Remove duplicative kernel

Author: orionpapadakis

src/main/java/com/example/tornadovm/Qwen3Kernels.java

@@ -296,70 +296,4 @@ private static void processHeadTornado(
         }
     }
 
-    public static void matrixVectorGenericWithResidual(
-            KernelContext context,
-            FloatArray v,
-            FloatArray out,
-            HalfFloatArray m,
-            int dim1,
-            int dim0,
-            int localWorkGroupSize) {
-
-        // One row per workgroup (not per thread)
-        int rowId = context.groupIdx;
-        int localId = context.localIdx;
-        int localSize = localWorkGroupSize;
-
-        // Early exit if this workgroup is beyond our output dimension
-        if (rowId >= dim0) {
-            return;
-        }
-
-        float sum = matrixVectorRowMajorOptimized(context, localSize, v, m, dim1, dim0);
-
-        // Thread 0 in each workgroup writes the final result
-        if (localId == 0) {
-            float result = out.get(rowId) + sum;
-            out.set(rowId, result);
-        }
-    }
-
-    public static float matrixVectorRowMajorOptimized(
-            KernelContext context,
-            int localSize,
-            FloatArray v,
-            HalfFloatArray m,
-            int dim1,
-            int dim0
-    ) {
-        int rowId = context.groupIdx;
-        int localId = context.localIdx;
-
-        // Allocate local memory for reduction
-        float[] localSum = context.allocateFloatLocalArray(localSize);
-
-        int rowOffset = rowId * dim1;
-
-        // Each thread calculates partial dot product
-        float partialSum = 0.0f;
-        for (int j = localId; j < dim1; j += localSize) {
-            int matrixIdx = rowOffset + j;
-            partialSum += m.get(matrixIdx).getFloat32() * v.get(j);
-        }
-
-        // Store partial sum in local memory
-        localSum[localId] = partialSum;
-        context.localBarrier();
-
-        // Parallel reduction within workgroup
-        for (int stride = localSize / 2; stride > 0; stride >>= 1) {
-            if (localId < stride) {
-                localSum[localId] += localSum[localId + stride];
-            }
-            context.localBarrier();
-        }
-
-        return localSum[0];
-    }
-
 }

src/main/java/com/example/tornadovm/Qwen3TornadoVMLayerPlanner.java

@@ -221,7 +221,7 @@ public Tuple2<List<ImmutableTaskGraph>, GridScheduler> setupTornadoForwardPlanLa
                     layerIndex,
                     config.contextLength());
 
-            unifiedLayer.task("matmul1", Qwen3Kernels::matrixVectorGenericWithResidual,
+            unifiedLayer.task("matmul1", TransformerComputeKernelsLayered::matrixVectorGenericWithResidual,
                     context,
                     state.wrapXb,                           // vector
                     state.wrapX,                            // out, should be [1024]