Prime testing using divisor tables depending on the last digit separated to its own class. PrimeTester switched back to standard endings.
This commit is contained in:
mlemanczyk committed
1 parent
a296e62d4d
commit
c311b27dcc
12 files changed
+526
-215
No files matched your search
@@ -16,10 +16,10 @@ public static class AcceleratorStreamPool
|
||||
|
||||
public static AcceleratorStream Rent(int acceleratorIndex)
|
||||
{
|
||||
var streamLock = _locks[acceleratorIndex];
|
||||
// var streamLock = _locks[acceleratorIndex];
|
||||
var queue = _streams[acceleratorIndex];
|
||||
|
||||
streamLock.Wait();
|
||||
// streamLock.Wait();
|
||||
|
||||
return queue.TryDequeue(out var stream)
|
||||
? stream
|
||||
@@ -28,10 +28,10 @@ public static class AcceleratorStreamPool
|
||||
|
||||
public static void Return(int acceleratorIndex, AcceleratorStream stream)
|
||||
{
|
||||
var streamLock = _locks[acceleratorIndex];
|
||||
// var streamLock = _locks[acceleratorIndex];
|
||||
var queue = _streams[acceleratorIndex];
|
||||
queue.Enqueue(stream);
|
||||
streamLock.Release();
|
||||
// streamLock.Release();
|
||||
}
|
||||
|
||||
public static void WarmUp(int acceleratorIndex)
|
||||
|
||||
@@ -115,7 +115,7 @@ public sealed class PrimeOrderCalculatorAccelerator
|
||||
public MemoryBuffer1D<ulong, Stride1D.Dense> OutputUlong;
|
||||
public readonly MemoryBuffer1D<ulong, Stride1D.Dense> OutputUlong2;
|
||||
public ArrayView1D<ulong, Stride1D.Dense> OutputUlongView;
|
||||
public readonly ArrayView1D<ulong, Stride1D.Dense> OutputUlongView2;
|
||||
public readonly ArrayView<ulong> OutputUlongView2;
|
||||
|
||||
public MemoryBuffer1D<int, Stride1D.Dense> OutputInt;
|
||||
public readonly MemoryBuffer1D<int, Stride1D.Dense> OutputInt2;
|
||||
@@ -159,7 +159,8 @@ public sealed class PrimeOrderCalculatorAccelerator
|
||||
|
||||
public readonly Action<AcceleratorStream, Index1D, ArrayView1D<ulong, Stride1D.Dense>, ulong, ulong, ulong, ulong, ulong, ArrayView1D<ulong, Stride1D.Dense>> ConvertToStandardKernelLauncher = KernelUtil.GetKernel(accelerator.LoadAutoGroupedStreamKernel<Index1D, ArrayView1D<ulong, Stride1D.Dense>, ulong, ulong, ulong, ulong, ulong, ArrayView1D<ulong, Stride1D.Dense>>(Pow2MontgomeryKernels.Pow2MontgomeryKernelConvertToStandard)).CreateLauncherDelegate<Action<AcceleratorStream, Index1D, ArrayView1D<ulong, Stride1D.Dense>, ulong, ulong, ulong, ulong, ulong, ArrayView1D<ulong, Stride1D.Dense>>>();
|
||||
|
||||
public readonly Action<AcceleratorStream, Index1D, byte, ArrayView<int>, ulong, ulong, HeuristicCombinedGpuViews> HeuristicCombinedTrialDivisionKernelLauncher = KernelUtil.GetKernel(accelerator.LoadAutoGroupedStreamKernel<Index1D, byte, ArrayView<int>, ulong, ulong, HeuristicCombinedGpuViews>(PrimeTesterKernels.HeuristicTrialCombinedDivisionKernel)).CreateLauncherDelegate<Action<AcceleratorStream, Index1D, byte, ArrayView<int>, ulong, ulong, HeuristicCombinedGpuViews>>();
