November 2017
Intermediate to advanced
398 pages
10h 14m
English
One of the common problems that may occur during a microbenchmark operation is the wrong use of the available memory. The problem occurs especially when testing out small code elements. These elements directly load into the fastest L1 cache, which ensures that the program immediately reaches the CPU on a preferential basis.
However, when the same code performs in a situation where the dataset contained within program objects is larger than the size of the L1 cache, the actual program runs much slower due to using the conventional memory. This means that the microbenchmark results will not depict the original program situation and optimization would not be the best solution.