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

Various optimizations to separate non-static data between threads and resolve read/write access issue.

Browse Source
This commit is contained in:
mlemanczyk committed 2025-11-04 20:45:56 +01:00
1 parent 5a17fa64d5
commit 64d615c402
26 files changed
+1181 -1117
Download Patch File Download Diff File Expand all files Collapse all files

No files matched your search

PerfectNumbers.Core/Gpu/Kernels/PrimeOrderSpecialMaxKernels.cs
+4 -5
View File
@@ -1,6 +1,5 @@
using ILGPU;
using ILGPU.Runtime;
using PerfectNumbers.Core;
namespace PerfectNumbers.Core.Gpu;
@@ -13,7 +12,7 @@ internal static partial class PrimeOrderGpuHeuristics
int factorCount,
MontgomeryDivisorData divisor,
ArrayView1D<ulong, Stride1D.Dense> candidates,
ArrayView1D<ushort, Stride1D.Dense> resultOut)
ArrayView1D<ulong, Stride1D.Dense> resultOut)
{
if (index != 0)
{
@@ -62,7 +61,7 @@ internal static partial class PrimeOrderGpuHeuristics
var stepper = new ExponentRemainderStepperGpu(divisorPartial);
if (stepper.InitializeIsUnityGpu(candidates[0]))
{
resultOut[0] = 0;
Atomic.Exchange(ref resultOut[0], 0);
return;
}
@@ -70,11 +69,11 @@ internal static partial class PrimeOrderGpuHeuristics
{
if (stepper.ComputeNextIsUnityGpu(candidates[i]))
{
resultOut[0] = 0;
Atomic.Exchange(ref resultOut[0], 0);
return;
}
}
resultOut[0] = 1;
Atomic.Exchange(ref resultOut[0], 1);
}
}
PerfectNumbers.Core/Gpu/Kernels/PrimeTesterKernels.cs
+13 -13
View File
@@ -18,19 +18,19 @@ internal static class PrimeTesterKernels
{
// GPU kernel: small-prime sieve only. Returns 1 if passes sieve (probable prime), 0 otherwise.
public static void SmallPrimeSieveKernel(
Index1D index,
ArrayView<ulong> numbers,
ArrayView<uint> smallPrimes,
ArrayView<uint> smallPrimesLastOne,
ArrayView<uint> smallPrimesLastSeven,
ArrayView<uint> smallPrimesLastThree,
ArrayView<uint> smallPrimesLastNine,
ArrayView<ulong> smallPrimesPow2,
ArrayView<ulong> smallPrimesPow2LastOne,
ArrayView<ulong> smallPrimesPow2LastSeven,
ArrayView<ulong> smallPrimesPow2LastThree,
ArrayView<ulong> smallPrimesPow2LastNine,
ArrayView<byte> results)
Index1D index,
ArrayView<ulong> numbers,
ArrayView<uint> smallPrimes,
ArrayView<uint> smallPrimesLastOne,
ArrayView<uint> smallPrimesLastSeven,
ArrayView<uint> smallPrimesLastThree,
ArrayView<uint> smallPrimesLastNine,
ArrayView<ulong> smallPrimesPow2,
ArrayView<ulong> smallPrimesPow2LastOne,
ArrayView<ulong> smallPrimesPow2LastSeven,
ArrayView<ulong> smallPrimesPow2LastThree,
ArrayView<ulong> smallPrimesPow2LastNine,
ArrayView<byte> results)
{
ulong n = numbers[index];
byte result = 1;