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

Remove temporary by-divisor review documents

Browse Source
This commit is contained in:
Marek Lemańczyk committed 2025-10-28 02:01:28 +01:00
1 parent bd0727131f
commit 6dd5cea79f
27 files changed
+498 -600
Download Patch File Download Diff File Expand all files Collapse all files

No files matched your search

EvenPerfectBitScanner/Program.cs
+2 -12
View File
@@ -104,14 +104,7 @@ internal static class Program
testPrimeCandidateLimit = Math.Max(1, threadCount) * 3;
}
if (useByDivisor)
{
_byDivisorStartPrime = startPrimeProvided ? currentP : 0UL;
}
else
{
_byDivisorStartPrime = 0UL;
}
_byDivisorStartPrime = useByDivisor && startPrimeProvided ? currentP : 0UL;
if (useByDivisor && !testMode && string.IsNullOrEmpty(filterFile))
{
@@ -177,10 +170,7 @@ internal static class Program
Console.WriteLine("Divisor cycles are ready");
if (useByDivisor)
{
blockSize = 1;
}
blockSize = useByDivisor ? 1 : blockSize;
CandidatesCalculator.Initialize(currentP);
PrimeTesters = new ThreadLocal<PrimeTester>(() => new PrimeTester(), trackAllValues: true);
PerfectNumbers.Core.Tests/Gpu/GpuPrimeLimiterTests.cs
+21 -10
View File
@@ -18,22 +18,33 @@ public class GpuPrimeLimiterTests
var startedSecond = new TaskCompletionSource<bool>();
var sw = new Stopwatch();
using var first = GpuPrimeWorkLimiter.Acquire();
var first = GpuPrimeWorkLimiter.Acquire();
sw.Start();
var secondTask = Task.Run(() =>
{
using var second = GpuPrimeWorkLimiter.Acquire();
startedSecond.TrySetResult(true);
var second = GpuPrimeWorkLimiter.Acquire();
try
{
startedSecond.TrySetResult(true);
}
finally
{
second.Dispose();
}
});
// Give secondTask a moment to attempt acquire (it should block).
await Task.Delay(50);
startedSecond.Task.IsCompleted.Should().BeFalse();
// Release first after short delay, allowing second to continue.
// (Dispose at scope end, but we want to end now.)
first.Dispose();
try
{
// Give secondTask a moment to attempt acquire (it should block).
await Task.Delay(50);
startedSecond.Task.IsCompleted.Should().BeFalse();
}
finally
{
// Release first after the initial wait so the second task can proceed.
first.Dispose();
}
await secondTask.WaitAsync(TimeSpan.FromSeconds(1));
sw.Stop();
PerfectNumbers.Core.Tests/GpuWorkLimiterTests.cs
+19 -7
View File
@@ -17,19 +17,31 @@ public class GpuWorkLimiterTests
var startedSecond = new TaskCompletionSource<bool>();
var sw = new Stopwatch();
using var first = GpuWorkLimiter.Acquire();
var first = GpuWorkLimiter.Acquire();
sw.Start();
var secondTask = Task.Run(() =>
{
using var second = GpuWorkLimiter.Acquire();
startedSecond.TrySetResult(true);
var second = GpuWorkLimiter.Acquire();
try
{
startedSecond.TrySetResult(true);
}
finally
{
second.Dispose();
}
});
await Task.Delay(50);
startedSecond.Task.IsCompleted.Should().BeFalse();
first.Dispose();
try
{
await Task.Delay(50);
startedSecond.Task.IsCompleted.Should().BeFalse();
}
finally
{
first.Dispose();
}
await secondTask.WaitAsync(TimeSpan.FromSeconds(1));
sw.Stop();
PerfectNumbers.Core.Tests/MersenneNumberDivisorCpuTesterTests.cs
+45 -24
View File
@@ -30,18 +30,25 @@ public class MersenneNumberDivisorCpuTesterTests
var tester = new MersenneNumberDivisorByDivisorCpuTester();
tester.ConfigureFromMaxPrime(13UL);
using var session = tester.CreateDivisorSession();
ulong[] primes = { 5UL, 7UL, 11UL, 13UL };
byte[] hits = new byte[primes.Length];
var session = tester.CreateDivisorSession();
try
{
ulong[] primes = { 5UL, 7UL, 11UL, 13UL };
byte[] hits = new byte[primes.Length];
ulong cycle23 = MersenneDivisorCycles.CalculateCycleLength(23UL, MontgomeryDivisorData.FromModulus(23UL));
session.CheckDivisor(23UL, MontgomeryDivisorData.FromModulus(23UL), cycle23, primes, hits);
hits.Should().ContainInOrder(new byte[] { 0, 0, 1, 0 });
ulong cycle23 = MersenneDivisorCycles.CalculateCycleLength(23UL, MontgomeryDivisorData.FromModulus(23UL));
session.CheckDivisor(23UL, MontgomeryDivisorData.FromModulus(23UL), cycle23, primes, hits);
hits.Should().ContainInOrder(new byte[] { 0, 0, 1, 0 });
Array.Fill(hits, (byte)0);
ulong cycle31 = MersenneDivisorCycles.CalculateCycleLength(31UL, MontgomeryDivisorData.FromModulus(31UL));
session.CheckDivisor(31UL, MontgomeryDivisorData.FromModulus(31UL), cycle31, primes, hits);
hits.Should().ContainInOrder(new byte[] { 1, 0, 0, 0 });
Array.Fill(hits, (byte)0);
ulong cycle31 = MersenneDivisorCycles.CalculateCycleLength(31UL, MontgomeryDivisorData.FromModulus(31UL));
session.CheckDivisor(31UL, MontgomeryDivisorData.FromModulus(31UL), cycle31, primes, hits);
hits.Should().ContainInOrder(new byte[] { 1, 0, 0, 0 });
}
finally
{
session.Dispose();
}
}
[Fact]
@@ -51,15 +58,22 @@ public class MersenneNumberDivisorCpuTesterTests
var tester = new MersenneNumberDivisorByDivisorCpuTester();
tester.ConfigureFromMaxPrime(43UL);
using var session = tester.CreateDivisorSession();
ulong[] exponents = { 6UL, 7UL, 9UL, 10UL, 12UL };
byte[] hits = new byte[exponents.Length];
var session = tester.CreateDivisorSession();
try
{
ulong[] exponents = { 6UL, 7UL, 9UL, 10UL, 12UL };
byte[] hits = new byte[exponents.Length];
MontgomeryDivisorData divisorData = MontgomeryDivisorData.FromModulus(7UL);
ulong cycle = MersenneDivisorCycles.CalculateCycleLength(7UL, divisorData);
session.CheckDivisor(7UL, divisorData, cycle, exponents, hits);
MontgomeryDivisorData divisorData = MontgomeryDivisorData.FromModulus(7UL);
ulong cycle = MersenneDivisorCycles.CalculateCycleLength(7UL, divisorData);
session.CheckDivisor(7UL, divisorData, cycle, exponents, hits);
hits.Should().Equal(new byte[] { 1, 0, 1, 0, 1 });
hits.Should().Equal(new byte[] { 1, 0, 1, 0, 1 });
}
finally
{
session.Dispose();
}
}
[Fact]
@@ -69,15 +83,22 @@ public class MersenneNumberDivisorCpuTesterTests
var tester = new MersenneNumberDivisorByDivisorCpuTester();
tester.ConfigureFromMaxPrime(61UL);
using var session = tester.CreateDivisorSession();
ulong[] exponents = { 10UL, 11UL, 20UL, 21UL, 30UL };
byte[] hits = new byte[exponents.Length];
var session = tester.CreateDivisorSession();
try
{
ulong[] exponents = { 10UL, 11UL, 20UL, 21UL, 30UL };
byte[] hits = new byte[exponents.Length];
MontgomeryDivisorData divisorData = MontgomeryDivisorData.FromModulus(11UL);
ulong cycle = MersenneDivisorCycles.CalculateCycleLength(11UL, divisorData);
session.CheckDivisor(11UL, divisorData, cycle, exponents, hits);
