, size_t src_step, uchar* dst, size_t dst_step, int width, int height, double alpha, double beta)
{
int vlmax = __riscv_vsetvlmax_e32m8();
@@ -164,6 +299,13 @@ int convertScale(const uchar* src, size_t src_step, uchar* dst, size_t dst_step,
if (!dst)
return CV_HAL_ERROR_OK;
if ((unsigned)ddepth >= CV_DEPTH_MAX)
return CV_HAL_ERROR_NOT_IMPLEMENTED;
// Column-shaped sources can target row-shaped continuous vectors.
if (width == 1 && height > 1 && dst_step != CV_ELEM_SIZE1(ddepth))
return CV_HAL_ERROR_NOT_IMPLEMENTED;
switch (sdepth)
{
case CV_8U:
@@ -171,6 +313,10 @@ int convertScale(const uchar* src, size_t src_step, uchar* dst, size_t dst_step,
{
case CV_8U:
return convertScale_8U8U(src, src_step, dst, dst_step, width, height, alpha, beta);
case CV_16U:
return convertScale_8U16U(src, src_step, dst, dst_step, width, height, alpha, beta);
case CV_16S:
return convertScale_8U16S(src, src_step, dst, dst_step, width, height, alpha, beta);
case CV_32F:
return convertScale_8U32F(src, src_step, dst, dst_step, width, height, alpha, beta);
}
@@ -178,6 +324,8 @@ int convertScale(const uchar* src, size_t src_step, uchar* dst, size_t dst_step,
case CV_16U:
switch (ddepth)
{
case CV_8U:
return convertScale_16U8U(src, src_step, dst, dst_step, width, height, alpha, beta);
case CV_32F:
return convertScale_16U32F(src, src_step, dst, dst_step, width, height, alpha, beta);
}