// Tencent is pleased to support the open source community by making ncnn available.
//
// Copyright (C) 2019 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.

#version 450

#if NCNN_fp16_storage
#extension GL_EXT_shader_16bit_storage: require
#endif
#if NCNN_fp16_arithmetic
#extension GL_EXT_shader_explicit_arithmetic_types_float16: require
#endif

layout (constant_id = 0) const int kernel_w = 1;
layout (constant_id = 1) const int kernel_h = 1;
layout (constant_id = 2) const int dilation_w = 1;
layout (constant_id = 3) const int dilation_h = 1;
layout (constant_id = 4) const int stride_w = 1;
layout (constant_id = 5) const int stride_h = 1;
layout (constant_id = 6) const int bias_term = 0;
layout (constant_id = 7) const int group = 1;
layout (constant_id = 8) const int activation_type = 0;
layout (constant_id = 9) const float activation_param_0 = 0;
layout (constant_id = 10) const float activation_param_1 = 0;

#define shape_constant_id_offset 11
layout (constant_id = shape_constant_id_offset + 0) const int dims = 0;
layout (constant_id = shape_constant_id_offset + 1) const int w = 0;
layout (constant_id = shape_constant_id_offset + 2) const int h = 0;
layout (constant_id = shape_constant_id_offset + 3) const int c = 0;
layout (constant_id = shape_constant_id_offset + 4) const int cstep = 0;

layout (constant_id = shape_constant_id_offset + 5) const int outdims = 0;
layout (constant_id = shape_constant_id_offset + 6) const int outw = 0;
layout (constant_id = shape_constant_id_offset + 7) const int outh = 0;
layout (constant_id = shape_constant_id_offset + 8) const int outc = 0;
layout (constant_id = shape_constant_id_offset + 9) const int outcstep = 0;

#if NCNN_image_shader
layout (binding = 0) uniform unfp sampler3D bottom_blob;
layout (binding = 1, imfmtc1) writeonly uniform unfp image3D top_blob;
layout (binding = 2) uniform unfp sampler3D weight_blob;
layout (binding = 3) uniform unfp sampler3D bias_blob;
#else
layout (binding = 0) readonly buffer bottom_blob { sfp bottom_blob_data[]; };
layout (binding = 1) writeonly buffer top_blob { sfp top_blob_data[]; };
layout (binding = 2) readonly buffer weight_blob { sfp weight_data[]; };
layout (binding = 3) readonly buffer bias_blob { sfp bias_data[]; };
#endif

layout (push_constant) uniform parameter
{
    int dims;
    int w;
    int h;
    int c;
    int cstep;

    int outdims;
    int outw;
    int outh;
    int outc;
    int outcstep;
} p;

void main()
{
    int gx = int(gl_GlobalInvocationID.x);
    int gy = int(gl_GlobalInvocationID.y);
    int gz = int(gl_GlobalInvocationID.z);

    if (gx >= psc(outw) || gy >= psc(outh) || gz >= psc(outc))
        return;

    afp sum;

    if (bias_term == 1)
    {
#if NCNN_image_shader
        sum = image3d_ld1(bias_blob, ivec3(gz, 0, 0));
#else
        sum = buffer_ld1(bias_data, gz);
#endif
    }
    else
    {
        sum = afp(0.f);
    }

    // depth-wise convolution
#if NCNN_image_shader
    int sy = gy * stride_h;
    int wx = 0;

    for (int y = 0; y < kernel_h; y++)
    {
        int sx = gx * stride_w;

        for (int x = 0; x < kernel_w; x++)
        {
            sum += image3d_ld1(weight_blob, ivec3(wx, gz, 0)) * image3d_ld1(bottom_blob, ivec3(sx, sy, gz));

            sx += dilation_w;
            wx += 1;
        }

        sy += dilation_h;
    }
#else
    int w_offset = gz * kernel_w * kernel_h;
    int v_offset = gz * psc(cstep) + gy * stride_h * psc(w) + gx * stride_w;

    for (int y = 0; y < kernel_h; y++)
    {
        for (int x = 0; x < kernel_w; x++)
        {
            sum += buffer_ld1(weight_data, w_offset + x) * buffer_ld1(bottom_blob_data, v_offset + x * dilation_w);
        }

        v_offset += dilation_h * psc(w);
        w_offset += kernel_w;
    }
#endif

    if (activation_type == 1)
    {
        sum = max(sum, afp(0.f));
    }
    if (activation_type == 2)
    {
        const afp slope = afp(activation_param_0);
        sum = sum < afp(0.f) ? sum * slope : sum;
    }
    if (activation_type == 3)
    {
        const afp const_min = afp(activation_param_0);
        const afp const_max = afp(activation_param_1);
        sum = clamp(sum, const_min, const_max);
    }
    if (activation_type == 4)
    {
        sum = afp(1.f) / (afp(1.f) + exp(-sum));
    }
    if (activation_type == 5)
    {
        sum = sum * tanh(log(exp(sum) + afp(1.f)));
    }

#if NCNN_image_shader
    image3d_st1(top_blob, ivec3(gx, gy, gz), sum);
#else
    const int gi = gz * psc(outcstep) + gy * psc(outw) + gx;

    buffer_st1(top_blob_data, gi, sum);
#endif
}