MontgomeryDivisorData divisorData = MontgomeryDivisorData.FromModulus(11UL);
ulong cycle = MersenneDivisorCycles.CalculateCycleLength(11UL, divisorData);
session.CheckDivisor(11UL, divisorData, cycle, exponents, hits);
hits.Should().Equal(new byte[] { 1, 0, 1, 0, 1 });
hits.Should().Equal(new byte[] { 1, 0, 1, 0, 1 });
}
finally
{
session.Dispose();
}
}
}
PerfectNumbers.Core.Tests/MersenneNumberDivisorGpuTesterTests.cs
+83 -41
View File
@@ -149,13 +149,20 @@ public class MersenneNumberDivisorGpuTesterTests
var tester = new MersenneNumberDivisorByDivisorGpuTester();
tester.ConfigureFromMaxPrime(43UL);
using var session = tester.CreateDivisorSession();
ulong[] primes = { 31UL, 37UL, 41UL };
byte[] hits = new byte[primes.Length];
var session = tester.CreateDivisorSession();
try
{
ulong[] primes = { 31UL, 37UL, 41UL };
byte[] hits = new byte[primes.Length];
session.CheckDivisor(223UL, MontgomeryDivisorData.FromModulus(223UL), 0UL, primes, hits);
session.CheckDivisor(223UL, MontgomeryDivisorData.FromModulus(223UL), 0UL, primes, hits);
hits.Should().ContainInOrder(new byte[] { 0, 1, 0 });
hits.Should().ContainInOrder(new byte[] { 0, 1, 0 });
}
finally
{
session.Dispose();
}
}
[Fact]
@@ -165,18 +172,25 @@ public class MersenneNumberDivisorGpuTesterTests
var tester = new MersenneNumberDivisorByDivisorGpuTester();
tester.ConfigureFromMaxPrime(43UL);
using var session = tester.CreateDivisorSession();
ulong[] primes = { 31UL, 37UL, 41UL, 43UL };
byte[] hits = new byte[primes.Length];
var session = tester.CreateDivisorSession();
try
{
ulong[] primes = { 31UL, 37UL, 41UL, 43UL };
byte[] hits = new byte[primes.Length];
ulong cycle223 = MersenneDivisorCycles.CalculateCycleLength(223UL, MontgomeryDivisorData.FromModulus(223UL));
session.CheckDivisor(223UL, MontgomeryDivisorData.FromModulus(223UL), cycle223, primes, hits);
hits.Should().ContainInOrder(new byte[] { 0, 1, 0, 0 });
ulong cycle223 = MersenneDivisorCycles.CalculateCycleLength(223UL, MontgomeryDivisorData.FromModulus(223UL));
session.CheckDivisor(223UL, MontgomeryDivisorData.FromModulus(223UL), cycle223, primes, hits);
hits.Should().ContainInOrder(new byte[] { 0, 1, 0, 0 });
Array.Fill(hits, (byte)0);
ulong cycle13367 = MersenneDivisorCycles.CalculateCycleLength(13367UL, MontgomeryDivisorData.FromModulus(13367UL));
session.CheckDivisor(13367UL, MontgomeryDivisorData.FromModulus(13367UL), cycle13367, primes, hits);
hits.Should().ContainInOrder(new byte[] { 0, 0, 1, 0 });
Array.Fill(hits, (byte)0);
ulong cycle13367 = MersenneDivisorCycles.CalculateCycleLength(13367UL, MontgomeryDivisorData.FromModulus(13367UL));
session.CheckDivisor(13367UL, MontgomeryDivisorData.FromModulus(13367UL), cycle13367, primes, hits);
hits.Should().ContainInOrder(new byte[] { 0, 0, 1, 0 });
}
finally
{
session.Dispose();
}
}
[Fact]
@@ -190,14 +204,21 @@ public class MersenneNumberDivisorGpuTesterTests
tester.ConfigureFromMaxPrime(47UL);
using var session = tester.CreateDivisorSession();
ulong[] primes = { 31UL, 37UL, 41UL, 43UL, 47UL };
byte[] hits = new byte[primes.Length];
var session = tester.CreateDivisorSession();
try
{
ulong[] primes = { 31UL, 37UL, 41UL, 43UL, 47UL };
byte[] hits = new byte[primes.Length];
ulong cycle223 = MersenneDivisorCycles.CalculateCycleLength(223UL, MontgomeryDivisorData.FromModulus(223UL));
session.CheckDivisor(223UL, MontgomeryDivisorData.FromModulus(223UL), cycle223, primes, hits);
ulong cycle223 = MersenneDivisorCycles.CalculateCycleLength(223UL, MontgomeryDivisorData.FromModulus(223UL));
session.CheckDivisor(223UL, MontgomeryDivisorData.FromModulus(223UL), cycle223, primes, hits);
hits.Should().ContainInOrder(new byte[] { 0, 1, 0, 0, 0 });
hits.Should().ContainInOrder(new byte[] { 0, 1, 0, 0, 0 });
}
finally
{
session.Dispose();
}
}
[Fact]
@@ -210,16 +231,23 @@ public class MersenneNumberDivisorGpuTesterTests
};
tester.ConfigureFromMaxPrime(43UL);
using var session = tester.CreateDivisorSession();
ulong[] exponents = { 6UL, 7UL, 9UL, 10UL, 12UL };
byte[] hits = new byte[exponents.Length];
var session = tester.CreateDivisorSession();
try
{
ulong[] exponents = { 6UL, 7UL, 9UL, 10UL, 12UL };
byte[] hits = new byte[exponents.Length];
MontgomeryDivisorData divisorData = MontgomeryDivisorData.FromModulus(7UL);
ulong cycle = MersenneDivisorCycles.CalculateCycleLength(7UL, divisorData);
MontgomeryDivisorData divisorData = MontgomeryDivisorData.FromModulus(7UL);
ulong cycle = MersenneDivisorCycles.CalculateCycleLength(7UL, divisorData);
session.CheckDivisor(7UL, divisorData, cycle, exponents, hits);
session.CheckDivisor(7UL, divisorData, cycle, exponents, hits);
hits.Should().Equal(new byte[] { 1, 0, 1, 0, 1 });
hits.Should().Equal(new byte[] { 1, 0, 1, 0, 1 });
}
finally
{
session.Dispose();
}
}
[Fact]
@@ -232,16 +260,23 @@ public class MersenneNumberDivisorGpuTesterTests
};
tester.ConfigureFromMaxPrime(61UL);
using var session = tester.CreateDivisorSession();
ulong[] exponents = { 10UL, 11UL, 20UL, 21UL, 30UL };
byte[] hits = new byte[exponents.Length];
var session = tester.CreateDivisorSession();
try
{
ulong[] exponents = { 10UL, 11UL, 20UL, 21UL, 30UL };
byte[] hits = new byte[exponents.Length];
MontgomeryDivisorData divisorData = MontgomeryDivisorData.FromModulus(11UL);
ulong cycle = MersenneDivisorCycles.CalculateCycleLength(11UL, divisorData);
MontgomeryDivisorData divisorData = MontgomeryDivisorData.FromModulus(11UL);
ulong cycle = MersenneDivisorCycles.CalculateCycleLength(11UL, divisorData);
session.CheckDivisor(11UL, divisorData, cycle, exponents, hits);
session.CheckDivisor(11UL, divisorData, cycle, exponents, hits);
hits.Should().Equal(new byte[] { 1, 0, 1, 0, 1 });
hits.Should().Equal(new byte[] { 1, 0, 1, 0, 1 });
}
finally
{
session.Dispose();
}
}
[Fact]
@@ -297,14 +332,21 @@ public class MersenneNumberDivisorGpuTesterTests
tester.ConfigureFromMaxPrime(47UL);
using var session = tester.CreateDivisorSession();
ulong[] primes = { 31UL, 37UL, 41UL, 43UL, 47UL };
byte[] hits = new byte[primes.Length];
var session = tester.CreateDivisorSession();
try
{
ulong[] primes = { 31UL, 37UL, 41UL, 43UL, 47UL };
byte[] hits = new byte[primes.Length];
ulong cycle223 = MersenneDivisorCycles.CalculateCycleLength(223UL, MontgomeryDivisorData.FromModulus(223UL));
session.CheckDivisor(223UL, MontgomeryDivisorData.FromModulus(223UL), cycle223, primes, hits);
