Movatterモバイル変換

[0]ホーム
URL:

Skip to content

Navigation Menu

Sign in
Appearance settings

Search code, repositories, users, issues, pull requests...

Provide feedback

We read every piece of feedback, and take your input very seriously.

Saved searches

Use saved searches to filter your results more quickly

 Sign up
Appearance settings

Add RISC-V HAL implementation for cv::dft and cv::dct#26865

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to ourterms of service andprivacy statement. We’ll occasionally send you account related emails.

Already on GitHub?Sign in to your account

Merged
asmorkalov merged 12 commits intoopencv:4.xfromamane-ame:dxt_hal_rvv
Mar 7, 2025
Merged
Show file tree
Hide file tree
Changes fromall commits
Commits
Show all changes
12 commits
Select commitHold shift + click to select a range
0f8e18f
Draft of HAL RVV cv::dft with radix 2/3/4/5, keeping large radix as s…
amane-ameJan 28, 2025
582f68b
Add HAL RVV for dxt large radix.
amane-ameJan 29, 2025
32f36b9
Convert the function into templates.
amane-ameJan 31, 2025
4503b1a
Merge branch '4.x' into dxt_hal_rvv
amane-ameJan 31, 2025
f41fbb7
Fix up about anonymous namespace error.
amane-ameFeb 1, 2025
589ffee
Change dftOcv to dft and add some comments.
amane-ameFeb 3, 2025
e5e1b54
Merge branch '4.x' into dxt_hal_rvv
amane-ameFeb 7, 2025
0d16cdd
Remove indent in namespaces and replaced CV_SWAP.
amane-ameFeb 19, 2025
2d3feab
Move functions into namespaces.
amane-ameFeb 19, 2025
ef84bd6
Remove AutoBuffer.
amane-ameFeb 23, 2025
a6b9071
Use segment load/store instead of strided.
amane-ameFeb 24, 2025
9c598af
Make clang 17 happy.
amane-ameFeb 24, 2025
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
1 change: 1 addition & 0 deletions3rdparty/hal_rvv/hal_rvv.hpp
View file
Open in desktop
Original file line numberDiff line numberDiff line change
Expand Up@@ -22,6 +22,7 @@
#if defined(__riscv_v) && __riscv_v == 1000000
#include "hal_rvv_1p0/merge.hpp" // core
#include "hal_rvv_1p0/mean.hpp" // core
#include "hal_rvv_1p0/dxt.hpp" // core
#include "hal_rvv_1p0/norm.hpp" // core
#include "hal_rvv_1p0/norm_diff.hpp" // core
#include "hal_rvv_1p0/convert_scale.hpp" // core
Expand Down
569 changes: 569 additions & 0 deletions3rdparty/hal_rvv/hal_rvv_1p0/dxt.hpp
View file
Open in desktop

Large diffs are not rendered by default.

105 changes: 51 additions & 54 deletions3rdparty/hal_rvv/hal_rvv_1p0/minmax.hpp
View file
Open in desktop
Original file line numberDiff line numberDiff line change
@@ -1,68 +1,65 @@
// This file is part of OpenCV project.
// It is subject to the license terms in the LICENSE file found in the top-level directory
// of this distribution and at http://opencv.org/license.html.
#ifndefOPENCV_HAL_RVV_MINMAXIDX_HPP_INCLUDED
#defineOPENCV_HAL_RVV_MINMAXIDX_HPP_INCLUDED
#ifndefOPENCV_HAL_RVV_MINMAX_HPP_INCLUDED
#defineOPENCV_HAL_RVV_MINMAX_HPP_INCLUDED

#include <riscv_vector.h>

namespace cv { namespace cv_hal_rvv {
namespace cv { namespace cv_hal_rvv { namespace minmax {

