1 /*
2 * By downloading, copying, installing or using the software you agree to this license.
3 * If you do not agree to this license, do not download, install,
4 * copy or use the software.
5 *
6 *
7 * License Agreement
8 * For Open Source Computer Vision Library
9 * (3-clause BSD License)
10 *
11 * Copyright (C) 2014, NVIDIA Corporation, all rights reserved.
12 * Third party copyrights are property of their respective owners.
13 *
14 * Redistribution and use in source and binary forms, with or without modification,
15 * are permitted provided that the following conditions are met:
16 *
17 * * Redistributions of source code must retain the above copyright notice,
18 * this list of conditions and the following disclaimer.
19 *
20 * * Redistributions in binary form must reproduce the above copyright notice,
21 * this list of conditions and the following disclaimer in the documentation
22 * and/or other materials provided with the distribution.
23 *
24 * * Neither the names of the copyright holders nor the names of the contributors
25 * may be used to endorse or promote products derived from this software
26 * without specific prior written permission.
27 *
28 * This software is provided by the copyright holders and contributors "as is" and
29 * any express or implied warranties, including, but not limited to, the implied
30 * warranties of merchantability and fitness for a particular purpose are disclaimed.
31 * In no event shall copyright holders or contributors be liable for any direct,
32 * indirect, incidental, special, exemplary, or consequential damages
33 * (including, but not limited to, procurement of substitute goods or services;
34 * loss of use, data, or profits; or business interruption) however caused
35 * and on any theory of liability, whether in contract, strict liability,
36 * or tort (including negligence or otherwise) arising in any way out of
37 * the use of this software, even if advised of the possibility of such damage.
38 */
39
40 #include "common.hpp"
41 #include "vtransform.hpp"
42
43 #include <cmath>
44
45 namespace CAROTENE_NS {
46
47 #ifdef CAROTENE_NEON
48
49 namespace {
50
51 struct Magnitude
52 {
53 typedef s16 type;
54
operator ()CAROTENE_NS::__anon831bdaa60111::Magnitude55 void operator() (const int16x8_t & v_src0, const int16x8_t & v_src1,
56 int16x8_t & v_dst) const
57 {
58 int16x4_t v_src0_p = vget_low_s16(v_src0), v_src1_p = vget_low_s16(v_src1);
59 float32x4_t v_sqr0 = vaddq_f32(vcvtq_f32_s32(vmull_s16(v_src0_p, v_src0_p)),
60 vcvtq_f32_s32(vmull_s16(v_src1_p, v_src1_p)));
61 v_src0_p = vget_high_s16(v_src0);
62 v_src1_p = vget_high_s16(v_src1);
63 float32x4_t v_sqr1 = vaddq_f32(vcvtq_f32_s32(vmull_s16(v_src0_p, v_src0_p)),
64 vcvtq_f32_s32(vmull_s16(v_src1_p, v_src1_p)));
65
66 int32x4_t v_sqrt0 = vcvtq_s32_f32(internal::vsqrtq_f32(v_sqr0));
67 int32x4_t v_sqrt1 = vcvtq_s32_f32(internal::vsqrtq_f32(v_sqr1));
68
69 v_dst = vcombine_s16(vqmovn_s32(v_sqrt0), vqmovn_s32(v_sqrt1));
70 }
71
operator ()CAROTENE_NS::__anon831bdaa60111::Magnitude72 void operator() (const int16x4_t & v_src0, const int16x4_t & v_src1,
73 int16x4_t & v_dst) const
74 {
75 float32x4_t v_tmp = vaddq_f32(vcvtq_f32_s32(vmull_s16(v_src0, v_src0)),
76 vcvtq_f32_s32(vmull_s16(v_src1, v_src1)));
77 int32x4_t v_sqrt = vcvtq_s32_f32(internal::vsqrtq_f32(v_tmp));
78 v_dst = vqmovn_s32(v_sqrt);
79 }
80
operator ()CAROTENE_NS::__anon831bdaa60111::Magnitude81 void operator() (const short * src0, const short * src1, short * dst) const
82 {
83 f32 src0val = (f32)src0[0], src1val = (f32)src1[0];
84 dst[0] = internal::saturate_cast<s16>((s32)sqrtf(src0val * src0val + src1val * src1val));
85 }
86 };
87
88 struct MagnitudeF32
89 {
90 typedef f32 type;
91
operator ()CAROTENE_NS::__anon831bdaa60111::MagnitudeF3292 void operator() (const float32x4_t & v_src0, const float32x4_t & v_src1,
93 float32x4_t & v_dst) const
94 {
95 v_dst = internal::vsqrtq_f32(vaddq_f32(vmulq_f32(v_src0, v_src0), vmulq_f32(v_src1, v_src1)));
96 }
97
operator ()CAROTENE_NS::__anon831bdaa60111::MagnitudeF3298 void operator() (const float32x2_t & v_src0, const float32x2_t & v_src1,
99 float32x2_t & v_dst) const
100 {
101 v_dst = internal::vsqrt_f32(vadd_f32(vmul_f32(v_src0, v_src0), vmul_f32(v_src1, v_src1)));
102 }
103
operator ()CAROTENE_NS::__anon831bdaa60111::MagnitudeF32104 void operator() (const f32 * src0, const f32 * src1, f32 * dst) const
105 {
106 dst[0] = sqrtf(src0[0] * src0[0] + src1[0] * src1[0]);
107 }
108 };
109
110 } // namespace
111
112 #endif
113
magnitude(const Size2D & size,const s16 * src0Base,ptrdiff_t src0Stride,const s16 * src1Base,ptrdiff_t src1Stride,s16 * dstBase,ptrdiff_t dstStride)114 void magnitude(const Size2D &size,
115 const s16 * src0Base, ptrdiff_t src0Stride,
116 const s16 * src1Base, ptrdiff_t src1Stride,
117 s16 * dstBase, ptrdiff_t dstStride)
118 {
119 internal::assertSupportedConfiguration();
120 #ifdef CAROTENE_NEON
121 internal::vtransform(size,
122 src0Base, src0Stride,
123 src1Base, src1Stride,
124 dstBase, dstStride,
125 Magnitude());
126 #else
127 (void)size;
128 (void)src0Base;
129 (void)src0Stride;
130 (void)src1Base;
131 (void)src1Stride;
132 (void)dstBase;
133 (void)dstStride;
134 #endif
135 }
136
magnitude(const Size2D & size,const f32 * src0Base,ptrdiff_t src0Stride,const f32 * src1Base,ptrdiff_t src1Stride,f32 * dstBase,ptrdiff_t dstStride)137 void magnitude(const Size2D &size,
138 const f32 * src0Base, ptrdiff_t src0Stride,
139 const f32 * src1Base, ptrdiff_t src1Stride,
140 f32 * dstBase, ptrdiff_t dstStride)
141 {
142 internal::assertSupportedConfiguration();
143 #ifdef CAROTENE_NEON
144 internal::vtransform(size,
145 src0Base, src0Stride,
146 src1Base, src1Stride,
147 dstBase, dstStride,
148 MagnitudeF32());
149 #else
150 (void)size;
151 (void)src0Base;
152 (void)src0Stride;
153 (void)src1Base;
154 (void)src1Stride;
155 (void)dstBase;
156 (void)dstStride;
157 #endif
158 }
159
160 } // namespace CAROTENE_NS
161