Merge branch 'main' of https://github.com/mlemanczyk/even-perfect-numbers-scanner

PerfectNumbers.Core/Gpu/GpuKernelPool.cs

@@ -542,9 +542,12 @@ public class GpuKernelPool
 	/// <param name="action">Action to run with (Accelerator, Stream).</param>
     public static void Run(Action<Accelerator, AcceleratorStream> action)
     {
-        using var lease = GetKernel(useGpuOrder: true);
+        var lease = GetKernel(useGpuOrder: true);
         var accelerator = lease.Accelerator;
-        using var stream = accelerator.CreateStream();
+        var stream = accelerator.CreateStream();
         action(accelerator, stream);
+        stream.Dispose();
+        lease.Dispose();
+
     }
 }

PerfectNumbers.Core/Gpu/MersenneNumberIncrementalGpuTester.cs

@@ -11,7 +11,7 @@ public class MersenneNumberIncrementalGpuTester(GpuKernelType kernelType, bool u
 
         public void Scan(ulong exponent, UInt128 twoP, bool lastIsSeven, UInt128 maxK, ref bool isPrime)
         {
-		using var gpuLease = GpuKernelPool.GetKernel(_useGpuOrder);
+                var gpuLease = GpuKernelPool.GetKernel(_useGpuOrder);
                 var accelerator = gpuLease.Accelerator;
                 int batchSize = GpuConstants.ScanBatchSize; // large batch improves GPU occupancy
                 UInt128 kStart = 1UL;
@@ -22,7 +22,7 @@ public class MersenneNumberIncrementalGpuTester(GpuKernelType kernelType, bool u
             ? gpuLease.Pow2ModKernel
             : gpuLease.IncrementalKernel;
 
-		using var orderBuffer = accelerator.Allocate1D<ulong>(batchSize);
+                var orderBuffer = accelerator.Allocate1D<ulong>(batchSize);
                 // Avoid giant stack allocations that can trigger StackOverflow when batchSize is large.
                 // Rent a reusable array from the shared pool instead.
                 ulong[] orderArray = ArrayPool<ulong>.Shared.Rent(batchSize);
@@ -73,5 +73,9 @@ public class MersenneNumberIncrementalGpuTester(GpuKernelType kernelType, bool u
                 {
                         ArrayPool<ulong>.Shared.Return(orderArray);
                 }
+
+                orderBuffer.Dispose();
+                gpuLease.Dispose();
+
         }
 }

PerfectNumbers.Core/Gpu/MersenneNumberLucasLehmerGpuTester.cs

@@ -55,30 +55,39 @@ public class MersenneNumberLucasLehmerGpuTester
             return false;
         }
 
-        using var limiter = GpuPrimeWorkLimiter.Acquire();
+        var limiter = GpuPrimeWorkLimiter.Acquire();
+        bool result;
         if (exponent >= 128UL)
         {
             if (!TryGetNttParameters(exponent, out var nttMod, out var primitiveRoot))
             {
+                limiter.Dispose();
                 throw new NotSupportedException("NTT parameters for the given exponent are not available.");
             }
 
-            return IsMersennePrimeNtt(exponent, nttMod, primitiveRoot, runOnGpu);
+            result = IsMersennePrimeNtt(exponent, nttMod, primitiveRoot, runOnGpu);
         }
-
-        using var gpu = GpuContextPool.RentPreferred(preferCpu: !runOnGpu);
+        else
+        {
+            var gpu = GpuContextPool.RentPreferred(preferCpu: !runOnGpu);
             var accelerator = gpu.Accelerator;
             var kernel = GetKernel(accelerator);
             var modulus = new GpuUInt128(((UInt128)1 << (int)exponent) - 1UL);
-        using var buffer = accelerator.Allocate1D<GpuUInt128>(1);
+            var buffer = accelerator.Allocate1D<GpuUInt128>(1);
             kernel(1, exponent, modulus, buffer.View);
             accelerator.Synchronize();
-        return buffer.GetAsArray1D()[0].IsZero;
+            result = buffer.GetAsArray1D()[0].IsZero;
+            buffer.Dispose();
+            gpu.Dispose();
+        }
+
+        limiter.Dispose();
+        return result;
     }
 
     private bool IsMersennePrimeNtt(ulong exponent, GpuUInt128 nttMod, GpuUInt128 primitiveRoot, bool runOnGpu)
     {
-        using var gpu = GpuContextPool.RentPreferred(preferCpu: !runOnGpu);
+        var gpu = GpuContextPool.RentPreferred(preferCpu: !runOnGpu);
         var accelerator = gpu.Accelerator;
         var addKernel = GetAddSmallKernel(accelerator);
         var subKernel = GetSubSmallKernel(accelerator);
@@ -92,7 +101,7 @@ public class MersenneNumberLucasLehmerGpuTester
         {
             transformLength <<= 1;
         }
-        using var stateBuffer = accelerator.Allocate1D<GpuUInt128>(transformLength);
+        var stateBuffer = accelerator.Allocate1D<GpuUInt128>(transformLength);
         stateBuffer.MemSetToZero();
         addKernel(1, 4UL, stateBuffer.View);
 