|
||||
// public readonly Action<AcceleratorStream, Index1D, byte, ArrayView<int>, ulong, ulong, HeuristicCombinedGpuViews> HeuristicCombinedTrialDivisionKernelLauncher = KernelUtil.GetKernel(accelerator.LoadAutoGroupedStreamKernel<Index1D, byte, ArrayView<int>, ulong, ulong, HeuristicCombinedGpuViews>(PrimeTesterKernels.HeuristicTrialCombinedDivisionKernel)).CreateLauncherDelegate<Action<AcceleratorStream, Index1D, byte, ArrayView<int>, ulong, ulong, HeuristicCombinedGpuViews>>();
|
||||
public readonly Action<AcceleratorStream, Index1D, ArrayView<int>, ulong, ulong, ArrayView<ulong>, ArrayView<ulong>> HeuristicCombinedTrialDivisionKernelLauncher = KernelUtil.GetKernel(accelerator.LoadAutoGroupedStreamKernel<Index1D, ArrayView<int>, ulong, ulong, ArrayView<ulong>, ArrayView<ulong>>(PrimeTesterKernels.HeuristicTrialCombinedDivisionKernel)).CreateLauncherDelegate<Action<AcceleratorStream, Index1D, ArrayView<int>, ulong, ulong, ArrayView<ulong>, ArrayView<ulong>>>();
|
||||
|
||||
public readonly Action<AcceleratorStream, Index1D, ArrayView1D<ulong, Stride1D.Dense>, ulong, ulong, ulong, ulong, ulong, ArrayView1D<ulong, Stride1D.Dense>> KeepMontgomeryKernelLauncher = KernelUtil.GetKernel(accelerator.LoadAutoGroupedStreamKernel<Index1D, ArrayView1D<ulong, Stride1D.Dense>, ulong, ulong, ulong, ulong, ulong, ArrayView1D<ulong, Stride1D.Dense>>(Pow2MontgomeryKernels.Pow2MontgomeryKernelKeepMontgomery)).CreateLauncherDelegate<Action<AcceleratorStream, Index1D, ArrayView1D<ulong, Stride1D.Dense>, ulong, ulong, ulong, ulong, ulong, ArrayView1D<ulong, Stride1D.Dense>>>();
|
||||
|
||||
@@ -173,17 +174,17 @@ public sealed class PrimeOrderCalculatorAccelerator
|
||||
|
||||
// public readonly Action<AcceleratorStream, Index1D, ArrayView<ulong>, ArrayView<byte>> SharesFactorKernelLauncher = KernelUtil.GetKernel(accelerator.LoadAutoGroupedStreamKernel<Index1D, ArrayView<ulong>, ArrayView<byte>>(PrimeTesterKernels.SharesFactorKernel)).CreateLauncherDelegate<Action<AcceleratorStream, Index1D, ArrayView<ulong>, ArrayView<byte>>>();
|
||||
|
||||
public readonly Action<AcceleratorStream, Index1D, ulong, uint, ArrayView1D<uint, Stride1D.Dense>, ArrayView1D<ulong, Stride1D.Dense>, int, ArrayView1D<ulong, Stride1D.Dense>, ArrayView1D<int, Stride1D.Dense>, ArrayView1D<int, Stride1D.Dense>, ArrayView1D<ulong, Stride1D.Dense>> SmallPrimeFactorKernelLauncher = KernelUtil.GetKernel(accelerator.LoadAutoGroupedStreamKernel<Index1D, ulong, uint, ArrayView1D<uint, Stride1D.Dense>, ArrayView1D<ulong, Stride1D.Dense>, int, ArrayView1D<ulong, Stride1D.Dense>, ArrayView1D<int, Stride1D.Dense>, ArrayView1D<int, Stride1D.Dense>, ArrayView1D<ulong, Stride1D.Dense>>(SmallPrimeFactorKernels.SmallPrimeFactorKernelScan)).CreateLauncherDelegate<Action<AcceleratorStream, Index1D, ulong, uint, ArrayView1D<uint, Stride1D.Dense>, ArrayView1D<ulong, Stride1D.Dense>, int, ArrayView1D<ulong, Stride1D.Dense>, ArrayView1D<int, Stride1D.Dense>, ArrayView1D<int, Stride1D.Dense>, ArrayView1D<ulong, Stride1D.Dense>>>();
