Collapse GPU by-divisor batching to a single launch

PerfectNumbers.Core/Gpu/Kernels/DivisorByDivisorKernels.cs

@@ -86,105 +86,6 @@ internal static class DivisorByDivisorKernels
         }
     }
 
-    public static void GenerateFilteredCandidatesKernel(
-        Index1D index,
-        ArrayView<ulong> candidates,
-        ArrayView<int> filteredCount,
-        ArrayView<GpuUInt128> nextStarts,
-        GpuUInt128 startValue,
-        GpuUInt128 stride,
-        byte lastIsSevenFlag)
-    {
-        int globalIndex = index;
-        int localIndex = Group.IdxX;
-        int groupSize = Group.Dimension.X;
-
-        ulong candidateOffset = (ulong)globalIndex;
-        GpuUInt128 candidateValue = stride;
-        candidateValue.Mul(candidateOffset);
-        candidateValue.Add(startValue);
-        ulong candidate = candidateValue.Low;
-        candidates[globalIndex] = candidate;
-
-        if (globalIndex == 0)
-        {
-            ulong candidateCount = (ulong)groupSize;
-            GpuUInt128 nextValue = stride;
-            nextValue.Mul(candidateCount);
-            nextValue.Add(startValue);
-            nextStarts[0] = nextValue;
-        }
-
-        bool lastIsSeven = lastIsSevenFlag != 0;
-        byte remainder10 = (byte)(candidate % 10UL);
-        bool acceptSeven = remainder10 == 3 || remainder10 == 7 || remainder10 == 9;
-        bool acceptNonSeven = remainder10 == 1 || remainder10 == 3 || remainder10 == 7 || remainder10 == 9;
-        bool accept10 = lastIsSeven ? acceptSeven : acceptNonSeven;
-        byte remainder8 = (byte)(candidate & 7UL);
-        bool accept8 = remainder8 == 1 || remainder8 == 7;
-        byte remainder5 = (byte)(candidate % 5UL);
-        bool accept5 = remainder5 != 0;
-        byte remainder3 = (byte)(candidate % 3UL);
-        bool accept3 = remainder3 != 0;
-        int accepted = (accept10 && accept8 && accept3 && accept5) ? 1 : 0;
-
-        var shared = SharedMemory.GetDynamic<int>();
-        int length = groupSize;
-        int prefixBase = 0;
-        int mappingBase = length;
-        int fallbackBase = length * 2;
-
-        shared[prefixBase + localIndex] = accepted;
-        shared[fallbackBase + localIndex] = globalIndex;
-
-        Group.Barrier();
-
-        int scanOffset = 1;
-        while (scanOffset < groupSize)
-        {
-            int addendIndex = prefixBase + localIndex - scanOffset;
-            int addend = localIndex >= scanOffset ? shared[addendIndex] : 0;
-
-            Group.Barrier();
-
-            int currentIndex = prefixBase + localIndex;
-            int current = shared[currentIndex];
-            int sum = current + addend;
-            int shouldUpdate = localIndex >= scanOffset ? 1 : 0;
-            shared[currentIndex] = shouldUpdate != 0 ? sum : current;
-
-            Group.Barrier();
-
-            scanOffset <<= 1;
-        }
-
-        int inclusivePrefix = shared[prefixBase + localIndex];
-        int exclusiveIndex = inclusivePrefix - accepted;
-        int targetBase = accepted != 0 ? mappingBase : fallbackBase;
-        int targetOffset = accepted != 0 ? exclusiveIndex : localIndex;
-        shared[targetBase + targetOffset] = globalIndex;
-
-        Group.Barrier();
-
-        int totalAccepted = shared[prefixBase + length - 1];
-        int writerMask = localIndex < totalAccepted ? 1 : 0;
-        int lookupBase = writerMask != 0 ? mappingBase : fallbackBase;
-        int sourceIndex = shared[lookupBase + localIndex];
-
-        ulong candidateValueToStore = candidates[sourceIndex];
-        if (writerMask != 0)
-        {
-            candidates[localIndex] = candidateValueToStore;
-        }
-
-        int lastLaneMask = globalIndex == length - 1 ? 1 : 0;
-        int countValue = shared[prefixBase + localIndex];
-        if (lastLaneMask != 0)
-        {
-            filteredCount[0] = countValue;
-        }
-    }
-
     public static void ConvertMontgomeryResultsToHitsKernel(
         Index1D index,
         ArrayView<ulong> results,