public static float[] RunOddEvenSort2(float[] a)
{
int N = a.Length;
bool evenArr = (N % 2) == 0 ? true : false;
//Create context and accelerator
var gpu = new CudaAccelerator(new Context());
//Create typed launcher
var oddEvenKernel = gpu.LoadAutoGroupedStreamKernel <
Index1,
ArrayView <float>,
VariableView <byte>,
int,
bool>(OddEvenSort2);
//Allocate memory
MemoryBuffer <float> d_a = gpu.Allocate <float>(N);
MemoryBuffer <byte> d_stopFlag = gpu.AllocateZero <byte>(1);
d_a.CopyFrom(a, 0, Index1.Zero, N);
//Run kernel
oddEvenKernel(N / 2, d_a, d_stopFlag.View.GetVariableView(0), N, evenArr);
gpu.Synchronize();
return(d_a.GetAsArray());
}
public static float[] RunOddEvenSort(float[] a, ref Stopwatch sw)
{
int N = a.Length;
bool evenArr = (N % 2) == 0 ? true : false;
bool stopFlag = false;
bool iterationEven = true;
//Create context and accelerator
var gpu = new CudaAccelerator(new Context());
//Create typed launcher
var oddEvenKernel = gpu.LoadAutoGroupedStreamKernel <
Index1,
ArrayView <float>,
VariableView <byte>,
bool,
int,
bool>(OddEvenSort);
//Allocate memory
MemoryBuffer <float> d_a = gpu.Allocate <float>(N);
MemoryBuffer <byte> d_stopFlag = gpu.AllocateZero <byte>(1);
d_a.CopyFrom(a, 0, Index1.Zero, N);
sw.Restart();
//Run kernel
byte[] zero_val = new byte[1];
zero_val[0] = 0;
while (true)
{
if (stopFlag)
{
break;
}
stopFlag = true;
d_stopFlag.CopyFrom(zero_val, 0, 0, 1);
oddEvenKernel(N / 2, d_a, d_stopFlag.View.GetVariableView(), iterationEven, N, evenArr);
gpu.Synchronize();
if (d_stopFlag.GetAsArray()[0] > 0)
{
stopFlag = false;
}
iterationEven = !iterationEven;
}
sw.Stop();
return(d_a.GetAsArray());
}
