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

BitContradiction GPU performance improvements. Local variables partially optimized to limit registry pressure.

Browse Source
This commit is contained in:
mlemanczyk committed 2026-01-26 15:02:59 +01:00
1 parent 45f21efa56
commit b31aab4d89
4 files changed
+265 -288
Download Patch File Download Diff File Expand all files Collapse all files

No files matched your search

PerfectNumbers.Core/Gpu/GpuKernelPool.cs
+10 -3
View File
@@ -125,13 +125,20 @@ public class GpuKernelPool
{
var loaded = accelerator.LoadAutoGroupedStreamKernel<
Index1D, ulong, ArrayView1D<ulong, Stride1D.Dense>, int, ArrayView1D<int, Stride1D.Dense>,
ArrayView1D<ulong, Stride1D.Dense>, ArrayView1D<int, Stride1D.Dense>, ArrayView1D<int, Stride1D.Dense>
#if DETAILED_LOG
ArrayView1D<ulong, Stride1D.Dense>,
#endif
ArrayView1D<int, Stride1D.Dense>, ArrayView1D<int, Stride1D.Dense>
>(BitContradictionKernels.BitContradictionKernelScanWithDeltaCache64);
var kernel = KernelUtil.GetKernel(loaded);
return kernel.CreateLauncherDelegate<Action<AcceleratorStream, Index1D, ulong, ArrayView1D<ulong, Stride1D.Dense>, int, ArrayView1D<int, Stride1D.Dense>,
ArrayView1D<ulong, Stride1D.Dense>, ArrayView1D<int, Stride1D.Dense>, ArrayView1D<int, Stride1D.Dense>>>();
return kernel.CreateLauncherDelegate<Action<AcceleratorStream, Index1D, ulong, ArrayView1D<ulong, Stride1D.Dense>, int,
ArrayView1D<int, Stride1D.Dense>,
#if DETAILED_LOG
ArrayView1D<ulong, Stride1D.Dense>,
#endif
ArrayView1D<int, Stride1D.Dense>, ArrayView1D<int, Stride1D.Dense>>>();
});
}