mlemanczyk/even-perfect-numbers-scanner
1
0 Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

Heuristic and small prime factor kernels restored in PrimeOrderAcceleratorTemplate.

Browse Source
This commit is contained in:
mlemanczyk committed 2025-12-28 00:12:39 +01:00
1 parent 6603f65993
commit 75a2017385
1 file changed
+8 -11
Download Patch File Download Diff File Expand all files Collapse all files
PerfectNumbers.Core/Gpu/Accelerators/PrimeOrderCalculatorAccelerator.cs
+8 -11
View File
@@ -16,7 +16,7 @@ public sealed class PrimeOrderCalculatorAccelerator
public static PrimeOrderCalculatorAccelerator Rent(int primeTesterCapacity) public static PrimeOrderCalculatorAccelerator Rent(int primeTesterCapacity)
{ {
var pool = _pool ??= new(PerfectNumberConstants.DefaultThreadPoolCapacity); var pool = _pool ??= new(PerfectNumberConstants.DefaultThreadPoolCapacity);
if (pool.Pop() is { } gpu) if (pool.TryPop(out var gpu))
{ {
gpu.EnsureCapacity(PerfectNumberConstants.DefaultFactorsBuffer, primeTesterCapacity); gpu.EnsureCapacity(PerfectNumberConstants.DefaultFactorsBuffer, primeTesterCapacity);
return gpu; return gpu;
@@ -67,8 +67,8 @@ public sealed class PrimeOrderCalculatorAccelerator
// TODO: Review which tables are needed when final execution path is defined. // TODO: Review which tables are needed when final execution path is defined.
LastDigitGpuTables.WarmUp(i, stream); LastDigitGpuTables.WarmUp(i, stream);
_kernels[i] = new CalculatorKernels(accelerator); _kernels[i] = new CalculatorKernels(accelerator);
// HeuristicCombinedGpuTables.WarmUp(i, stream); HeuristicCombinedGpuTables.WarmUp(i, stream);
// SmallPrimeFactorGpuTables.WarmUp(i, stream); SmallPrimeFactorGpuTables.WarmUp(i, stream);
// SharedHeuristicGpuTables.EnsureStaticTables(accelerator, stream); // SharedHeuristicGpuTables.EnsureStaticTables(accelerator, stream);
// _ = GpuKernelPool.GetOrAddKernels(accelerator, stream, KernelType.None); // _ = GpuKernelPool.GetOrAddKernels(accelerator, stream, KernelType.None);
// KernelContainer kernels = GpuKernelPool.GetOrAddKernels(accelerator, stream); // KernelContainer kernels = GpuKernelPool.GetOrAddKernels(accelerator, stream);
@@ -170,7 +170,7 @@ public sealed class PrimeOrderCalculatorAccelerator
// public readonly Action<AcceleratorStream, Index1D, ArrayView<ulong>, ArrayView<byte>> SharesFactorKernelLauncher = KernelUtil.GetKernel(accelerator.LoadAutoGroupedStreamKernel<Index1D, ArrayView<ulong>, ArrayView<byte>>(PrimeTesterKernels.SharesFactorKernel)).CreateLauncherDelegate<Action<AcceleratorStream, Index1D, ArrayView<ulong>, ArrayView<byte>>>(); // public readonly Action<AcceleratorStream, Index1D, ArrayView<ulong>, ArrayView<byte>> SharesFactorKernelLauncher = KernelUtil.GetKernel(accelerator.LoadAutoGroupedStreamKernel<Index1D, ArrayView<ulong>, ArrayView<byte>>(PrimeTesterKernels.SharesFactorKernel)).CreateLauncherDelegate<Action<AcceleratorStream, Index1D, ArrayView<ulong>, ArrayView<byte>>>();
