In software development there is an unavoidable trend in which applications become larger and more complex. For compiled programming languages one of the consequences is that their compilation takes longer.

Today I want to talk about using distcc to speed C/C++ compilation using different nodes in a scientific cluster.

I think we have enough pieces of machinery working already that we can start with the most exciting part of this journey: autovectorisation!

In the last installment I mentioned we could start looking at enabling the vectoriser in the compiler. However when I did that I realised some benchmarks were giving weird results. I had made a mistake with copies, so let’s remediate this.

We finished the last installment of this series mentioning that the compiler cannot copy, load or store a vector. Today we will address this.

There is an issue we have mentioned several times in earlier installments: the value of the vector length at function boundaries. This is, when entering or leaving a function. We will address this question today.