#undef cv_hal_minMaxIdx
#define cv_hal_minMaxIdx cv::cv_hal_rvv::minMaxIdx
#define cv_hal_minMaxIdx cv::cv_hal_rvv::minmax::minMaxIdx
#undef cv_hal_minMaxIdxMaskStep
#define cv_hal_minMaxIdxMaskStep cv::cv_hal_rvv::minMaxIdx
#define cv_hal_minMaxIdxMaskStep cv::cv_hal_rvv::minmax::minMaxIdx

namespace
{
template<typename T> struct rvv;
template<typename T> struct rvv;

#define HAL_RVV_GENERATOR(T, EEW, TYPE, IS_U, EMUL, M_EMUL, B_LEN) \
template<> struct rvv<T> \
{ \
using vec_t = v##IS_U##int##EEW##EMUL##_t; \
using bool_t = vbool##B_LEN##_t; \
static inline size_t vsetvlmax() { return __riscv_vsetvlmax_e##EEW##EMUL(); } \
static inline size_t vsetvl(size_t a) { return __riscv_vsetvl_e##EEW##EMUL(a); } \
static inline vec_t vmv_v_x(T a, size_t b) { return __riscv_vmv_v_x_##TYPE##EMUL(a, b); } \
static inline vec_t vle(const T* a, size_t b) { return __riscv_vle##EEW##_v_##TYPE##EMUL(a, b); } \
static inline vuint8##M_EMUL##_t vle_mask(const uchar* a, size_t b) { return __riscv_vle8_v_u8##M_EMUL(a, b); } \
static inline vec_t vmin_tu(vec_t a, vec_t b, vec_t c, size_t d) { return __riscv_vmin##IS_U##_tu(a, b, c, d); } \
static inline vec_t vmax_tu(vec_t a, vec_t b, vec_t c, size_t d) { return __riscv_vmax##IS_U##_tu(a, b, c, d); } \
static inline vec_t vmin_tumu(bool_t a, vec_t b, vec_t c, vec_t d, size_t e) { return __riscv_vmin##IS_U##_tumu(a, b, c, d, e); } \
static inline vec_t vmax_tumu(bool_t a, vec_t b, vec_t c, vec_t d, size_t e) { return __riscv_vmax##IS_U##_tumu(a, b, c, d, e); } \
static inline vec_t vredmin(vec_t a, vec_t b, size_t c) { return __riscv_vredmin##IS_U(a, b, c); } \
static inline vec_t vredmax(vec_t a, vec_t b, size_t c) { return __riscv_vredmax##IS_U(a, b, c); } \
};
HAL_RVV_GENERATOR(uchar , 8 , u8 , u, m1, m1 , 8 )
HAL_RVV_GENERATOR(schar , 8 , i8 , , m1, m1 , 8 )
HAL_RVV_GENERATOR(ushort, 16, u16, u, m1, mf2, 16)
HAL_RVV_GENERATOR(short , 16, i16, , m1, mf2, 16)
#undef HAL_RVV_GENERATOR
#define HAL_RVV_GENERATOR(T, EEW, TYPE, IS_U, EMUL, M_EMUL, B_LEN) \
template<> struct rvv<T> \
{ \
using vec_t = v##IS_U##int##EEW##EMUL##_t; \
using bool_t = vbool##B_LEN##_t; \
static inline size_t vsetvlmax() { return __riscv_vsetvlmax_e##EEW##EMUL(); } \
static inline size_t vsetvl(size_t a) { return __riscv_vsetvl_e##EEW##EMUL(a); } \
static inline vec_t vmv_v_x(T a, size_t b) { return __riscv_vmv_v_x_##TYPE##EMUL(a, b); } \
static inline vec_t vle(const T* a, size_t b) { return __riscv_vle##EEW##_v_##TYPE##EMUL(a, b); } \
static inline vuint8##M_EMUL##_t vle_mask(const uchar* a, size_t b) { return __riscv_vle8_v_u8##M_EMUL(a, b); } \
static inline vec_t vmin_tu(vec_t a, vec_t b, vec_t c, size_t d) { return __riscv_vmin##IS_U##_tu(a, b, c, d); } \
static inline vec_t vmax_tu(vec_t a, vec_t b, vec_t c, size_t d) { return __riscv_vmax##IS_U##_tu(a, b, c, d); } \
static inline vec_t vmin_tumu(bool_t a, vec_t b, vec_t c, vec_t d, size_t e) { return __riscv_vmin##IS_U##_tumu(a, b, c, d, e); } \
static inline vec_t vmax_tumu(bool_t a, vec_t b, vec_t c, vec_t d, size_t e) { return __riscv_vmax##IS_U##_tumu(a, b, c, d, e); } \
static inline vec_t vredmin(vec_t a, vec_t b, size_t c) { return __riscv_vredmin##IS_U(a, b, c); } \
static inline vec_t vredmax(vec_t a, vec_t b, size_t c) { return __riscv_vredmax##IS_U(a, b, c); } \
};
HAL_RVV_GENERATOR(uchar , 8 , u8 , u, m1, m1 , 8 )
HAL_RVV_GENERATOR(schar , 8 , i8 , , m1, m1 , 8 )
HAL_RVV_GENERATOR(ushort, 16, u16, u, m1, mf2, 16)
HAL_RVV_GENERATOR(short , 16, i16, , m1, mf2, 16)
#undef HAL_RVV_GENERATOR

