Which one works faster for downsampling?

Hello everyone. I am processing a long audio data in small durations. In the first step of each iteration, signal is downsampled. In the code, raw audio sample is cropped and stored in variable x. Then LPF and downsampling is applied to x. After downsampling, downsampled signal is stored in variable x. That is to say, in each iteration, size of variable x grows and shrinks. Instead of this, I can create separate variables for raw audio sample and downsampled audio sample. Which method is more efficient? Which one do you prefer? Thanks in advance for your answers.