Remove redundant zeroing

[danielzgtg]

std::vector::resize already zeroes the new samples, so we don't need to use std::fill. See https://stackoverflow.com/a/73333009/10477326

However, we pass a second parameter to explicitly copy-init with zero
and to not default-init. Benchmarks showed the performance difference
between the two to be statistically insignificant, despite default-init
having memset.

[adamstark]

Hi there, thanks for this! I think it depends slightly because the type of the vector depends on the class template type T, which is very very likely to be a float or double, or an integer of some type, but in theory one could use it with a different parameter for T (as long as that parameter fulfilled the requirements for being assigned values etc).

Perhaps the cleaner way to do this is to use the second parameter of resize? I.e.

samples[i].resize (numSamples, static_cast (0));

just to explicitly make clear we want to zero-initialise any new samples?

[danielzgtg]

@adamstark Fixed

Despite my initial assumption default-init would be faster because of memset, benchmarking your method showed no difference.

[adamstark]

Thanks for this! Merging in now...