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

Merge pull request #198 from mlemanczyk/codex/napraw-bedy-kompilacji-kerneli

Browse Source 
Fix kernel integration compile errors
This commit is contained in:
Marek Lemańczyk authored and GitHub committed 2025-10-18 18:02:05 +02:00
2 parents 9232778173 8adeb59acc
commit 2ba193bd65
4 files changed
+9 -22
Download Patch File Download Diff File Expand all files Collapse all files

No files matched your search

PerfectNumbers.Core/Gpu/Kernels/PrimeOrderKernels.cs
+1 -1
View File
@@ -7,7 +7,7 @@ using ILGPU.Runtime;
namespace PerfectNumbers.Core.Gpu; namespace PerfectNumbers.Core.Gpu;
internal static class PrimeOrderKernels internal static partial class PrimeOrderGpuHeuristics
{ {
internal static void PartialFactorKernel( internal static void PartialFactorKernel(
Index1D index, Index1D index,
PerfectNumbers.Core/Gpu/PrimeOrderGpuHeuristics.cs
+7 -20
View File
@@ -123,7 +123,7 @@ internal static partial class PrimeOrderGpuHeuristics
{ {
return PartialFactorKernelCache.GetOrAdd(accelerator, static accel => return PartialFactorKernelCache.GetOrAdd(accelerator, static accel =>
{ {
var loaded = accel.LoadAutoGroupedStreamKernel<Index1D, ArrayView1D<uint, Stride1D.Dense>, ArrayView1D<ulong, Stride1D.Dense>, int, int, ulong, uint, ArrayView1D<ulong, Stride1D.Dense>, ArrayView1D<int, Stride1D.Dense>, ArrayView1D<int, Stride1D.Dense>, ArrayView1D<ulong, Stride1D.Dense>, ArrayView1D<byte, Stride1D.Dense>>(PrimeOrderKernels.PartialFactorKernel); var loaded = accel.LoadAutoGroupedStreamKernel<Index1D, ArrayView1D<uint, Stride1D.Dense>, ArrayView1D<ulong, Stride1D.Dense>, int, int, ulong, uint, ArrayView1D<ulong, Stride1D.Dense>, ArrayView1D<int, Stride1D.Dense>, ArrayView1D<int, Stride1D.Dense>, ArrayView1D<ulong, Stride1D.Dense>, ArrayView1D<byte, Stride1D.Dense>>(PartialFactorKernel);
var kernel = KernelUtil.GetKernel(loaded); var kernel = KernelUtil.GetKernel(loaded);
return kernel.CreateLauncherDelegate<Action<AcceleratorStream, Index1D, ArrayView1D<uint, Stride1D.Dense>, ArrayView1D<ulong, Stride1D.Dense>, int, int, ulong, uint, ArrayView1D<ulong, Stride1D.Dense>, ArrayView1D<int, Stride1D.Dense>, ArrayView1D<int, Stride1D.Dense>, ArrayView1D<ulong, Stride1D.Dense>, ArrayView1D<byte, Stride1D.Dense>>>(); return kernel.CreateLauncherDelegate<Action<AcceleratorStream, Index1D, ArrayView1D<uint, Stride1D.Dense>, ArrayView1D<ulong, Stride1D.Dense>, int, int, ulong, uint, ArrayView1D<ulong, Stride1D.Dense>, ArrayView1D<int, Stride1D.Dense>, ArrayView1D<int, Stride1D.Dense>, ArrayView1D<ulong, Stride1D.Dense>, ArrayView1D<byte, Stride1D.Dense>>>();
}); });
@@ -420,15 +420,15 @@ internal static partial class PrimeOrderGpuHeuristics
byte status = 0; byte status = 0;
statusBuffer.View.CopyToCPU(ref status, 1); statusBuffer.View.CopyToCPU(ref status, 1);
PrimeOrderKernels.PrimeOrderKernelStatus kernelStatus = (PrimeOrderKernels.PrimeOrderKernelStatus)status; PrimeOrderKernelStatus kernelStatus = (PrimeOrderKernelStatus)status;
if (kernelStatus == PrimeOrderKernels.PrimeOrderKernelStatus.Fallback) if (kernelStatus == PrimeOrderKernelStatus.Fallback)
{ {
DisposeResources(); DisposeResources();
lease.Dispose(); lease.Dispose();
return false; return false;
} }
if (kernelStatus == PrimeOrderKernels.PrimeOrderKernelStatus.PollardOverflow) if (kernelStatus == PrimeOrderKernelStatus.PollardOverflow)
{ {
DisposeResources(); DisposeResources();
lease.Dispose(); lease.Dispose();
@@ -437,7 +437,7 @@ internal static partial class PrimeOrderGpuHeuristics
resultBuffer.View.CopyToCPU(ref order, 1); resultBuffer.View.CopyToCPU(ref order, 1);