|
||||
// public readonly Action<AcceleratorStream, Index1D, ulong, uint, ArrayView1D<uint, Stride1D.Dense>, ArrayView1D<ulong, Stride1D.Dense>, int, ArrayView1D<ulong, Stride1D.Dense>, ArrayView1D<int, Stride1D.Dense>, ArrayView1D<int, Stride1D.Dense>, ArrayView1D<ulong, Stride1D.Dense>> SmallPrimeFactorKernelLauncher = KernelUtil.GetKernel(accelerator.LoadAutoGroupedStreamKernel<Index1D, ulong, uint, ArrayView1D<uint, Stride1D.Dense>, ArrayView1D<ulong, Stride1D.Dense>, int, ArrayView1D<ulong, Stride1D.Dense>, ArrayView1D<int, Stride1D.Dense>, ArrayView1D<int, Stride1D.Dense>, ArrayView1D<ulong, Stride1D.Dense>>(SmallPrimeFactorKernels.SmallPrimeFactorKernelScan)).CreateLauncherDelegate<Action<AcceleratorStream, Index1D, ulong, uint, ArrayView1D<uint, Stride1D.Dense>, ArrayView1D<ulong, Stride1D.Dense>, int, ArrayView1D<ulong, Stride1D.Dense>, ArrayView1D<int, Stride1D.Dense>, ArrayView1D<int, Stride1D.Dense>, ArrayView1D<ulong, Stride1D.Dense>>>();
|
||||
|
||||
public readonly Action<AcceleratorStream, Index1D, ArrayView<ulong>, ArrayView<uint>, ArrayView<uint>, ArrayView<uint>, ArrayView<uint>, ArrayView<ulong>, ArrayView<ulong>, ArrayView<ulong>, ArrayView<ulong>, ArrayView<byte>> SmallPrimeSieveKernelLauncher = KernelUtil.GetKernel(accelerator.LoadAutoGroupedStreamKernel<Index1D, ArrayView<ulong>, ArrayView<uint>, ArrayView<uint>, ArrayView<uint>, ArrayView<uint>, ArrayView<ulong>, ArrayView<ulong>, ArrayView<ulong>, ArrayView<ulong>, ArrayView<byte>>(PrimeTesterKernels.SmallPrimeSieveKernel)).CreateLauncherDelegate<Action<AcceleratorStream, Index1D, ArrayView<ulong>, ArrayView<uint>, ArrayView<uint>, ArrayView<uint>, ArrayView<uint>, ArrayView<ulong>, ArrayView<ulong>, ArrayView<ulong>, ArrayView<ulong>, ArrayView<byte>>>();
|
||||
// public readonly Action<AcceleratorStream, Index1D, ArrayView<ulong>, ArrayView<uint>, ArrayView<uint>, ArrayView<uint>, ArrayView<uint>, ArrayView<ulong>, ArrayView<ulong>, ArrayView<ulong>, ArrayView<ulong>, ArrayView<byte>> SmallPrimeSieveKernelLauncher = KernelUtil.GetKernel(accelerator.LoadAutoGroupedStreamKernel<Index1D, ArrayView<ulong>, ArrayView<uint>, ArrayView<uint>, ArrayView<uint>, ArrayView<uint>, ArrayView<ulong>, ArrayView<ulong>, ArrayView<ulong>, ArrayView<ulong>, ArrayView<byte>>(PrimeTesterKernels.SmallPrimeSieveKernel)).CreateLauncherDelegate<Action<AcceleratorStream, Index1D, ArrayView<ulong>, ArrayView<uint>, ArrayView<uint>, ArrayView<uint>, ArrayView<uint>, ArrayView<ulong>, ArrayView<ulong>, ArrayView<ulong>, ArrayView<ulong>, ArrayView<byte>>>();
