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

Merge pull request #32 from mlemanczyk/codex/review-evenperfectbitscanner-for-overflow-checks

Browse Source 
Add overflow checks to GPU residue and divisor scanners
This commit is contained in:
Marek Lemańczyk authored and GitHub committed 2025-09-20 03:22:05 +02:00
2 parents ff2c516a0e 9f2cf12ce1
commit cf4f93fbfe
2 files changed
+66 -59
Download Patch File Download Diff File Expand all files Collapse all files

No files matched your search

PerfectNumbers.Core/Gpu/MersenneNumberDivisorGpuTester.cs
+2 -1
View File
@@ -95,7 +95,8 @@ public sealed class MersenneNumberDivisorGpuTester
for (k2 = 1UL; k2 <= limit; k2++) for (k2 = 1UL; k2 <= limit; k2++)
{ {
kMul2 = (UInt128)k2 << 1; kMul2 = (UInt128)k2 << 1;
d = kMul2 * p + UInt128.One; UInt128 candidate = checked(kMul2 * p);
d = checked(candidate + UInt128.One);
if (p < 64UL && d == ((UInt128)1 << (int)p) - UInt128.One) if (p < 64UL && d == ((UInt128)1 << (int)p) - UInt128.One)
{ {
continue; continue;
PerfectNumbers.Core/Gpu/MersenneNumberResidueGpuTester.cs
+64 -58
View File
@@ -41,72 +41,78 @@ public class MersenneNumberResidueGpuTester(bool useGpuOrder)
ulong step5 = ((exponent % 5UL) << 1) % 5UL; ulong step5 = ((exponent % 5UL) << 1) % 5UL;
GpuUInt128 twoPGpu = (GpuUInt128)twoP; GpuUInt128 twoPGpu = (GpuUInt128)twoP;
var orderBuffer = accelerator.Allocate1D<ulong>(batchSize); var orderBuffer = accelerator.Allocate1D<ulong>(batchSize);
ulong[] orderArray = ArrayPool<ulong>.Shared.Rent(batchSize); ulong[] orderArray = ArrayPool<ulong>.Shared.Rent(batchSize);
UInt128 batchSize128 = (UInt128)batchSize; UInt128 batchSize128 = (UInt128)batchSize;
UInt128 limitInclusive = overallLimit + UInt128.One; UInt128 limitInclusive = overallLimit == UInt128.MaxValue
Span<ulong> orders = orderArray.AsSpan(0, batchSize); ? overallLimit
try : checked(overallLimit + UInt128.One);
{ Span<ulong> orders = orderArray.AsSpan(0, batchSize);
for (UInt128 setIndex = UInt128.Zero; setIndex < setCount && Volatile.Read(ref isPrime); setIndex++) try
{ {
UInt128 setOffset = perSetLimit * setIndex; for (ulong setIndex = 0; setIndex < setCount && Volatile.Read(ref isPrime); setIndex++)
UInt128 setStart = setOffset + UInt128.One; {
if (setStart >= limitInclusive) UInt128 setOffset = checked(perSetLimit * (UInt128)setIndex);
{ UInt128 setStart = checked(setOffset + UInt128.One);
break; if (setStart >= limitInclusive)
} {
break;
}
UInt128 setLimitExclusive = setOffset + perSetLimit + UInt128.One; UInt128 setLimitExclusive = checked(setOffset + perSetLimit);
if (setLimitExclusive > limitInclusive) setLimitExclusive = setLimitExclusive >= limitInclusive
{ ? limitInclusive
setLimitExclusive = limitInclusive; : checked(setLimitExclusive + UInt128.One);
} if (setLimitExclusive > limitInclusive)
{
setLimitExclusive = limitInclusive;
}
UInt128 kStart = setStart; UInt128 kStart = setStart;
while (kStart < setLimitExclusive && Volatile.Read(ref isPrime)) while (kStart < setLimitExclusive && Volatile.Read(ref isPrime))
{ {
UInt128 remaining = setLimitExclusive - kStart; UInt128 remaining = checked(setLimitExclusive - kStart);
int currentSize = remaining > batchSize128 ? batchSize : (int)remaining; int currentSize = remaining > batchSize128 ? batchSize : (int)remaining;
if (currentSize <= 0) if (currentSize <= 0)
{ {
break; break;
} }
orders = orderArray.AsSpan(0, currentSize); orders = orderArray.AsSpan(0, currentSize);
UInt128 q = twoP * kStart + UInt128.One; UInt128 q = checked(twoP * kStart);
q.Mod10_8_5_3(out ulong q0m10, out ulong q0m8, out ulong q0m5, out ulong q0m3); q = checked(q + UInt128.One);
var ra = new ResidueAutomatonArgs(q0m10, step10, q0m8, step8, q0m3, step3, q0m5, step5); q.Mod10_8_5_3(out ulong q0m10, out ulong q0m8, out ulong q0m5, out ulong q0m3);
var ra = new ResidueAutomatonArgs(q0m10, step10, q0m8, step8, q0m3, step3, q0m5, step5);
kernel(currentSize, exponent, twoPGpu, (GpuUInt128)kStart, last, 0UL, kernel(currentSize, exponent, twoPGpu, (GpuUInt128)kStart, last, 0UL,
ra, orderBuffer.View, smallCyclesView, primeViews.LastOne, primeViews.LastSeven, primeViews.LastOnePow2, primeViews.LastSevenPow2); ra, orderBuffer.View, smallCyclesView, primeViews.LastOne, primeViews.LastSeven, primeViews.LastOnePow2, primeViews.LastSevenPow2);
accelerator.Synchronize(); accelerator.Synchronize();
orderBuffer.View.CopyToCPU(ref MemoryMarshal.GetReference(orders), currentSize); orderBuffer.View.CopyToCPU(ref MemoryMarshal.GetReference(orders), currentSize);
if (!Volatile.Read(ref isPrime)) if (!Volatile.Read(ref isPrime))
{ {
break; break;
} }
for (int i = 0; i < currentSize; i++) for (int i = 0; i < currentSize; i++)
{ {
if (orders[i] != 0UL) if (orders[i] != 0UL)
{ {
Volatile.Write(ref isPrime, false); Volatile.Write(ref isPrime, false);
break; break;
} }
} }
kStart += batchSize128; kStart = checked(kStart + batchSize128);
} }
} }
} }
finally finally
{ {
ArrayPool<ulong>.Shared.Return(orderArray); ArrayPool<ulong>.Shared.Return(orderArray);
orderBuffer.Dispose(); orderBuffer.Dispose();
gpuLease.Dispose(); gpuLease.Dispose();
} }
divisorsExhausted = true; divisorsExhausted = true;
} }