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

Add GPU divisor tester

Browse Source
This commit is contained in:
Marek Lemańczyk committed 2025-09-10 22:03:17 +02:00
1 parent fda058cde0
commit 596bb7e38f
4 files changed
+238 -113
Download Patch File Download Diff File Expand all files Collapse all files

No files matched your search

EvenPerfectBitScanner.Tests/ResultsFileNameTests.cs
+2 -1
View File
@@ -19,7 +19,7 @@ public class ResultsFileNameTests
types: new[] types: new[]
{ {
typeof(bool), typeof(int), typeof(int), typeof(global::PerfectNumbers.Core.GpuKernelType), typeof(bool), typeof(int), typeof(int), typeof(global::PerfectNumbers.Core.GpuKernelType),
typeof(bool), typeof(bool), typeof(bool), typeof(bool), typeof(bool), typeof(NttBackend), typeof(int), typeof(int), typeof(int), typeof(ulong), typeof(ModReductionMode), typeof(string), typeof(string), typeof(string) typeof(bool), typeof(bool), typeof(bool), typeof(bool), typeof(bool), typeof(bool), typeof(NttBackend), typeof(int), typeof(int), typeof(int), typeof(ulong), typeof(ModReductionMode), typeof(string), typeof(string), typeof(string)
}, },
modifiers: null)!; modifiers: null)!;
@@ -30,6 +30,7 @@ public class ResultsFileNameTests
1, // block 1, // block
global::PerfectNumbers.Core.GpuKernelType.Incremental, global::PerfectNumbers.Core.GpuKernelType.Incremental,
false, // useLucasFlag false, // useLucasFlag
false, // useDivisorFlag
true, // mersenneOnGpu (unused inside name content, devices passed explicitly below) true, // mersenneOnGpu (unused inside name content, devices passed explicitly below)
false, // useOrder false, // useOrder
false, // useModWorkaround false, // useModWorkaround
EvenPerfectBitScanner/Program.cs
+159 -112
View File
@@ -25,7 +25,10 @@ internal static class Program
private static int _primeCount; private static int _primeCount;
private static bool _primeFoundAfterInit; private static bool _primeFoundAfterInit;
private static bool _useGcdFilter; private static bool _useGcdFilter;
private static bool _useResidue; private static bool _useResidue;
private static bool _useDivisor;
private static ulong _divisor;
private static MersenneNumberDivisorGpuTester? _divisorTester;
private static ulong? _orderWarmupLimitOverride; private static ulong? _orderWarmupLimitOverride;
private static unsafe delegate*<ulong, ref ulong, ulong> _transformP; private static unsafe delegate*<ulong, ref ulong, ulong> _transformP;
private static long _state; private static long _state;
@@ -56,8 +59,10 @@ internal static class Program
bool useOrder = false; bool useOrder = false;
bool showHelp = false; bool showHelp = false;
bool useBitTransform = false; bool useBitTransform = false;
bool useLucas = false; bool useLucas = false;
bool useResidue = false; // M_p test via residue divisors (replaces LL/incremental) bool useResidue = false; // M_p test via residue divisors (replaces LL/incremental)
bool useDivisor = false; // M_p divisibility by specific divisor
ulong divisor = 0UL;
// Device routing // Device routing
bool useGpuCycles = true; bool useGpuCycles = true;
bool mersenneOnGpu = true; // controls Lucas/incremental/pow2mod device bool mersenneOnGpu = true; // controls Lucas/incremental/pow2mod device
@@ -100,24 +105,35 @@ internal static class Program
{ {
int eq = arg.IndexOf('='); int eq = arg.IndexOf('=');
ReadOnlySpan<char> value = arg.AsSpan(eq + 1); ReadOnlySpan<char> value = arg.AsSpan(eq + 1);
if (value.Equals("pow2mod", StringComparison.OrdinalIgnoreCase)) if (value.Equals("pow2mod", StringComparison.OrdinalIgnoreCase))
{ {
kernelType = GpuKernelType.Pow2Mod; kernelType = GpuKernelType.Pow2Mod;
} }
else if (value.Equals("lucas", StringComparison.OrdinalIgnoreCase)) else if (value.Equals("lucas", StringComparison.OrdinalIgnoreCase))
{ {
useLucas = true; useLucas = true;
} }
else if (value.Equals("residue", StringComparison.OrdinalIgnoreCase)) else if (value.Equals("residue", StringComparison.OrdinalIgnoreCase))
{ {
useResidue = true; useResidue = true;
useLucas = false; useLucas = false;
} }
else else if (value.Equals("divisor", StringComparison.OrdinalIgnoreCase))
{ {
kernelType = GpuKernelType.Incremental; useDivisor = true;