// public readonly Action<AcceleratorStream, Index1D, ulong, uint, ArrayView1D<uint, Stride1D.Dense>, ArrayView1D<ulong, Stride1D.Dense>, int, ArrayView1D<ulong, Stride1D.Dense>, ArrayView1D<int, Stride1D.Dense>, ArrayView1D<int, Stride1D.Dense>, ArrayView1D<ulong, Stride1D.Dense>> SmallPrimeFactorKernelLauncher = KernelUtil.GetKernel(accelerator.LoadAutoGroupedStreamKernel<Index1D, ulong, uint, ArrayView1D<uint, Stride1D.Dense>, ArrayView1D<ulong, Stride1D.Dense>, int, ArrayView1D<ulong, Stride1D.Dense>, ArrayView1D<int, Stride1D.Dense>, ArrayView1D<int, Stride1D.Dense>, ArrayView1D<ulong, Stride1D.Dense>>(SmallPrimeFactorKernels.SmallPrimeFactorKernelScan)).CreateLauncherDelegate<Action<AcceleratorStream, Index1D, ulong, uint, ArrayView1D<uint, Stride1D.Dense>, ArrayView1D<ulong, Stride1D.Dense>, int, ArrayView1D<ulong, Stride1D.Dense>, ArrayView1D<int, Stride1D.Dense>, ArrayView1D<int, Stride1D.Dense>, ArrayView1D<ulong, Stride1D.Dense>>>(); public readonly Action<AcceleratorStream, Index1D, ulong, uint, ArrayView1D<uint, Stride1D.Dense>, ArrayView1D<ulong, Stride1D.Dense>, int, ArrayView1D<ulong, Stride1D.Dense>, ArrayView1D<int, Stride1D.Dense>, ArrayView1D<int, Stride1D.Dense>, ArrayView1D<ulong, Stride1D.Dense>> SmallPrimeFactorKernelLauncher = KernelUtil.GetKernel(accelerator.LoadAutoGroupedStreamKernel<Index1D, ulong, uint, ArrayView1D<uint, Stride1D.Dense>, ArrayView1D<ulong, Stride1D.Dense>, int, ArrayView1D<ulong, Stride1D.Dense>, ArrayView1D<int, Stride1D.Dense>, ArrayView1D<int, Stride1D.Dense>, ArrayView1D<ulong, Stride1D.Dense>>(SmallPrimeFactorKernels.SmallPrimeFactorKernelScan)).CreateLauncherDelegate<Action<AcceleratorStream, Index1D, ulong, uint, ArrayView1D<uint, Stride1D.Dense>, ArrayView1D<ulong, Stride1D.Dense>, int, ArrayView1D<ulong, Stride1D.Dense>, ArrayView1D<int, Stride1D.Dense>, ArrayView1D<int, Stride1D.Dense>, ArrayView1D<ulong, Stride1D.Dense>>>();
// public readonly Action<AcceleratorStream, Index1D, ArrayView<ulong>, ArrayView<uint>, ArrayView<uint>, ArrayView<uint>, ArrayView<uint>, ArrayView<ulong>, ArrayView<ulong>, ArrayView<ulong>, ArrayView<ulong>, ArrayView<byte>> SmallPrimeSieveKernelLauncher = KernelUtil.GetKernel(accelerator.LoadAutoGroupedStreamKernel<Index1D, ArrayView<ulong>, ArrayView<uint>, ArrayView<uint>, ArrayView<uint>, ArrayView<uint>, ArrayView<ulong>, ArrayView<ulong>, ArrayView<ulong>, ArrayView<ulong>, ArrayView<byte>>(PrimeTesterKernels.SmallPrimeSieveKernel)).CreateLauncherDelegate<Action<AcceleratorStream, Index1D, ArrayView<ulong>, ArrayView<uint>, ArrayView<uint>, ArrayView<uint>, ArrayView<uint>, ArrayView<ulong>, ArrayView<ulong>, ArrayView<ulong>, ArrayView<ulong>, ArrayView<byte>>>(); // public readonly