ulong cycle223 = MersenneDivisorCycles.CalculateCycleLength(223UL, MontgomeryDivisorData.FromModulus(223UL));
session.CheckDivisor(223UL, MontgomeryDivisorData.FromModulus(223UL), cycle223, primes, hits);
hits.Should().ContainInOrder(new byte[] { 0, 1, 0, 0, 0 });
hits.Should().ContainInOrder(new byte[] { 0, 1, 0, 0, 0 });
}
finally
{
session.Dispose();
}
}
}
PerfectNumbers.Core/Cpu/MersenneCpuDivisorScanSession.cs
-9
View File
@@ -2,11 +2,8 @@ namespace PerfectNumbers.Core.Cpu;
internal sealed class MersenneCpuDivisorScanSession : IMersenneNumberDivisorByDivisorTester.IDivisorScanSession
{
private bool _disposed;
public void Reset()
{
_disposed = false;
}
public void CheckDivisor(
@@ -67,12 +64,6 @@ internal sealed class MersenneCpuDivisorScanSession : IMersenneNumberDivisorByDi
public void Dispose()
{
if (_disposed)
{
throw new ObjectDisposedException(nameof(MersenneCpuDivisorScanSession), "Divisor scan session disposed twice.");
}
_disposed = true;
ThreadStaticPools.ReturnMersenneCpuDivisorSession(this);
}
}
PerfectNumbers.Core/Cpu/MersenneNumberDivisorByDivisorCpuTester.cs
+3 -28
View File
@@ -145,14 +145,12 @@ public sealed class MersenneNumberDivisorByDivisorCpuTester : IMersenneNumberDiv
byte step10;
byte step8;
byte step5;
byte step3;
byte step7;
byte step11;
byte remainder10;
byte remainder8;
byte remainder5;
byte remainder3;
byte remainder7;
byte remainder11;
@@ -161,14 +159,12 @@ public sealed class MersenneNumberDivisorByDivisorCpuTester : IMersenneNumberDiv
{
step10 = (byte)(stepLow % 10UL);
step8 = (byte)(stepLow & 7UL);
step5 = (byte)(stepLow % 5UL);
step3 = (byte)(stepLow % 3UL);
step7 = (byte)(stepLow % 7UL);
step11 = (byte)(stepLow % 11UL);
remainder10 = (byte)(divisorLow % 10UL);
remainder8 = (byte)(divisorLow & 7UL);
remainder5 = (byte)(divisorLow % 5UL);
remainder3 = (byte)(divisorLow % 3UL);
remainder7 = (byte)(divisorLow % 7UL);
remainder11 = (byte)(divisorLow % 11UL);
@@ -181,13 +177,11 @@ public sealed class MersenneNumberDivisorByDivisorCpuTester : IMersenneNumberDiv
decimalMask,
step10,
step8,
step5,
step3,
step7,
step11,
remainder10,
remainder8,
remainder5,
remainder3,
remainder7,
remainder11,
@@ -197,21 +191,19 @@ public sealed class MersenneNumberDivisorByDivisorCpuTester : IMersenneNumberDiv
ulong divisorHigh = divisor.High;
step10 = (byte)((((stepHigh % 10UL) * 6UL) + (stepLow % 10UL)) % 10UL);
step8 = (byte)(stepLow % 8UL);
step5 = (byte)(((stepHigh % 5UL) + (stepLow % 5UL)) % 5UL);
step3 = (byte)(((stepHigh % 3UL) + (stepLow % 3UL)) % 3UL);
step7 = (byte)((((stepHigh % 7UL) * 2UL) + (stepLow % 7UL)) % 7UL);
step11 = (byte)((((stepHigh % 11UL) * 5UL) + (stepLow % 11UL)) % 11UL);
remainder10 = (byte)((((divisorHigh % 10UL) * 6UL) + (divisorLow % 10UL)) % 10UL);
remainder8 = (byte)(divisorLow % 8UL);
remainder5 = (byte)(((divisorHigh % 5UL) + (divisorLow % 5UL)) % 5UL);
remainder3 = (byte)(((divisorHigh % 3UL) + (divisorLow % 3UL)) % 3UL);
remainder7 = (byte)((((divisorHigh % 7UL) * 2UL) + (divisorLow % 7UL)) % 7UL);
remainder11 = (byte)((((divisorHigh % 11UL) * 5UL) + (divisorLow % 11UL)) % 11UL);
while (divisor.CompareTo(limit) <= 0)
{
bool passesSmallModuli = remainder3 != 0 && remainder5 != 0 && remainder7 != 0 && remainder11 != 0;
bool passesSmallModuli = remainder3 != 0 && remainder7 != 0 && remainder11 != 0;
if (passesSmallModuli && (remainder8 == 1 || remainder8 == 7) && ((decimalMask >> remainder10) & 1) != 0)
{
ulong candidate = divisor.Low;
@@ -261,7 +253,6 @@ public sealed class MersenneNumberDivisorByDivisorCpuTester : IMersenneNumberDiv
divisor.Add(step);
remainder10 = AddMod10(remainder10, step10);
remainder8 = AddMod8(remainder8, step8);
remainder5 = AddMod5(remainder5, step5);
remainder3 = AddMod3(remainder3, step3);
remainder7 = AddMod7(remainder7, step7);
remainder11 = AddMod11(remainder11, step11);
@@ -278,13 +269,11 @@ public sealed class MersenneNumberDivisorByDivisorCpuTester : IMersenneNumberDiv
ushort decimalMask,
byte step10,
byte step8,
byte step5,
byte step3,
byte step7,
byte step11,
byte remainder10,
byte remainder8,
byte remainder5,
byte remainder3,
byte remainder7,
byte remainder11,
@@ -297,7 +286,7 @@ public sealed class MersenneNumberDivisorByDivisorCpuTester : IMersenneNumberDiv
{
if (divisor <= limit)
{
if (remainder3 != 0 && remainder5 != 0 && remainder7 != 0 && remainder11 != 0 && (remainder8 == 1 || remainder8 == 7) && ((decimalMask >> remainder10) & 1) != 0)
if (remainder3 != 0 && remainder7 != 0 && remainder11 != 0 && (remainder8 == 1 || remainder8 == 7) && ((decimalMask >> remainder10) & 1) != 0)
{
MontgomeryDivisorData divisorData = MontgomeryDivisorData.FromModulus(divisor);
ulong divisorCycle;
@@ -345,7 +334,7 @@ public sealed class MersenneNumberDivisorByDivisorCpuTester : IMersenneNumberDiv
while (true)
{
bool passesSmallModuli = remainder3 != 0 && remainder5 != 0 && remainder7 != 0 && remainder11 != 0;
bool passesSmallModuli = remainder3 != 0 && remainder7 != 0 && remainder11 != 0;
if (passesSmallModuli && (remainder8 == 1 || remainder8 == 7) && ((decimalMask >> remainder10) & 1) != 0)
{
MontgomeryDivisorData divisorData = MontgomeryDivisorData.FromModulus(divisor);
@@ -389,7 +378,6 @@ public sealed class MersenneNumberDivisorByDivisorCpuTester : IMersenneNumberDiv
divisor += step;
remainder10 = AddMod10(remainder10, step10);
remainder8 = AddMod8(remainder8, step8);
remainder5 = AddMod5(remainder5, step5);
remainder3 = AddMod3(remainder3, step3);
remainder7 = AddMod7(remainder7, step7);
remainder11 = AddMod11(remainder11, step11);
@@ -416,19 +404,6 @@ public sealed class MersenneNumberDivisorByDivisorCpuTester : IMersenneNumberDiv
return (byte)sum;
}
[MethodImpl(MethodImplOptions.AggressiveInlining)]
private static byte AddMod5(byte value, byte delta)
{
const int Modulus = 5;
int sum = value + delta;
if (sum >= Modulus)
{
sum -= Modulus;
}
return (byte)sum;
}
[MethodImpl(MethodImplOptions.AggressiveInlining)]
private static byte AddMod7(byte value, byte delta)
PerfectNumbers.Core/Gpu/GpuContextPool.cs
-13
View File
@@ -14,8 +14,6 @@ public static class GpuContextPool
internal sealed class PooledContext
{
private bool disposed;
public Context Context { get; }
public Accelerator Accelerator { get; }
public bool IsCpu { get; }
@@ -34,10 +32,6 @@ public static class GpuContextPool
public void Dispose()
{
if (disposed)
{
return;
}
// Ensure all cached GPU buffers for this accelerator are released.
