Guard residue GPU scanning with execution lock
This commit is contained in:
1 parent
5dd9eb636d
commit
819cb7b00b
1 file changed
+17
-5
@@ -17,11 +17,16 @@ public class MersenneNumberResidueGpuTester(bool useGpuOrder)
|
|||||||
public void Scan(ulong exponent, UInt128 twoP, bool lastIsSeven, UInt128 maxK, ref bool isPrime)
|
public void Scan(ulong exponent, UInt128 twoP, bool lastIsSeven, UInt128 maxK, ref bool isPrime)
|
||||||
{
|
{
|
||||||
var gpuLease = GpuKernelPool.GetKernel(_useGpuOrder);
|
var gpuLease = GpuKernelPool.GetKernel(_useGpuOrder);
|
||||||
|
var execution = gpuLease.EnterExecutionScope();
|
||||||
|
try
|
||||||
|
{
|
||||||
var accelerator = gpuLease.Accelerator;
|
var accelerator = gpuLease.Accelerator;
|
||||||
var stream = gpuLease.Stream;
|
var stream = gpuLease.Stream;
|
||||||
|
|
||||||
// Ensure device has small cycles and primes tables for in-kernel lookup
|
// Ensure device has small cycles and primes tables for in-kernel lookup
|
||||||
var smallCyclesView = GpuKernelPool.EnsureSmallCyclesOnDevice(accelerator);
|
var smallCyclesView = GpuKernelPool.EnsureSmallCyclesOnDevice(accelerator);
|
||||||
ResiduePrimeViews primeViews = GpuKernelPool.EnsureSmallPrimesOnDevice(accelerator);
|
ResiduePrimeViews primeViews = GpuKernelPool.EnsureSmallPrimesOnDevice(accelerator);
|
||||||
|
|
||||||
int batchSize = GpuConstants.ScanBatchSize;
|
int batchSize = GpuConstants.ScanBatchSize;
|
||||||
UInt128 kStart = 1UL;
|
UInt128 kStart = 1UL;
|
||||||
UInt128 limit = maxK + UInt128.One;
|
UInt128 limit = maxK + UInt128.One;
|
||||||
@@ -41,13 +46,15 @@ public class MersenneNumberResidueGpuTester(bool useGpuOrder)
|
|||||||
// testers such as the order kernels.
|
// testers such as the order kernels.
|
||||||
orderBuffer.MemSetToZero();
|
orderBuffer.MemSetToZero();
|
||||||
ulong[] orderArray = ArrayPool<ulong>.Shared.Rent(batchSize);
|
ulong[] orderArray = ArrayPool<ulong>.Shared.Rent(batchSize);
|
||||||
UInt128 batchSize128 = (UInt128)batchSize;
|
|
||||||
UInt128 fullBatchStep = twoP * batchSize128;
|
|
||||||
UInt128 q = twoP * kStart + UInt128.One;
|
|
||||||
// TODO: Replace the direct UInt128 multiply/add updates below with the residue-cycle stepping helper so that q advances reuse cached cycle increments rather than recomputing twoP multiples on every iteration.
|
|
||||||
|
|
||||||
try
|
try
|
||||||
{
|
{
|
||||||
|
UInt128 batchSize128 = (UInt128)batchSize;
|
||||||
|
UInt128 fullBatchStep = twoP * batchSize128;
|
||||||
|
UInt128 q = twoP * kStart + UInt128.One;
|
||||||
|
// TODO: Replace the direct UInt128 multiply/add updates below with the residue-cycle stepping helper so that q advances
|
||||||
|
// reuse cached cycle increments rather than recomputing twoP multiples on every iteration.
|
||||||
|
|
||||||
while (kStart < limit && Volatile.Read(ref isPrime))
|
while (kStart < limit && Volatile.Read(ref isPrime))
|
||||||
{
|
{
|
||||||
UInt128 remaining = limit - kStart;
|
UInt128 remaining = limit - kStart;
|
||||||
@@ -101,8 +108,13 @@ public class MersenneNumberResidueGpuTester(bool useGpuOrder)
|
|||||||
}
|
}
|
||||||
finally
|
finally
|
||||||
{
|
{
|
||||||
ArrayPool<ulong>.Shared.Return(orderArray);
|
ArrayPool<ulong>.Shared.Return(orderArray, clearArray: false);
|
||||||
orderBuffer.Dispose();
|
orderBuffer.Dispose();
|
||||||
|
}
|
||||||
|
}
|
||||||
|
finally
|
||||||
|
{
|
||||||
|
execution.Dispose();
|
||||||
gpuLease.Dispose();
|
gpuLease.Dispose();
|
||||||
}
|
}
|
||||||
}
|
}
|
||||||
|
|||||||
Reference in new issue
Block a user