} useLucas = false;
} useResidue = false;
}
else
{
kernelType = GpuKernelType.Incremental;
}
}
else if (arg.StartsWith("--divisor=", StringComparison.OrdinalIgnoreCase))
{
int eq = arg.IndexOf('=');
divisor = ulong.Parse(arg.AsSpan(eq + 1));
}
else if (arg.StartsWith("--residue-max-k=", StringComparison.OrdinalIgnoreCase)) else if (arg.StartsWith("--residue-max-k=", StringComparison.OrdinalIgnoreCase))
{ {
int eq = arg.IndexOf('='); int eq = arg.IndexOf('=');
@@ -335,11 +351,17 @@ internal static class Program
} }
} }
if (showHelp) if (showHelp)
{ {
PrintHelp(); PrintHelp();
return; return;
} }
if (useDivisor && divisor == 0UL)
{
Console.WriteLine("Divisor mode requires --divisor=<value>.");
return;
}
// Apply GPU prime sieve runtime configuration // Apply GPU prime sieve runtime configuration
GpuPrimeWorkLimiter.SetLimit(gpuPrimeThreads); GpuPrimeWorkLimiter.SetLimit(gpuPrimeThreads);
@@ -386,34 +408,43 @@ internal static class Program
MersenneDivisorCycles.Shared.LoadFrom(cyclesPath); MersenneDivisorCycles.Shared.LoadFrom(cyclesPath);
} }
Console.WriteLine("Divisor cycles are ready"); Console.WriteLine("Divisor cycles are ready");
_useResidue = useResidue; _useResidue = useResidue;
_transformP = useBitTransform ? &TransformPBit : &TransformPAdd; _useDivisor = useDivisor;
PResidue = new ThreadLocal<ModResidueTracker>(() => new ModResidueTracker(ResidueModel.Identity, initialNumber: currentP, initialized: true), trackAllValues: true); _divisor = divisor;
PrimeTesters = new ThreadLocal<PrimeTester>(() => new PrimeTester(), trackAllValues: true); _transformP = useBitTransform ? &TransformPBit : &TransformPAdd;
// Note: --primes-device controls default device for library kernels; p primality remains CPU here. PResidue = new ThreadLocal<ModResidueTracker>(() => new ModResidueTracker(ResidueModel.Identity, initialNumber: currentP, initialized: true), trackAllValues: true);
// Initialize per-thread p residue tracker (Identity model) at currentP PrimeTesters = new ThreadLocal<PrimeTester>(() => new PrimeTester(), trackAllValues: true);
MersenneTesters = new ThreadLocal<MersenneNumberTester>(() => // Note: --primes-device controls default device for library kernels; p primality remains CPU here.
{ // Initialize per-thread p residue tracker (Identity model) at currentP
var tester = new MersenneNumberTester( if (!useDivisor)
useIncremental: !useLucas, {
useOrderCache: false, MersenneTesters = new ThreadLocal<MersenneNumberTester>(() =>
kernelType: kernelType, {
useModuloWorkaround: useModuloWorkaround, var tester = new MersenneNumberTester(
useOrder: useOrder, useIncremental: !useLucas,
useGpuLucas: mersenneOnGpu, useOrderCache: false,
useGpuScan: mersenneOnGpu, kernelType: kernelType,
useGpuOrder: orderOnGpu, useModuloWorkaround: useModuloWorkaround,
useResidue: useResidue, useOrder: useOrder,
maxK: residueKMax); useGpuLucas: mersenneOnGpu,
if (!useLucas) useGpuScan: mersenneOnGpu,
{ useGpuOrder: orderOnGpu,
tester.WarmUpOrders(currentP, _orderWarmupLimitOverride ?? 5_000_000UL); useResidue: useResidue,
} maxK: residueKMax);
if (!useLucas)
{
tester.WarmUpOrders(currentP, _orderWarmupLimitOverride ?? 5_000_000UL);
}
return tester; return tester;
}, trackAllValues: true); }, trackAllValues: true);
}
else
{
_divisorTester = new MersenneNumberDivisorGpuTester();
}
// Load RLE blacklist (optional) // Load RLE blacklist (optional)
if (!string.IsNullOrEmpty(_rleBlacklistPath)) if (!string.IsNullOrEmpty(_rleBlacklistPath))
@@ -439,25 +470,26 @@ internal static class Program
Console.WriteLine("Initialization..."); Console.WriteLine("Initialization...");
// Compose a results file name that encodes configuration (before opening file) // Compose a results file name that encodes configuration (before opening file)
var builtName = BuildResultsFileName( var builtName = BuildResultsFileName(
useBitTransform, useBitTransform,
threadCount, threadCount,
blockSize, blockSize,
kernelType, kernelType,
useLucas, useLucas,
mersenneOnGpu, useDivisor,
useOrder, mersenneOnGpu,
useModuloWorkaround, useOrder,
_useGcdFilter, useModuloWorkaround,
NttGpuMath.GpuTransformBackend, _useGcdFilter,
gpuPrimeThreads, NttGpuMath.GpuTransformBackend,