|
||||
|
||||
public readonly Action<AcceleratorStream, Index1D, ulong, ArrayView1D<ulong, Stride1D.Dense>, int, ulong, ArrayView1D<ulong, Stride1D.Dense>, ArrayView1D<ulong, Stride1D.Dense>> SpecialMaxKernelLauncher = KernelUtil.GetKernel(accelerator.LoadAutoGroupedStreamKernel<Index1D, ulong, ArrayView1D<ulong, Stride1D.Dense>, int, ulong, ArrayView1D<ulong, Stride1D.Dense>, ArrayView1D<ulong, Stride1D.Dense>>(PrimeOrderGpuHeuristics.EvaluateSpecialMaxCandidatesKernel)).CreateLauncherDelegate<Action<AcceleratorStream, Index1D, ulong, ArrayView1D<ulong, Stride1D.Dense>, int, ulong, ArrayView1D<ulong, Stride1D.Dense>, ArrayView1D<ulong, Stride1D.Dense>>>();
|
||||
public readonly Action<AcceleratorStream, Index1D, ulong, ArrayView<ulong>, int, ulong, ArrayView<ulong>> SpecialMaxKernelLauncher = KernelUtil.GetKernel(accelerator.LoadAutoGroupedStreamKernel<Index1D, ulong, ArrayView<ulong>, int, ulong, ArrayView<ulong>>(PrimeOrderGpuHeuristics.EvaluateSpecialMaxCandidatesKernel)).CreateLauncherDelegate<Action<AcceleratorStream, Index1D, ulong, ArrayView<ulong>, int, ulong, ArrayView<ulong>>>();
|
||||
}
|
||||
|
||||
public readonly Action<AcceleratorStream, Index1D, ulong, CalculateOrderKernelConfig, ulong, ulong, ulong, ulong, ulong, ArrayView1D<uint, Stride1D.Dense>, ArrayView1D<ulong, Stride1D.Dense>, int, OrderKernelBuffers> CalculateOrderKernelLauncher;
|
||||
public readonly Kernel CheckFactorsKernel;
|
||||
public readonly Action<AcceleratorStream, Index1D, ArrayView1D<ulong, Stride1D.Dense>, ulong, ulong, ulong, ulong, ulong, ArrayView1D<ulong, Stride1D.Dense>> ConvertToStandardKernelLauncher;
|
||||
public readonly Action<AcceleratorStream, Index1D, byte, ArrayView<int>, ulong, ulong, HeuristicCombinedGpuViews> HeuristicCombinedTrialDivisionKernelLauncher;
|
||||
public readonly Action<AcceleratorStream, Index1D, ArrayView<int>, ulong, ulong, ArrayView<ulong>, ArrayView<ulong>> HeuristicCombinedTrialDivisionKernelLauncher;
|
||||
public readonly Action<AcceleratorStream, Index1D, ArrayView1D<ulong, Stride1D.Dense>, ulong, ulong, ulong, ulong, ulong, ArrayView1D<ulong, Stride1D.Dense>> KeepMontgomeryKernelLauncher;
|
||||
public readonly 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>> PartialFactorKernelLauncher;
|
||||
public readonly Kernel PollardRhoKernel;
|
||||
@@ -192,7 +193,7 @@ public sealed class PrimeOrderCalculatorAccelerator
|
||||
public readonly Action<AcceleratorStream, Index1D, ArrayView<ulong>, ArrayView<byte>> SharesFactorKernelLauncher;
|
||||
