要获得运行每个矩阵转换所需的时间，您可以设置一个表达式列表并使用eval运行这些表达式。这对我来说有点快速和肮脏的方法-但它可能是一个解决方案...

list.of.functions = list(
 matrix1 = expression(cor(matrix(rnorm(100), ncol = 10))), # Quick
 matrix2 = expression(cor(matrix(rnorm(1000), ncol = 100))), # Slow
 matrix3 = expression(cor(matrix(rnorm(10000), ncol = 1000))) # Slower
)
elapsed = list.of.functions # Copy the size and anes (R takes care of casting to the new data type)
for(f in names(list.of.functions)){
    elapsed[[f]] = system.time({eval(list.of.functions[[f]])})["elapsed"]
}
do.call("rbind", elapsed)

如果将这样的输出保存为data.frame，每次迭代都有一列，则可以在其上运行rowMeans，以获得每个矩阵的平均运行时间。

# Set up empty data.frame for the timing results
elapsed.df = as.data.frame(matrix(NA, nrow = length(list.of.functions), ncol = n))
rownames(elapsed.df) = names(list.of.functions)
# Outer loop of iterations
for(i in seq_len(n)){
    # Inner loop of function evaluations
    for(f in names(list.of.functions)){
        elapsed.df[f, i] = system.time({eval(list.of.functions[[f]])})["elapsed"]
    }
}
rowMeans(elapsed.df)

没有最少的例子很难更详细地解释。