Analyzing Data Reusability of Raytrace Application in Splash2 Benchmark

When designing a chip multiprocessors, we use Splash2 to estimate its performance. This benchmark contains eleven applications. The performance when running them is similar, except Raytrace. We analyse it to clarity why the performance is not good. We discover, in theory, Raytrace never reuses data. This leads the fact that the performance is not good due to the low hit ratio in data cache.