Optimize x86 DeformableConv2D

src/layer/x86/deformableconv2d_pack1to16.h

@@ -0,0 +1,195 @@
+// Tencent is pleased to support the open source community by making ncnn available.
+//
+// Copyright (C) 2022 THL A29 Limited, a Tencent company. All rights reserved.
+//
+// Licensed under the BSD 3-Clause License (the "License"); you may not use this file except
+// in compliance with the License. You may obtain a copy of the License at
+//
+// https://opensource.org/licenses/BSD-3-Clause
+//
+// Unless required by applicable law or agreed to in writing, software distributed
+// under the License is distributed on an "AS IS" BASIS, WITHOUT WARRANTIES OR
+// CONDITIONS OF ANY KIND, either express or implied. See the License for the
+// specific language governing permissions and limitations under the License.
+
+static void deformableconv2d_pack1to16_avx512(const std::vector<Mat>& bottom_blobs, Mat& top_blob, const Mat& weight_data_packed, const Mat& bias_data, int kernel_w, int kernel_h, int dilation_w, int dilation_h, int stride_w, int stride_h, int pad_left, int pad_top, int activation_type, const Mat& activation_params, const Option& opt)
+{
+    const Mat& bottom_blob = bottom_blobs[0];
+    const Mat& offset = bottom_blobs[1];
+    const bool has_mask = (bottom_blobs.size() == 3);
+    const bool offset_not_pack = offset.elempack == 1;
+    const bool mask_not_pack = has_mask ? bottom_blobs[2].elempack == 1 : true;
+
+    int w = bottom_blob.w;
+    int h = bottom_blob.h;
+    int inch = bottom_blob.c;
+
+    int outw = top_blob.w;
+    int outh = top_blob.h;
+    int outch = top_blob.c;
+    const int size = outw * outh;
+    const int maxk = kernel_w * kernel_h;
+
+    const float* bias_data_ptr = bias_data;
+    const int elempack = 1;
+    const int out_elempack = 16;
+    const int wstep = out_elempack * elempack;
+    const float zeros[out_elempack] = {0.f, 0.f, 0.f, 0.f, 0.f, 0.f, 0.f, 0.f, 0.f, 0.f, 0.f, 0.f, 0.f, 0.f, 0.f, 0.f};
+    const float* zeros_ptr = zeros;
+
+    #pragma omp parallel for num_threads(opt.num_threads)
+    for (int h_col = 0; h_col < outh; h_col++)
+    {
+        for (int w_col = 0; w_col < outw; w_col++)
+        {
+            int h_in = h_col * stride_h - pad_top;
+            int w_in = w_col * stride_w - pad_left;
+            for (int oc = 0; oc < outch; oc++)
+            {
+                const float* kptr = weight_data_packed.channel(oc);
+                float* outptr = top_blob.channel(oc);
+                __m512 _sum = _mm512_setzero_ps();
+                if (bias_data_ptr)
+                    _sum = _mm512_loadu_ps(bias_data_ptr + oc * out_elempack);
+                for (int i = 0; i < kernel_h; i++)
+                {
+                    for (int j = 0; j < kernel_w; j++)
+                    {
+                        float offset_h = 0.f;
+                        float offset_w = 0.f;
+                        float mask_ = 1.f;
+                        if (offset_not_pack)
+                        {
+                            offset_h = offset.channel((i * kernel_w + j) * 2).row(h_col)[w_col];
+                            offset_w = offset.channel((i * kernel_w + j) * 2 + 1).row(h_col)[w_col];
+                        }
+                        else
+                        {
+                            const int y_c = (i * kernel_w + j) * 2;
+                            const int x_c = (i * kernel_w + j) * 2 + 1;
+                            offset_h = offset.channel(y_c / offset.elempack).row(h_col)[w_col * offset.elempack + y_c % offset.elempack];
+                            offset_w = offset.channel(x_c / offset.elempack).row(h_col)[w_col * offset.elempack + x_c % offset.elempack];
+                        }
+                        if (has_mask)
+                        {
+                            const Mat& mask = bottom_blobs[2];
+                            if (mask_not_pack)
+                            {
+                                mask_ = mask.channel(i * kernel_w + j).row(h_col)[w_col];
+                            }
+                            else
+                            {
+                                const int m_c = i * kernel_w + j;
+                                mask_ = mask.channel(m_c / mask.elempack).row(h_col)[w_col * mask.elempack + m_c % mask.elempack];
+                            }
+                        }
+                        const float h_im = h_in + i * dilation_h + offset_h;
+                        const float w_im = w_in + j * dilation_w + offset_w;
+
+                        // Bilinear
+                        const bool cond = h_im > -1 && w_im > -1 && h_im < h && w_im < w;
+                        float w1 = 0.f;
+                        float w2 = 0.f;
+                        float w3 = 0.f;
+                        float w4 = 0.f;
+                        bool v1_cond = false;
+                        bool v2_cond = false;
+                        bool v3_cond = false;
+                        bool v4_cond = false;
+                        int v1_pos = 0;
+                        int v2_pos = 0;
+                        int v3_pos = 0;
+                        int v4_pos = 0;
+                        if (cond)
+                        {
