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

Merge remote-tracking branch 'github/codex/optimize-evenperfectbitscanner-for-static-arrays-khd9w8' into work

Browse Source 
# Conflicts:
#	PerfectNumbers.Core/Gpu/GpuContextPool.cs
#	PerfectNumbers.Core/Gpu/GpuKernelLease.cs
#	PerfectNumbers.Core/Gpu/GpuKernelPool.cs
#	PerfectNumbers.Core/Gpu/GpuStaticTableInitializer.cs
#	PerfectNumbers.Core/Gpu/Kernels/PrimeOrderSpecialMaxKernels.cs
#	PerfectNumbers.Core/PrimeOrderCalculator.Cpu.cs
#	PerfectNumbers.Core/PrimeOrderCalculator.Gpu.cs
#	PerfectNumbers.Core/PrimeTester.cs
This commit is contained in:
Marek Lemańczyk committed 2025-11-03 23:59:12 +01:00
1 parent 507daaddb9
commit 4c4d4243fb
23 files changed
+858 -633
Download Patch File Download Diff File Expand all files Collapse all files

No files matched your search

PerfectNumbers.Core/Gpu/GpuKernelLease.cs
+31 -19
View File
@@ -2,6 +2,7 @@ using System;
using System.Collections.Concurrent;
using ILGPU;
using ILGPU.Runtime;
using PerfectNumbers.Core;
using static PerfectNumbers.Core.Gpu.GpuContextPool;
namespace PerfectNumbers.Core.Gpu;
@@ -10,25 +11,24 @@ public sealed class GpuKernelLease
{
private static readonly ConcurrentQueue<GpuKernelLease> Pool = new();
private Action? _releaseLimiter;
private GpuContextLease? _gpu;
private KernelContainer? _kernels;
public KernelContainer Kernels;
private AcceleratorStream? _stream;
private GpuKernelLease()
{
}
private GpuKernelLease(KernelContainer kernels)
{
Kernels = kernels;
}
internal static GpuKernelLease Rent(GpuPrimeWorkLimiter.Lease limiter, GpuContextLease gpu, KernelContainer kernels)
internal static GpuKernelLease Rent(GpuContextLease gpu, KernelContainer kernels)
{
if (!Pool.TryDequeue(out var lease))
{
lease = new GpuKernelLease();
lease = new GpuKernelLease(kernels);
}
lease._releaseLimiter = limiter.Dispose;
lease._gpu = gpu;
lease._kernels = kernels;
lease.Kernels = kernels;
lease._stream = null;
return lease;
}
@@ -56,7 +56,7 @@ public sealed class GpuKernelLease
get
{
var accel = Accelerator; // avoid capturing 'this' in lambda
return KernelContainer.InitOnce(ref _kernels!.Order, () =>
return KernelContainer.InitOnce(ref Kernels!.Order, () =>
{
var loaded = accel.LoadAutoGroupedStreamKernel<Index1D, ulong, ulong, ArrayView<GpuUInt128>, ArrayView<ulong>>(OrderKernels.OrderKernelScan);
var kernel = KernelUtil.GetKernel(loaded);
@@ -70,7 +70,7 @@ public sealed class GpuKernelLease
get
{
var accel = Accelerator;
return KernelContainer.InitOnce(ref _kernels!.Pow2Mod, () =>
return KernelContainer.InitOnce(ref Kernels!.Pow2Mod, () =>
{
var loaded = accel.LoadAutoGroupedStreamKernel<Index1D, ulong, GpuUInt128, GpuUInt128, byte, ulong, ResidueAutomatonArgs, ArrayView<ulong>, ArrayView1D<ulong, Stride1D.Dense>, ArrayView1D<uint, Stride1D.Dense>, ArrayView1D<uint, Stride1D.Dense>, ArrayView1D<ulong, Stride1D.Dense>, ArrayView1D<ulong, Stride1D.Dense>>(Pow2ModKernels.Pow2ModKernelScan);
var kernel = KernelUtil.GetKernel(loaded);