#define HAL_RVV_GENERATOR(T, NAME, EEW, TYPE, IS_F, F_OR_S, F_OR_X, EMUL, M_EMUL, P_EMUL, B_LEN) \
template<> struct rvv<T> \
{ \
using vec_t = v##NAME##EEW##EMUL##_t; \
using bool_t = vbool##B_LEN##_t; \
static inline size_t vsetvlmax() { return __riscv_vsetvlmax_e##EEW##EMUL(); } \
static inline size_t vsetvl(size_t a) { return __riscv_vsetvl_e##EEW##EMUL(a); } \
static inline vec_t vmv_v_x(T a, size_t b) { return __riscv_v##IS_F##mv_v_##F_OR_X##_##TYPE##EMUL(a, b); } \
static inline vuint32##P_EMUL##_t vid(size_t a) { return __riscv_vid_v_u32##P_EMUL(a); } \
static inline vuint32##P_EMUL##_t vundefined() { return __riscv_vundefined_u32##P_EMUL(); } \
static inline vec_t vle(const T* a, size_t b) { return __riscv_vle##EEW##_v_##TYPE##EMUL(a, b); } \
static inline vuint8##M_EMUL##_t vle_mask(const uchar* a, size_t b) { return __riscv_vle8_v_u8##M_EMUL(a, b); } \
static inline bool_t vmlt(vec_t a, vec_t b, size_t c) { return __riscv_vm##F_OR_S##lt(a, b, c); } \
static inline bool_t vmgt(vec_t a, vec_t b, size_t c) { return __riscv_vm##F_OR_S##gt(a, b, c); } \
static inline bool_t vmlt_mu(bool_t a, bool_t b, vec_t c, vec_t d, size_t e) { return __riscv_vm##F_OR_S##lt##_mu(a, b, c, d, e); } \
static inline bool_t vmgt_mu(bool_t a, bool_t b, vec_t c, vec_t d, size_t e) { return __riscv_vm##F_OR_S##gt##_mu(a, b, c, d, e); } \
static inline T vmv_x_s(vec_t a) { return __riscv_v##IS_F##mv_##F_OR_X(a); } \
};
HAL_RVV_GENERATOR(int , int , 32, i32, , s, x, m4, m1 , m4, 8 )
HAL_RVV_GENERATOR(float , float, 32, f32, f, f, f, m4, m1 , m4, 8 )
HAL_RVV_GENERATOR(double, float, 64, f64, f, f, f, m4, mf2, m2, 16)
#undef HAL_RVV_GENERATOR
}
#define HAL_RVV_GENERATOR(T, NAME, EEW, TYPE, IS_F, F_OR_S, F_OR_X, EMUL, M_EMUL, P_EMUL, B_LEN) \
template<> struct rvv<T> \
{ \
using vec_t = v##NAME##EEW##EMUL##_t; \
using bool_t = vbool##B_LEN##_t; \
static inline size_t vsetvlmax() { return __riscv_vsetvlmax_e##EEW##EMUL(); } \
static inline size_t vsetvl(size_t a) { return __riscv_vsetvl_e##EEW##EMUL(a); } \
static inline vec_t vmv_v_x(T a, size_t b) { return __riscv_v##IS_F##mv_v_##F_OR_X##_##TYPE##EMUL(a, b); } \
static inline vuint32##P_EMUL##_t vid(size_t a) { return __riscv_vid_v_u32##P_EMUL(a); } \
static inline vuint32##P_EMUL##_t vundefined() { return __riscv_vundefined_u32##P_EMUL(); } \
static inline vec_t vle(const T* a, size_t b) { return __riscv_vle##EEW##_v_##TYPE##EMUL(a, b); } \
static inline vuint8##M_EMUL##_t vle_mask(const uchar* a, size_t b) { return __riscv_vle8_v_u8##M_EMUL(a, b); } \
static inline bool_t vmlt(vec_t a, vec_t b, size_t c) { return __riscv_vm##F_OR_S##lt(a, b, c); } \
static inline bool_t vmgt(vec_t a, vec_t b, size_t c) { return __riscv_vm##F_OR_S##gt(a, b, c); } \
static inline bool_t vmlt_mu(bool_t a, bool_t b, vec_t c, vec_t d, size_t e) { return __riscv_vm##F_OR_S##lt##_mu(a, b, c, d, e); } \
static inline bool_t vmgt_mu(bool_t a, bool_t b, vec_t c, vec_t d, size_t e) { return __riscv_vm##F_OR_S##gt##_mu(a, b, c, d, e); } \
static inline T vmv_x_s(vec_t a) { return __riscv_v##IS_F##mv_##F_OR_X(a); } \
};
HAL_RVV_GENERATOR(int , int , 32, i32, , s, x, m4, m1 , m4, 8 )
HAL_RVV_GENERATOR(float , float, 32, f32, f, f, f, m4, m1 , m4, 8 )
HAL_RVV_GENERATOR(double, float, 64, f64, f, f, f, m4, mf2, m2, 16)
#undef HAL_RVV_GENERATOR

