3D Convolution

题面

Implement a program that performs a 3D convolution operation with "valid" 边界（无 padding）。输入为体数据与三维 kernel，均按行优先然后按 depth 展平为一维数组。输出尺寸：
output_depth = input_depth - kernel_depth + 1
output_rows = input_rows - kernel_rows + 1
output_cols = input_cols - kernel_cols + 1

Implementation Requirements

• Use only native features (external libraries are not permitted)
• The solve function signature must remain unchanged
• The final result must be stored in output

Examples

示例见站点页面（包含 3×3×3 与 2×2×2 的体数据例）。

Constraints

• 1 ≤ input_depth, input_rows, input_cols ≤ 256
• 1 ≤ kernel_depth, kernel_rows, kernel_cols ≤ 5
• kernel_depth ≤ input_depth
• kernel_rows ≤ input_rows
• kernel_cols ≤ input_cols
• Performance is measured with input_cols = 128, input_rows = 128, kernel_cols = 5, kernel_rows = 5