struct CodeGenerator::CUDA::Preferences¶

Overview¶

#include <backend.h>

struct Preferences: public CodeGenerator::PreferencesBase
{
    // fields

    bool showPtxInfo;
    DeviceSelect deviceSelectMethod;
    unsigned int manualDeviceID;
    BlockSizeSelect blockSizeSelectMethod;
    KernelBlockSize manualBlockSizes;
    std::string userNvccFlags;

    // methods

    Preferences();
};

Inherited Members¶

public:
    // fields

    bool optimizeCode;
    bool debugCode;
    std::string userCxxFlagsGNU;
    std::string userNvccFlagsGNU;
    plog::Severity logLevel;

Detailed Documentation¶

Preferences for CUDA backend.

Fields¶

bool showPtxInfo

Should PTX assembler information be displayed for each CUDA kernel during compilation.

DeviceSelect deviceSelectMethod

How to select GPU device.

unsigned int manualDeviceID

If device select method is set to DeviceSelect::MANUAL, id of device to use.

BlockSizeSelect blockSizeSelectMethod

How to select CUDA blocksize.

KernelBlockSize manualBlockSizes

If block size select method is set to BlockSizeSelect::MANUAL, block size to use for each kernel.

std::string userNvccFlags

NVCC compiler options for all GPU code.