NttGpuMath.ClearCaches(Accelerator);
@@ -48,7 +42,6 @@ public static class GpuContextPool
Accelerator.Dispose();
Context.Dispose();
disposed = true;
}
}
@@ -131,7 +124,6 @@ public static class GpuContextPool
public struct GpuContextLease
{
private readonly PooledContext _ctx;
private bool disposed;
internal GpuContextLease(PooledContext ctx, object kernelInitLock)
{
@@ -150,13 +142,8 @@ public static class GpuContextPool
public void Dispose()
{
if (disposed)
{
return;
}
Return(_ctx);
disposed = true;
}
}
}
PerfectNumbers.Core/Gpu/GpuKernelLease.cs
+17 -32
View File
@@ -1,3 +1,4 @@
using System;
using System.Collections.Concurrent;
using ILGPU;
using ILGPU.Runtime;
@@ -9,30 +10,28 @@ public sealed class GpuKernelLease
{
private static readonly ConcurrentQueue<GpuKernelLease> Pool = new();
private IDisposable? _limiter;
private Action? _releaseLimiter;
private GpuContextLease? _gpu;
private KernelContainer? _kernels;
private AcceleratorStream? _stream;
private object? _executionLock;
private bool _disposed;
private GpuKernelLease()
{
}
internal static GpuKernelLease Rent(IDisposable limiter, GpuContextLease gpu, KernelContainer kernels)
internal static GpuKernelLease Rent(GpuPrimeWorkLimiter.Lease limiter, GpuContextLease gpu, KernelContainer kernels)
{
if (!Pool.TryDequeue(out var lease))
{
lease = new GpuKernelLease();
}
lease._limiter = limiter;
lease._releaseLimiter = limiter.Dispose;
lease._gpu = gpu;
lease._kernels = kernels;
lease._stream = null;
lease._executionLock = gpu.ExecutionLock;
lease._disposed = false;
return lease;
}
@@ -158,35 +157,21 @@ public sealed class GpuKernelLease
}
}
private void Dispose(bool disposing)
{
if (_disposed)
{
return;
}
if (disposing)
{
_stream?.Dispose();
_stream = null;
_gpu?.Dispose();
_gpu = null;
_executionLock = null;
_limiter?.Dispose();
_limiter = null;
}
_kernels = null;
_disposed = true;
Pool.Enqueue(this);
}
public void Dispose()
{
Dispose(disposing: false);
_stream?.Dispose();
_stream = null;
_gpu?.Dispose();
_gpu = null;
_executionLock = null;
_releaseLimiter?.Invoke();
_releaseLimiter = null;
_kernels = null;
Pool.Enqueue(this);
}
public readonly struct ExecutionScope
PerfectNumbers.Core/Gpu/GpuPrimeWorkLimiter.cs
+5 -12
View File
@@ -6,13 +6,13 @@ public static class GpuPrimeWorkLimiter
private static readonly object _lock = new();
private static int _currentLimit = 1;
public static IDisposable Acquire()
public static Lease Acquire()
{
// TODO: Inline the pooled limiter guard from the GpuLimiterThroughputBenchmarks so prime-sieve
// GPU jobs stop allocating Releaser instances on every acquisition.
// GPU jobs stop allocating Lease instances on every acquisition.
var sem = _semaphore;
sem.Wait();
return new Releaser(sem);
return new Lease(sem);
}
public static void SetLimit(int value)
@@ -36,25 +36,18 @@ public static class GpuPrimeWorkLimiter
}
}
private sealed class Releaser : IDisposable
public sealed class Lease
{
private readonly SemaphoreSlim _sem;
private bool _disposed;
public Releaser(SemaphoreSlim sem)
internal Lease(SemaphoreSlim sem)
{
_sem = sem;
}
public void Dispose()
{
if (_disposed)
{
return;
}
_sem.Release();
_disposed = true;
}
}
}
PerfectNumbers.Core/Gpu/GpuWorkLimiter.cs
+6 -13
View File
@@ -6,14 +6,14 @@ public static class GpuWorkLimiter
private static readonly object _lock = new();
private static int _currentLimit = 1;
public static IDisposable Acquire()
public static Lease Acquire()
{
// TODO: Replace the per-call Releaser allocation with the pooled struct-based guard from the
// TODO: Replace the per-call Lease allocation with the pooled struct-based guard from the
// GpuLimiterThroughputBenchmarks so limiter acquisition does not allocate when we enter the
// GPU scanning hot path.
var sem = _semaphore;
sem.Wait();
return new Releaser(sem);
return new Lease(sem);
}
public static void SetLimit(int value)
@@ -32,7 +32,7 @@ public static class GpuWorkLimiter
// Swap to a new semaphore for the new limit. We intentionally
// do not dispose the old semaphore to avoid releasing handles
// that may still be in use by existing Releaser instances.
// that may still be in use by existing Lease instances.
// TODO: Consolidate with GpuPrimeWorkLimiter so both limiters share a pooled SemaphoreSlim and avoid rebuilding the
// limiter state whenever limits are adjusted from the CLI.
_semaphore = new SemaphoreSlim(value, value);
@@ -40,25 +40,18 @@ public static class GpuWorkLimiter
}
}
private sealed class Releaser : IDisposable
public sealed class Lease
{
private readonly SemaphoreSlim _sem;
private bool _disposed;
public Releaser(SemaphoreSlim sem)
internal Lease(SemaphoreSlim sem)
{
_sem = sem;
}
public void Dispose()
{
if (_disposed)
{
return;
}
_sem.Release();
_disposed = true;
}
}
}
PerfectNumbers.Core/Gpu/MersenneNumberDivisorByDivisorGpuTester.cs
+7 -47
View File
@@ -220,7 +220,6 @@ public sealed class MersenneNumberDivisorByDivisorGpuTester : IMersenneNumberDiv
ulong stepValue = (ulong)twoP128;
byte step10 = (byte)(stepValue % 10UL);
byte step8 = (byte)(stepValue & 7UL);
byte step5 = (byte)(stepValue % 5UL);
byte step3 = (byte)(stepValue % 3UL);
byte step7 = (byte)(stepValue % 7UL);
byte step11 = (byte)(stepValue % 11UL);
@@ -229,7 +228,6 @@ public sealed class MersenneNumberDivisorByDivisorGpuTester : IMersenneNumberDiv
ulong currentDivisorValue = (ulong)currentDivisor128;
byte remainder10 = (byte)(currentDivisorValue % 10UL);
byte remainder8 = (byte)(currentDivisorValue & 7UL);
byte remainder5 = (byte)(currentDivisorValue % 5UL);
byte remainder3 = (byte)(currentDivisorValue % 3UL);
byte remainder7 = (byte)(currentDivisorValue % 7UL);
byte remainder11 = (byte)(currentDivisorValue % 11UL);
@@ -275,14 +273,13 @@ public sealed class MersenneNumberDivisorByDivisorGpuTester : IMersenneNumberDiv
byte localRemainder10 = remainder10;
byte localRemainder8 = remainder8;
byte localRemainder5 = remainder5;
byte localRemainder3 = remainder3;
byte localRemainder7 = remainder7;
byte localRemainder11 = remainder11;
for (int i = 0; i < chunkCount; i++)
{
bool passesSmallModuli = localRemainder3 != 0 && localRemainder5 != 0 && localRemainder7 != 0 && localRemainder11 != 0;
bool passesSmallModuli = localRemainder3 != 0 && localRemainder7 != 0 && localRemainder11 != 0;
if (passesSmallModuli && (localRemainder8 == 1 || localRemainder8 == 7) && ((decimalMask >> localRemainder10) & 1) != 0)
{
ulong divisorValue = (ulong)nextDivisor128;
@@ -290,21 +287,16 @@ public sealed class MersenneNumberDivisorByDivisorGpuTester : IMersenneNumberDiv
ulong divisorCycle = ResolveDivisorCycle(divisorValue, prime, in montgomeryData);
if (divisorCycle == prime)
{
int targetIndex = admissibleCount;
divisorSpan[targetIndex] = divisorValue;
exponentSpan[targetIndex] = prime;
divisorDataSpan[targetIndex] = new GpuDivisorPartialData(divisorValue);
offsetSpan[targetIndex] = targetIndex;
countSpan[targetIndex] = 1;
cycleSpan[targetIndex] = divisorCycle;
admissibleCount++;
processedCount += (ulong)(i + 1);
lastProcessed = divisorValue;
coveredRange = true;
return true;
}
}
nextDivisor128 += twoP128;
localRemainder10 = AddMod10(localRemainder10, step10);
localRemainder8 = AddMod8(localRemainder8, step8);
localRemainder5 = AddMod5(localRemainder5, step5);
localRemainder3 = AddMod3(localRemainder3, step3);
localRemainder7 = AddMod7(localRemainder7, step7);
localRemainder11 = AddMod11(localRemainder11, step11);
@@ -312,7 +304,6 @@ public sealed class MersenneNumberDivisorByDivisorGpuTester : IMersenneNumberDiv
remainder10 = localRemainder10;
remainder8 = localRemainder8;
remainder5 = localRemainder5;
remainder3 = localRemainder3;
remainder7 = localRemainder7;
remainder11 = localRemainder11;
@@ -380,19 +371,6 @@ public sealed class MersenneNumberDivisorByDivisorGpuTester : IMersenneNumberDiv
return (byte)sum;
}
[MethodImpl(MethodImplOptions.AggressiveInlining)]
private static byte AddMod5(byte value, byte delta)
{
const int Modulus = 5;
int sum = value + delta;
if (sum >= Modulus)
{
sum -= Modulus;
}
return (byte)sum;
}
[MethodImpl(MethodImplOptions.AggressiveInlining)]
private static byte AddMod7(byte value, byte delta)
@@ -490,7 +468,6 @@ public sealed class MersenneNumberDivisorByDivisorGpuTester : IMersenneNumberDiv
private MemoryBuffer1D<byte, Stride1D.Dense> _hitBuffer = null!;
private ulong[] _hostBuffer = null!;
private int _capacity;
private bool _disposed;
internal DivisorScanSession(MersenneNumberDivisorByDivisorGpuTester owner)
{
@@ -502,7 +479,6 @@ public sealed class MersenneNumberDivisorByDivisorGpuTester : IMersenneNumberDiv
internal void Reset()
{
_disposed = false;
}
private void EnsureExecutionResourcesLocked(int requiredCapacity)
@@ -549,11 +525,6 @@ public sealed class MersenneNumberDivisorByDivisorGpuTester : IMersenneNumberDiv
public void CheckDivisor(ulong divisor, in MontgomeryDivisorData divisorData, ulong divisorCycle, in ReadOnlySpan<ulong> primes, Span<byte> hits)
{
if (_disposed)
{
throw new ObjectDisposedException(nameof(DivisorScanSession));
}
int length = primes.Length;
// EvenPerfectBitScanner always supplies at least one exponent per divisor check, so the guard stays commented out.
