Provide enough parallelism for benchmarking.

test/Benchmark/Benchmark.cc

@@ -52,7 +52,7 @@ Benchmark::Benchmark()
 template <typename SampleType, typename VisibilityType> void Benchmark::doTest(unsigned nrBits, unsigned nrReceiversPerBlock, unsigned nrReceivers)
 {
   constexpr double   measureTime         = 3; // seconds
-  constexpr unsigned nrChannels          = 480;
+	    unsigned nrChannels          = 4 * device.getAttribute<CU_DEVICE_ATTRIBUTE_MULTIPROCESSOR_COUNT>(); // provide enough parallelism
   constexpr unsigned nrPolarizations     = 2;
   constexpr unsigned nrSamplesPerChannel = 3072;
   constexpr bool     addVisibilities     = false;