# 產生一百萬個常態變數值
num <- rnorm(1e6)
# 建一個要輸出的matrix dimention的list
struct <- list(c(500, 900), c(500, 100), c(250, 350), c(550, 750))

st <- proc.time()
# 先把數字切成你要的length (#rows * #columns)
out <- split(num, rep(seq_along(struct), sapply(struct, Reduce, f = `*`)))
# 把每個list改成你要的dimension
for (i in seq_along(struct))
  dim(out[[i]]) <- struct[[i]]
proc.time() - st
# user  system elapsed
# 0.17    0.01    0.18

# 確認結果跟原PO想要的一樣
all(unlist(out) == num) # TRUE