template<typename T>
inline int minMaxIdxReadTwice(const uchar* src_data, size_t src_step, int width, int height, double* minVal, double* maxVal,
Expand DownExpand Up@@ -330,6 +327,6 @@ inline int minMaxIdx(const uchar* src_data, size_t src_step, int width, int heig
return CV_HAL_ERROR_NOT_IMPLEMENTED;
}

}}
}}}

#endif
6 changes: 3 additions & 3 deletions3rdparty/hal_rvv/hal_rvv_1p0/norm.hpp
View file
Open in desktop
Original file line numberDiff line numberDiff line change
Expand Up@@ -6,10 +6,10 @@

#include <riscv_vector.h>

namespace cv { namespace cv_hal_rvv {
namespace cv { namespace cv_hal_rvv { namespace norm {

#undef cv_hal_norm
#define cv_hal_norm cv::cv_hal_rvv::norm
#define cv_hal_norm cv::cv_hal_rvv::norm::norm

inline int normInf_8UC1(const uchar* src, size_t src_step, const uchar* mask, size_t mask_step, int width, int height, double* result)
{
Expand DownExpand Up@@ -512,6 +512,6 @@ inline int norm(const uchar* src, size_t src_step, const uchar* mask, size_t mas
return CV_HAL_ERROR_NOT_IMPLEMENTED;
}

}}
}}}

#endif
8 changes: 4 additions & 4 deletions3rdparty/hal_rvv/hal_rvv_1p0/norm_diff.hpp
View file
Open in desktop
Original file line numberDiff line numberDiff line change
Expand Up@@ -6,10 +6,10 @@

#include <riscv_vector.h>

namespace cv { namespace cv_hal_rvv {
namespace cv { namespace cv_hal_rvv { namespace norm_diff {

