mirror/opencv
1
0
Fork 0
mirror of https://github.com/opencv/opencv.git synced 2024-11-24 03:00:14 +08:00
Code Issues Actions Packages Projects Releases Wiki Activity

build: fix AVX2/AVX512 builds failed due to intrinsics operator usage

Browse Source
This commit is contained in:
Maksim Shabunin 2024-09-20 13:38:59 +03:00
parent f503890c2b
commit 6ef357fd54
2 changed files with 13 additions and 13 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

10
modules/gapi/src/backends/fluid/gfluidimgproc_simd_avx2.hpp
View File

@ -85,15 +85,15 @@ CV_ALWAYS_INLINE void calcRowLinear32FC1Impl(float *dst[],
v_deinterleave(low1, high1, s00, s01); v_deinterleave(low1, high1, s00, s01);
// v_float32 res0 = s00*alpha0 + s01*alpha1; // v_float32 res0 = s00*alpha0 + s01*alpha1;
v_float32x8 res0 = v_fma(s00 - s01, alpha0, s01); v_float32x8 res0 = v_fma(v_sub(s00, s01), alpha0, s01);
v_gather_pairs(src1[line], &mapsx[x], low2, high2); v_gather_pairs(src1[line], &mapsx[x], low2, high2);
v_deinterleave(low2, high2, s10, s11); v_deinterleave(low2, high2, s10, s11);
// v_float32 res1 = s10*alpha0 + s11*alpha1; // v_float32 res1 = s10*alpha0 + s11*alpha1;
v_float32x8 res1 = v_fma(s10 - s11, alpha0, s11); v_float32x8 res1 = v_fma(v_sub(s10, s11), alpha0, s11);
// v_float32 d = res0*beta0 + res1*beta1; // v_float32 d = res0*beta0 + res1*beta1;
v_float32x8 d = v_fma(res0 - res1, v_beta0, res1); v_float32x8 d = v_fma(v_sub(res0, res1), v_beta0, res1);
v_store(&dst[line][x], d); v_store(&dst[line][x], d);
} }
@ -126,7 +126,7 @@ CV_ALWAYS_INLINE void calcRowLinear32FC1Impl(float *dst[],
v_deinterleave(low, high, s00, s01); v_deinterleave(low, high, s00, s01);
// v_float32 d = s00*alpha0 + s01*alpha1; // v_float32 d = s00*alpha0 + s01*alpha1;
v_float32x8 d = v_fma(s00 - s01, alpha0, s01); v_float32x8 d = v_fma(v_sub(s00, s01), alpha0, s01);
v_store(&dst[line][x], d); v_store(&dst[line][x], d);
} }
@ -157,7 +157,7 @@ CV_ALWAYS_INLINE void calcRowLinear32FC1Impl(float *dst[],
v_float32x8 s1 = v256_load(&src1[line][x]); v_float32x8 s1 = v256_load(&src1[line][x]);
// v_float32 d = s0*beta0 + s1*beta1; // v_float32 d = s0*beta0 + s1*beta1;
v_float32x8 d = v_fma(s0 - s1, v_beta0, s1); v_float32x8 d = v_fma(v_sub(s0, s1), v_beta0, s1);
v_store(&dst[line][x], d); v_store(&dst[line][x], d);
} }

