BuildDivisorCandidates moved to MersenneNumberDivisorGpuTester. Micro-optimizations.
This commit is contained in:
Marek Lemańczyk committed
1 parent
82737b99fd
commit
0535ad2b7b
3 files changed
+216
-224
No files matched your search
@@ -12,6 +12,26 @@ public sealed class MersenneNumberDivisorGpuTester
|
||||
private Action<Index1D, ulong, ulong, ArrayView<byte>> GetKernel(Accelerator accelerator) =>
|
||||
_kernelCache.GetOrAdd(accelerator, acc => acc.LoadAutoGroupedStreamKernel<Index1D, ulong, ulong, ArrayView<byte>>(Kernel));
|
||||
|
||||
public static void BuildDivisorCandidates()
|
||||
{
|
||||
uint[] snapshot = MersenneDivisorCycles.Shared.ExportSmallCyclesSnapshot();
|
||||
Span<(ulong divisor, uint cycle)> list = stackalloc (ulong divisor, uint cycle)[snapshot.Length / 2];
|
||||
uint cycle;
|
||||
int count = 0, i, snapshotLength = snapshot.Length;
|
||||
for (i = 3; i < snapshotLength; i += 2)
|
||||
{
|
||||
cycle = snapshot[i];
|
||||
if (cycle == 0U)
|
||||
{
|
||||
continue;
|
||||
}
|
||||
|
||||
list[count++] = ((ulong)i, cycle);
|
||||
}
|
||||
|
||||
_divisorCandidates = list[..count].ToArray();
|
||||
}
|
||||
|
||||
public bool IsDivisible(ulong exponent, ulong divisor)
|
||||
{
|
||||
var gpu = GpuContextPool.RentPreferred(preferCpu: false);
|
||||
@@ -26,6 +46,71 @@ public sealed class MersenneNumberDivisorGpuTester
|
||||
return divisible;
|
||||
}
|
||||
|
||||
private static (ulong divisor, uint cycle)[]? _divisorCandidates;
|
||||
|
||||
public bool IsPrime(ulong p, ulong d, int divisorCyclesSearchLimit, out bool detailedCheck)
|
||||
{
|
||||
if (d != 0UL)
|
||||
{
|
||||
if (IsDivisible(p, d))
|
||||
{
|
||||
detailedCheck = true;
|
||||
return false;
|
||||
}
|
||||
|
||||
detailedCheck = false;
|
||||
return true;
|
||||
}
|
||||
|
||||
var candidates = _divisorCandidates!;
|
||||
int k, len = candidates.Length;
|
||||
uint cycle;
|
||||
|
||||
// k is simple iterator here
|
||||
for (k = 0; k < len; k++)
|
||||
{
|
||||
(d, cycle) = candidates[k];
|
||||
if (p % cycle != 0UL)
|
||||
{
|
||||
continue;
|
||||
}
|
||||
|
||||
if (IsDivisible(p, d))
|
||||
{
|
||||
detailedCheck = true;
|
||||
return false;
|
||||
}
|
||||
}
|
||||
|
||||
ulong kMul2;
|
||||
var divisorCycles = MersenneDivisorCycles.Shared;
|
||||
for (k = 1; k <= divisorCyclesSearchLimit; k++)
|
||||
{
|
||||
kMul2 = (ulong)k << 1;
|
||||
if (p > ulong.MaxValue / kMul2)
|
||||
{
|
||||
break;
|
||||
}
|
||||
|
||||
d = kMul2 * p + 1UL;
|
||||
// kkMul2 now becomes cycle. We're reusing existing variable for the best performance
|
||||
kMul2 = divisorCycles.GetCycle(d);
|
||||
if (p % kMul2 != 0UL)
|
||||
{
|
||||
continue;
|
||||
}
|
||||
|
||||
if (IsDivisible(p, d))
|
||||
{
|
||||
detailedCheck = true;
|
||||
return false;
|
||||
}
|
||||
}
|
||||
|
||||
detailedCheck = false;
|
||||
return true;
|
||||
}
|
||||
|
||||
private static void Kernel(Index1D _, ulong exponent, ulong divisor, ArrayView<byte> result)
|
||||
{
|
||||
if (divisor <= 1UL)
|
||||
|
||||
@@ -6,82 +6,80 @@ namespace PerfectNumbers.Core.Gpu;
|
||||
|
||||
public class MersenneNumberResidueGpuTester(bool useGpuOrder)
|
||||
{
|
||||
private readonly bool _useGpuOrder = useGpuOrder;
|
||||
// TODO: Integrate MersenneDivisorCycles.Shared to consult cycle lengths for small q (<= 4M) and fast-reject
|
||||
// candidates without launching heavy kernels. Consider a device-visible constant buffer per-accelerator
|
||||
// via GpuKernelPool to avoid host round-trips.