@@ -121,12 +130,16 @@ public class MersenneNumberLucasLehmerGpuTester
 
         reduceKernel(1, exponent, stateBuffer.View);
 
-        using var resultBuffer = accelerator.Allocate1D<byte>(1);
+        var resultBuffer = accelerator.Allocate1D<byte>(1);
         zeroKernel(1, stateBuffer.View, resultBuffer.View);
         accelerator.Synchronize();
         byte[] result = new byte[1];
         resultBuffer.View.CopyToCPU(result);
-        return result[0] != 0;
+        bool isPrime = result[0] != 0;
+        resultBuffer.Dispose();
+        stateBuffer.Dispose();
+        gpu.Dispose();
+        return isPrime;
     }
 
     private static readonly ConcurrentDictionary<int, (GpuUInt128 Modulus, GpuUInt128 PrimitiveRoot)> NttParameterCache = new()

PerfectNumbers.Core/Gpu/MersenneNumberOrderGpuTester.cs

@@ -9,7 +9,7 @@ public class MersenneNumberOrderGpuTester(GpuKernelType kernelType, bool useGpuO
 
         public void Scan(ulong exponent, UInt128 twoP, bool lastIsSeven, UInt128 maxK, ref bool isPrime)
         {
-		using var gpuLease = GpuKernelPool.GetKernel(_useGpuOrder);
+                var gpuLease = GpuKernelPool.GetKernel(_useGpuOrder);
                 var accelerator = gpuLease.Accelerator;
                 int batchSize = GpuConstants.ScanBatchSize; // large batch improves GPU occupancy and avoids TDR
                 UInt128 kStart = UInt128.One;
@@ -22,7 +22,7 @@ public class MersenneNumberOrderGpuTester(GpuKernelType kernelType, bool useGpuO
             _ => gpuLease.IncrementalOrderKernel,
         };
 
-		using var foundBuffer = accelerator.Allocate1D<int>(1);
+                var foundBuffer = accelerator.Allocate1D<int>(1);
                 UInt128 remaining;
                 int currentSize;
                 int found = 0;
@@ -59,5 +59,9 @@ public class MersenneNumberOrderGpuTester(GpuKernelType kernelType, bool useGpuO
 
                         kStart += (UInt128)currentSize;
                 }
+
+                foundBuffer.Dispose();
+                gpuLease.Dispose();
+
         }
 }

PerfectNumbers.Core/Gpu/MersenneNumberResidueGpuTester.cs

@@ -14,7 +14,7 @@ public class MersenneNumberResidueGpuTester(bool useGpuOrder)
 	// GPU residue variant: check 2^p % q == 1 for q = 2*p*k + 1.
     public void Scan(ulong exponent, UInt128 twoP, bool lastIsSeven, UInt128 maxK, ref bool isPrime)
     {
-        using var gpuLease = GpuKernelPool.GetKernel(_useGpuOrder);
+        var gpuLease = GpuKernelPool.GetKernel(_useGpuOrder);
         var accelerator = gpuLease.Accelerator;
         // Ensure device has small cycles table for in-kernel lookup
         var smallCyclesView = GpuKernelPool.EnsureSmallCyclesOnDevice(accelerator);
@@ -23,7 +23,7 @@ public class MersenneNumberResidueGpuTester(bool useGpuOrder)
                 byte last = lastIsSeven ? (byte)1 : (byte)0;
                 var kernel = gpuLease.Pow2ModKernel;
 