@@ -84,7 +84,7 @@ public sealed class GpuKernelLease
get
{
var accel = Accelerator;
return KernelContainer.InitOnce(ref _kernels!.Incremental, () =>
return KernelContainer.InitOnce(ref Kernels!.Incremental, () =>
{
var loaded = accel.LoadAutoGroupedStreamKernel<Index1D, ulong, GpuUInt128, GpuUInt128, byte, ulong, ulong, ulong, ulong, ulong, ArrayView<ulong>, ArrayView1D<ulong, Stride1D.Dense>>(IncrementalKernels.IncrementalKernelScan);
var kernel = KernelUtil.GetKernel(loaded);
@@ -98,7 +98,7 @@ public sealed class GpuKernelLease
get
{
var accel = Accelerator;
return KernelContainer.InitOnce(ref _kernels!.IncrementalOrder, () =>
return KernelContainer.InitOnce(ref Kernels!.IncrementalOrder, () =>
{
var loaded = accel.LoadAutoGroupedStreamKernel<Index1D, ulong, GpuUInt128, GpuUInt128, byte, ulong, ResidueAutomatonArgs, ArrayView<int>, ArrayView1D<ulong, Stride1D.Dense>>(IncrementalKernels.IncrementalOrderKernelScan);
var kernel = KernelUtil.GetKernel(loaded);
@@ -112,7 +112,7 @@ public sealed class GpuKernelLease
get
{
var accel = Accelerator;
return KernelContainer.InitOnce(ref _kernels!.Pow2ModOrder, () =>
return KernelContainer.InitOnce(ref Kernels!.Pow2ModOrder, () =>
{
var loaded = accel.LoadAutoGroupedStreamKernel<Index1D, ulong, GpuUInt128, GpuUInt128, byte, ulong, ResidueAutomatonArgs, ArrayView<int>, ArrayView1D<ulong, Stride1D.Dense>>(Pow2ModKernels.Pow2ModOrderKernelScan);
var kernel = KernelUtil.GetKernel(loaded);
@@ -121,12 +121,26 @@ public sealed class GpuKernelLease
}
}
public Action<AcceleratorStream, Index1D, ulong, ArrayView1D<ulong, Stride1D.Dense>, int, MontgomeryDivisorData, ArrayView1D<ulong, Stride1D.Dense>, ArrayView1D<ushort, Stride1D.Dense>> SpecialMaxKernel
{
get
{
var accel = Accelerator;
return KernelContainer.InitOnce(ref Kernels!.SpecialMax, () =>
{
var loaded = accel.LoadAutoGroupedStreamKernel<Index1D, ulong, ArrayView1D<ulong, Stride1D.Dense>, int, MontgomeryDivisorData, ArrayView1D<ulong, Stride1D.Dense>, ArrayView1D<ushort, Stride1D.Dense>>(PrimeOrderGpuHeuristics.EvaluateSpecialMaxCandidatesKernel);
var kernel = KernelUtil.GetKernel(loaded);
return kernel.CreateLauncherDelegate<Action<AcceleratorStream, Index1D, ulong, ArrayView1D<ulong, Stride1D.Dense>, int, MontgomeryDivisorData, ArrayView1D<ulong, Stride1D.Dense>, ArrayView1D<ushort, Stride1D.Dense>>>();
});
}
}
public 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>> SmallPrimeFactorKernel
{
get
{
var accel = Accelerator;
return KernelContainer.InitOnce(ref _kernels!.SmallPrimeFactor, () =>
return KernelContainer.InitOnce(ref Kernels!.SmallPrimeFactor, () =>
{
var loaded = accel.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);
var kernel = KernelUtil.GetKernel(loaded);
@@ -143,10 +157,8 @@ public sealed class GpuKernelLease
_gpu?.Dispose();
_gpu = null;
_releaseLimiter?.Invoke();
_releaseLimiter = null;
_kernels = null;
Kernels?.Dispose();
GpuPrimeWorkLimiter.Release();
Pool.Enqueue(this);
}