|
||||
private readonly bool _useGpuOrder = useGpuOrder;
|
||||
// TODO: Integrate MersenneDivisorCycles.Shared to consult cycle lengths for small q (<= 4M) and fast-reject
|
||||
// candidates without launching heavy kernels. Consider a device-visible constant buffer per-accelerator
|
||||
// via GpuKernelPool to avoid host round-trips.
|
||||
|
||||
// GPU residue variant: check 2^p % q == 1 for q = 2*p*k + 1.
|
||||
public void Scan(ulong exponent, UInt128 twoP, bool lastIsSeven, UInt128 maxK, ref bool isPrime)
|
||||
{
|
||||
var gpuLease = GpuKernelPool.GetKernel(_useGpuOrder);
|
||||
var accelerator = gpuLease.Accelerator;
|
||||
// Ensure device has small cycles table for in-kernel lookup
|
||||
var smallCyclesView = GpuKernelPool.EnsureSmallCyclesOnDevice(accelerator);
|
||||
int batchSize = GpuConstants.ScanBatchSize;
|
||||
UInt128 kStart = 1UL;
|
||||
byte last = lastIsSeven ? (byte)1 : (byte)0;
|
||||
var kernel = gpuLease.Pow2ModKernel;
|
||||
public void Scan(ulong exponent, UInt128 twoP, bool lastIsSeven, UInt128 maxK, ref bool isPrime)
|
||||
{
|
||||
var gpuLease = GpuKernelPool.GetKernel(_useGpuOrder);
|
||||
var accelerator = gpuLease.Accelerator;
|
||||
// Ensure device has small cycles table for in-kernel lookup
|
||||
var smallCyclesView = GpuKernelPool.EnsureSmallCyclesOnDevice(accelerator);
|
||||
int batchSize = GpuConstants.ScanBatchSize;
|
||||
UInt128 kStart = 1UL;
|
||||
byte last = lastIsSeven ? (byte)1 : (byte)0;
|
||||
var kernel = gpuLease.Pow2ModKernel;
|
||||
|
||||
var orderBuffer = accelerator.Allocate1D<ulong>(batchSize);
|
||||
ulong[] orderArray = ArrayPool<ulong>.Shared.Rent(batchSize);
|
||||
UInt128 remaining;
|
||||
int currentSize;
|
||||
int i;
|
||||
UInt128 batchSize128 = (UInt128)batchSize,
|
||||
q = UInt128.Zero;
|
||||
Span<ulong> orders = orderArray.AsSpan(0, batchSize);
|
||||
try
|
||||
{
|
||||
while (kStart < maxK && Volatile.Read(ref isPrime))
|
||||
{
|
||||
remaining = maxK - kStart;
|
||||
if (remaining > batchSize128)
|
||||
{
|
||||
currentSize = batchSize;
|
||||
}
|
||||
else
|
||||
{
|
||||
currentSize = (int)remaining;
|
||||
orders = orderArray.AsSpan(0, currentSize);
|
||||
}
|
||||
var orderBuffer = accelerator.Allocate1D<ulong>(batchSize);
|
||||
ulong[] orderArray = ArrayPool<ulong>.Shared.Rent(batchSize);
|
||||
UInt128 remaining;
|
||||
int currentSize;
|
||||
int i;
|
||||
UInt128 batchSize128 = (UInt128)batchSize, q;
|
||||
Span<ulong> orders = orderArray.AsSpan(0, batchSize);
|
||||
try
|
||||
{
|
||||
while (kStart < maxK && Volatile.Read(ref isPrime))