public readonly Action<AcceleratorStream, Index1D, ulong, uint, ArrayView1D<uint, Stride1D.Dense>, ArrayView1D<ulong, Stride1D.Dense>, int, ArrayView1D<ulong, Stride1D.Dense>, ArrayView1D<int, Stride1D.Dense>, ArrayView1D<int, Stride1D.Dense>, ArrayView1D<ulong, Stride1D.Dense>> SmallPrimeFactorKernelLauncher;
|
||||
public readonly Action<AcceleratorStream, Index1D, ArrayView<ulong>, ArrayView<uint>, ArrayView<uint>, ArrayView<uint>, ArrayView<uint>, ArrayView<ulong>, ArrayView<ulong>, ArrayView<ulong>, ArrayView<ulong>, ArrayView<byte>> SmallPrimeSieveKernelLauncher;
|
||||
public readonly Action<AcceleratorStream, Index1D, ulong, ArrayView1D<ulong, Stride1D.Dense>, int, ulong, ArrayView1D<ulong, Stride1D.Dense>, ArrayView1D<ulong, Stride1D.Dense>> SpecialMaxKernelLauncher;
|
||||
public readonly Action<AcceleratorStream, Index1D, ulong, ArrayView<ulong>, int, ulong, ArrayView<ulong>> SpecialMaxKernelLauncher;
|
||||
|
||||
public void EnsureCalculateOrderWideCapacity(int newSize)
|
||||
{
|
||||
@@ -442,14 +443,14 @@ public sealed class PrimeOrderCalculatorAccelerator
|
||||
CheckFactorsKernel = kernels.CheckFactorsKernel;
|
||||
ConvertToStandardKernelLauncher = kernels.ConvertToStandardKernelLauncher;
|
||||
HeuristicCombinedTrialDivisionKernelLauncher = kernels.HeuristicCombinedTrialDivisionKernelLauncher;
|
||||
KeepMontgomeryKernelLauncher = kernels.KeepMontgomeryKernelLauncher;
|
||||
// KeepMontgomeryKernelLauncher = kernels.KeepMontgomeryKernelLauncher;
|
||||
// PartialFactorKernelLauncher = kernels.PartialFactorKernelLauncher;
|
||||
// PollardRhoKernel = kernels.PollardRhoKernel;
|
||||
Pow2ModKernelLauncher = kernels.Pow2ModKernelLauncher;
|
||||
Pow2ModWideKernelLauncher = kernels.Pow2ModWideKernelLauncher;
|
||||
// SharesFactorKernelLauncher = kernels.SharesFactorKernelLauncher;
|
||||
SmallPrimeFactorKernelLauncher = kernels.SmallPrimeFactorKernelLauncher;
|
||||
SmallPrimeSieveKernelLauncher = kernels.SmallPrimeSieveKernelLauncher;
|
||||
// SmallPrimeFactorKernelLauncher = kernels.SmallPrimeFactorKernelLauncher;
|
||||
// SmallPrimeSieveKernelLauncher = kernels.SmallPrimeSieveKernelLauncher;
|
||||
SpecialMaxKernelLauncher = kernels.SpecialMaxKernelLauncher;
|
||||
// }
|
||||
}
|
||||
|
||||
@@ -8,72 +8,49 @@ internal static partial class PrimeOrderGpuHeuristics
|
||||
internal static void EvaluateSpecialMaxCandidatesKernel(
|
||||
Index1D index,
|
||||
ulong phi,
|
||||
ArrayView1D<ulong, Stride1D.Dense> factors,
|
||||
ArrayView<ulong> factors,
|
||||
int factorCount,
|
||||
ulong divisorModulus,
|
||||
ArrayView1D<ulong, Stride1D.Dense> candidates,
|
||||
ArrayView1D<ulong, Stride1D.Dense> resultOut)
|
||||
ArrayView<ulong> resultOut)
|
||||
{
|
||||
if (index != 0)
|
||||
{
|
||||
return;
|
||||
}
|
||||
|
||||
int actual = 0;
|
||||
for (int i = 0; i < factorCount; i++)
|
||||
{
|
||||
ulong factor = factors[i];
|
||||
if (factor <= 1UL)
