Streams added to all kernels and GPU operations to separate threads from each other.

PerfectNumbers.Core/Gpu/Kernels/LucasLehmerKernels.cs

@@ -305,7 +305,7 @@ internal static class LucasLehmerKernels
         states[idx] = s;
     }
 
-    public static void Kernel(Index1D index, ulong exponent, GpuUInt128 modulus, ArrayView<GpuUInt128> state)
+    public static void Kernel(Index1D _, ulong exponent, GpuUInt128 modulus, ArrayView<GpuUInt128> state)
     {
         // Lucas–Lehmer iteration in the field GF(2^p-1).
         // For p < 128, use a Mersenne-specific squaring + reduction to avoid

PerfectNumbers.Core/Gpu/Kernels/NttBarrettKernels.cs

@@ -1,5 +1,6 @@
 using System.Runtime.CompilerServices;
 using ILGPU;
+using ILGPU.Runtime;
 
 namespace PerfectNumbers.Core.Gpu;
 
@@ -32,7 +33,7 @@ internal static class NttBarrettKernels
     }
 
     // Barrett-based squaring: v = v^2 mod n for 128-bit n
-    public static void SquareBarrett128Kernel(Index1D index, ArrayView<GpuUInt128> data, ulong modHigh, ulong modLow, ulong muHigh, ulong muLow)
+    public static void SquareBarrett128Kernel(Index1D index, AcceleratorStream stream, ArrayView<GpuUInt128> data, ulong modHigh, ulong modLow, ulong muHigh, ulong muLow)
     {
         var v = data[index];
         Mul128Full(v.High, v.Low, v.High, v.Low, out var z3, out var z2, out var z1, out var z0);

PerfectNumbers.Core/Gpu/Kernels/NttButterflyKernels.cs

@@ -1,4 +1,5 @@
 using ILGPU;
+using ILGPU.Runtime;
 
 namespace PerfectNumbers.Core.Gpu;
 

PerfectNumbers.Core/Gpu/Kernels/NttMontgomeryKernels.cs

@@ -1,5 +1,6 @@
 using System.Runtime.CompilerServices;
 using ILGPU;
+using ILGPU.Runtime;
 
 namespace PerfectNumbers.Core.Gpu;
 
@@ -29,7 +30,7 @@ internal static class NttMontgomeryKernels
         data[i2] = new GpuUInt128(0UL, diff);
     }
 
-    public static void ToMont64Kernel(Index1D index, ArrayView<GpuUInt128> data, ulong modulus, ulong nPrime, ulong r2)
+    public static void ToMont64Kernel(Index1D index, AcceleratorStream stream, ArrayView<GpuUInt128> data, ulong modulus, ulong nPrime, ulong r2)
     {
         ulong a = data[index].Low;
         ulong v = MontMul64(a, r2, modulus, nPrime);
@@ -43,7 +44,7 @@ internal static class NttMontgomeryKernels
         data[index] = new GpuUInt128(0UL, v);
     }
 
-    public static void SquareMont64Kernel(Index1D index, ArrayView<GpuUInt128> data, ulong modulus, ulong nPrime)
+    public static void SquareMont64Kernel(Index1D index, AcceleratorStream stream, ArrayView<GpuUInt128> data, ulong modulus, ulong nPrime)
     {
         ulong aR = data[index].Low;
         ulong v = MontMul64(aR, aR, modulus, nPrime);

PerfectNumbers.Core/Gpu/Kernels/NttPointwiseKernels.cs

@@ -1,4 +1,5 @@
 using ILGPU;
+using ILGPU.Runtime;
 
 namespace PerfectNumbers.Core.Gpu;
 

PerfectNumbers.Core/Gpu/Kernels/NttTransformKernels.cs

@@ -1,4 +1,5 @@
 using ILGPU;
+using ILGPU.Runtime;
 
 namespace PerfectNumbers.Core.Gpu;