// if (length == 0)
@@ -622,12 +593,7 @@ public sealed class MersenneNumberDivisorByDivisorGpuTester : IMersenneNumberDiv
int sentinel = int.MaxValue;
firstHitView.CopyFromCPU(ref sentinel, 1);
var kernel = _kernel;
if (kernel is null)
{
Monitor.Exit(_lease.ExecutionLock);
throw new InvalidOperationException("GPU divisor kernel was not initialized.");
}
var kernel = _kernel!;
kernel(1, divisorView, offsetView, countView, exponentView, cycleView, hitView, firstHitView);
@@ -640,12 +606,6 @@ public sealed class MersenneNumberDivisorByDivisorGpuTester : IMersenneNumberDiv
public void Dispose()
{
if (_disposed)
{
return;
}
_disposed = true;
_owner._sessionPool.Add(this);
}
}
@@ -695,7 +655,7 @@ public sealed class MersenneNumberDivisorByDivisorGpuTester : IMersenneNumberDiv
private void ReturnBatchResources(BatchResources resources) => _resourcePool.Add(resources);
private sealed class BatchResources : IDisposable
private sealed class BatchResources
{
private Accelerator? _accelerator;
private MemoryBuffer1D<GpuDivisorPartialData, Stride1D.Dense>? _divisorDataBuffer;
PerfectNumbers.Core/Gpu/MersenneNumberIncrementalGpuTester.cs
+74 -79
View File
@@ -47,88 +47,83 @@ public class MersenneNumberIncrementalGpuTester(GpuKernelType kernelType, bool u
int currentSize;
int i;
UInt128 q = UInt128.Zero;
try
while (kStart <= maxK && Volatile.Read(ref isPrime))
{
while (kStart <= maxK && Volatile.Read(ref isPrime))
remaining = maxK - kStart + UInt128.One;
currentSize = remaining > (UInt128)batchSize ? batchSize : (int)remaining;
Span<ulong> orders = orderArray.AsSpan(0, currentSize);
// Precompute residue automaton bases for this batch
UInt128 q0 = twoP * kStart + UInt128.One;
if (_kernelType == GpuKernelType.Pow2Mod)
{
remaining = maxK - kStart + UInt128.One;
currentSize = remaining > (UInt128)batchSize ? batchSize : (int)remaining;
Span<ulong> orders = orderArray.AsSpan(0, currentSize);
// Precompute residue automaton bases for this batch
UInt128 q0 = twoP * kStart + UInt128.One;
if (_kernelType == GpuKernelType.Pow2Mod)
{
q0.Mod10_8_5_3(out ulong q0m10, out ulong q0m8, out ulong q0m5, out ulong q0m3);
var kernelArgs = new ResidueAutomatonArgs(q0m10, step10, q0m8, step8, q0m3, step3, q0m5, step5);
pow2Kernel(
stream,
currentSize,
exponent,
twoPGpu,
(GpuUInt128)kStart,
last,
divMul,
kernelArgs,
orderBuffer.View,
smallCyclesView,
primeViews.LastOne,
primeViews.LastSeven,
primeViews.LastOnePow2,
primeViews.LastSevenPow2);
}
else
{
q0.Mod10_8_5_3(out ulong q0m10, out ulong q0m8, out ulong q0m5, out ulong q0m3);
incKernel(
stream,
currentSize,
exponent,
twoPGpu,
(GpuUInt128)kStart,
last,
divMul,
q0m10,
q0m8,
q0m3,
q0m5,
orderBuffer.View,
smallCyclesView);
}
stream.Synchronize();
orderBuffer.View.CopyToCPU(ref MemoryMarshal.GetReference(orders), currentSize);
if (_kernelType == GpuKernelType.Pow2Mod)
{
for (i = 0; i < currentSize; i++)
{
if (orders[i] != 0UL)
{
Volatile.Write(ref isPrime, false);
break;
}
}
}
else
{
q = twoP * kStart + UInt128.One;
for (i = 0; i < currentSize && Volatile.Read(ref isPrime); i++, q += twoP)
{
if (orders[i] != 0UL && q.IsPrimeCandidate())
{
Volatile.Write(ref isPrime, false);
}
}
}
kStart += (UInt128)currentSize;
q0.Mod10_8_5_3(out ulong q0m10, out ulong q0m8, out ulong q0m5, out ulong q0m3);
var kernelArgs = new ResidueAutomatonArgs(q0m10, step10, q0m8, step8, q0m3, step3, q0m5, step5);
pow2Kernel(
stream,
currentSize,
exponent,
twoPGpu,
(GpuUInt128)kStart,
last,
divMul,
kernelArgs,
orderBuffer.View,
smallCyclesView,
primeViews.LastOne,
primeViews.LastSeven,
primeViews.LastOnePow2,
primeViews.LastSevenPow2);
}
else
{
q0.Mod10_8_5_3(out ulong q0m10, out ulong q0m8, out ulong q0m5, out ulong q0m3);
incKernel(
stream,
currentSize,
exponent,
twoPGpu,
(GpuUInt128)kStart,
last,
divMul,
q0m10,
q0m8,
q0m3,
q0m5,
orderBuffer.View,
smallCyclesView);
}
stream.Synchronize();
orderBuffer.View.CopyToCPU(ref MemoryMarshal.GetReference(orders), currentSize);
if (_kernelType == GpuKernelType.Pow2Mod)
{
for (i = 0; i < currentSize; i++)
{
if (orders[i] != 0UL)
{
Volatile.Write(ref isPrime, false);
break;
}
}
}
else
{
q = twoP * kStart + UInt128.One;
for (i = 0; i < currentSize && Volatile.Read(ref isPrime); i++, q += twoP)
{
if (orders[i] != 0UL && q.IsPrimeCandidate())
{
Volatile.Write(ref isPrime, false);
}
}
}
kStart += (UInt128)currentSize;
}
finally
{
ArrayPool<ulong>.Shared.Return(orderArray);
orderBuffer.Dispose();
execution.Dispose();
gpuLease.Dispose();
}
ArrayPool<ulong>.Shared.Return(orderArray);
orderBuffer.Dispose();
execution.Dispose();
gpuLease.Dispose();
}
}
PerfectNumbers.Core/Gpu/MersenneNumberLucasLehmerGpuTester.cs
+4 -10
View File
@@ -152,18 +152,12 @@ public class MersenneNumberLucasLehmerGpuTester
}
GpuUInt128[] buffer = ArrayPool<GpuUInt128>.Shared.Rent(count);
try
ComputeResidues(exponents, buffer.AsSpan(0, count));
for (int i = 0; i < count; i++)
{
ComputeResidues(exponents, buffer.AsSpan(0, count));
for (int i = 0; i < count; i++)
{
results[i] = buffer[i].IsZero;
}
}
finally
{
ArrayPool<GpuUInt128>.Shared.Return(buffer);
results[i] = buffer[i].IsZero;
}
ArrayPool<GpuUInt128>.Shared.Return(buffer);
}
private bool IsMersennePrimeNtt(ulong exponent, GpuUInt128 nttMod, GpuUInt128 primitiveRoot, bool runOnGpu)
PerfectNumbers.Core/Gpu/MersenneNumberOrderGpuTester.cs
+45 -51
View File
@@ -4,19 +4,19 @@ namespace PerfectNumbers.Core.Gpu;
public class MersenneNumberOrderGpuTester(GpuKernelType kernelType, bool useGpuOrder)
{
private readonly GpuKernelType _kernelType = kernelType;
private readonly bool _useGpuOrder = useGpuOrder;
private readonly GpuKernelType _kernelType = kernelType;
private readonly bool _useGpuOrder = useGpuOrder;
public void Scan(ulong exponent, UInt128 twoP, LastDigit lastDigit, UInt128 maxK, ref bool isPrime)
{
var gpuLease = GpuKernelPool.GetKernel(_useGpuOrder);
var execution = gpuLease.EnterExecutionScope();
var accelerator = gpuLease.Accelerator;
var stream = gpuLease.Stream;
int batchSize = GpuConstants.ScanBatchSize; // large batch improves GPU occupancy and avoids TDR
UInt128 kStart = UInt128.One;
ulong divMul = (ulong)((((UInt128)1 << 64) - UInt128.One) / exponent) + 1UL;
byte last = lastDigit == LastDigit.Seven ? (byte)1 : (byte)0; // ILGPU kernels do not support bool parameters
public void Scan(ulong exponent, UInt128 twoP, LastDigit lastDigit, UInt128 maxK, ref bool isPrime)
{
var gpuLease = GpuKernelPool.GetKernel(_useGpuOrder);
var execution = gpuLease.EnterExecutionScope();
var accelerator = gpuLease.Accelerator;
var stream = gpuLease.Stream;
int batchSize = GpuConstants.ScanBatchSize; // large batch improves GPU occupancy and avoids TDR
UInt128 kStart = UInt128.One;
ulong divMul = (ulong)((((UInt128)1 << 64) - UInt128.One) / exponent) + 1UL;
byte last = lastDigit == LastDigit.Seven ? (byte)1 : (byte)0; // ILGPU kernels do not support bool parameters
var kernel = _kernelType switch
{
@@ -26,48 +26,42 @@ public class MersenneNumberOrderGpuTester(GpuKernelType kernelType, bool useGpuO
var foundBuffer = accelerator.Allocate1D<int>(1); // TODO: Replace this allocation with the pooled device buffer from GpuOrderKernelBenchmarks so repeated scans reuse the pinned staging memory instead of allocating per run.
