From ec574cff822a1238138c0aa69b3d1459bdc3dfa8 Mon Sep 17 00:00:00 2001 From: Johannes Ranke Date: Fri, 19 Jun 2015 17:46:11 +0200 Subject: Use odeintr instead of ccSolve for compiling models --- vignettes/compiled_models.html | 63 +++++++++++++++++++----------------------- 1 file changed, 28 insertions(+), 35 deletions(-) (limited to 'vignettes/compiled_models.html') diff --git a/vignettes/compiled_models.html b/vignettes/compiled_models.html index 2f2a6edb..efdbe20d 100644 --- a/vignettes/compiled_models.html +++ b/vignettes/compiled_models.html @@ -77,37 +77,30 @@ img { -->

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 greater than 0.9-36 and the ccSolve package is installed and functional, you will get a message that the model is being compiled when defining a model using mkinmod.

+

This evaluation is taken from the example section of mkinfit. When using an mkin version greater or equal than 0.9-36 and the C++ compiler g++ is installed and functional (on Windows, install Rtools), you will get a message that the model is being compiled 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...
+ m1 = list(type = "SFO"), odeintr_compile = "yes") +
## 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")
+
smb.1 <- summary(mb.1)[-1]
+rownames(smb.1) <- c("deSolve, not compiled", "Eigenvalue based", "odeintr, compiled")
 print(smb.1)

 
##                             min        lq      mean    median        uq
-## deSolve, not compiled 6192.0125 6195.3470 6211.0309 6198.6816 6220.5401
-## Eigenvalue based       956.7604 1008.7224 1026.2572 1060.6844 1061.0055
-## deSolve, compiled      869.6880  871.9315  883.4929  874.1751  890.3953
+## deSolve, not compiled 5254.1030 5261.3501 5277.1074 5268.5973 5288.6096
+## Eigenvalue based       897.1575  921.6935  930.9546  946.2296  947.8531
+## odeintr, compiled      693.6001  709.0719  719.5530  724.5438  732.5295
 ##                             max neval
-## deSolve, not compiled 6242.3986     3
-## Eigenvalue based      1061.3266     3
-## deSolve, compiled      906.6155     3

-
We see that using the compiled model is almost a factor of 8 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"]

+## deSolve, not compiled 5308.6218     3
+## Eigenvalue based       949.4766     3
+## odeintr, compiled      740.5151     3
+

We see that using the compiled model is more than a factor of 7 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["odeintr, compiled", "median"]
##                         median
-## deSolve, not compiled 7.120877
-## Eigenvalue based      1.205328
-## deSolve, compiled     1.000000
+## deSolve, not compiled 7.290796 +## Eigenvalue based 1.370242 +## odeintr, compiled 1.000000

Benchmark for a model that can not be solved with Eigenvalues

@@ -115,25 +108,25 @@ print(smb.1) 
FOMC_SFO <- mkinmod(
   parent = list(type = "FOMC", to = "m1", sink = TRUE),
   m1 = list(type = "SFO"))
-
## Compiling differential equation model from auto-generated C code...
+
## 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),
+  mkinfit(FOMC_SFO, FOCUS_2006_D, solution_type = "deSolve", quiet = TRUE),
+  mkinfit(FOMC_SFO, FOCUS_2006_D, solution_type = "odeintr", quiet = TRUE),
   times = 3, control = list(warmup = 1))
 smb.2 <- summary(mb.2)[-1]
-rownames(smb.2) <- c("deSolve, not compiled", "deSolve, compiled")
+rownames(smb.2) <- c("deSolve, not compiled", "odeintr, compiled")
 print(smb.2)
##                             min        lq      mean    median        uq
-## deSolve, not compiled 13.297283 13.427702 13.481155 13.558121 13.573092
-## deSolve, compiled      1.486926  1.526887  1.546851  1.566848  1.576813
+## deSolve, not compiled 11.243675 11.324875 11.382415 11.406074 11.451785
+## odeintr, compiled      1.207114  1.209908  1.239989  1.212703  1.256426
 ##                             max neval
-## deSolve, not compiled 13.588063     3
-## deSolve, compiled      1.586778     3
-
smb.2["median"]/smb.2["deSolve, compiled", "median"]
+## deSolve, not compiled 11.497496 3 +## odeintr, compiled 1.300149 3 +
smb.2["median"]/smb.2["odeintr, compiled", "median"]
##                         median
-## deSolve, not compiled 8.653119
-## deSolve, compiled     1.000000
-

Here we get a performance benefit of more than a factor of 8 using the version of the differential equation model compiled from C code using the ccSolve package!

+## deSolve, not compiled 9.405494 +## odeintr, compiled 1.000000 +

Here we get a performance benefit of more than a factor of 8 using the version of the differential equation model compiled from C++ code using the odeintr package!

-- cgit v1.2.1