static public FSharpOption <double> Calc(CalcPIParam param)
{
if (CanDoGPUCalc)
{
var worker = Worker.Default;
var pi = worker.EvalFunc(() => _Calc(worker, param));
return(FSharpOption <double> .Some(pi));
}
return(FSharpOption <double> .None);
}
static private double _Calc(Worker worker, CalcPIParam param)
{
var seed = param.Seed;
var numStreams = param.NumStreams;
const int numDimentions = 2;
var numPoints = param.NumPoints;
var numSMs = worker.Device.Attributes.MULTIPROCESSOR_COUNT;
var startStreamId = param.StartStreamId;
var stopStreamId = param.StopStreamId;
var random = param.GetRandom(seed, numStreams, numDimentions);
using (var reduce = DeviceSumModuleI32.Default.Create(numPoints))
using (var points = random.AllocCUDAStreamBuffer(numPoints))
using (var numPointsInside = worker.Malloc <int>(numPoints))
{
var pointsX = points.Ptr;
var pointsY = points.Ptr + numPoints;
var lp = new LaunchParam(numSMs * 8, 256);
var countStream = stopStreamId + 1 - startStreamId;
var pi =
Enumerable.Range(startStreamId, countStream).Select(streamId =>
{
random.Fill(streamId, numPoints, points);
worker.Launch(KernelCountInside, lp, pointsX, pointsY, numPoints, numPointsInside.Ptr);
var numPointsInsideH = reduce.Reduce(numPointsInside.Ptr, numPoints);
return(4.0 * (double)numPointsInsideH / (double)numPoints / (double)countStream);
}).Aggregate((a, b) => a + b);
Console.WriteLine("Streams({0}-{1}/{2}) Points({3}) : {4}",
startStreamId + 1,
stopStreamId + 1,
numStreams,
numPoints,
pi);
return(pi);
}
}