|
||||
{
|
||||
remaining = maxK - kStart;
|
||||
if (remaining > batchSize128)
|
||||
{
|
||||
currentSize = batchSize;
|
||||
}
|
||||
else
|
||||
{
|
||||
currentSize = (int)remaining;
|
||||
orders = orderArray.AsSpan(0, currentSize);
|
||||
}
|
||||
|
||||
q = twoP * kStart + UInt128.One;
|
||||
q.Mod10_8_5_3(out ulong q0m10, out ulong q0m8, out ulong q0m5, out ulong q0m3);
|
||||
ulong step10 = ((exponent % 10UL) << 1) % 10UL;
|
||||
ulong step8 = ((exponent & 7UL) << 1) & 7UL;
|
||||
ulong step3 = ((exponent % 3UL) << 1) % 3UL;
|
||||
ulong step5 = ((exponent % 5UL) << 1) % 5UL;
|
||||
q = twoP * kStart + UInt128.One;
|
||||
q.Mod10_8_5_3(out ulong q0m10, out ulong q0m8, out ulong q0m5, out ulong q0m3);
|
||||
ulong step10 = ((exponent % 10UL) << 1) % 10UL;
|
||||
ulong step8 = ((exponent & 7UL) << 1) & 7UL;
|
||||
ulong step3 = ((exponent % 3UL) << 1) % 3UL;
|
||||
ulong step5 = ((exponent % 5UL) << 1) % 5UL;
|
||||
|
||||
kernel(currentSize, exponent, (GpuUInt128)twoP, (GpuUInt128)kStart, last, 0UL,
|
||||
q0m10, q0m8, q0m3, q0m5, orderBuffer.View, smallCyclesView);
|
||||
kernel(currentSize, exponent, (GpuUInt128)twoP, (GpuUInt128)kStart, last, 0UL,
|
||||
q0m10, q0m8, q0m3, q0m5, orderBuffer.View, smallCyclesView);
|
||||
|
||||
accelerator.Synchronize();
|
||||
orderBuffer.View.CopyToCPU(ref MemoryMarshal.GetReference(orders), currentSize);
|
||||
accelerator.Synchronize();
|
||||
if (!Volatile.Read(ref isPrime))
|
||||
{
|
||||
break;
|
||||
}
|
||||
accelerator.Synchronize();
|
||||
orderBuffer.View.CopyToCPU(ref MemoryMarshal.GetReference(orders), currentSize);
|
||||
accelerator.Synchronize();
|
||||
if (!Volatile.Read(ref isPrime))
|
||||
{
|
||||
break;
|
||||
}
|
||||
|
||||
for (i = 0; i < currentSize; i++, q += twoP)
|
||||
{
|
||||
if (orders[i] != 0UL && q.IsPrimeCandidate())
|
||||
{
|
||||
Volatile.Write(ref isPrime, false);
|
||||
break;
|
||||
}
|
||||
}
|
||||
for (i = 0; i < currentSize; i++, q += twoP)
|
||||
{
|
||||
if (orders[i] != 0UL && q.IsPrimeCandidate())
|
||||
{
|
||||
Volatile.Write(ref isPrime, false);
|
||||
break;
|
||||
}
|
||||
}
|
||||
|
||||
kStart += batchSize128;
|
||||
}
|
||||
}
|
||||
finally
|
||||
{
|
||||
ArrayPool<ulong>.Shared.Return(orderArray);
|
||||
orderBuffer.Dispose();
|
||||
gpuLease.Dispose();
|
||||
}
|
||||
|
||||
}
|
||||
kStart += batchSize128;
|
||||
}
|
||||
}
|
||||
finally
|
||||
{
|
||||
ArrayPool<ulong>.Shared.Return(orderArray);
|
||||
orderBuffer.Dispose();
|
||||
gpuLease.Dispose();
|
||||
}
|
||||
}
|
||||
}
|
||||
Reference in new issue
Block a user