Action<AcceleratorStream, Index1D, ArrayView<ulong>, ArrayView<uint>, ArrayView<uint>, ArrayView<uint>, ArrayView<uint>, ArrayView<ulong>, ArrayView<ulong>, ArrayView<ulong>, ArrayView<ulong>, ArrayView<byte>> SmallPrimeSieveKernelLauncher = KernelUtil.GetKernel(accelerator.LoadAutoGroupedStreamKernel<Index1D, ArrayView<ulong>, ArrayView<uint>, ArrayView<uint>, ArrayView<uint>, ArrayView<uint>, ArrayView<ulong>, ArrayView<ulong>, ArrayView<ulong>, ArrayView<ulong>, ArrayView<byte>>(PrimeTesterKernels.SmallPrimeSieveKernel)).CreateLauncherDelegate<Action<AcceleratorStream, Index1D, ArrayView<ulong>, ArrayView<uint>, ArrayView<uint>, ArrayView<uint>, ArrayView<uint>, ArrayView<ulong>, ArrayView<ulong>, ArrayView<ulong>, ArrayView<ulong>, ArrayView<byte>>>();
@@ -360,8 +360,6 @@ public sealed class PrimeOrderCalculatorAccelerator
Accelerator = accelerator; Accelerator = accelerator;
AcceleratorIndex = acceleratorIndex; AcceleratorIndex = acceleratorIndex;
// lock (accelerator)
// {
Input = accelerator.Allocate1D<ulong>(Math.Max(Math.Max(specialMaxFactorCapacity, primeTesterCapacity), PrimeOrderConstants.MaxGpuBatchSize)); Input = accelerator.Allocate1D<ulong>(Math.Max(Math.Max(specialMaxFactorCapacity, primeTesterCapacity), PrimeOrderConstants.MaxGpuBatchSize));
InputView = Input.View; InputView = Input.View;
@@ -423,9 +421,9 @@ public sealed class PrimeOrderCalculatorAccelerator
DevicePrimesPow2LastThree = lastDigitSharedTables.DevicePrimesPow2LastThree.View; DevicePrimesPow2LastThree = lastDigitSharedTables.DevicePrimesPow2LastThree.View;
DevicePrimesPow2LastNine = lastDigitSharedTables.DevicePrimesPow2LastNine.View; DevicePrimesPow2LastNine = lastDigitSharedTables.DevicePrimesPow2LastNine.View;
// DivisorTables = HeuristicCombinedGpuTables DivisorTables = HeuristicCombinedGpuTables
// .GetStaticTables(acceleratorIndex) .GetStaticTables(acceleratorIndex)
// .CreateViews(); .CreateViews();
var smallPrimeFactorTables = SmallPrimeFactorGpuTables.GetStaticTables(acceleratorIndex); var smallPrimeFactorTables = SmallPrimeFactorGpuTables.GetStaticTables(acceleratorIndex);
@@ -448,10 +446,9 @@ public sealed class PrimeOrderCalculatorAccelerator
Pow2ModKernelLauncher = kernels.Pow2ModKernelLauncher; Pow2ModKernelLauncher = kernels.Pow2ModKernelLauncher;
Pow2ModWideKernelLauncher = kernels.Pow2ModWideKernelLauncher; Pow2ModWideKernelLauncher = kernels.Pow2ModWideKernelLauncher;
// SharesFactorKernelLauncher = kernels.SharesFactorKernelLauncher; // SharesFactorKernelLauncher = kernels.SharesFactorKernelLauncher;
// SmallPrimeFactorKernelLauncher = kernels.SmallPrimeFactorKernelLauncher; SmallPrimeFactorKernelLauncher = kernels.SmallPrimeFactorKernelLauncher;
// SmallPrimeSieveKernelLauncher = kernels.SmallPrimeSieveKernelLauncher; // SmallPrimeSieveKernelLauncher = kernels.SmallPrimeSieveKernelLauncher;
SpecialMaxKernelLauncher = kernels.SpecialMaxKernelLauncher; SpecialMaxKernelLauncher = kernels.SpecialMaxKernelLauncher;
// }
} }
public void Dispose() public void Dispose()