1 /*
2 * Copyright (C) 2010-2018 Arm Limited or its affiliates. All rights reserved.
3 *
4 * SPDX-License-Identifier: Apache-2.0
5 *
6 * Licensed under the Apache License, Version 2.0 (the License); you may
7 * not use this file except in compliance with the License.
8 * You may obtain a copy of the License at
9 *
10 * www.apache.org/licenses/LICENSE-2.0
11 *
12 * Unless required by applicable law or agreed to in writing, software
13 * distributed under the License is distributed on an AS IS BASIS, WITHOUT
14 * WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
15 * See the License for the specific language governing permissions and
16 * limitations under the License.
17 */
18
19 #include "ref_functions.h"
20
arm_convolve_HWC_q7_ref_nonsquare(const q7_t * Im_in,const uint16_t dim_im_in_x,const uint16_t dim_im_in_y,const uint16_t ch_im_in,const q7_t * wt,const uint16_t ch_im_out,const uint16_t dim_kernel_x,const uint16_t dim_kernel_y,const uint16_t padding_x,const uint16_t padding_y,const uint16_t stride_x,const uint16_t stride_y,const q7_t * bias,const uint16_t bias_shift,const uint16_t out_shift,q7_t * Im_out,const uint16_t dim_im_out_x,const uint16_t dim_im_out_y,q15_t * bufferA,q7_t * bufferB)21 void arm_convolve_HWC_q7_ref_nonsquare(const q7_t * Im_in, // input image
22 const uint16_t dim_im_in_x, // input image dimention x
23 const uint16_t dim_im_in_y, // input image dimention y
24 const uint16_t ch_im_in, // number of input image channels
25 const q7_t * wt, // kernel weights
26 const uint16_t ch_im_out, // number of filters, i.e., output image channels
27 const uint16_t dim_kernel_x, // filter kernel size x
28 const uint16_t dim_kernel_y, // filter kernel size y
29 const uint16_t padding_x, // padding sizes x
30 const uint16_t padding_y, // padding sizes y
31 const uint16_t stride_x, // stride x
32 const uint16_t stride_y, // stride y
33 const q7_t * bias, // bias
34 const uint16_t bias_shift, const uint16_t out_shift, q7_t * Im_out, // output image
35 const uint16_t dim_im_out_x, // output image dimension x
36 const uint16_t dim_im_out_y, // output image dimension y
37 q15_t * bufferA, //buffer space for input
38 q7_t * bufferB //buffer space for output
39 )
40 {
41 int i, j, k, l, m, n;
42 int conv_out;
43 int in_row, in_col;
44
45 for (i = 0; i < ch_im_out; i++)
46 {
47 for (j = 0; j < dim_im_out_y; j++)
48 {
49 for (k = 0; k < dim_im_out_x; k++)
50 {
51 #ifndef ARM_NN_TRUNCATE
52 conv_out = ((q31_t) (bias[i]) << bias_shift) + (0x1 << (out_shift - 1));
53 #else
54 conv_out = bias[i] << bias_shift;
55 #endif
56 for (m = 0; m < dim_kernel_y; m++)
57 {
58 for (n = 0; n < dim_kernel_x; n++)
59 {
60 // if-for implementation
61 in_row = stride_y * j + m - padding_y;
62 in_col = stride_x * k + n - padding_x;
63 if (in_row >= 0 && in_col >= 0 && in_row < dim_im_in_y && in_col < dim_im_in_x)
64 {
65 for (l = 0; l < ch_im_in; l++)
66 {
67 conv_out += Im_in[(in_row * dim_im_in_x + in_col) * ch_im_in + l] *
68 wt[i * ch_im_in * dim_kernel_y * dim_kernel_x + (m * dim_kernel_x + n) * ch_im_in +
69 l];
70 }
71 }
72 }
73 }
74 Im_out[i + (j * dim_im_out_x + k) * ch_im_out] = (q7_t) __SSAT((conv_out >> out_shift), 8);
75 }
76 }
77 }
78 }
79