The speed of memory: the numbers
Now that we have our benchmarking code to measure the speed of reading and writing into memory, we can collect the results and see how we can get the best performance when accessing data in memory. We begin with random access, where the location of each value we read or write is unpredictable.
The speed of random memory access
The measurements are likely to be fairly noisy unless you run this benchmark many times and average the results (the benchmark library can do that for you). For a reasonable run time (minutes), you will likely see the results that look something like this:
Figure 4.3 – Random read speed as a function of memory size
The benchmark results in Figure 4.3 show the number of words read from memory per second (in billions, on any reasonable PC or workstation you can find today), where the word is a 64-bit integer or a 265-bit integer (long or __m256i, respectively). The same measurements can be alternatively...
