The convolution tools have polymorphic instances for 1D and 2D data.
The 2D instances have a very useful input to control the output size (full, size X, compact).
The 1D instances don't have this input (Why?!). In the vast majority of my 1D convolution applications, I would prefer a "size x" behavior for the output, making the output size identical to the input spectrum, no matter the size of the convolution kernel. If the 1D convolutions would accept this option, performance and inplaceness could possibly be optimized compared to a manual trimming later.
All convolution instances need an "output size" input, not just the 2D versions.