Utility.cuh

#pragma once

#include <string>
#include <cstdint>
#include <iostream>
#include <iomanip>  

#define int_32 std::int_fast32_t

using std::string;
using std::cout;
enum class UnitMeasure:size_t { B=1, KB=1024, MB=1048576, GB=1073741824, TB=1099511627776 };
class GpuConfig final {

	//all mem size is in bytes by defaults
private:
	//default members
	string	_name{""};
	int_32	_id{-1}; 
	int_32	_cuda_version{0};
	size_t	_total_memory{0};
	size_t	_shared_mem_per_block{0};
	size_t	_total_const_mem{0};
	int_32	_registers_per_block{0};
	int_32	_warp_size{0};
	int_32	_max_threads_per_block{0};
	int	    _max_threads_dim[3]{};
	int  	_max_grid_size[3]{};
	int_32	_multiprocessor_count{0};

private:
	//singleton
	GpuConfig();
	//template<typename T,size_t sz>
	void __copy(int (& from)[], int (&dest)[], size_t num_of_elemets) {
		//simple copy 
		//no need for boundry check in this usecase
		for (size_t i{ 0 }; i < num_of_elemets; ++i)
			dest[i] = from[i];
	}
	
	float extract_size(size_t bytes, const UnitMeasure sz) noexcept;
    
public:
	string		 Name()					     const;
	int_32		 Id()						 const;
	int_32		 CudaVersion()				 const;
	size_t		 TotalDeviceMem() 			 const;
	size_t		 TotalConstMem()			 const;
	size_t		 SharedBlockMem() 			 const;
	int_32		 RegistersPerBlock()		 const;
	int_32		 WarpSize()				 	 const;
	int_32		 MaxThreadPerBlock()		 const;
	const int*   MaxThreadDim();
	const int*   MaxGridSize();
	int_32		 MultiprocessorCount()		 const;
	void		 PrintInfo();
	float		 GetApproxSize(size_t bytes,const UnitMeasure ms);

	static GpuConfig& GetInstance();
	GpuConfig(const GpuConfig&) = delete;
	void operator=(const GpuConfig&) = delete;
	~GpuConfig() = default;
};