Generate multivariate normal correlated variables from random number input.

I would like a PC based app to allow me to input random numbers and generate a multivariate normal distribution such that the variables have intercorrelations that I specify. I would like to specify the desired intercorrelation between each of the pairs of variables and have the app generate multivariate normal variables with specified intercorrelations. I would like to specify the intercorrelations using an excel or CSV file. For example, I would input 1 million lines with random numbers for 5 variables, perhaps from [login to view URL], and specify that the correlation between var 1 and var 2 is .3, the correlation between var 1 and var 3 is .4, and the correlation between var 2 and var 3 is .1, and all other intercorrelations are zero. The random numbers from [login to view URL] cost less if they are uniform but I need normal. So it would be best if the app would offer the option of inputting uniform dist or normal dist while always outputting multivariate normal. I expect to be using data sets up to 10 million cases, perhaps a little more, and up to 10 variables. I would like a description of the mathematics that the program uses. The attached pdf page has some information on how to do this, statistically speaking.