-		using var orderBuffer = accelerator.Allocate1D<ulong>(batchSize);
+                var orderBuffer = accelerator.Allocate1D<ulong>(batchSize);
                 ulong[] orderArray = ArrayPool<ulong>.Shared.Rent(batchSize);
                 UInt128 remaining;
                 int currentSize;
@@ -60,7 +60,7 @@ public class MersenneNumberResidueGpuTester(bool useGpuOrder)
                                 orderBuffer.View.CopyToCPU(ref MemoryMarshal.GetReference(orders), currentSize);
                                 if (!Volatile.Read(ref isPrime))
                                 {
-					return;
+                                        break;
                                 }
 
                 for (i = 0; i < currentSize; i++, q += twoP)
@@ -79,5 +79,9 @@ public class MersenneNumberResidueGpuTester(bool useGpuOrder)
                 {
                         ArrayPool<ulong>.Shared.Return(orderArray);
                 }
+
+                orderBuffer.Dispose();
+                gpuLease.Dispose();
+
     }
 }

PerfectNumbers.Core/Gpu/NttGpuMath.cs

@@ -804,19 +804,22 @@ public static class NttGpuMath
     {
         int n = values.Length;
         int bits = (int)Math.Log2(n);
-        using var gpu = GpuContextPool.Rent();
+        var gpu = GpuContextPool.Rent();
         var accelerator = gpu.Accelerator;
         var kernel = GetBitReverseKernel(accelerator);
         var pool = ArrayPool<GpuUInt128>.Shared;
         var array = pool.Rent(n);
         values.CopyTo(array);
-        using var buffer = accelerator.Allocate1D<GpuUInt128>(n);
+        var buffer = accelerator.Allocate1D<GpuUInt128>(n);
         buffer.View.CopyFromCPU(ref array[0], n);
         kernel(n, buffer.View, bits);
         accelerator.Synchronize();
         buffer.View.CopyToCPU(ref array[0], n);
         array.AsSpan(0, n).CopyTo(values);
         pool.Return(array, clearArray: true);
+        buffer.Dispose();
+        gpu.Dispose();
+
     }
 
     public static void Forward(Span<GpuUInt128> values, GpuUInt128 modulus, GpuUInt128 primitiveRoot)
@@ -863,15 +866,15 @@ public static class NttGpuMath
     private static void ForwardGpuReference(Span<GpuUInt128> values, GpuUInt128 modulus, GpuUInt128 primitiveRoot)
     {
         int n = (int)values.Length;
-        using var gpu = GpuContextPool.Rent();
+        var gpu = GpuContextPool.Rent();
         var accelerator = gpu.Accelerator;
         var kernel = GetForwardKernel(accelerator);
         var pool = ArrayPool<GpuUInt128>.Shared;
         var inputArray = pool.Rent(n);
         var outputArray = pool.Rent(n);
         values.CopyTo(inputArray);
-        using var inputBuffer = accelerator.Allocate1D<GpuUInt128>(n);
-        using var outputBuffer = accelerator.Allocate1D<GpuUInt128>(n);
+        var inputBuffer = accelerator.Allocate1D<GpuUInt128>(n);
+        var outputBuffer = accelerator.Allocate1D<GpuUInt128>(n);
         inputBuffer.View.CopyFromCPU(ref inputArray[0], n);
         UInt128 modValue = modulus;
         UInt128 expBase = (modValue - 1UL) / (ulong)n;
@@ -883,6 +886,10 @@ public static class NttGpuMath
         outputArray.AsSpan(0, n).CopyTo(values);
         pool.Return(inputArray, clearArray: true);
         pool.Return(outputArray, clearArray: true);
+        outputBuffer.Dispose();
+        inputBuffer.Dispose();
+        gpu.Dispose();
+
     }
 
     public static void Inverse(Span<GpuUInt128> values, GpuUInt128 modulus, GpuUInt128 primitiveRoot)
@@ -954,15 +961,15 @@ public static class NttGpuMath
     private static void InverseGpuReference(Span<GpuUInt128> values, GpuUInt128 modulus, GpuUInt128 primitiveRoot)
     {
         int n = values.Length;
-        using var gpu = GpuContextPool.Rent();
+        var gpu = GpuContextPool.Rent();
         var accelerator = gpu.Accelerator;
         var kernel = GetInverseKernel(accelerator);
         var pool = ArrayPool<GpuUInt128>.Shared;
         var inputArray = pool.Rent(n);
         var outputArray = pool.Rent(n);
         values.CopyTo(inputArray);
-        using var inputBuffer = accelerator.Allocate1D<GpuUInt128>(n);
-        using var outputBuffer = accelerator.Allocate1D<GpuUInt128>(n);
+        var inputBuffer = accelerator.Allocate1D<GpuUInt128>(n);
+        var outputBuffer = accelerator.Allocate1D<GpuUInt128>(n);
         inputBuffer.View.CopyFromCPU(ref inputArray[0], n);
         UInt128 modValue = modulus;
         UInt128 expBase = (modValue - 1UL) / (ulong)n;
@@ -993,30 +1000,40 @@ public static class NttGpuMath
         outputArray.AsSpan(0, n).CopyTo(values);
         pool.Return(inputArray, clearArray: true);
         pool.Return(outputArray, clearArray: true);
+        outputBuffer.Dispose();
+        inputBuffer.Dispose();
+        gpu.Dispose();
+
     }
 