sliceSize, gpuPrimeThreads,
scanBatchSize, sliceSize,
_orderWarmupLimitOverride ?? 5_000_000UL, scanBatchSize,
NttGpuMath.ReductionMode, _orderWarmupLimitOverride ?? 5_000_000UL,
mersenneOnGpu ? "gpu" : "cpu", NttGpuMath.ReductionMode,
(GpuContextPool.ForceCpu ? "cpu" : "gpu"), mersenneOnGpu ? "gpu" : "cpu",
orderOnGpu ? "gpu" : "cpu"); (GpuContextPool.ForceCpu ? "cpu" : "gpu"),
orderOnGpu ? "gpu" : "cpu");
if (!string.IsNullOrEmpty(_resultsPrefix)) if (!string.IsNullOrEmpty(_resultsPrefix))
{ {
@@ -517,34 +549,38 @@ internal static class Program
PerfectNumbers.Core.PrimeTester.GpuBatchSize = gpuPrimeBatch; PerfectNumbers.Core.PrimeTester.GpuBatchSize = gpuPrimeBatch;
// Compose a results file name that encodes configuration // Compose a results file name that encodes configuration
ResultsFileName = BuildResultsFileName( ResultsFileName = BuildResultsFileName(
useBitTransform, useBitTransform,
threadCount, threadCount,
blockSize, blockSize,
kernelType, kernelType,
useLucas, useLucas,
mersenneOnGpu, useDivisor,
useOrder, mersenneOnGpu,
useModuloWorkaround, useOrder,
_useGcdFilter, useModuloWorkaround,
NttGpuMath.GpuTransformBackend, _useGcdFilter,
gpuPrimeThreads, NttGpuMath.GpuTransformBackend,
sliceSize, gpuPrimeThreads,
scanBatchSize, sliceSize,
_orderWarmupLimitOverride ?? 5_000_000UL, scanBatchSize,
NttGpuMath.ReductionMode, _orderWarmupLimitOverride ?? 5_000_000UL,
mersenneOnGpu ? "gpu" : "cpu", NttGpuMath.ReductionMode,
GpuContextPool.ForceCpu ? "cpu" : "gpu", mersenneOnGpu ? "gpu" : "cpu",
orderOnGpu ? "gpu" : "cpu"); GpuContextPool.ForceCpu ? "cpu" : "gpu",
orderOnGpu ? "gpu" : "cpu");
Console.WriteLine("Warming up orders..."); if (!useDivisor)
{
Console.WriteLine("Warming up orders...");
threadCount = Math.Max(1, threadCount); threadCount = Math.Max(1, threadCount);
_ = MersenneTesters.Value; _ = MersenneTesters.Value;
if (!useLucas) if (!useLucas)
{ {
MersenneTesters.Value!.WarmUpOrders(currentP, _orderWarmupLimitOverride ?? 5_000_000UL); MersenneTesters.Value!.WarmUpOrders(currentP, _orderWarmupLimitOverride ?? 5_000_000UL);
} }
}
Console.WriteLine("Starting scan..."); Console.WriteLine("Starting scan...");
_state = ((long)currentP << 3) | (long)remainder; _state = ((long)currentP << 3) | (long)remainder;
@@ -653,11 +689,12 @@ internal static class Program
Console.WriteLine(" --increment=bit|add exponent increment method"); Console.WriteLine(" --increment=bit|add exponent increment method");
Console.WriteLine(" --threads=<value> number of worker threads"); Console.WriteLine(" --threads=<value> number of worker threads");
Console.WriteLine(" --block-size=<value> values processed per thread batch"); Console.WriteLine(" --block-size=<value> values processed per thread batch");
Console.WriteLine(" --mersenne=pow2mod|incremental|lucas|residue Mersenne test method"); Console.WriteLine(" --mersenne=pow2mod|incremental|lucas|residue|divisor Mersenne test method");
Console.WriteLine(" --residue-max-k=<value> max k for residue Mersenne test (q = 2*p*k + 1)"); Console.WriteLine(" --divisor=<value> divisor for --mersenne=divisor mode");
Console.WriteLine(" --mersenne-device=cpu|gpu Device for Mersenne method (default gpu)"); Console.WriteLine(" --residue-max-k=<value> max k for residue Mersenne test (q = 2*p*k + 1)");
Console.WriteLine(" --primes-device=cpu|gpu Device for prime-scan kernels (default gpu)"); Console.WriteLine(" --mersenne-device=cpu|gpu Device for Mersenne method (default gpu)");
Console.WriteLine(" --gpu-prime-batch=<n> Batch size for GPU primality sieve (default 262144)"); Console.WriteLine(" --primes-device=cpu|gpu Device for prime-scan kernels (default gpu)");
Console.WriteLine(" --gpu-prime-batch=<n> Batch size for GPU primality sieve (default 262144)");
Console.WriteLine(" --order-device=cpu|gpu Device for order computations (default gpu)"); Console.WriteLine(" --order-device=cpu|gpu Device for order computations (default gpu)");
Console.WriteLine(" --ntt=reference|staged GPU NTT backend (default staged)"); Console.WriteLine(" --ntt=reference|staged GPU NTT backend (default staged)");