|
||||
{
|
||||
continue;
|
||||
}
|
||||
|
||||
ulong reduced = phi / factor;
|
||||
if (reduced == 0UL)
|
||||
{
|
||||
continue;
|
||||
}
|
||||
|
||||
candidates[actual] = reduced;
|
||||
actual++;
|
||||
factors[i] = phi / factors[i];
|
||||
}
|
||||
|
||||
if (actual == 0)
|
||||
for (int i = 1; i < factorCount; i++)
|
||||
{
|
||||
resultOut[0] = 1;
|
||||
return;
|
||||
}
|
||||
|
||||
for (int i = 1; i < actual; i++)
|
||||
{
|
||||
ulong current = candidates[i];
|
||||
ulong current = factors[i];
|
||||
int insert = i - 1;
|
||||
while (insert >= 0 && candidates[insert] > current)
|
||||
while (insert >= 0 && factors[insert] > current)
|
||||
{
|
||||
candidates[insert + 1] = candidates[insert];
|
||||
factors[insert + 1] = factors[insert];
|
||||
insert--;
|
||||
}
|
||||
|
||||
candidates[insert + 1] = current;
|
||||
factors[insert + 1] = current;
|
||||
}
|
||||
|
||||
var divisorPartial = new GpuDivisorPartialData(divisorModulus);
|
||||
var stepper = new ExponentRemainderStepperGpu(divisorPartial);
|
||||
if (stepper.InitializeIsUnityGpu(candidates[0]))
|
||||
if (stepper.InitializeIsUnityGpu(factors[0]))
|
||||
{
|
||||
Atomic.Exchange(ref resultOut[0], 0);
|
||||
resultOut[0] = 0UL;
|
||||
// Atomic.Exchange(ref resultOut[0], 0);
|
||||
return;
|
||||
}
|
||||
|
||||
for (int i = 1; i < actual; i++)
|
||||
for (int i = 1; i < factorCount; i++)
|
||||
{
|
||||
if (stepper.ComputeNextIsUnityGpu(candidates[i]))
|
||||
if (stepper.ComputeNextIsUnityGpu(factors[i]))
|
||||
{
|
||||
Atomic.Exchange(ref resultOut[0], 0);
|
||||
resultOut[0] = 0UL;
|
||||
// Atomic.Exchange(ref resultOut[0], 0);
|
||||
return;
|
||||
}
|
||||
}
|
||||
|
||||
Atomic.Exchange(ref resultOut[0], 1);
|
||||
resultOut[0] = 1UL;
|
||||
// Atomic.Exchange(ref resultOut[0], 1);
|
||||
}
|
||||
}
|
||||
@@ -80,21 +80,17 @@ internal static class PrimeTesterKernels
|
||||
|
||||
public static void HeuristicTrialCombinedDivisionKernel(
|
||||
Index1D index,
|
||||
byte nMod10,
|
||||
// byte nMod10,
|
||||
ArrayView<int> resultFlag,
|
||||
ulong n,
|
||||
ulong maxDivisorSquare,
|
||||
HeuristicCombinedGpuViews tables)
|
||||
ArrayView<ulong> divisors,
|
||||
ArrayView<ulong> divisorSquares)
|
||||
{
|
||||
var (divisors, divisorSquares) = tables.SelectDivisorsAndSquares(nMod10);
|
||||
// var (divisors, divisorSquares) = tables.SelectDivisorsAndSquares(nMod10);
|
||||
|
||||
int threadIndex = index;
|
||||
if (divisorSquares[threadIndex] > maxDivisorSquare)
|
||||
{
|
||||
return;
|
||||
}
|
||||
|
||||
if (n % divisors[threadIndex] == 0UL)
|
||||
if (divisorSquares[threadIndex] <= maxDivisorSquare && (n % divisors[threadIndex] == 0UL))
|
||||
{
|
||||
Atomic.Exchange(ref resultFlag[0], 1);
|
||||
}
|
||||
|
||||
Reference in new issue
Block a user