TODOs added

PerfectNumbers.Core/Cpu/MersenneNumberDivisorByDivisorCpuTester.cs

@@ -8,6 +8,7 @@ using MontgomeryDivisorData = PerfectNumbers.Core.Gpu.MersenneNumberDivisorByDiv
 
 namespace PerfectNumbers.Core.Cpu;
 
+// TODO: The current implementation is over-complicated for CPU. We shouldn't need all the hassle with intermediate buffers, checking the status there etc.
 public sealed class MersenneNumberDivisorByDivisorCpuTester : IMersenneNumberDivisorByDivisorTester
 {
 	private readonly object _sync = new();

PerfectNumbers.Core/Gpu/MersenneNumberDivisorByDivisorGpuTester.cs

@@ -301,6 +301,7 @@ public sealed class MersenneNumberDivisorByDivisorGpuTester : IMersenneNumberDiv
 		return Math.Min((1UL << (int)(prime - 1UL)) - 1UL, divisorLimit);
 	}
 
+	// TODO: Since lastProcessed is unused, let's remove it.
 	private void UpdateStatusUnsafe(ulong lastProcessed, ulong processedCount)
 	{
 		if (processedCount == 0UL)
@@ -315,10 +316,9 @@ public sealed class MersenneNumberDivisorByDivisorGpuTester : IMersenneNumberDiv
 			return;
 		}
 
+		// Can we avoid the modulo calculation here? For the higher divisors it may become very expensive. Do we need to calculate the total, too?
 		ulong total = _lastStatusDivisor + processedCount;
 		_lastStatusDivisor = total % interval;
-
-		// Removed noisy console output; status is tracked internally only.
 	}
 
 	private static void CheckKernel(Index1D index, ulong prime, byte useCycleChecks, ArrayView<MontgomeryDivisorData> divisors, ArrayView<byte> hits)
@@ -340,7 +340,7 @@ public sealed class MersenneNumberDivisorByDivisorGpuTester : IMersenneNumberDiv
 				hits[index] = 0;
 				return;
 			}
-
+			// TODO: Can we move this back before the GPU call and let CPU handle it? We significantly slowed down after this was moved to GPU.
 			ulong remainder = prime % cycle;
 			if (remainder != 0UL)
 			{
@@ -471,6 +471,7 @@ public sealed class MersenneNumberDivisorByDivisorGpuTester : IMersenneNumberDiv
 
 				if (cycleEnabled)
 				{
+					// TODO: Can we move this back before the GPU call and let CPU handle it? We significantly slowed down after this was moved to GPU.
 					cycleMaskSpan = _cycleMaskHost.AsSpan(0, batchSize);
 					ArrayView1D<ulong, Stride1D.Dense> primeInputView = _resultsBuffer.View.SubView(0, batchSize);
 					primeInputView.CopyFromCPU(ref MemoryMarshal.GetReference(primesSlice), batchSize);
@@ -488,6 +489,7 @@ public sealed class MersenneNumberDivisorByDivisorGpuTester : IMersenneNumberDiv
 						continue;
 					}
 
+					// TODO: Avoid one-time use variables like prime here.
 					ulong prime = primesSlice[i];
 					primeSpan[computeCount] = prime;
 					positionSpan[computeCount] = i;
@@ -537,6 +539,7 @@ public sealed class MersenneNumberDivisorByDivisorGpuTester : IMersenneNumberDiv
 
 					for (int i = 0; i < computeCount; i++)
 					{
+						// TODO: Avoid one-time use variables like prime and position here.
 						int position = positionSpan[i];
 						ulong stepResidue = exponentSlice[i];
 						ulong prime = primeSpan[i];