16
modules/imgproc/src/resize.cpp
View File

@ -2536,7 +2536,7 @@ public:
#elif CV_SIMD_WIDTH == 64 #elif CV_SIMD_WIDTH == 64
v_zip(t0, t3, s0, s1); v_zip(t1, t4, s2, s3); v_zip(t2, t5, s4, s5); v_zip(t0, t3, s0, s1); v_zip(t1, t4, s2, s3); v_zip(t2, t5, s4, s5);
v_zip(s0, s3, t0, t1); v_zip(s1, s4, t2, t3); v_zip(s2, s5, t4, t5); v_zip(s0, s3, t0, t1); v_zip(s1, s4, t2, t3); v_zip(s2, s5, t4, t5);
bl = t0 + t3; gl = t1 + t4; rl = t2 + t5; bl = v_add(t0, t3); gl = v_add(t1, t4); rl = v_add(t2, t5);
#endif #endif
s0 = v_add(vx_load_expand(S0 + 6 * VTraits<v_uint16>::vlanes()), vx_load_expand(S1 + 6 * VTraits<v_uint16>::vlanes())); s0 = v_add(vx_load_expand(S0 + 6 * VTraits<v_uint16>::vlanes()), vx_load_expand(S1 + 6 * VTraits<v_uint16>::vlanes()));
s1 = v_add(vx_load_expand(S0 + 7 * VTraits<v_uint16>::vlanes()), vx_load_expand(S1 + 7 * VTraits<v_uint16>::vlanes())); s1 = v_add(vx_load_expand(S0 + 7 * VTraits<v_uint16>::vlanes()), vx_load_expand(S1 + 7 * VTraits<v_uint16>::vlanes()));
@ -2556,7 +2556,7 @@ public:
#elif CV_SIMD_WIDTH == 64 #elif CV_SIMD_WIDTH == 64
v_zip(t0, t3, s0, s1); v_zip(t1, t4, s2, s3); v_zip(t2, t5, s4, s5); v_zip(t0, t3, s0, s1); v_zip(t1, t4, s2, s3); v_zip(t2, t5, s4, s5);
v_zip(s0, s3, t0, t1); v_zip(s1, s4, t2, t3); v_zip(s2, s5, t4, t5); v_zip(s0, s3, t0, t1); v_zip(s1, s4, t2, t3); v_zip(s2, s5, t4, t5);
bh = t0 + t3; gh = t1 + t4; rh = t2 + t5; bh = v_add(t0, t3); gh = v_add(t1, t4); rh = v_add(t2, t5);
#endif #endif
v_store_interleave(D, v_rshr_pack<2>(bl, bh), v_rshr_pack<2>(gl, gh), v_rshr_pack<2>(rl, rh)); v_store_interleave(D, v_rshr_pack<2>(bl, bh), v_rshr_pack<2>(gl, gh), v_rshr_pack<2>(rl, rh));
} }
@ -2643,7 +2643,7 @@ public:
bl = v_add(t0, t3); gl = v_add(t1, t4); rl = v_add(t2, t5); bl = v_add(t0, t3); gl = v_add(t1, t4); rl = v_add(t2, t5);
#else //CV_SIMD_WIDTH == 64 #else //CV_SIMD_WIDTH == 64
v_zip(t0, t3, s0, s1); v_zip(t1, t4, s2, s3); v_zip(t2, t5, s4, s5); v_zip(t0, t3, s0, s1); v_zip(t1, t4, s2, s3); v_zip(t2, t5, s4, s5);
bl = s0 + s3; gl = s1 + s4; rl = s2 + s5; bl = v_add(s0, s3); gl = v_add(s1, s4); rl = v_add(s2, s5);
#endif #endif
s0 = v_add(vx_load_expand(S0 + 6 * VTraits<v_uint32>::vlanes()), vx_load_expand(S1 + 6 * VTraits<v_uint32>::vlanes())); s0 = v_add(vx_load_expand(S0 + 6 * VTraits<v_uint32>::vlanes()), vx_load_expand(S1 + 6 * VTraits<v_uint32>::vlanes()));
s1 = v_add(vx_load_expand(S0 + 7 * VTraits<v_uint32>::vlanes()), vx_load_expand(S1 + 7 * VTraits<v_uint32>::vlanes())); s1 = v_add(vx_load_expand(S0 + 7 * VTraits<v_uint32>::vlanes()), vx_load_expand(S1 + 7 * VTraits<v_uint32>::vlanes()));
@ -2659,7 +2659,7 @@ public:
bh = v_add(t0, t3); gh = v_add(t1, t4); rh = v_add(t2, t5); bh = v_add(t0, t3); gh = v_add(t1, t4); rh = v_add(t2, t5);
#else //CV_SIMD_WIDTH == 64 #else //CV_SIMD_WIDTH == 64
v_zip(t0, t3, s0, s1); v_zip(t1, t4, s2, s3); v_zip(t2, t5, s4, s5); v_zip(t0, t3, s0, s1); v_zip(t1, t4, s2, s3); v_zip(t2, t5, s4, s5);
bh = s0 + s3; gh = s1 + s4; rh = s2 + s5; bh = v_add(s0, s3); gh = v_add(s1, s4); rh = v_add(s2, s5);
#endif #endif
v_store_interleave(D, v_rshr_pack<2>(bl, bh), v_rshr_pack<2>(gl, gh), v_rshr_pack<2>(rl, rh)); v_store_interleave(D, v_rshr_pack<2>(bl, bh), v_rshr_pack<2>(gl, gh), v_rshr_pack<2>(rl, rh));
