riscv v optimization for flatten

src/layer/riscv/flatten_riscv.cpp

@@ -0,0 +1,480 @@
+// Tencent is pleased to support the open source community by making ncnn available.
+//
+// Copyright (C) 2021 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.
+
+#include "flatten_riscv.h"
+
+#if __riscv_vector
+#ifdef RVV_SPEC_0_7
+#include "riscv_v_071_fix.h"
+#else
+#include <riscv_vector.h>
+#endif
+#endif // __riscv_vector
+
+#include "riscv_usability.h"
+
+namespace ncnn {
+
+Flatten_riscv::Flatten_riscv()
+{
+#if __riscv_vector
+    support_packing = true;
+#if __riscv_zfh
+    support_fp16_storage = true;
+#endif
+#endif // __riscv_vector
+
+    support_bf16_storage = true;
+}
+
+int Flatten_riscv::forward(const Mat& bottom_blob, Mat& top_blob, const Option& opt) const
+{
+    int elembits = bottom_blob.elembits();
+
+    if (elembits == 8)
+        return forward_int8(bottom_blob, top_blob, opt);
+
+#if __riscv_vector && __riscv_zfh
+    if (opt.use_fp16_storage && elembits == 16)
+        return forward_bf16s_fp16s(bottom_blob, top_blob, opt);
+#endif
+
+    if (opt.use_bf16_storage && elembits == 16)
+        return forward_bf16s_fp16s(bottom_blob, top_blob, opt);
+
+    int dims = bottom_blob.dims;
+
+    if (dims == 1)
+    {
+        top_blob = bottom_blob;
+        return 0;
+    }
+
+#if __riscv_vector
+    const int packn = csrr_vlenb() / 4;
+#endif
+
+    int w = bottom_blob.w;
+    int h = bottom_blob.h;
+    int channels = bottom_blob.c;
+    size_t elemsize = bottom_blob.elemsize;
+    int elempack = bottom_blob.elempack;
+    int size = w * h;
+
+    int total = size * channels * elempack;
+
+    int out_elempack = 1;
+#if __riscv_vector
+    if (opt.use_packing_layout)
+    {
+        out_elempack = total % packn == 0 ? packn : 1;
+    }
+#endif
+    size_t out_elemsize = elemsize / elempack * out_elempack;
+
+    if (out_elempack == 1)
+    {
+        return Flatten::forward(bottom_blob, top_blob, opt);
+    }
+
+    if (dims == 2 && elempack == 1) // out_elempack == packn
+    {
+        top_blob = bottom_blob;
+        top_blob.dims = 1;
+        top_blob.w = total / out_elempack;
+        top_blob.h = 1;
+        top_blob.cstep = top_blob.w;
+        top_blob.elemsize = out_elemsize;
+        top_blob.elempack = out_elempack;
+        return 0;
+    }
+
+    top_blob.create(total / out_elempack, out_elemsize, out_elempack, opt.blob_allocator);
+    if (top_blob.empty())
+        return -100;
+
+    if (dims == 2)
+    {
+#if __riscv_vector
+        if (elempack == packn) // out_elempack == packn
+        {
+            #pragma omp parallel for num_threads(opt.num_threads)
+            for (int i = 0; i < h; i++)
+            {
+                const float* ptr = bottom_blob.row(i);
+                float* outptr = (float*)top_blob + w * i * packn;
+
+                int n = w * elempack;
+                while (n > 0)
+                {
+                    word_type vl = vsetvl_e32m1(n);
+
+                    vfloat32m1_t _p = vle32_v_f32m1(ptr, vl);
+                    vsse32_v_f32m1(outptr, w * sizeof(float), _p, vl);
+
+                    ptr += vl;
+                    outptr += 1;
+                    n -= vl;
+                }
+            }
+        }
+#endif // __riscv_vector
+    }
+
+    if (dims == 3)
+    {
+#if __riscv_vector
+        if (elempack == packn) // out_elempack == packn
+        {
+            #pragma omp parallel for num_threads(opt.num_threads)
+            for (int q = 0; q < channels; q++)
+            {
+                const float* ptr = bottom_blob.channel(q);
+                float* outptr = (float*)top_blob + size * q * packn;
+
+                int n = size * elempack;
+                while (n > 0)
+                {
+                    word_type vl = vsetvl_e32m1(n);
+
+                    vfloat32m1_t _p = vle32_v_f32m1(ptr, vl);
+                    vsse32_v_f32m1(outptr, size * sizeof(float), _p, vl);
+
+                    ptr += vl;
+                    outptr += 1;
+                    n -= vl;
+                }
+            }
+        }
+#endif // __riscv_vector
+
+        if (elempack == 1) // out_elempack == packn
+        {
+            #pragma omp parallel for num_threads(opt.num_threads)
+            for (int q = 0; q < channels; q++)
+            {
+                const float* ptr = bottom_blob.channel(q);
+                float* outptr = (float*)top_blob + size * q;
+
+#if __riscv_vector
+                int n = size * elempack;
+                while (n > 0)
+                {
+                    word_type vl = vsetvl_e32m8(n);
+
+                    vfloat32m8_t _p = vle32_v_f32m8(ptr, vl);
