Remove linearized index calculation

example/convolution1D/src/convolution1D.cpp

@@ -55,24 +55,22 @@ struct ConvolutionKernel
         Vec const globalThreadIdx = alpaka::getIdx<alpaka::Grid, alpaka::Threads>(acc);
         Vec const globalThreadExtent = alpaka::getWorkDiv<alpaka::Grid, alpaka::Threads>(acc);
 
-        // Map the three dimensional thread index into a
-        // one dimensional thread index space
-        auto const linearizedGlobalThreadIdx = alpaka::mapIdx<1u>(globalThreadIdx, globalThreadExtent)[0];
-
-        if(linearizedGlobalThreadIdx < static_cast<unsigned>(inputSize))
+        // Since the kernel is launched 1-D calculating linearizedGlobalThreadIdx line is unnecessary.
+        // globalThreadIdx[0] can be used to map all the threads.
+        if(globalThreadIdx[0] < static_cast<unsigned>(inputSize))
         {
             uint32_t const halfFilterSize = filterSize / 2;
             TElem result = 0.0f;
             // Calculate sum of multiplications of corresponding elements
             for(size_t i = 0; i < filterSize; ++i)
             {
-                int inputIndex = linearizedGlobalThreadIdx - halfFilterSize + i;
+                int inputIndex = globalThreadIdx[0] - halfFilterSize + i;
                 if(inputIndex >= 0 && inputIndex < inputSize)
                 {
                     result += input[inputIndex] * filter[i];
                 }
             }
-            output[linearizedGlobalThreadIdx] = result;
+            output[globalThreadIdx[0]] = result;
         }
     }
 };
@@ -178,7 +176,7 @@ auto main() -> int
     alpaka::wait(queue);
 
     // Allocate memory on host
-    BufAcc resultGpuHost = alpaka::allocBuf<DataType, Idx>(devHost, InputSize);
+    auto resultGpuHost = alpaka::allocBuf<DataType, Idx>(devHost, InputSize);
     // Copy from device memory to host
     alpaka::memcpy(queue, resultGpuHost, outputDeviceMemory, InputSize);