     private static void ForwardGpuStaged(Span<GpuUInt128> values, GpuUInt128 modulus, GpuUInt128 primitiveRoot)
     {
         int n = values.Length;
-        using var gpu = GpuContextPool.Rent();
+        var gpu = GpuContextPool.Rent();
         var accelerator = gpu.Accelerator;
-        using var buffer = accelerator.Allocate1D<GpuUInt128>(n);
+        var buffer = accelerator.Allocate1D<GpuUInt128>(n);
         buffer.View.CopyFromCPU(ref values[0], n);
         var cache = GetSquareCache(accelerator, n, modulus, primitiveRoot);
         ForwardDevice(accelerator, buffer.View, n, modulus, cache);
         buffer.View.CopyToCPU(ref values[0], n);
+        buffer.Dispose();
+        gpu.Dispose();
+
     }
 
     private static void InverseGpuStaged(Span<GpuUInt128> values, GpuUInt128 modulus, GpuUInt128 primitiveRoot)
     {
         int n = values.Length;
-        using var gpu = GpuContextPool.Rent();
+        var gpu = GpuContextPool.Rent();
         var accelerator = gpu.Accelerator;
-        using var buffer = accelerator.Allocate1D<GpuUInt128>(n);
+        var buffer = accelerator.Allocate1D<GpuUInt128>(n);
         buffer.View.CopyFromCPU(ref values[0], n);
         var cache = GetSquareCache(accelerator, n, modulus, primitiveRoot);
         InverseDevice(accelerator, buffer.View, n, modulus, cache);
         buffer.View.CopyToCPU(ref values[0], n);
+        buffer.Dispose();
+        gpu.Dispose();
+
     }
 
     private static void ForwardDevice(Accelerator accelerator, ArrayView<GpuUInt128> data, int length, GpuUInt128 modulus, SquareCacheEntry cache)
@@ -1221,7 +1238,7 @@ public static class NttGpuMath
 
     public static void PointwiseMultiply(Span<GpuUInt128> left, Span<GpuUInt128> right, GpuUInt128 modulus)
     {
-        using var gpu = GpuContextPool.Rent();
+        var gpu = GpuContextPool.Rent();
         var accelerator = gpu.Accelerator;
         var kernel = GetMulKernel(accelerator);
         var pool = ArrayPool<GpuUInt128>.Shared;
@@ -1229,8 +1246,8 @@ public static class NttGpuMath
         var rightArray = pool.Rent(right.Length);
         left.CopyTo(leftArray);
         right.CopyTo(rightArray);
-        using var leftBuffer = accelerator.Allocate1D<GpuUInt128>(left.Length);
-        using var rightBuffer = accelerator.Allocate1D<GpuUInt128>(right.Length);
+        var leftBuffer = accelerator.Allocate1D<GpuUInt128>(left.Length);
+        var rightBuffer = accelerator.Allocate1D<GpuUInt128>(right.Length);
         leftBuffer.View.CopyFromCPU(ref leftArray[0], left.Length);
         rightBuffer.View.CopyFromCPU(ref rightArray[0], right.Length);
         kernel(left.Length, leftBuffer.View, rightBuffer.View, modulus);
@@ -1241,6 +1258,10 @@ public static class NttGpuMath
         rightArray.AsSpan(0, right.Length).CopyTo(right);
         pool.Return(leftArray, clearArray: true);
         pool.Return(rightArray, clearArray: true);
+        rightBuffer.Dispose();
+        leftBuffer.Dispose();
+        gpu.Dispose();
+
     }
 
     private static ulong[] GenerateSmallPrimes(int limit)

PerfectNumbers.Core/MersenneDivisorCycles.cs

@@ -225,9 +225,9 @@ public class MersenneDivisorCycles
 
                 divisors = pool.Rent(count);
                 outCycles = pool.Rent(count);
-		using var lease = GpuKernelPool.GetKernel(useGpuOrder: true);
+                var lease = GpuKernelPool.GetKernel(useGpuOrder: true);
                 var accelerator = lease.Accelerator;
-		using var stream = accelerator.CreateStream();
+                var stream = accelerator.CreateStream();
                 var kernel = accelerator.LoadAutoGroupedStreamKernel<
                         Index1D,
                         ArrayView1D<ulong, Stride1D.Dense>,
@@ -314,6 +314,10 @@ public class MersenneDivisorCycles
                         pool.Return(divisors, clearArray: false);
                         pool.Return(outCycles, clearArray: false);
                 }
+
+                stream.Dispose();
+                lease.Dispose();
+
         }
 
 	const byte ByteZero = 0;

