Initial project import

examples/compute_demo/compute_demo.pas

@@ -0,0 +1,61 @@
+program AgxComputeDemo;
+
+{$mode ObjFPC}{$H+}
+
+uses
+  SysUtils, AgxOpenTypes, AgxOpenRuntime;
+
+const
+  N = 8;
+
+type
+  TFloatArray = array[0..N - 1] of Single;
+
+procedure VectorAdd(const Inputs, Outputs: array of TAgxBuffer;
+  const ElementCount: NativeUInt);
+var
+  A, B, C: TFloatArray;
+  I: NativeUInt;
+begin
+  Inputs[0].Download(@A[0], ElementCount * SizeOf(Single));
+  Inputs[1].Download(@B[0], ElementCount * SizeOf(Single));
+  for I := 0 to ElementCount - 1 do
+    C[I] := A[I] + B[I];
+  Outputs[0].Upload(@C[0], ElementCount * SizeOf(Single));
+end;
+
+var
+  Device: TAgxDevice;
+  ABuffer, BBuffer, CBuffer: TAgxBuffer;
+  Kernel: TAgxKernel;
+  A, B, C: TFloatArray;
+  I: Integer;
+begin
+  for I := 0 to N - 1 do
+  begin
+    A[I] := I;
+    B[I] := I * 10;
+  end;
+
+  Device := TAgxDevice.Create;
+  ABuffer := nil; BBuffer := nil; CBuffer := nil; Kernel := nil;
+  try
+    Device.CreateBuffer(SizeOf(A), agxBufferStorage, ABuffer);
+    Device.CreateBuffer(SizeOf(B), agxBufferStorage, BBuffer);
+    Device.CreateBuffer(SizeOf(C), agxBufferStorage, CBuffer);
+    ABuffer.Upload(@A[0], SizeOf(A));
+    BBuffer.Upload(@B[0], SizeOf(B));
+    Device.CreateCPUKernel('vector_add', @VectorAdd, Kernel);
+    Kernel.Dispatch([ABuffer, BBuffer], [CBuffer], N);
+    CBuffer.Download(@C[0], SizeOf(C));
+
+    for I := 0 to N - 1 do
+      WriteLn(Format('%0.0f + %0.0f = %0.0f', [A[I], B[I], C[I]]));
+  finally
+    Kernel.Free;
+    CBuffer.Free;
+    BBuffer.Free;
+    ABuffer.Free;
+    Device.Free;
+  end;
+end.