/// <summary>
/// Setups all required settings.
/// </summary>
private void SetupAccelerator()
{
Bind();
CudaException.ThrowIfFailed(
CurrentAPI.GetDeviceName(out string name, DeviceId));
Name = name;
DefaultStream = new CudaStream(this, IntPtr.Zero);
CudaException.ThrowIfFailed(
CurrentAPI.GetDeviceComputeCapability(out int major, out int minor, DeviceId));
Architecture = PTXBackend.GetArchitecture(major, minor);
CudaException.ThrowIfFailed(
CurrentAPI.GetTotalDeviceMemory(out long total, DeviceId));
MemorySize = total;
// Resolve max grid size
MaxGridSize = new Index3(
CurrentAPI.GetDeviceAttribute(
DeviceAttribute.CU_DEVICE_ATTRIBUTE_MAX_GRID_DIM_X, DeviceId),
CurrentAPI.GetDeviceAttribute(
DeviceAttribute.CU_DEVICE_ATTRIBUTE_MAX_GRID_DIM_Y, DeviceId),
CurrentAPI.GetDeviceAttribute(
DeviceAttribute.CU_DEVICE_ATTRIBUTE_MAX_GRID_DIM_Z, DeviceId));
// Resolve max group size
MaxGroupSize = new Index3(
CurrentAPI.GetDeviceAttribute(
DeviceAttribute.CU_DEVICE_ATTRIBUTE_MAX_BLOCK_DIM_X, DeviceId),
CurrentAPI.GetDeviceAttribute(
DeviceAttribute.CU_DEVICE_ATTRIBUTE_MAX_BLOCK_DIM_Y, DeviceId),
CurrentAPI.GetDeviceAttribute(
DeviceAttribute.CU_DEVICE_ATTRIBUTE_MAX_BLOCK_DIM_Z, DeviceId));
// Resolve max threads per group
MaxNumThreadsPerGroup = CurrentAPI.GetDeviceAttribute(
DeviceAttribute.CU_DEVICE_ATTRIBUTE_MAX_THREADS_PER_BLOCK, DeviceId);
// Resolve max shared memory per block
MaxSharedMemoryPerGroup = CurrentAPI.GetDeviceAttribute(
DeviceAttribute.CU_DEVICE_ATTRIBUTE_MAX_SHARED_MEMORY_PER_BLOCK, DeviceId);
// Resolve total constant memory
MaxConstantMemory = CurrentAPI.GetDeviceAttribute(
DeviceAttribute.CU_DEVICE_ATTRIBUTE_TOTAL_CONSTANT_MEMORY, DeviceId);
// Resolve clock rate
ClockRate = CurrentAPI.GetDeviceAttribute(
DeviceAttribute.CU_DEVICE_ATTRIBUTE_CLOCK_RATE, DeviceId);
// Resolve warp size
WarpSize = CurrentAPI.GetDeviceAttribute(
DeviceAttribute.CU_DEVICE_ATTRIBUTE_WARP_SIZE, DeviceId);
// Resolve number of multiprocessors
NumMultiprocessors = CurrentAPI.GetDeviceAttribute(
DeviceAttribute.CU_DEVICE_ATTRIBUTE_MULTIPROCESSOR_COUNT, DeviceId);
// Result max number of threads per multiprocessor
MaxNumThreadsPerMultiprocessor = CurrentAPI.GetDeviceAttribute(
DeviceAttribute.CU_DEVICE_ATTRIBUTE_MAX_THREADS_PER_MULTIPROCESSOR, DeviceId);
// Resolve cache configuration
CudaException.ThrowIfFailed(
CurrentAPI.GetSharedMemoryConfig(out sharedMemoryConfiguration));
CudaException.ThrowIfFailed(
CurrentAPI.GetCacheConfig(out cacheConfiguration));
}