PerfectNumbers.Core/MersenneNumberTester.cs

@@ -141,7 +141,7 @@ public sealed class MersenneNumberTester(
 
                 UInt128[] qs = qsBuffer;
 
-                using var gpuLease = GpuKernelPool.GetKernel(_useGpuOrder);
+                var gpuLease = GpuKernelPool.GetKernel(_useGpuOrder);
                 var accelerator = gpuLease.Accelerator;
         var orderKernel = gpuLease.OrderKernel!;
 
@@ -153,8 +153,8 @@ public sealed class MersenneNumberTester(
                 while (offset < idx)
                 {
             int count = Math.Min(batchSize, idx - offset);
-            using var qBuffer = accelerator.Allocate1D<PerfectNumbers.Core.Gpu.GpuUInt128>(count);
-            using var orderBuffer = accelerator.Allocate1D<ulong>(count);
+            var qBuffer = accelerator.Allocate1D<PerfectNumbers.Core.Gpu.GpuUInt128>(count);
+            var orderBuffer = accelerator.Allocate1D<ulong>(count);
             // Convert to device-friendly GpuUInt128 on the fly
             var tmp = System.Buffers.ArrayPool<PerfectNumbers.Core.Gpu.GpuUInt128>.Shared.Rent(count);
             for (int i = 0; i < count; i++)
@@ -183,8 +183,11 @@ public sealed class MersenneNumberTester(
                         }
 
                         offset += count;
+            orderBuffer.Dispose();
+            qBuffer.Dispose();
                 }
 
+                gpuLease.Dispose();
                 ArrayPool<UInt128>.Shared.Return(qs);
 	}
 

PerfectNumbers.Core/PrimeTester.cs

@@ -81,8 +81,8 @@ public sealed class PrimeTester(bool useInternal = false)
     public static void IsPrimeBatchGpu(ReadOnlySpan<ulong> values, Span<byte> results)
     {
         // Limit concurrency and declare variables outside loops for performance and reuse
-        using var limiter = GpuPrimeWorkLimiter.Acquire();
-        using var gpu = GpuContextPool.Rent();
+        var limiter = GpuPrimeWorkLimiter.Acquire();
+        var gpu = GpuContextPool.Rent();
         var accelerator = gpu.Accelerator;
 
         // Get per-accelerator cached kernel and device primes buffer.
@@ -123,6 +123,8 @@ public sealed class PrimeTester(bool useInternal = false)
 
         ArrayPool<ulong>.Shared.Return(temp);
         state.ReturnScratch(scratch);
+        gpu.Dispose();
+        limiter.Dispose();
     }
 
     // Per-accelerator GPU state for prime sieve (kernel + uploaded primes).
@@ -283,13 +285,13 @@ public sealed class PrimeTester(bool useInternal = false)
 
     internal static void SharesFactorWithMaxExponentBatch(ReadOnlySpan<ulong> values, Span<byte> results)
     {
-        using var gpu = GpuContextPool.Rent();
+        var gpu = GpuContextPool.Rent();
         var accelerator = gpu.Accelerator;
         var kernel = accelerator.LoadAutoGroupedStreamKernel<Index1D, ArrayView<ulong>, ArrayView<byte>>(SharesFactorKernel);
 
         int length = values.Length;
-        using var inputBuffer = accelerator.Allocate1D<ulong>(length);
-        using var resultBuffer = accelerator.Allocate1D<byte>(length);
+        var inputBuffer = accelerator.Allocate1D<ulong>(length);
+        var resultBuffer = accelerator.Allocate1D<byte>(length);
 
         ulong[] temp = ArrayPool<ulong>.Shared.Rent(length);
         values.CopyTo(temp);
@@ -298,6 +300,9 @@ public sealed class PrimeTester(bool useInternal = false)
         accelerator.Synchronize();
         resultBuffer.View.CopyToCPU(ref results[0], length);
         ArrayPool<ulong>.Shared.Return(temp);
+        resultBuffer.Dispose();
+        inputBuffer.Dispose();
+        gpu.Dispose();
     }
 
     [MethodImpl(MethodImplOptions.AggressiveInlining)]