+                            int h_low = floor(h_im);
+                            int w_low = floor(w_im);
+                            int h_high = h_low + 1;
+                            int w_high = w_low + 1;
+
+                            float lh = h_im - h_low;
+                            float lw = w_im - w_low;
+                            float hh = 1 - lh;
+                            float hw = 1 - lw;
+
+                            v1_cond = (h_low >= 0 && w_low >= 0);
+                            v2_cond = (h_low >= 0 && w_high <= w - 1);
+                            v3_cond = (h_high <= h - 1 && w_low >= 0);
+                            v4_cond = (h_high <= h - 1 && w_high <= w - 1);
+                            if (v1_cond)
+                                v1_pos = h_low * w + w_low;
+                            if (v2_cond)
+                                v2_pos = h_low * w + w_high;
+                            if (v3_cond)
+                                v3_pos = h_high * w + w_low;
+                            if (v4_cond)
+                                v4_pos = h_high * w + w_high;
+
+                            w1 = hh * hw;
+                            w2 = hh * lw;
+                            w3 = lh * hw;
+                            w4 = lh * lw;
+                        }
+                        const float w1s[out_elempack] = {w1, w1, w1, w1, w1, w1, w1, w1, w1, w1, w1, w1, w1, w1, w1, w1};
+                        const float* w1_ptr = w1s;
+                        const float w2s[out_elempack] = {w2, w2, w2, w2, w2, w2, w2, w2, w2, w2, w2, w2, w2, w2, w2, w2};
+                        const float* w2_ptr = w2s;
+                        const float w3s[out_elempack] = {w3, w3, w3, w3, w3, w3, w3, w3, w3, w3, w3, w3, w3, w3, w3, w3};
+                        const float* w3_ptr = w3s;
+                        const float w4s[out_elempack] = {w4, w4, w4, w4, w4, w4, w4, w4, w4, w4, w4, w4, w4, w4, w4, w4};
+                        const float* w4_ptr = w4s;
+                        const float masks[out_elempack] = {mask_, mask_, mask_, mask_, mask_, mask_, mask_, mask_, mask_, mask_, mask_, mask_, mask_, mask_, mask_, mask_};
+                        const float* mask_ptr = masks;
+
+                        for (int ic = 0; ic < inch; ic++)
+                        {
+                            const float* data_im_ptr = bottom_blob.channel(ic);
+                            __m512 _val_channel0 = _mm512_loadu_ps(zeros_ptr);
+                            if (cond)
+                            {
+                                __m512 _v1_channel0 = _val_channel0;
+                                __m512 _v2_channel0 = _val_channel0;
+                                __m512 _v3_channel0 = _val_channel0;
+                                __m512 _v4_channel0 = _val_channel0;
+                                if (v1_cond)
+                                {
+                                    _v1_channel0 = _mm512_set1_ps(data_im_ptr[v1_pos]);
+                                }
+                                if (v2_cond)
+                                {
+                                    _v2_channel0 = _mm512_set1_ps(data_im_ptr[v2_pos]);
+                                }
+                                if (v3_cond)
+                                {
+                                    _v3_channel0 = _mm512_set1_ps(data_im_ptr[v3_pos]);
+                                }
+                                if (v4_cond)
+                                {
+                                    _v4_channel0 = _mm512_set1_ps(data_im_ptr[v4_pos]);
+                                }
+                                __m512 _w1 = _mm512_loadu_ps(w1_ptr);
+                                __m512 _w2 = _mm512_loadu_ps(w2_ptr);
+                                __m512 _w3 = _mm512_loadu_ps(w3_ptr);
+                                __m512 _w4 = _mm512_loadu_ps(w4_ptr);
+                                _val_channel0 = _mm512_fmadd_ps(_v1_channel0, _w1, _val_channel0);
+                                _val_channel0 = _mm512_fmadd_ps(_v2_channel0, _w2, _val_channel0);
+                                _val_channel0 = _mm512_fmadd_ps(_v3_channel0, _w3, _val_channel0);
+                                _val_channel0 = _mm512_fmadd_ps(_v4_channel0, _w4, _val_channel0);
+                            }
+                            if (has_mask)
+                            {
+                                __m512 _mask = _mm512_loadu_ps(mask_ptr);
+                                _val_channel0 = _mm512_mul_ps(_val_channel0, _mask);
+                            }
+                            __m512 _conv_w0 = _mm512_load_ps(kptr);
+                            _sum = _mm512_fmadd_ps(_val_channel0, _conv_w0, _sum);
+                            kptr += wstep;
+                        }
+                    }
+                }
+                _sum = activation_avx512(_sum, activation_type, activation_params);
+                _mm512_store_ps(outptr + (h_col * outw + w_col) * out_elempack, _sum);
+            }
+        }
+    }
+}

src/layer/x86/deformableconv2d_pack1to4.h

@@ -0,0 +1,195 @@
+// Tencent is pleased to support the open source community by making ncnn available.