+                    vse32_v_f32m8(outptr, _p, vl);
+
+                    ptr += vl;
+                    outptr += vl;
+                    n -= vl;
+                }
+#else  // __riscv_vector
+                for (int i = 0; i < size; i++)
+                {
+                    *outptr++ = *ptr++;
+                }
+#endif // __riscv_vector
+            }
+        }
+    }
+
+    return 0;
+}
+
+int Flatten_riscv::forward_bf16s_fp16s(const Mat& bottom_blob, Mat& top_blob, const Option& opt) const
+{
+    int dims = bottom_blob.dims;
+
+    if (dims == 1)
+    {
+        top_blob = bottom_blob;
+        return 0;
+    }
+
+#if __riscv_vector
+    const int packn = csrr_vlenb() / 2;
+#endif
+
+    int w = bottom_blob.w;
+    int h = bottom_blob.h;
+    int channels = bottom_blob.c;
+    size_t elemsize = bottom_blob.elemsize;
+    int elempack = bottom_blob.elempack;
+    int size = w * h;
+
+    int total = size * channels * elempack;
+
+    int out_elempack = 1;
+#if __riscv_vector
+    if (opt.use_packing_layout)
+    {
+        out_elempack = total % packn == 0 ? packn : 1;
+    }
+#endif
+    size_t out_elemsize = elemsize / elempack * out_elempack;
+
+    if (out_elempack == 1)
+    {
+        return Flatten::forward(bottom_blob, top_blob, opt);
+    }
+
+    if (dims == 2 && elempack == 1) // out_elempack == packn
+    {
+        top_blob = bottom_blob;
+        top_blob.dims = 1;
+        top_blob.w = total / out_elempack;
+        top_blob.h = 1;
+        top_blob.cstep = top_blob.w;
+        top_blob.elemsize = out_elemsize;
+        top_blob.elempack = out_elempack;
+        return 0;
+    }
+
+    top_blob.create(total / out_elempack, out_elemsize, out_elempack, opt.blob_allocator);
+    if (top_blob.empty())
+        return -100;
+
+    if (dims == 2)
+    {
+#if __riscv_vector
+        if (elempack == packn) // out_elempack == packn
+        {
+            #pragma omp parallel for num_threads(opt.num_threads)
+            for (int i = 0; i < h; i++)
+            {
+                const unsigned short* ptr = bottom_blob.row<unsigned short>(i);
+                unsigned short* outptr = (unsigned short*)top_blob + w * i * packn;
+
+                int n = w * elempack;
+                while (n > 0)
+                {
+                    word_type vl = vsetvl_e16m1(n);
+
+                    vuint16m1_t _p = vle16_v_u16m1(ptr, vl);
+                    vsse16_v_u16m1(outptr, w * sizeof(unsigned short), _p, vl);
+
+                    ptr += vl;
+                    outptr += 1;
+                    n -= vl;
+                }
+            }
+        }
+#endif // __riscv_vector
+    }
+
+    if (dims == 3)
+    {
+#if __riscv_vector
+        if (elempack == packn) // out_elempack == packn
+        {
+            #pragma omp parallel for num_threads(opt.num_threads)
+            for (int q = 0; q < channels; q++)
+            {
+                const unsigned short* ptr = bottom_blob.channel(q);
+                unsigned short* outptr = (unsigned short*)top_blob + size * q * packn;
+
+                int n = size * elempack;
+                while (n > 0)
+                {
+                    word_type vl = vsetvl_e16m1(n);
+
+                    vuint16m1_t _p = vle16_v_u16m1(ptr, vl);
+                    vsse16_v_u16m1(outptr, size * sizeof(unsigned short), _p, vl);
+
+                    ptr += vl;
+                    outptr += 1;
+                    n -= vl;
+                }
+            }
+        }
+#endif // __riscv_vector
+
+        if (elempack == 1) // out_elempack == packn
+        {
+            #pragma omp parallel for num_threads(opt.num_threads)
+            for (int q = 0; q < channels; q++)
+            {
+                const unsigned short* ptr = bottom_blob.channel(q);
+                unsigned short* outptr = (unsigned short*)top_blob + size * q;
+
+#if __riscv_vector
+                int n = size * elempack;
+                while (n > 0)
+                {
+                    word_type vl = vsetvl_e16m8(n);
+
+                    vuint16m8_t _p = vle16_v_u16m8(ptr, vl);
+                    vse16_v_u16m8(outptr, _p, vl);
+
+                    ptr += vl;
+                    outptr += vl;
+                    n -= vl;
+                }
+#else  // __riscv_vector
+                for (int i = 0; i < size; i++)
+                {
+                    *outptr++ = *ptr++;
+                }
+#endif // __riscv_vector
+            }
+        }
+    }
+
+    return 0;
+}
+
+int Flatten_riscv::forward_int8(const Mat& bottom_blob, Mat& top_blob, const Option& opt) const
+{
+    int dims = bottom_blob.dims;
+
+    if (dims == 1)
