In the previous section, we first saw the difference between the native compilation and cross-compilation of a user-space program. In the preceding example, it's quite obvious that the native compilation wins, but as already stated, there are some circumstances where the cross-compilation is a must.

One of these cases is the kernel compilation.

In the older days of embedded programming, when the embedded systems were really small in the sense of limited resources (from 4 MB to 32 MB of RAM memory and 32 MB to 128 MB mass storage memory), recompiling the kernel natively was considered foolish. However, day by day and year by year, the embedded systems became more powerful, and now we have 1 GB or 2 GB of RAM memory and 4 GB to 32 GB of mass storage, which makes it easier to recompile, even if its not much of a common practice. So, nowadays, considering recompiling the whole kernel can be possible, but we've to take into account that compiling the BeagleBone Black's kernel...