Console.WriteLine(" --mod-reduction=auto|uint128|mont64|barrett128 staged NTT reduction (default auto)"); Console.WriteLine(" --mod-reduction=auto|uint128|mont64|barrett128 staged NTT reduction (default auto)");
@@ -670,8 +707,12 @@ internal static class Program
Console.WriteLine(" --rle-only-last7=true|false apply RLE only when p % 10 == 7 (default true)"); Console.WriteLine(" --rle-only-last7=true|false apply RLE only when p % 10 == 7 (default true)");
Console.WriteLine(" --zero-hard=<f> hard reject if zero_fraction(p) > f (default off)"); Console.WriteLine(" --zero-hard=<f> hard reject if zero_fraction(p) > f (default off)");
Console.WriteLine(" --zero-conj=<f>:<r> hard reject if zero_fraction(p) > f AND max_zero_block >= r (default off)"); Console.WriteLine(" --zero-conj=<f>:<r> hard reject if zero_fraction(p) > f AND max_zero_block >= r (default off)");
Console.WriteLine(" --results-dir=<path> directory for results file"); Console.WriteLine(" --results-dir=<path> directory for results file");
Console.WriteLine(" --results-prefix=<text> prefix to prepend to results filename"); Console.WriteLine(" --results-prefix=<text> prefix to prepend to results filename");
Console.WriteLine(" --divisor-cycles=<path> divisor cycles data file");
Console.WriteLine(" --divisor-cycles-device=cpu|gpu device for cycles generation (default gpu)");
Console.WriteLine(" --divisor-cycles-batch-size=<value> batch size for cycles generation (default 512)");
Console.WriteLine(" --divisor-cycles-continue continue divisor cycles generation");
Console.WriteLine(" --use-order test primality via q order"); Console.WriteLine(" --use-order test primality via q order");
Console.WriteLine(" --workaround-mod avoid '%' operator on the GPU"); Console.WriteLine(" --workaround-mod avoid '%' operator on the GPU");
@@ -682,10 +723,10 @@ internal static class Program
Console.WriteLine(" --help, -help, --?, -?, /? show this help message"); Console.WriteLine(" --help, -help, --?, -?, /? show this help message");
} }
private static string BuildResultsFileName(bool bitInc, int threads, int block, GpuKernelType kernelType, bool useLucasFlag, bool mersenneOnGpu, bool useOrder, bool useModWorkaround, bool useGcd, NttBackend nttBackend, int gpuPrimeThreads, int llSlice, int gpuScanBatch, ulong warmupLimit, ModReductionMode reduction, string mersenneDevice, string primesDevice, string orderDevice) private static string BuildResultsFileName(bool bitInc, int threads, int block, GpuKernelType kernelType, bool useLucasFlag, bool useDivisorFlag, bool mersenneOnGpu, bool useOrder, bool useModWorkaround, bool useGcd, NttBackend nttBackend, int gpuPrimeThreads, int llSlice, int gpuScanBatch, ulong warmupLimit, ModReductionMode reduction, string mersenneDevice, string primesDevice, string orderDevice)
{ {
string inc = bitInc ? "bit" : "add"; string inc = bitInc ? "bit" : "add";
string mers = useLucasFlag ? "lucas" : (kernelType == GpuKernelType.Pow2Mod ? "pow2mod" : "incremental"); string mers = useDivisorFlag ? "divisor" : (useLucasFlag ? "lucas" : (kernelType == GpuKernelType.Pow2Mod ? "pow2mod" : "incremental"));
string ntt = nttBackend == NttBackend.Staged ? "staged" : "reference"; string ntt = nttBackend == NttBackend.Staged ? "staged" : "reference";
string red = reduction switch { ModReductionMode.Mont64 => "mont64", ModReductionMode.Barrett128 => "barrett128", ModReductionMode.GpuUInt128 => "uint128", _ => "auto" }; string red = reduction switch { ModReductionMode.Mont64 => "mont64", ModReductionMode.Barrett128 => "barrett128", ModReductionMode.GpuUInt128 => "uint128", _ => "auto" };
string order = useOrder ? "order-on" : "order-off"; string order = useOrder ? "order-on" : "order-off";
@@ -928,10 +969,16 @@ internal static class Program
return false; return false;
} }
searchedMersenne = true; searchedMersenne = true;
detailedCheck = MersenneTesters.Value!.IsMersennePrime(p); if (_useDivisor)
return detailedCheck; {
} detailedCheck = _divisorTester!.IsDivisible(p, _divisor);
return detailedCheck;
}
detailedCheck = MersenneTesters.Value!.IsMersennePrime(p);
return detailedCheck;
}
// Use ModResidueTracker with a small set of primes to prefilter composite p. // Use ModResidueTracker with a small set of primes to prefilter composite p.
