Benchmark for a model that can also be solved with Eigenvalues

This evaluation is taken from the example section of mkinfit. When using an mkin version equal to or greater than 0.9-36 and a compiler (gcc) is installed, you will see a message that the model is being compiled from autogenerated C code when defining a model using mkinmod.

library("mkin")
SFO_SFO <- mkinmod(
  parent = list(type = "SFO", to = "m1", sink = TRUE),
  m1 = list(type = "SFO"))
## Compiling differential equation model from auto-generated C code...

We can compare the performance of the Eigenvalue based solution against the compiled version and the R implementation of the differential equations using the microbenchmark package.

library("microbenchmark")
mb.1 <- microbenchmark(
  mkinfit(SFO_SFO, FOCUS_2006_D, solution_type = "deSolve", use_compiled = FALSE, 
          quiet = TRUE),
  mkinfit(SFO_SFO, FOCUS_2006_D, solution_type = "eigen", quiet = TRUE),
  mkinfit(SFO_SFO, FOCUS_2006_D, solution_type = "deSolve", quiet = TRUE),
  times = 3, control = list(warmup = 1))
smb.1 <- summary(mb.1)[-1]
rownames(smb.1) <- c("deSolve, not compiled", "Eigenvalue based", "deSolve, compiled")
print(smb.1)
##                             min        lq      mean    median        uq
## deSolve, not compiled 6980.8002 6996.4739 7024.5611 7012.1476 7046.4415
## Eigenvalue based       925.3350  928.9405  951.8405  932.5460  965.0932
## deSolve, compiled      747.2635  761.9405  771.4339  776.6174  783.5191
##                             max neval
## deSolve, not compiled 7080.7354     3
## Eigenvalue based       997.6404     3
## deSolve, compiled      790.4207     3

We see that using the compiled model is by a factor of 9 faster than using the R version with the default ode solver, and it is even faster than the Eigenvalue based solution implemented in R which does not need iterative solution of the ODEs:

smb.1["median"]/smb.1["deSolve, compiled", "median"]
##                         median
## deSolve, not compiled 9.029089
## Eigenvalue based      1.200779
## deSolve, compiled     1.000000

Benchmark for a model that can not be solved with Eigenvalues

This evaluation is also taken from the example section of mkinfit.

FOMC_SFO <- mkinmod(
  parent = list(type = "FOMC", to = "m1", sink = TRUE),
  m1 = list(type = "SFO"))
## Compiling differential equation model from auto-generated C code...
mb.2 <- microbenchmark(
  mkinfit(FOMC_SFO, FOCUS_2006_D, use_compiled = FALSE, quiet = TRUE),
  mkinfit(FOMC_SFO, FOCUS_2006_D, quiet = TRUE),
  times = 3, control = list(warmup = 1))
smb.2 <- summary(mb.2)[-1]
rownames(smb.2) <- c("deSolve, not compiled", "deSolve, compiled")
print(smb.2)
##                             min        lq      mean    median        uq
## deSolve, not compiled 14.127630 14.245064 14.298201 14.362497 14.383486
## deSolve, compiled      1.354744  1.362167  1.366362  1.369589  1.372171
##                             max neval
## deSolve, not compiled 14.404474     3
## deSolve, compiled      1.374752     3
smb.2["median"]/smb.2["deSolve, compiled", "median"]
##                         median
## deSolve, not compiled 10.48672
## deSolve, compiled      1.00000

Here we get a performance benefit of a factor of 10.5 using the version of the differential equation model compiled from C code using the inline package!