You can approximate it and throw away some samples to get the true distribution. See https://en.wikipedia.org/wiki/Ziggurat_algorithm https://en.wikipedia.org/w

You can approximate it and throw away some samples to get the true distribution. See https://en.wikipedia.org/wiki/Ziggurat_algorithm https://en.wikipedia.org/wiki/Ziggurat_algorithm for one method which I believe statistical languages like R currently use to generate normal random variables.