private static bool IsCompositeByResidues(ulong p) private static bool IsCompositeByResidues(ulong p)
PerfectNumbers.Core.Tests/MersenneNumberDivisorGpuTesterTests.cs
+20
View File
@@ -0,0 +1,20 @@
using FluentAssertions;
using PerfectNumbers.Core.Gpu;
using Xunit;
namespace PerfectNumbers.Core.Tests;
public class MersenneNumberDivisorGpuTesterTests
{
[Theory]
[InlineData(3UL, 7UL, true)]
[InlineData(11UL, 23UL, true)]
[InlineData(7UL, 35UL, false)]
[InlineData(13UL, 23UL, false)]
[Trait("Category", "Fast")]
public void IsDivisible_returns_expected(ulong exponent, ulong divisor, bool expected)
{
var tester = new MersenneNumberDivisorGpuTester();
tester.IsDivisible(exponent, divisor).Should().Be(expected);
}
}
PerfectNumbers.Core/Gpu/MersenneNumberDivisorGpuTester.cs
+57
View File
@@ -0,0 +1,57 @@
using System.Collections.Concurrent;
using ILGPU;
using ILGPU.Algorithms;
using ILGPU.Runtime;
namespace PerfectNumbers.Core.Gpu;
public sealed class MersenneNumberDivisorGpuTester
{
private readonly ConcurrentDictionary<Accelerator, Action<Index1D, ulong, ulong, ArrayView<byte>>> _kernelCache = new();
private Action<Index1D, ulong, ulong, ArrayView<byte>> GetKernel(Accelerator accelerator) =>
_kernelCache.GetOrAdd(accelerator, acc => acc.LoadAutoGroupedStreamKernel<Index1D, ulong, ulong, ArrayView<byte>>(Kernel));
public bool IsDivisible(ulong exponent, ulong divisor)
{
var gpu = GpuContextPool.RentPreferred(preferCpu: false);
var accelerator = gpu.Accelerator;
var kernel = GetKernel(accelerator);
using var resultBuffer = accelerator.Allocate1D<byte>(1);
kernel(1, exponent, divisor, resultBuffer.View);
accelerator.Synchronize();
bool divisible = resultBuffer.GetAsArray1D()[0] != 0;
gpu.Dispose();
return divisible;
}
private static void Kernel(Index1D _, ulong exponent, ulong divisor, ArrayView<byte> result)
{
if (divisor <= 1UL)
{
result[0] = 0;
return;
}
int x = 63 - XMath.LeadingZeroCount(divisor);
if (x == 0)
{
var mod = new GpuUInt128(0UL, divisor);
var temp = GpuUInt128.Pow2Mod(exponent, mod);
result[0] = temp.High == 0UL && temp.Low == 1UL ? (byte)1 : (byte)0;
return;
}
ulong ux = (ulong)x;
ulong q = exponent / ux;
ulong r = exponent % ux;
ulong pow2x = 1UL << x;
ulong y = divisor - pow2x;
var modVal = new GpuUInt128(0UL, divisor);
var baseVal = new GpuUInt128(0UL, pow2x); // 2^x ≡ -y (mod divisor)
var part1 = baseVal.ModPow(q, modVal);
var part2 = GpuUInt128.Pow2Mod(r, modVal);
var rem = part1.MulMod(part2, modVal);
result[0] = rem.High == 0UL && rem.Low == 1UL ? (byte)1 : (byte)0;
}
}