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

Document GPU kernel audit progress

Browse Source
This commit is contained in:
Marek Lemańczyk committed 2025-10-18 17:42:33 +02:00
1 parent 1e25ea0ccf
commit 6217ea7888
31 files changed
+3496 -3246
Download Patch File Download Diff File Expand all files Collapse all files

No files matched your search

PerfectNumbers.Core/MersenneDivisorCycles.cs
+3 -88
View File
@@ -847,7 +847,7 @@ public class MersenneDivisorCycles
var kernel = accelerator.LoadAutoGroupedStreamKernel<
Index1D,
ArrayView1D<ulong, Stride1D.Dense>,
ArrayView1D<ulong, Stride1D.Dense>>(GpuDivisorCycleKernel);
ArrayView1D<ulong, Stride1D.Dense>>(DivisorCycleKernels.GpuDivisorCycleKernel);
using Stream outputStream = new FileStream(path, FileMode.OpenOrCreate, FileAccess.ReadWrite, FileShare.Read, BufferSize10M, useAsync: true);
@@ -941,93 +941,8 @@ public class MersenneDivisorCycles
}
const byte ByteZero = 0;
const byte ByteOne = 1;
// GPU kernel for divisor cycle calculation
static void GpuDivisorCycleKernel(
Index1D index,
ArrayView1D<ulong, Stride1D.Dense> divisors,
ArrayView1D<ulong, Stride1D.Dense> outCycles)
{
int i = index.X;
outCycles[i] = CalculateCycleLengthGpu(divisors[i]);
}
// GPU-friendly version of cycle length calculation
/// <summary>
/// GPU-friendly cycle calculator that unrolls sixteen doubling steps; it wins the 8,388,607 benchmark and stays within ~2% of the octo loop at divisor 131,071.
/// </summary>
public static ulong CalculateCycleLengthGpu(ulong divisor)
{
if ((divisor & (divisor - 1UL)) == 0UL)
return 1UL;
ulong order = 1UL;
ulong pow = 2UL;
while (true)
{
if (GpuStep(ref pow, divisor, ref order))
return order;
if (GpuStep(ref pow, divisor, ref order))
return order;
if (GpuStep(ref pow, divisor, ref order))
return order;
if (GpuStep(ref pow, divisor, ref order))
return order;
if (GpuStep(ref pow, divisor, ref order))
return order;
if (GpuStep(ref pow, divisor, ref order))
return order;
if (GpuStep(ref pow, divisor, ref order))
return order;
if (GpuStep(ref pow, divisor, ref order))
return order;
if (GpuStep(ref pow, divisor, ref order))
return order;
if (GpuStep(ref pow, divisor, ref order))
return order;
if (GpuStep(ref pow, divisor, ref order))
return order;
if (GpuStep(ref pow, divisor, ref order))
return order;
if (GpuStep(ref pow, divisor, ref order))
return order;
if (GpuStep(ref pow, divisor, ref order))
return order;
if (GpuStep(ref pow, divisor, ref order))
return order;
if (GpuStep(ref pow, divisor, ref order))
return order;
}
}
[MethodImpl(MethodImplOptions.AggressiveInlining)]
private static bool GpuStep(ref ulong pow, ulong divisor, ref ulong order)
{
pow += pow;
if (pow >= divisor)
pow -= divisor;
order++;
return pow == 1UL;
}
public static ulong CalculateCycleLengthGpu(ulong divisor) => DivisorCycleKernels.CalculateCycleLengthGpu(divisor);
public static ulong CalculateCycleLength(ulong divisor, in MontgomeryDivisorData divisorData, bool skipPrimeOrderHeuristic = false)
{