+//
+// Copyright (C) 2022 THL A29 Limited, a Tencent company. All rights reserved.
+//
+// Licensed under the BSD 3-Clause License (the "License"); you may not use this file except
+// in compliance with the License. You may obtain a copy of the License at
+//
+// https://opensource.org/licenses/BSD-3-Clause
+//
+// Unless required by applicable law or agreed to in writing, software distributed
+// under the License is distributed on an "AS IS" BASIS, WITHOUT WARRANTIES OR
+// CONDITIONS OF ANY KIND, either express or implied. See the License for the
+// specific language governing permissions and limitations under the License.
+
+static void deformableconv2d_pack1to4_sse(const std::vector<Mat>& bottom_blobs, Mat& top_blob, const Mat& weight_data_packed, const Mat& bias_data, int kernel_w, int kernel_h, int dilation_w, int dilation_h, int stride_w, int stride_h, int pad_left, int pad_top, int activation_type, const Mat& activation_params, const Option& opt)
+{
+    const Mat& bottom_blob = bottom_blobs[0];
+    const Mat& offset = bottom_blobs[1];
+    const bool has_mask = (bottom_blobs.size() == 3);
+    const bool offset_not_pack = offset.elempack == 1;
+    const bool mask_not_pack = has_mask ? bottom_blobs[2].elempack == 1 : true;
+
+    int w = bottom_blob.w;
+    int h = bottom_blob.h;
+    int inch = bottom_blob.c;
+
+    int outw = top_blob.w;
+    int outh = top_blob.h;
+    int outch = top_blob.c;
+    const int size = outw * outh;
+    const int maxk = kernel_w * kernel_h;
+
+    const float* bias_data_ptr = bias_data;
+    const int elempack = 1;
+    const int out_elempack = 4;
+    const int wstep = out_elempack * elempack;
+    const float zeros[out_elempack] = {0.f, 0.f, 0.f, 0.f};
+    const float* zeros_ptr = zeros;
+
+    #pragma omp parallel for num_threads(opt.num_threads)
+    for (int h_col = 0; h_col < outh; h_col++)
+    {
+        for (int w_col = 0; w_col < outw; w_col++)
+        {
+            int h_in = h_col * stride_h - pad_top;
+            int w_in = w_col * stride_w - pad_left;
+            for (int oc = 0; oc < outch; oc++)
+            {
+                const float* kptr = weight_data_packed.channel(oc);
+                float* outptr = top_blob.channel(oc);
+                __m128 _sum = _mm_setzero_ps();
+                if (bias_data_ptr)
+                    _sum = _mm_loadu_ps(bias_data_ptr + oc * out_elempack);
+                for (int i = 0; i < kernel_h; i++)
+                {
+                    for (int j = 0; j < kernel_w; j++)
+                    {
+                        float offset_h = 0.f;
+                        float offset_w = 0.f;
+                        float mask_ = 1.f;
+                        if (offset_not_pack)
+                        {
+                            offset_h = offset.channel((i * kernel_w + j) * 2).row(h_col)[w_col];
+                            offset_w = offset.channel((i * kernel_w + j) * 2 + 1).row(h_col)[w_col];
+                        }
+                        else
+                        {
+                            const int y_c = (i * kernel_w + j) * 2;
+                            const int x_c = (i * kernel_w + j) * 2 + 1;
+                            offset_h = offset.channel(y_c / offset.elempack).row(h_col)[w_col * offset.elempack + y_c % offset.elempack];
+                            offset_w = offset.channel(x_c / offset.elempack).row(h_col)[w_col * offset.elempack + x_c % offset.elempack];
+                        }
+                        if (has_mask)
+                        {
+                            const Mat& mask = bottom_blobs[2];
+                            if (mask_not_pack)
+                            {
+                                mask_ = mask.channel(i * kernel_w + j).row(h_col)[w_col];
+                            }
+                            else
+                            {
+                                const int m_c = i * kernel_w + j;
+                                mask_ = mask.channel(m_c / mask.elempack).row(h_col)[w_col * mask.elempack + m_c % mask.elempack];
+                            }
+                        }
+                        const float h_im = h_in + i * dilation_h + offset_h;
+                        const float w_im = w_in + j * dilation_w + offset_w;
+
+                        // Bilinear
+                        const bool cond = h_im > -1 && w_im > -1 && h_im < h && w_im < w;
+                        float w1 = 0.f;
+                        float w2 = 0.f;
+                        float w3 = 0.f;
+                        float w4 = 0.f;
+                        bool v1_cond = false;
+                        bool v2_cond = false;
+                        bool v3_cond = false;
+                        bool v4_cond = false;
+                        int v1_pos = 0;
+                        int v2_pos = 0;
+                        int v3_pos = 0;