exponent.Mod10_8_5_3Steps(out ulong step10, out ulong step8, out ulong step5, out ulong step3);
try
UInt128 remaining;
int currentSize;
int found = 0;
while (kStart <= maxK && Volatile.Read(ref isPrime))
{
UInt128 remaining;
int currentSize;
int found = 0;
while (kStart <= maxK && Volatile.Read(ref isPrime))
{
remaining = maxK - kStart + 1UL;
currentSize = remaining > (UInt128)batchSize ? batchSize : (int)remaining;
// This clear is required because the result may not always be set. Remove it after the kernels are modified to always set the result.
foundBuffer.MemSetToZero();
// Precompute residue automaton bases for this batch
UInt128 q0 = twoP * kStart + 1UL;
// TODO: Replace these modulo recomputations with the divisor-cycle stepping deltas captured in
// MersenneDivisorCycleLengthGpuBenchmarks so each batch advances via cached residues from the
// single snapshot block instead of recalculating mod 10/8/5/3 in every iteration.
q0.Mod10_8_5_3(out ulong q0m10, out ulong q0m8, out ulong q0m5, out ulong q0m3);
var kernelArgs = new ResidueAutomatonArgs(q0m10, step10, q0m8, step8, q0m3, step3, q0m5, step5);
remaining = maxK - kStart + 1UL;
currentSize = remaining > (UInt128)batchSize ? batchSize : (int)remaining;
// This clear is required because the result may not always be set. Remove it after the kernels are modified to always set the result.
foundBuffer.MemSetToZero();
// Precompute residue automaton bases for this batch
UInt128 q0 = twoP * kStart + 1UL;
// TODO: Replace these modulo recomputations with the divisor-cycle stepping deltas captured in
// MersenneDivisorCycleLengthGpuBenchmarks so each batch advances via cached residues from the
// single snapshot block instead of recalculating mod 10/8/5/3 in every iteration.
q0.Mod10_8_5_3(out ulong q0m10, out ulong q0m8, out ulong q0m5, out ulong q0m3);
var kernelArgs = new ResidueAutomatonArgs(q0m10, step10, q0m8, step8, q0m3, step3, q0m5, step5);
// Ensure device has small cycles table for early rejections in order kernels
var smallCyclesView = GpuKernelPool.EnsureSmallCyclesOnDevice(accelerator);
kernel(stream, currentSize, exponent, (GpuUInt128)twoP, (GpuUInt128)kStart, last, divMul,
kernelArgs, foundBuffer.View, smallCyclesView);
// Ensure device has small cycles table for early rejections in order kernels
var smallCyclesView = GpuKernelPool.EnsureSmallCyclesOnDevice(accelerator);
kernel(stream, currentSize, exponent, (GpuUInt128)twoP, (GpuUInt128)kStart, last, divMul,
kernelArgs, foundBuffer.View, smallCyclesView);
stream.Synchronize();
foundBuffer.View.CopyToCPU(ref found, 1);
if (found != 0)
{
Volatile.Write(ref isPrime, false);
break;
}
stream.Synchronize();
foundBuffer.View.CopyToCPU(ref found, 1);
if (found != 0)
{
Volatile.Write(ref isPrime, false);
break;
}
kStart += (UInt128)currentSize;
}
kStart += (UInt128)currentSize;
}
}
finally
{
foundBuffer.Dispose();
execution.Dispose();
gpuLease.Dispose();
}
}
foundBuffer.Dispose();
execution.Dispose();
gpuLease.Dispose();
}
}
PerfectNumbers.Core/Gpu/MersenneNumberResidueGpuTester.cs
+1 -1
View File
@@ -161,7 +161,7 @@ public class MersenneNumberResidueGpuTester(bool useGpuOrder)
private void ReturnResources(Accelerator accelerator, ResidueResources resources) =>
_resourcePools.GetOrAdd(accelerator, static _ => new ConcurrentBag<ResidueResources>()).Add(resources);
private sealed class ResidueResources : IDisposable
private sealed class ResidueResources
{
internal ResidueResources(Accelerator accelerator, int capacity)
{
PerfectNumbers.Core/Gpu/NttGpuMath.cs
+1 -1
View File
@@ -47,7 +47,7 @@ public static class NttGpuMath
private static readonly ConcurrentDictionary<Accelerator, Action<Index1D, ArrayView<GpuUInt128>, int>> BitReverseKernelCache = new(); // TODO: Precompute bit-reversal kernels during boot so we can remove the concurrent lookup overhead flagged in the benchmarks.
private sealed class SquareCacheEntry : IDisposable
private sealed class SquareCacheEntry
{
public MemoryBuffer1D<GpuUInt128, Stride1D.Dense> Buffer { get; }
public GpuUInt128 Root { get; }
PerfectNumbers.Core/IMersenneNumberDivisorByDivisorTester.cs
+3 -1
View File
@@ -16,8 +16,10 @@ public interface IMersenneNumberDivisorByDivisorTester
void PrepareCandidates(in ReadOnlySpan<ulong> primes, Span<ulong> allowedMaxValues);
public interface IDivisorScanSession : IDisposable
public interface IDivisorScanSession
{
void CheckDivisor(ulong divisor, in MontgomeryDivisorData divisorData, ulong divisorCycle, in ReadOnlySpan<ulong> primes, Span<byte> hits);
void Dispose();
}
}
PerfectNumbers.Core/MersenneDivisorCycles.cs
+20 -9
View File
@@ -137,20 +137,31 @@ public class MersenneDivisorCycles
{
// Binary pairs: (ulong divisor, ulong cycle)
using var reader = new BinaryReader(compressor, Encoding.UTF8, leaveOpen: true);
Stream stream = reader.BaseStream;
byte[] buffer = new byte[16];
while (true)
{
ulong d, c;
try
int read = 0;
Span<byte> span = buffer;
while (read < span.Length)
{
d = reader.ReadUInt64();
c = reader.ReadUInt64();
}
catch (EndOfStreamException)
{
yield break;
int chunk = stream.Read(span[read..]);
if (chunk == 0)
{
if (read == 0)
{
yield break;
}
throw new EndOfStreamException("Incomplete divisor-cycle entry.");
}
read += chunk;
}
yield return (d, c);
ulong divisor = BinaryPrimitives.ReadUInt64LittleEndian(span[..8]);
ulong cycle = BinaryPrimitives.ReadUInt64LittleEndian(span[8..]);
yield return (divisor, cycle);
}
}
PerfectNumbers.Core/MersenneNumberDivisorByDivisorTester.cs
+4 -8
View File
@@ -167,14 +167,10 @@ public static class MersenneNumberDivisorByDivisorTester
if (primesToTest.Count == 0)
{
if (applyStartPrime)
{
Console.WriteLine($"No primes greater than or equal to {startPrime.ToString(CultureInfo.InvariantCulture)} were found for --mersenne=bydivisor.");
}
else
{
Console.WriteLine("No prime candidates remain for --mersenne=bydivisor after filtering.");
}
string message = applyStartPrime
? $"No primes greater than or equal to {startPrime.ToString(CultureInfo.InvariantCulture)} were found for --mersenne=bydivisor."