} }
@ -2697,7 +2697,7 @@ public:
v_expand(v_reinterpret_as_u16(r01), r01l, r01h); v_expand(v_reinterpret_as_u16(r01), r01l, r01h);
v_expand(v_reinterpret_as_u16(r10), r10l, r10h); v_expand(v_reinterpret_as_u16(r10), r10l, r10h);
v_expand(v_reinterpret_as_u16(r11), r11l, r11h); v_expand(v_reinterpret_as_u16(r11), r11l, r11h);
v_store(D, v_rshr_pack<2>(r00l + r01l + r10l + r11l, r00h + r01h + r10h + r11h)); v_store(D, v_rshr_pack<2>(v_add(r00l, r01l, r10l, r11l), v_add(r00h, r01h, r10h, r11h)));
} }
#else #else
for ( ; dx <= w - VTraits<v_uint32>::vlanes(); dx += VTraits<v_uint32>::vlanes(), S0 += VTraits<v_uint16>::vlanes(), S1 += VTraits<v_uint16>::vlanes(), D += VTraits<v_uint32>::vlanes()) for ( ; dx <= w - VTraits<v_uint32>::vlanes(); dx += VTraits<v_uint32>::vlanes(), S0 += VTraits<v_uint16>::vlanes(), S1 += VTraits<v_uint16>::vlanes(), D += VTraits<v_uint32>::vlanes())
@ -2773,7 +2773,7 @@ public:
bl = v_add(t0, t3); gl = v_add(t1, t4); rl = v_add(t2, t5); bl = v_add(t0, t3); gl = v_add(t1, t4); rl = v_add(t2, t5);
#else //CV_SIMD_WIDTH == 64 #else //CV_SIMD_WIDTH == 64
v_zip(t0, t3, s0, s1); v_zip(t1, t4, s2, s3); v_zip(t2, t5, s4, s5); v_zip(t0, t3, s0, s1); v_zip(t1, t4, s2, s3); v_zip(t2, t5, s4, s5);
bl = s0 + s3; gl = s1 + s4; rl = s2 + s5; bl = v_add(s0, s3); gl = v_add(s1, s4); rl = v_add(s2, s5);
#endif #endif
s0 = v_add(vx_load_expand(S0 + 6 * VTraits<v_int32>::vlanes()), vx_load_expand(S1 + 6 * VTraits<v_int32>::vlanes())); s0 = v_add(vx_load_expand(S0 + 6 * VTraits<v_int32>::vlanes()), vx_load_expand(S1 + 6 * VTraits<v_int32>::vlanes()));
s1 = v_add(vx_load_expand(S0 + 7 * VTraits<v_int32>::vlanes()), vx_load_expand(S1 + 7 * VTraits<v_int32>::vlanes())); s1 = v_add(vx_load_expand(S0 + 7 * VTraits<v_int32>::vlanes()), vx_load_expand(S1 + 7 * VTraits<v_int32>::vlanes()));
@ -2789,7 +2789,7 @@ public:
bh = v_add(t0, t3); gh = v_add(t1, t4); rh = v_add(t2, t5); bh = v_add(t0, t3); gh = v_add(t1, t4); rh = v_add(t2, t5);
#else //CV_SIMD_WIDTH == 64 #else //CV_SIMD_WIDTH == 64
v_zip(t0, t3, s0, s1); v_zip(t1, t4, s2, s3); v_zip(t2, t5, s4, s5); v_zip(t0, t3, s0, s1); v_zip(t1, t4, s2, s3); v_zip(t2, t5, s4, s5);
bh = s0 + s3; gh = s1 + s4; rh = s2 + s5; bh = v_add(s0, s3); gh = v_add(s1, s4); rh = v_add(s2, s5);
#endif #endif
v_store_interleave(D, v_rshr_pack<2>(bl, bh), v_rshr_pack<2>(gl, gh), v_rshr_pack<2>(rl, rh)); v_store_interleave(D, v_rshr_pack<2>(bl, bh), v_rshr_pack<2>(gl, gh), v_rshr_pack<2>(rl, rh));
} }
@ -2826,7 +2826,7 @@ public:
v_expand(v_reinterpret_as_s16(r01), r01l, r01h); v_expand(v_reinterpret_as_s16(r01), r01l, r01h);
v_expand(v_reinterpret_as_s16(r10), r10l, r10h); v_expand(v_reinterpret_as_s16(r10), r10l, r10h);
v_expand(v_reinterpret_as_s16(r11), r11l, r11h); v_expand(v_reinterpret_as_s16(r11), r11l, r11h);
v_store(D, v_rshr_pack<2>(r00l + r01l + r10l + r11l, r00h + r01h + r10h + r11h)); v_store(D, v_rshr_pack<2>(v_add(r00l, r01l, r10l, r11l), v_add(r00h, r01h, r10h, r11h)));
#else #else
v_int32 r0, r1, r2, r3; v_int32 r0, r1, r2, r3;
r0 = v_add(vx_load_expand(S0), vx_load_expand(S1)); r0 = v_add(vx_load_expand(S0), vx_load_expand(S1));