#undef cv_hal_normDiff
#define cv_hal_normDiff cv::cv_hal_rvv::normDiff
#define cv_hal_normDiff cv::cv_hal_rvv::norm_diff::normDiff

inline int normDiffInf_8UC1(const uchar* src1, size_t src1_step, const uchar* src2, size_t src2_step, const uchar* mask, size_t mask_step, int width, int height, double* result)
{
Expand DownExpand Up@@ -590,7 +590,7 @@ inline int normDiff(const uchar* src1, size_t src1_step, const uchar* src2, size
if(ret == CV_HAL_ERROR_OK && (norm_type & NORM_RELATIVE))
{
double result_;
ret = cv::cv_hal_rvv::norm(src2, src2_step, mask, mask_step, width, height, type, norm_type & ~NORM_RELATIVE, &result_);
ret = cv::cv_hal_rvv::norm::norm(src2, src2_step, mask, mask_step, width, height, type, norm_type & ~NORM_RELATIVE, &result_);
if(ret == CV_HAL_ERROR_OK)
{
*result /= result_ + DBL_EPSILON;
Expand All@@ -600,6 +600,6 @@ inline int normDiff(const uchar* src1, size_t src1_step, const uchar* src2, size
return ret;
}

}}
}}}

#endif
11 changes: 11 additions & 0 deletionsmodules/core/src/dxt.cpp
View file
Open in desktop
Original file line numberDiff line numberDiff line change
Expand Up@@ -844,6 +844,17 @@ DFT(const OcvDftOptions & c, const Complex<T>* src, Complex<T>* dst)
Complex<T> t;
T scale = (T)c.scale;

if(typeid(T) == typeid(float))
{
CALL_HAL(dft, cv_hal_dft, reinterpret_cast<const uchar*>(src), reinterpret_cast<uchar*>(dst), CV_32F,
c.nf, c.factors, c.scale, c.itab, c.wave, c.tab_size, c.n, c.isInverse, c.noPermute);
}
if(typeid(T) == typeid(double))
{
CALL_HAL(dft, cv_hal_dft, reinterpret_cast<const uchar*>(src), reinterpret_cast<uchar*>(dst), CV_64F,
c.nf, c.factors, c.scale, c.itab, c.wave, c.tab_size, c.n, c.isInverse, c.noPermute);
}

if( c.useIpp )
{
#ifdef USE_IPP_DFT
Expand Down
18 changes: 18 additions & 0 deletionsmodules/core/src/hal_replacement.hpp
View file
Open in desktop
Original file line numberDiff line numberDiff line change
Expand Up@@ -756,10 +756,28 @@ inline int hal_ni_dft1D(cvhalDFT *context, const uchar *src, uchar *dst) { retur
*/
inline int hal_ni_dftFree1D(cvhalDFT *context) { return CV_HAL_ERROR_NOT_IMPLEMENTED; }

/**
@param src source data
@param dst destination data
@param depth depth of source
@param nf OcvDftOptions data
@param factors OcvDftOptions data
@param scale OcvDftOptions data
@param itab OcvDftOptions data
@param wave OcvDftOptions data
@param tab_size OcvDftOptions data
@param n OcvDftOptions data
@param isInverse OcvDftOptions data
@param noPermute OcvDftOptions data
*/
inline int hal_ni_dft(const uchar* src, uchar* dst, int depth, int nf, int *factors, double scale, int* itab, void* wave,
int tab_size, int n, bool isInverse, bool noPermute) { return CV_HAL_ERROR_NOT_IMPLEMENTED; }

//! @cond IGNORED
#define cv_hal_dftInit1D hal_ni_dftInit1D
#define cv_hal_dft1D hal_ni_dft1D
#define cv_hal_dftFree1D hal_ni_dftFree1D
#define cv_hal_dft hal_ni_dft
//! @endcond

/**
Expand Down
Loading
[8]ページ先頭
©2009-2025 Movatter.jp