: "No prime candidates remain for --mersenne=bydivisor after filtering.";
Console.WriteLine(message);
return;
}
PerfectNumbers.Core/MersenneNumberTester.cs
+41 -46
View File
@@ -152,56 +152,51 @@ public sealed class MersenneNumberTester(
var stream = gpuLease.Stream;
var orderKernel = gpuLease.OrderKernel!;
try
// Guard long-running kernels by chunking the warm-up across batches.
// Reuse the same ScanBatchSize knob used for scanning.
int batchSize = GpuConstants.ScanBatchSize;
ulong divMul = (ulong)((((UInt128)1 << 64) - 1UL) / exponent) + 1UL;
int offset = 0;
while (offset < idx)
{
// Guard long-running kernels by chunking the warm-up across batches.
// Reuse the same ScanBatchSize knob used for scanning.
int batchSize = GpuConstants.ScanBatchSize;
ulong divMul = (ulong)((((UInt128)1 << 64) - 1UL) / exponent) + 1UL;
int offset = 0;
while (offset < idx)
int count = Math.Min(batchSize, idx - offset);
var qBuffer = accelerator.Allocate1D<PerfectNumbers.Core.Gpu.GpuUInt128>(count);
var orderBuffer = accelerator.Allocate1D<ulong>(count);
// Convert to device-friendly GpuUInt128 on the fly
var tmp = System.Buffers.ArrayPool<PerfectNumbers.Core.Gpu.GpuUInt128>.Shared.Rent(count);
for (int i = 0; i < count; i++)
{
int count = Math.Min(batchSize, idx - offset);
var qBuffer = accelerator.Allocate1D<PerfectNumbers.Core.Gpu.GpuUInt128>(count);
var orderBuffer = accelerator.Allocate1D<ulong>(count);
// Convert to device-friendly GpuUInt128 on the fly
var tmp = System.Buffers.ArrayPool<PerfectNumbers.Core.Gpu.GpuUInt128>.Shared.Rent(count);
for (int i = 0; i < count; i++)
{
tmp[i] = (PerfectNumbers.Core.Gpu.GpuUInt128)qs[offset + i];
}
// Copy only the populated elements to the device buffer to avoid overrun on pooled arrays.
qBuffer.View.CopyFromCPU(ref tmp[0], count);
System.Buffers.ArrayPool<PerfectNumbers.Core.Gpu.GpuUInt128>.Shared.Return(tmp);
orderKernel(stream, count, exponent, divMul, qBuffer.View, orderBuffer.View);
stream.Synchronize();
ulong[] orders = orderBuffer.GetAsArray1D();
for (int i = 0; i < count; i++)
{
ulong order = orders[i];
if (order == 0UL)
{
order = qs[offset + i].CalculateOrder();
}
if (useOrderCache)
{
OrderCache[qs[offset + i]] = order;
}
}
offset += count;
orderBuffer.Dispose();
qBuffer.Dispose();
tmp[i] = (PerfectNumbers.Core.Gpu.GpuUInt128)qs[offset + i];
}
// Copy only the populated elements to the device buffer to avoid overrun on pooled arrays.
qBuffer.View.CopyFromCPU(ref tmp[0], count);
System.Buffers.ArrayPool<PerfectNumbers.Core.Gpu.GpuUInt128>.Shared.Return(tmp);
orderKernel(stream, count, exponent, divMul, qBuffer.View, orderBuffer.View);
stream.Synchronize();
ulong[] orders = orderBuffer.GetAsArray1D();
for (int i = 0; i < count; i++)
{
ulong order = orders[i];
if (order == 0UL)
{
order = qs[offset + i].CalculateOrder();
}
if (useOrderCache)
{
OrderCache[qs[offset + i]] = order;
}
}
offset += count;
orderBuffer.Dispose();
qBuffer.Dispose();
}
finally
{
execution.Dispose();
gpuLease.Dispose();
ArrayPool<UInt128>.Shared.Return(qs);
}
execution.Dispose();
gpuLease.Dispose();
ArrayPool<UInt128>.Shared.Return(qs);
}
public bool IsMersennePrime(ulong exponent)
PerfectNumbers.Core/MersennePrimeFactorTester.cs
+1 -1
View File
@@ -14,7 +14,7 @@ public static class MersennePrimeFactorTester
[MethodImpl(MethodImplOptions.AggressiveInlining)]
public static bool IsPrimeFactor(ulong p, ulong q, CancellationToken ct)
{
if (q < 2UL || (q & 1UL) == 0)
if (q < 2UL)
{
return false;
}
PerfectNumbers.Core/PrimeOrderCalculator.Gpu.cs
+50 -71
View File
@@ -9,86 +9,65 @@ namespace PerfectNumbers.Core;
internal static partial class PrimeOrderCalculator
{
private static bool IsGpuPow2Allowed => s_pow2ModeInitialized && s_allowGpuPow2;
private static bool IsGpuPow2Allowed => s_pow2ModeInitialized && s_allowGpuPow2;
private const int GpuSmallPrimeFactorSlots = 64;
private const int GpuSmallPrimeFactorSlots = 64;
private static bool TryPopulateSmallPrimeFactorsGpu(ulong value, uint limit, Dictionary<ulong, int> counts, out int factorCount, out ulong remaining)
{
var primeBufferArray = ThreadStaticPools.UlongPool.Rent(GpuSmallPrimeFactorSlots);
var exponentBufferArray = ThreadStaticPools.IntPool.Rent(GpuSmallPrimeFactorSlots);
Span<ulong> primeBuffer = primeBufferArray.AsSpan(0, GpuSmallPrimeFactorSlots);
Span<int> exponentBuffer = exponentBufferArray.AsSpan(0, GpuSmallPrimeFactorSlots);
// primeBuffer.Clear();
// exponentBuffer.Clear();
remaining = value;
private static bool TryPopulateSmallPrimeFactorsGpu(ulong value, uint limit, Dictionary<ulong, int> counts, out int factorCount, out ulong remaining)
{
var primeBufferArray = ThreadStaticPools.UlongPool.Rent(GpuSmallPrimeFactorSlots);
var exponentBufferArray = ThreadStaticPools.IntPool.Rent(GpuSmallPrimeFactorSlots);
Span<ulong> primeBuffer = primeBufferArray.AsSpan(0, GpuSmallPrimeFactorSlots);
Span<int> exponentBuffer = exponentBufferArray.AsSpan(0, GpuSmallPrimeFactorSlots);
remaining = value;
try
{
var lease = GpuKernelPool.GetKernel(useGpuOrder: true);
var execution = lease.EnterExecutionScope();
try
{
Accelerator accelerator = lease.Accelerator;
AcceleratorStream stream = lease.Stream;
var lease = GpuKernelPool.GetKernel(useGpuOrder: true);
var execution = lease.EnterExecutionScope();
Accelerator accelerator = lease.Accelerator;
AcceleratorStream stream = lease.Stream;
SmallPrimeFactorTables tables = GpuKernelPool.EnsureSmallPrimeFactorTables(accelerator);
SmallPrimeFactorScratch scratch = GpuKernelPool.EnsureSmallPrimeFactorScratch(accelerator, GpuSmallPrimeFactorSlots);
// scratch.Clear();
SmallPrimeFactorTables tables = GpuKernelPool.EnsureSmallPrimeFactorTables(accelerator);
SmallPrimeFactorScratch scratch = GpuKernelPool.EnsureSmallPrimeFactorScratch(accelerator, GpuSmallPrimeFactorSlots);
var kernel = lease.SmallPrimeFactorKernel;
kernel(
stream,
1,
value,
limit,
tables.PrimesView,
tables.SquaresView,
tables.Count,
scratch.PrimeSlots.View,
scratch.ExponentSlots.View,
scratch.CountSlot.View,
scratch.RemainingSlot.View);
var kernel = lease.SmallPrimeFactorKernel;
kernel(
stream,
1,
value,
limit,
tables.PrimesView,
tables.SquaresView,
tables.Count,
scratch.PrimeSlots.View,
scratch.ExponentSlots.View,
scratch.CountSlot.View,
scratch.RemainingSlot.View);
stream.Synchronize();
stream.Synchronize();
factorCount = 0;
scratch.CountSlot.View.CopyToCPU(ref factorCount, 1);
factorCount = Math.Min(factorCount, GpuSmallPrimeFactorSlots);
factorCount = 0;
scratch.CountSlot.View.CopyToCPU(ref factorCount, 1);
factorCount = Math.Min(factorCount, GpuSmallPrimeFactorSlots);
if (factorCount > 0)
{
scratch.PrimeSlots.View.CopyToCPU(ref MemoryMarshal.GetReference(primeBuffer), factorCount);
scratch.ExponentSlots.View.CopyToCPU(ref MemoryMarshal.GetReference(exponentBuffer), factorCount);
}
