potok/opencv
1
0
Fork 0
mirror of https://github.com/opencv/opencv.git synced 2026-07-30 07:43:03 +04:00
Code Packages Releases Wiki Activity

Merge branch 4.x

Browse Source
This commit is contained in:
Alexander Smorkalov
2026-05-25 16:29:45 +03:00
parent c2594b41bf e0d604378f
commit d8263a9899
54 changed files with 523 additions and 414 deletions
Download Patch File Download Diff File Expand all files Collapse all files
modules/core/include/opencv2/core/cvdef.h
+9 -3
View File
@@ -708,9 +708,15 @@ __CV_ENUM_FLAGS_BITWISE_XOR_EQ (EnumType, EnumType)
#endif
#if defined(__clang__) || defined(__GNUC__)
#define CV_DISABLE_UBSAN __attribute__((no_sanitize("undefined")))
#else
#define CV_DISABLE_UBSAN
# if defined(__has_attribute)
# if __has_attribute(no_sanitize)
# define CV_DISABLE_UBSAN __attribute__((no_sanitize("undefined")))
# endif
# endif
#endif
#ifndef CV_DISABLE_UBSAN
# define CV_DISABLE_UBSAN
#endif
/****************************************************************************************\
modules/core/include/opencv2/core/hal/intrin_rvv_scalable.hpp
+13 -11
View File
@@ -2620,23 +2620,25 @@ inline v_float32 v_matmul(const v_float32& v, const v_float32& mat0,
const v_float32& mat1, const v_float32& mat2,
const v_float32& mat3)
{
vfloat32m2_t res;
res = __riscv_vfmul_vf_f32m2(mat0, v_extract_n(v, 0), VTraits<v_float32>::vlanes());
res = __riscv_vfmacc_vf_f32m2(res, v_extract_n(v, 1), mat1, VTraits<v_float32>::vlanes());
res = __riscv_vfmacc_vf_f32m2(res, v_extract_n(v, 2), mat2, VTraits<v_float32>::vlanes());
res = __riscv_vfmacc_vf_f32m2(res, v_extract_n(v, 3), mat3, VTraits<v_float32>::vlanes());
return res;
const int vl = VTraits<v_float32>::vlanes();
vuint32m2_t idx = __riscv_vand(__riscv_vid_v_u32m2(vl), 0xfffffffc, vl);
v_float32 v0 = __riscv_vrgather(v, idx, vl);
v_float32 v1 = __riscv_vrgather(v, __riscv_vadd(idx, 1, vl), vl);
v_float32 v2 = __riscv_vrgather(v, __riscv_vadd(idx, 2, vl), vl);
v_float32 v3 = __riscv_vrgather(v, __riscv_vadd(idx, 3, vl), vl);
return v_fma(v0, mat0, v_fma(v1, mat1, v_fma(v2, mat2, v_mul(v3, mat3))));
}
// TODO: only 128 bit now.
inline v_float32 v_matmuladd(const v_float32& v, const v_float32& mat0,
const v_float32& mat1, const v_float32& mat2,
const v_float32& a)
{
vfloat32m2_t res = __riscv_vfmul_vf_f32m2(mat0, v_extract_n(v,0), VTraits<v_float32>::vlanes());
res = __riscv_vfmacc_vf_f32m2(res, v_extract_n(v,1), mat1, VTraits<v_float32>::vlanes());
res = __riscv_vfmacc_vf_f32m2(res, v_extract_n(v,2), mat2, VTraits<v_float32>::vlanes());
return __riscv_vfadd(res, a, VTraits<v_float32>::vlanes());
const int vl = VTraits<v_float32>::vlanes();
vuint32m2_t idx = __riscv_vand(__riscv_vid_v_u32m2(vl), 0xfffffffc, vl);
v_float32 v0 = __riscv_vrgather(v, idx, vl);
v_float32 v1 = __riscv_vrgather(v, __riscv_vadd(idx, 1, vl), vl);
v_float32 v2 = __riscv_vrgather(v, __riscv_vadd(idx, 2, vl), vl);
return v_fma(v0, mat0, v_fma(v1, mat1, v_fma(v2, mat2, a)));
}
inline void v_cleanup() {}