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

Merge pull request #46 from mlemanczyk/codex/investigate-evenperfectbitscanner-calculation-errors

Browse Source 
Harden GPU divisor scan buffer handling
This commit is contained in:
Marek Lemańczyk authored and GitHub committed 2025-09-20 21:29:36 +02:00
2 parents 2f07b776c6 8e686fbe6e
commit 633c94906a
1 file changed
+20 -10
Download Patch File Download Diff File Expand all files Collapse all files
PerfectNumbers.Core/Gpu/MersenneNumberDivisorGpuTester.cs
+20 -10
View File
@@ -1,3 +1,4 @@
using System;
using System.Collections.Concurrent; using System.Collections.Concurrent;
using System.Numerics; using System.Numerics;
using ILGPU; using ILGPU;
@@ -31,7 +32,7 @@ public sealed class MersenneNumberDivisorGpuTester
list[count++] = ((ulong)i, cycle); list[count++] = ((ulong)i, cycle);
} }
_divisorCandidates = list[..count].ToArray(); _divisorCandidates = count == 0 ? Array.Empty<(ulong divisor, uint cycle)>() : list[..count].ToArray();
} }
public bool IsDivisible(ulong exponent, UInt128 divisor) public bool IsDivisible(ulong exponent, UInt128 divisor)
@@ -40,14 +41,20 @@ public sealed class MersenneNumberDivisorGpuTester
var accelerator = gpu.Accelerator; var accelerator = gpu.Accelerator;
var kernel = GetKernel(accelerator); var kernel = GetKernel(accelerator);
var resultBuffer = _resultBuffers.GetOrAdd(accelerator, acc => acc.Allocate1D<byte>(1)); var resultBuffer = _resultBuffers.GetOrAdd(accelerator, acc => acc.Allocate1D<byte>(1));
resultBuffer.MemSetToZero();
kernel(1, exponent, (GpuUInt128)divisor, resultBuffer.View); kernel(1, exponent, (GpuUInt128)divisor, resultBuffer.View);
accelerator.Synchronize(); accelerator.Synchronize();
bool divisible = resultBuffer.GetAsArray1D()[0] != 0; Span<byte> result = stackalloc byte[1];
resultBuffer.View.CopyToCPU(ref result[0], 1);
bool divisible = result[0] != 0;
result[0] = 0;
resultBuffer.View.CopyFromCPU(ref result[0], 1);
accelerator.Synchronize();
gpu.Dispose(); gpu.Dispose();
return divisible; return divisible;
} }
private static (ulong divisor, uint cycle)[]? _divisorCandidates; private static (ulong divisor, uint cycle)[]? _divisorCandidates = Array.Empty<(ulong divisor, uint cycle)>();
public bool IsPrime(ulong p, UInt128 d, ulong divisorCyclesSearchLimit, out bool divisorsExhausted) public bool IsPrime(ulong p, UInt128 d, ulong divisorCyclesSearchLimit, out bool divisorsExhausted)
{ {
@@ -63,11 +70,11 @@ public sealed class MersenneNumberDivisorGpuTester
return true; return true;
} }
var candidates = _divisorCandidates!; if (_divisorCandidates is { Length: > 0 } candidates)
int k, len = candidates.Length; {
int len = candidates.Length;
uint cycle; uint cycle;
for (int k = 0; k < len; k++)
for (k = 0; k < len; k++)
{ {
(ulong dSmall, cycle) = candidates[k]; (ulong dSmall, cycle) = candidates[k];
if (p % cycle != 0UL) if (p % cycle != 0UL)
@@ -81,6 +88,7 @@ public sealed class MersenneNumberDivisorGpuTester
return false; return false;
} }
} }
}
UInt128 kMul2; UInt128 kMul2;
var divisorCycles = MersenneDivisorCycles.Shared; var divisorCycles = MersenneDivisorCycles.Shared;
@@ -149,10 +157,12 @@ public sealed class MersenneNumberDivisorGpuTester
baseVal = new GpuUInt128(0UL, 1UL << x); baseVal = new GpuUInt128(0UL, 1UL << x);
} }
baseVal.ModPow(exponent / ux, mod); GpuUInt128 pow = baseVal;
pow.ModPow(exponent / ux, mod);
var part2 = GpuUInt128.Pow2Mod(exponent % ux, mod); var part2 = GpuUInt128.Pow2Mod(exponent % ux, mod);
baseVal.MulMod(part2, mod); GpuUInt128 product = pow;
result[0] = baseVal.High == 0UL && baseVal.Low == 1UL ? (byte)1 : (byte)0; product.MulMod(part2, mod);
result[0] = product.High == 0UL && product.Low == 1UL ? (byte)1 : (byte)0;
} }
} }