+    {
+        top_blob = bottom_blob;
+        return 0;
+    }
+
+#if __riscv_vector
+    const int packn = csrr_vlenb() / 1;
+#endif
+
+    int w = bottom_blob.w;
+    int h = bottom_blob.h;
+    int channels = bottom_blob.c;
+    size_t elemsize = bottom_blob.elemsize;
+    int elempack = bottom_blob.elempack;
+    int size = w * h;
+
+    int total = size * channels * elempack;
+
+    int out_elempack = 1;
+#if __riscv_vector
+    if (opt.use_packing_layout)
+    {
+        out_elempack = total % packn == 0 ? packn : 1;
+    }
+#endif
+    size_t out_elemsize = elemsize / elempack * out_elempack;
+
+    if (out_elempack == 1)
+    {
+        return Flatten::forward(bottom_blob, top_blob, opt);
+    }
+
+    if (dims == 2 && elempack == 1) // out_elempack == packn
+    {
+        top_blob = bottom_blob;
+        top_blob.dims = 1;
+        top_blob.w = total / out_elempack;
+        top_blob.h = 1;
+        top_blob.cstep = top_blob.w;
+        top_blob.elemsize = out_elemsize;
+        top_blob.elempack = out_elempack;
+        return 0;
+    }
+
+    top_blob.create(total / out_elempack, out_elemsize, out_elempack, opt.blob_allocator);
+    if (top_blob.empty())
+        return -100;
+
+    if (dims == 2)
+    {
+#if __riscv_vector
+        if (elempack == packn) // out_elempack == packn
+        {
+            #pragma omp parallel for num_threads(opt.num_threads)
+            for (int i = 0; i < h; i++)
+            {
+                const signed char* ptr = bottom_blob.row<signed char>(i);
+                signed char* outptr = (signed char*)top_blob + w * i * packn;
+
+                int n = w * elempack;
+                while (n > 0)
+                {
+                    word_type vl = vsetvl_e8m1(n);
+
+                    vint8m1_t _p = vle8_v_i8m1(ptr, vl);
+                    vsse8_v_i8m1(outptr, w * sizeof(unsigned char), _p, vl);
+
+                    ptr += vl;
+                    outptr += 1;
+                    n -= vl;
+                }
+            }
+        }
+#endif // __riscv_vector
+    }
+
+    if (dims == 3)
+    {
+#if __riscv_vector
+        if (elempack == packn) // out_elempack == packn
+        {
+            #pragma omp parallel for num_threads(opt.num_threads)
+            for (int q = 0; q < channels; q++)
+            {
+                const signed char* ptr = bottom_blob.channel(q);
+                signed char* outptr = (signed char*)top_blob + size * q * packn;
+
+                int n = size * elempack;
+                while (n > 0)
+                {
+                    word_type vl = vsetvl_e8m1(n);
+
+                    vint8m1_t _p = vle8_v_i8m1(ptr, vl);
+                    vsse8_v_i8m1(outptr, size * sizeof(signed char), _p, vl);
+
+                    ptr += vl;
+                    outptr += 1;
+                    n -= vl;
+                }
+            }
+        }
+#endif // __riscv_vector
+
+        if (elempack == 1) // out_elempack == packn
+        {
+            #pragma omp parallel for num_threads(opt.num_threads)
+            for (int q = 0; q < channels; q++)
+            {
+                const signed char* ptr = bottom_blob.channel(q);
+                signed char* outptr = (signed char*)top_blob + size * q;
+
+#if __riscv_vector
+                int n = size * elempack;
+                while (n > 0)
+                {
+                    word_type vl = vsetvl_e8m8(n);
+
+                    vint8m8_t _p = vle8_v_i8m8(ptr, vl);
+                    vse8_v_i8m8(outptr, _p, vl);
+
+                    ptr += vl;
+                    outptr += vl;
+                    n -= vl;
+                }
+#else  // __riscv_vector
+                for (int i = 0; i < size; i++)
+                {
+                    *outptr++ = *ptr++;
+                }
+#endif // __riscv_vector
+            }
+        }
+    }
+
+    return 0;
+}
+
+} // namespace ncnn

src/layer/riscv/flatten_riscv.h

@@ -0,0 +1,36 @@
+// Tencent is pleased to support the open source community by making ncnn available.
+//
+// Copyright (C) 2021 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.
+
+#ifndef LAYER_FLATTEN_RISCV_H
+#define LAYER_FLATTEN_RISCV_H
+
+#include "flatten.h"
+
+namespace ncnn {
+
+class Flatten_riscv : virtual public Flatten
+{
+public:
+    Flatten_riscv();
+
+    virtual int forward(const Mat& bottom_blob, Mat& top_blob, const Option& opt) const;
+
+protected:
+    int forward_bf16s_fp16s(const Mat& bottom_blob, Mat& top_blob, const Option& opt) const;
+    int forward_int8(const Mat& bottom_blob, Mat& top_blob, const Option& opt) const;
+};
+
+} // namespace ncnn
+
+#endif // LAYER_FLATTEN_RISCV_H