+                        int v4_pos = 0;
+                        if (cond)
+                        {
+                            int h_low = floor(h_im);
+                            int w_low = floor(w_im);
+                            int h_high = h_low + 1;
+                            int w_high = w_low + 1;
+
+                            float lh = h_im - h_low;
+                            float lw = w_im - w_low;
+                            float hh = 1 - lh;
+                            float hw = 1 - lw;
+
+                            v1_cond = (h_low >= 0 && w_low >= 0);
+                            v2_cond = (h_low >= 0 && w_high <= w - 1);
+                            v3_cond = (h_high <= h - 1 && w_low >= 0);
+                            v4_cond = (h_high <= h - 1 && w_high <= w - 1);
+                            if (v1_cond)
+                                v1_pos = h_low * w + w_low;
+                            if (v2_cond)
+                                v2_pos = h_low * w + w_high;
+                            if (v3_cond)
+                                v3_pos = h_high * w + w_low;
+                            if (v4_cond)
+                                v4_pos = h_high * w + w_high;
+
+                            w1 = hh * hw;
+                            w2 = hh * lw;
+                            w3 = lh * hw;
+                            w4 = lh * lw;
+                        }
+                        const float w1s[out_elempack] = {w1, w1, w1, w1};
+                        const float* w1_ptr = w1s;
+                        const float w2s[out_elempack] = {w2, w2, w2, w2};
+                        const float* w2_ptr = w2s;
+                        const float w3s[out_elempack] = {w3, w3, w3, w3};
+                        const float* w3_ptr = w3s;
+                        const float w4s[out_elempack] = {w4, w4, w4, w4};
+                        const float* w4_ptr = w4s;
+                        const float masks[out_elempack] = {mask_, mask_, mask_, mask_};
+                        const float* mask_ptr = masks;
+
+                        for (int ic = 0; ic < inch; ic++)
+                        {
+                            const float* data_im_ptr = bottom_blob.channel(ic);
+                            __m128 _val_channel0 = _mm_loadu_ps(zeros_ptr);
+                            if (cond)
+                            {
+                                __m128 _v1_channel0 = _val_channel0;
+                                __m128 _v2_channel0 = _val_channel0;
+                                __m128 _v3_channel0 = _val_channel0;
+                                __m128 _v4_channel0 = _val_channel0;
+                                if (v1_cond)
+                                {
+                                    _v1_channel0 = _mm_load1_ps(data_im_ptr + v1_pos);
+                                }
+                                if (v2_cond)
+                                {
+                                    _v2_channel0 = _mm_load1_ps(data_im_ptr + v2_pos);
+                                }
+                                if (v3_cond)
+                                {
+                                    _v3_channel0 = _mm_load1_ps(data_im_ptr + v3_pos);
+                                }
+                                if (v4_cond)
+                                {
+                                    _v4_channel0 = _mm_load1_ps(data_im_ptr + v4_pos);
+                                }
+                                __m128 _w1 = _mm_loadu_ps(w1_ptr);
+                                __m128 _w2 = _mm_loadu_ps(w2_ptr);
+                                __m128 _w3 = _mm_loadu_ps(w3_ptr);
+                                __m128 _w4 = _mm_loadu_ps(w4_ptr);
+                                _val_channel0 = _mm_comp_fmadd_ps(_v1_channel0, _w1, _val_channel0);
+                                _val_channel0 = _mm_comp_fmadd_ps(_v2_channel0, _w2, _val_channel0);
+                                _val_channel0 = _mm_comp_fmadd_ps(_v3_channel0, _w3, _val_channel0);
+                                _val_channel0 = _mm_comp_fmadd_ps(_v4_channel0, _w4, _val_channel0);
+                            }
+                            if (has_mask)
+                            {
+                                __m128 _mask = _mm_loadu_ps(mask_ptr);
+                                _val_channel0 = _mm_mul_ps(_val_channel0, _mask);
+                            }
+                            __m128 _conv_w0 = _mm_load_ps(kptr);
+                            _sum = _mm_comp_fmadd_ps(_val_channel0, _conv_w0, _sum);
+                            kptr += wstep;
+                        }
+                    }
+                }
+                _sum = activation_sse(_sum, activation_type, activation_params);
+                _mm_storeu_ps(outptr + (h_col * outw + w_col) * out_elempack, _sum);
+            }
+        }
+    }
+}