Recursive forecasting alternative to looping.

I am working on some out-of-sample forecasts where i utilize a recursive forecast method.

Currently I conduct the forecast using a loop, where i save the forecast values in a seperate vector, and add one observation everytime the loop runs.

In the code I run regressions for 3 return measures, and 7 seperate variables, yeilding 21 regressions.

However this is computational very heavy, since I also conduct bootstrapping to obtain p-values for my test statistics.

My in-sample tests runs fairly quickly in a similar code, so I expect the bottleneck is the loops that conducts the recursive forecasts.

Do any one have suggestions on how I could do this differently?

Is there an alternative way, instead of a loop, to conduct recursive foresating