if (factorCount > 0)
{
scratch.PrimeSlots.View.CopyToCPU(ref MemoryMarshal.GetReference(primeBuffer), factorCount);
scratch.ExponentSlots.View.CopyToCPU(ref MemoryMarshal.GetReference(exponentBuffer), factorCount);
}
scratch.RemainingSlot.View.CopyToCPU(ref remaining, 1);
scratch.RemainingSlot.View.CopyToCPU(ref remaining, 1);
for (int i = 0; i < factorCount; i++)
{
ulong primeValue = primeBuffer[i];
int exponent = exponentBuffer[i];
// This will never happen in production code
// if (primeValue == 0UL || exponent == 0)
// {
// continue;
// }
for (int i = 0; i < factorCount; i++)
{
ulong primeValue = primeBuffer[i];
int exponent = exponentBuffer[i];
counts.Add(primeValue, exponent);
}
counts.Add(primeValue, exponent);
}
return true;
}
finally
{
execution.Dispose();
lease.Dispose();
}
}
finally
{
ThreadStaticPools.UlongPool.Return(primeBufferArray);
ThreadStaticPools.IntPool.Return(exponentBufferArray);
}
}
execution.Dispose();
lease.Dispose();
ThreadStaticPools.UlongPool.Return(primeBufferArray);
ThreadStaticPools.IntPool.Return(exponentBufferArray);
return true;
}
}
PerfectNumbers.Core/PrimeOrderCalculator.Shared.cs
+33 -39
View File
@@ -185,55 +185,49 @@ internal static partial class PrimeOrderCalculator
private static UInt128 ExponentLoweringWide(UInt128 order, in UInt128 prime, in MontgomeryDivisorData divisorData, in PartialFactorResult128 factors)
{
ArrayPool<FactorEntry128> pool = ThreadStaticPools.FactorEntry128Pool;
ReadOnlySpan<FactorEntry128> factorSpan = factors.Factors;
int length = factors.Count;
// TODO: This should never trigger from production code - check
// if (length == 0)
// {
// return order;
// }
FactorEntry128[]? tempArray = pool.Rent(length + 1);
try
ArrayPool<FactorEntry128> pool = ThreadStaticPools.FactorEntry128Pool;
ReadOnlySpan<FactorEntry128> factorSpan = factors.Factors;
int length = factors.Count;
// TODO: This should never trigger from production code - check
// if (length == 0)
// {
// return order;
// }
FactorEntry128[] tempArray = pool.Rent(length + 1);
Span<FactorEntry128> buffer = tempArray.AsSpan(0, length);
factorSpan.CopyTo(buffer);
if (!factors.FullyFactored && factors.Cofactor > UInt128.One && IsPrimeWide(factors.Cofactor))
{
buffer[length] = new FactorEntry128(factors.Cofactor, 1);
length++;
}
Span<FactorEntry128> buffer = tempArray.AsSpan(0, length);
factorSpan.CopyTo(buffer);
buffer.Slice(0, length).Sort(static (a, b) => a.Value.CompareTo(b.Value));
if (!factors.FullyFactored && factors.Cofactor > UInt128.One && IsPrimeWide(factors.Cofactor))
for (int i = 0; i < length; i++)
{
UInt128 primeFactor = buffer[i].Value;
int exponent = buffer[i].Exponent;
for (int iteration = 0; iteration < exponent; iteration++)
{
buffer[length] = new FactorEntry128(factors.Cofactor, 1);
length++;
}
buffer.Slice(0, length).Sort(static (a, b) => a.Value.CompareTo(b.Value));
for (int i = 0; i < length; i++)
{
UInt128 primeFactor = buffer[i].Value;
int exponent = buffer[i].Exponent;
for (int iteration = 0; iteration < exponent; iteration++)
if ((order % primeFactor) == UInt128.Zero)
{
if ((order % primeFactor) == UInt128.Zero)
UInt128 reduced = order / primeFactor;
if (Pow2ModWide(reduced, prime, divisorData) == UInt128.One)
{
UInt128 reduced = order / primeFactor;
if (Pow2ModWide(reduced, prime, divisorData) == UInt128.One)
{
order = reduced;
continue;
}
order = reduced;
continue;
}
break;
}
}
return order;
}
finally
{
pool.Return(tempArray, clearArray: false);
break;
}
}
pool.Return(tempArray, clearArray: false);
return order;
}
private static bool TryConfirmOrderWide(in UInt128 prime, in UInt128 order, in MontgomeryDivisorData divisorData, in PrimeOrderSearchConfig config)
PerfectNumbers.Core/PrimeTester.cs
+4 -15
View File
@@ -149,9 +149,8 @@ public sealed class PrimeTester
internal static class PrimeTesterGpuContextPool
{
internal sealed class PooledContext : IDisposable
internal sealed class PooledContext
{
private bool _disposed;
public Context Context { get; }
@@ -167,15 +166,9 @@ public sealed class PrimeTester
public void Dispose()
{
if (_disposed)
{
return;
}
PrimeTester.ClearGpuCaches(Accelerator);
Accelerator.Dispose();
Context.Dispose();
_disposed = true;
}
}
@@ -209,7 +202,7 @@ public sealed class PrimeTester
Pool.Enqueue(ctx);
}
internal struct PrimeTesterGpuContextLease : IDisposable
internal struct PrimeTesterGpuContextLease
{
private PooledContext? _ctx;
@@ -224,11 +217,7 @@ public sealed class PrimeTester
public void Dispose()
{
if (_ctx is { } ctx)
{
Return(ctx);
_ctx = null;
}
Return(_ctx ?? throw new InvalidOperationException("GPU context lease is not initialized."));
}
}
}
@@ -257,7 +246,7 @@ public sealed class PrimeTester
DevicePrimes.View.CopyFromCPU(primes);
}
internal sealed class ScratchBuffers : IDisposable
internal sealed class ScratchBuffers
{
public MemoryBuffer1D<ulong, Stride1D.Dense> Input { get; private set; }
public MemoryBuffer1D<byte, Stride1D.Dense> Output { get; private set; }
PerfectNumbers.Core/TaskThreadPool.cs
+8 -19
View File
@@ -6,7 +6,7 @@ using System.Threading.Tasks;
namespace PerfectNumbers.Core;
internal sealed class TaskThreadPool : IDisposable
internal sealed class TaskThreadPool
{
private readonly struct WorkItem
{
@@ -22,8 +22,6 @@ internal sealed class TaskThreadPool : IDisposable
private readonly BlockingCollection<WorkItem> _pendingTasks;
private readonly Thread[] _threads;
private readonly Action<Task> _taskExecutor;
private int _disposed;
public TaskThreadPool(int minimumThreads, Action<Task> taskExecutor)
{
if (minimumThreads <= 0)
@@ -57,22 +55,18 @@ internal sealed class TaskThreadPool : IDisposable
throw new ArgumentNullException(nameof(task));
}
if (Volatile.Read(ref _disposed) != 0)
{
throw new ObjectDisposedException(nameof(TaskThreadPool));
}
try
{
_pendingTasks.Add(new WorkItem(task));
}
catch (InvalidOperationException ex) when (Volatile.Read(ref _disposed) != 0)
catch (Exception ex)
{
throw new ObjectDisposedException(nameof(TaskThreadPool), ex);
}
catch (ObjectDisposedException ex)
{
throw new ObjectDisposedException(nameof(TaskThreadPool), ex);
if (ex is InvalidOperationException || ex is ObjectDisposedException)
{
throw new ObjectDisposedException(nameof(TaskThreadPool), ex);
}
throw;
}
}
@@ -97,11 +91,6 @@ internal sealed class TaskThreadPool : IDisposable
public void Dispose()
{
if (Interlocked.Exchange(ref _disposed, 1) != 0)
{
return;
}
_pendingTasks.CompleteAdding();
foreach (Thread thread in _threads)
PerfectNumbers.Core/TextFileWriter.cs
+1 -1
View File
@@ -6,7 +6,7 @@ namespace PerfectNumbers.Core;
/// <summary>
/// Provides thread-safe file writing for both text and binary data.
/// </summary>
public sealed class TextFileWriter : IDisposable
public sealed class TextFileWriter
{
private readonly FileStream _stream;
private StreamWriter? _textWriter;