aboutsummaryrefslogtreecommitdiff
path: root/vignettes/compiled_models.Rmd
blob: 3b24acbbf968e7571b7cbe15f63c22ec93d1a253 (plain) (blame)
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
---
title: "Performance benefit by using compiled model definitions in mkin"
output:
  html_document:
    css: mkin_vignettes.css
    toc: true
    mathjax: null
    theme: united
---
<!--
%\VignetteEngine{knitr::rmarkdown}
%\VignetteIndexEntry{Performance benefit by using compiled model definitions in mkin}
-->

```{r, include = FALSE}
library(knitr)
opts_chunk$set(tidy = FALSE, cache = TRUE)
```

# 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.

```{r create_SFO_SFO}
library("mkin")
SFO_SFO <- mkinmod(
  parent = list(type = "SFO", to = "m1", sink = TRUE),
  m1 = list(type = "SFO"))
```

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.


```{r benchmark_SFO_SFO, echo=-(1:2)}
# Redefining the model, in order not to confuse the knitr cache which leads to segfaults
suppressMessages(SFO_SFO <- mkinmod(
  parent = list(type = "SFO", to = "m1", sink = TRUE),
  m1 = list(type = "SFO")))
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)
```

We see that using the compiled model is by a factor of 
`r round(smb.1["deSolve, not compiled", "median"]/smb.1["deSolve, compiled", "median"], 1)` 
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:

```{r}
smb.1["median"]/smb.1["deSolve, compiled", "median"]
```

# Benchmark for a model that can not be solved with Eigenvalues

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

```{r benchmark_FOMC_SFO}
FOMC_SFO <- mkinmod(
  parent = list(type = "FOMC", to = "m1", sink = TRUE),
  m1 = list(type = "SFO"))

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)
smb.2["median"]/smb.2["deSolve, compiled", "median"]

```

Here we get a performance benefit of a factor of 
`r round(smb.2["deSolve, not compiled", "median"]/smb.2["deSolve, compiled", "median"], 1)` 
using the version of the differential equation model compiled from C code using
the inline package!

Contact - Imprint