library(microbenchmark)
library(tensorEVD)
library(fastmatrix)
library(ggplot2)

# 1. Simulating a small Kronecker product
m = 50; n = 75
p = 40; q = 60
A <- matrix(rnorm(m*n), ncol=n)
B <- matrix(rnorm(p*q), ncol=q)

res1 <- microbenchmark(
         'kronecker\n(base)'            = kronecker(A,B),
         'kronecker.prod\n(fastmatrix)' = kronecker.prod(A,B),
         'Kronecker\n(tensorEVD)'       = Kronecker(A,B),
       times = 30)

tt1 <- paste0("'Kronecker('*A[",m,"*'x'*",n,"]*', '*B[",p,"*'x'*",q,"]*')'==K[",m*p,"*'x'*",n*q,"]")

# 2. Simulating a large Kronecker product
m = 100; n = 150
p = 80; q = 120
A <- matrix(rnorm(m*n), ncol=n)
B <- matrix(rnorm(p*q), ncol=q)

res2 <- microbenchmark(
         'kronecker\n(base)'            = kronecker(A,B),
         'kronecker.prod\n(fastmatrix)' = kronecker.prod(A,B),
         'Kronecker\n(tensorEVD)'       = Kronecker(A,B),
      times = 30)

tt2 <- paste0("'Kronecker('*A[",m,"*'x'*",n,"]*', '*B[",p,"*'x'*",q,"]*')'==K[",m*p,"*'x'*",n*q,"]")

# 3. Making a barplot
res <- rbind(data.frame(group=tt1, expr=res1$expr, time=res1$time/1E9),
             data.frame(group=tt2, expr=res2$expr, time=res2$time/1E9))
dat <- aggregate(time~expr+group, data=res, FUN=median)

pp <- ggplot(dat, aes(expr,time)) + theme_bw() +
      geom_bar(stat="identity", fill='lightyellow3', width=0.6) +
      labs(x=NULL, y="Time (seconds)") +
      facet_wrap(~group, scales="free", labeller=label_parsed)
print(pp)