Merge remote-tracking branch 'origin/ILGpu' into ILGpu

SDFMapCreator/Program.cs

@@ -148,7 +148,7 @@ public class Program {
         int iterCount = 0;
         var sw = new Stopwatch();
         sw.Start();
-        Vector3[,] temp = new Vector3[width, height];
+        /*Vector3[,] temp = new Vector3[width, height];
 
         var min = MAX;
         var max = MIN;
@@ -169,11 +169,12 @@ public class Program {
 
             if (gradient < min) min = gradient;
             if (gradient > max) max = gradient;
-        });
+        });*/
 
+        kernels.Gradient(mask.Mask, sdfA.SDF, sdfB.SDF, out var temp);
         Console.WriteLine(
             $"Gradient Time: {sw.Elapsed.TotalSeconds:N4}s ({iterCount / (float)sw.Elapsed.TotalSeconds:N0} pixels/s)");
-        Console.WriteLine($"Min: {min} | Max: {max}");
+        //Console.WriteLine($"Min: {min} | Max: {max}");
         return temp;
     }
 

SDFMapCreator/SdfKernels.Kernels.cs

@@ -69,5 +69,15 @@ public partial class SdfKernels {
        
         sdf[index] = new(minDist);
     }
-    
+
-}
+    private static void GradientKernel(Index2D index,
+        ArrayView2D<Vector3, Stride2D.DenseX> mask,
+        ArrayView2D<Vector3, Stride2D.DenseX> sdfa,
+        ArrayView2D<Vector3, Stride2D.DenseX> sdfb,
+        ArrayView2D<Vector3, Stride2D.DenseX> gradient
+    ) { //early exit if not on mask
+        if(mask[index].X == 0f || mask[index].Y == 0f) return;
+        float a = sdfa[index].X; float b = sdfb[index].X;
+        gradient[index] = new (a / (a + b));
+    }
+}

SDFMapCreator/SdfKernels.cs

@@ -128,6 +128,37 @@ public partial class SdfKernels {
         sdf = buffer.GetAsArray2D();
     }
     
+    public void Gradient(ref Vector3[,] mask, ref Vector3[,] sdfa, ref Vector3[,] sdfb, out Vector3[,] gradient) {
+        var dev = gpuContext.GetPreferredDevice(preferCPU:false);
+        int width = mask.GetLength(0);
+        int height = mask.GetLength(1);
+        gradient = new Vector3[width, height];
+        using Accelerator accelerator = dev.CreateAccelerator(gpuContext);
+        
+        using var bufferMask = accelerator.Allocate2DDenseX<Vector3>(new (width, height));
+        bufferMask.CopyFromCPU(mask);
+        
+        using var bufferSdfa = accelerator.Allocate2DDenseX<Vector3>(new (width, height));
+        bufferSdfa.CopyFromCPU(sdfa);
+        
+        using var bufferSdfb = accelerator.Allocate2DDenseX<Vector3>(new (width, height));
+        bufferSdfb.CopyFromCPU(sdfb);
+        
+        using var bufferGradient = accelerator.Allocate2DDenseX<Vector3>(new (width, height));
+        
+        var gradientKernel = accelerator.LoadAutoGroupedStreamKernel<Index2D,
+            ArrayView2D<Vector3, Stride2D.DenseX>,
+            ArrayView2D<Vector3, Stride2D.DenseX>,
+            ArrayView2D<Vector3, Stride2D.DenseX>,
+            ArrayView2D<Vector3, Stride2D.DenseX>>(GradientKernel);
+
+        gradientKernel(new (width, height), bufferMask.View, bufferSdfa.View, bufferSdfb.View, bufferGradient.View);
+        
+        accelerator.Synchronize();
+
+        gradient = bufferMask.GetAsArray2D();
+    }
+    
     private static string GetInfoString(Accelerator a)
     {
         StringWriter infoString = new StringWriter();