public static void IlGpu(CudaAccelerator gpu, Real[] result, Real[] left, Real[] right, int n)
{
using (var cudaResult = gpu.Allocate(result))
using (var cudaLeft = gpu.Allocate(left))
using (var cudaRight = gpu.Allocate(right))
{
using var blas = new CuBlas(gpu, CuBlasAPIVersion.V11);
var timer = Stopwatch.StartNew();
blas.Gemm(
CuBlasOperation.NonTranspose,
CuBlasOperation.NonTranspose,
n, n, n,
1, cudaLeft.View, n,
cudaRight.View, n, 0,
cudaResult.View, n);
gpu.Synchronize();
PrintPerformance(timer, "MatrixMultiplication.IlGpu.cuBLAS", n, n, n);
cudaResult.CopyTo(result, 0, 0, result.Length);
}
}
