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

Most device buffers integrated into PrimeOrderCalculatorAccelerator. PrimeOrderCalculator switched to partial factoring on GPU.

Browse Source
This commit is contained in:
mlemanczyk committed 2025-11-26 03:16:39 +01:00
1 parent 5b48a89448
commit 13498dc460
17 files changed
+483 -432
Download Patch File Download Diff File Expand all files Collapse all files

No files matched your search

PerfectNumbers.Core/Gpu/Pow2MontgomeryGpuCalculator.cs
+10 -10
View File
@@ -15,18 +15,18 @@ public static class Pow2MontgomeryGpuCalculator
ulong result = 0UL;
int acceleratorIndex = gpu.AcceleratorIndex;
Accelerator? accelerator = gpu.Accelerator;
MemoryBuffer1D<ulong, Stride1D.Dense> exponentBuffer = gpu.Input;
MemoryBuffer1D<ulong, Stride1D.Dense> resultBuffer = gpu.OutputUlong;
var exponentBufferView = gpu.InputView;
var resultBufferView = gpu.OutputUlongView;
AcceleratorStream stream = AcceleratorStreamPool.Rent(acceleratorIndex);
exponentBuffer.View.CopyFromCPU(stream, ref exponent, 1);
exponentBufferView.CopyFromCPU(stream, ref exponent, 1);
// We don't need to worry about any left-overs here.
// resultBuffer.MemSetToZero(stream);
var kernelLauncher = gpu.KeepMontgomeryKernelLauncher;
kernelLauncher(stream, 1, exponentBuffer.View, divisor.Modulus, divisor.NPrime, divisor.MontgomeryOne, divisor.MontgomeryTwo, divisor.MontgomeryTwoSquared, resultBuffer.View);
kernelLauncher(stream, 1, exponentBufferView, divisor.Modulus, divisor.NPrime, divisor.MontgomeryOne, divisor.MontgomeryTwo, divisor.MontgomeryTwoSquared, resultBufferView);
resultBuffer.View.CopyToCPU(stream, ref result, 1);
resultBufferView.CopyToCPU(stream, ref result, 1);
stream.Synchronize();
AcceleratorStreamPool.Return(acceleratorIndex, stream);
// GpuPrimeWorkLimiter.Release();
@@ -38,21 +38,21 @@ public static class Pow2MontgomeryGpuCalculator
public static ulong CalculateConvert(PrimeOrderCalculatorAccelerator gpu, ulong exponent, in MontgomeryDivisorData divisor)
{
ulong result = 0UL;
MemoryBuffer1D<ulong, Stride1D.Dense> exponentBuffer = gpu.Input;
MemoryBuffer1D<ulong, Stride1D.Dense> resultBuffer = gpu.OutputUlong;
var exponentBufferView = gpu.InputView;
var resultBufferView = gpu.OutputUlongView;
int acceleratorIndex = gpu.AcceleratorIndex;
AcceleratorStream stream = AcceleratorStreamPool.Rent(acceleratorIndex);
exponentBuffer.View.CopyFromCPU(stream, ref exponent, 1);
exponentBufferView.CopyFromCPU(stream, ref exponent, 1);
// We don't need to worry about any left-overs here.
// resultBuffer.MemSetToZero(stream);
var accelerator = gpu.Accelerator;
var kernelLauncher = gpu.ConvertToStandardKernelLauncher;
kernelLauncher(stream, 1, exponentBuffer.View, divisor.Modulus, divisor.NPrime, divisor.MontgomeryOne, divisor.MontgomeryTwo, divisor.MontgomeryTwoSquared, resultBuffer.View);
kernelLauncher(stream, 1, exponentBufferView, divisor.Modulus, divisor.NPrime, divisor.MontgomeryOne, divisor.MontgomeryTwo, divisor.MontgomeryTwoSquared, resultBufferView);
resultBuffer.View.CopyToCPU(stream, ref result, 1);
resultBufferView.CopyToCPU(stream, ref result, 1);
stream.Synchronize();
AcceleratorStreamPool.Return(acceleratorIndex, stream);