if (kernelStatus == PrimeOrderKernels.PrimeOrderKernelStatus.FactoringFailure) if (kernelStatus == PrimeOrderKernelStatus.FactoringFailure)
{ {
order = 0UL; order = 0UL;
} }
@@ -467,7 +467,7 @@ internal static partial class PrimeOrderGpuHeuristics
{ {
return OrderKernelCache.GetOrAdd(accelerator, static accel => return OrderKernelCache.GetOrAdd(accelerator, static accel =>
{ {
var loaded = accel.LoadAutoGroupedStreamKernel<Index1D, ulong, OrderKernelConfig, MontgomeryDivisorData, ArrayView1D<uint, Stride1D.Dense>, ArrayView1D<ulong, Stride1D.Dense>, int, OrderKernelBuffers>(PrimeOrderKernels.CalculateOrderKernel); var loaded = accel.LoadAutoGroupedStreamKernel<Index1D, ulong, OrderKernelConfig, MontgomeryDivisorData, ArrayView1D<uint, Stride1D.Dense>, ArrayView1D<ulong, Stride1D.Dense>, int, OrderKernelBuffers>(CalculateOrderKernel);
var kernel = KernelUtil.GetKernel(loaded); var kernel = KernelUtil.GetKernel(loaded);
return kernel.CreateLauncherDelegate<OrderKernelLauncher>(); return kernel.CreateLauncherDelegate<OrderKernelLauncher>();
}); });
@@ -658,27 +658,14 @@ internal static partial class PrimeOrderGpuHeuristics
{ {
return Pow2ModKernelCache.GetOrAdd(accelerator, static accel => return Pow2ModKernelCache.GetOrAdd(accelerator, static accel =>
{ {
var loaded = accel.LoadAutoGroupedStreamKernel<Index1D, ArrayView1D<ulong, Stride1D.Dense>, MontgomeryDivisorData, ArrayView1D<ulong, Stride1D.Dense>>(PrimeOrderKernels.Pow2ModKernel); var loaded = accel.LoadAutoGroupedStreamKernel<Index1D, ArrayView1D<ulong, Stride1D.Dense>, MontgomeryDivisorData, ArrayView1D<ulong, Stride1D.Dense>>(Pow2ModKernel);
var kernel = KernelUtil.GetKernel(loaded); var kernel = KernelUtil.GetKernel(loaded);
return kernel.CreateLauncherDelegate<Action<AcceleratorStream, Index1D, ArrayView1D<ulong, Stride1D.Dense>, MontgomeryDivisorData, ArrayView1D<ulong, Stride1D.Dense>>>(); return kernel.CreateLauncherDelegate<Action<AcceleratorStream, Index1D, ArrayView1D<ulong, Stride1D.Dense>, MontgomeryDivisorData, ArrayView1D<ulong, Stride1D.Dense>>>();
}); });
} }
private static Action<AcceleratorStream, Index1D, ArrayView1D<GpuUInt128, Stride1D.Dense>, GpuUInt128, ArrayView1D<GpuUInt128, Stride1D.Dense>> GetPow2ModWideKernel(Accelerator accelerator)
{
return Pow2ModKernelWideCache.GetOrAdd(accelerator, static accel =>
{
var loaded = accel.LoadAutoGroupedStreamKernel<Index1D, ArrayView1D<GpuUInt128, Stride1D.Dense>, GpuUInt128, ArrayView1D<GpuUInt128, Stride1D.Dense>>(PrimeOrderKernels.Pow2ModKernelWide);
var kernel = KernelUtil.GetKernel(loaded);
return kernel.CreateLauncherDelegate<Action<AcceleratorStream, Index1D, ArrayView1D<GpuUInt128, Stride1D.Dense>, GpuUInt128, ArrayView1D<GpuUInt128, Stride1D.Dense>>>();
});
}
private static GpuUInt128 Pow2ModKernelCore(GpuUInt128 exponent, GpuUInt128 modulus)
{
return PrimeOrderKernels.Pow2ModKernelCore(exponent, modulus);
}
internal readonly record struct PrimeOrderGpuCapability(int ModulusBits, int ExponentBits) internal readonly record struct PrimeOrderGpuCapability(int ModulusBits, int ExponentBits)
{ {
PerfectNumbers.Core/PrimeTester.cs
-1
View File
@@ -335,4 +335,3 @@ public sealed class PrimeTester(bool useInternal = false)
return u << shift; return u << shift;
} }
} }
}
PerfectNumbers.Core/ULongExtensions.cs
+1
View File
@@ -574,6 +574,7 @@ public static class ULongExtensions
} }
}
[MethodImpl(MethodImplOptions.AggressiveInlining)] [MethodImpl(MethodImplOptions.AggressiveInlining)]
private static ulong Pow2MontgomeryModSingleBit(ulong exponent, in MontgomeryDivisorData divisor, bool keepMontgomery) private static ulong Pow2MontgomeryModSingleBit(ulong exponent, in MontgomeryDivisorData divisor, bool keepMontgomery)
{ {