run times

[Basquill]

hi @rvalavi. Not sure if you want feedback on this tool. Any insight would be appreciated.

• Note, I have openMP on my desktop. It's in a folder for the Rcpp package. That package isn't a dependency for gdmtools though, so I don't know if openMP is being called upon.

I ran 2-3 trials today. One minor error when I used a terra::vect() to bring in a protected area shapefile. That kicked an error when I ran par_index()

• Error in par_index(x, pa, mod, n = NULL, ncores = -1, filename = "test") :
• pa must be a vector, raster or a data frame of GDM transform values for protected areas.

So, I created a dataframe of GDM-transformed values from the pa, and that seemed to work.

To test this function, I simplified my data.

---

First, I first created both 1 & 3 layer subsets of GDM-transformed rasters. My full GDM model had been fit with 6 layers. I then created pa object with 1 and 3 protected areas.

[I have 20 cores on my computer, but have found it's best not to call them all.]

My first trial (3 layer raster; 3 protected areas)

• I checked the memory require to process raster
  mem_info()

Memory (GB) (for raster)

check threshold : 1 (memmin)
available : 19.77
allowed (50%) : 9.88
needed (n=1) : 34.16
proc in memory : FALSE
nr chunks : 7

Then I stipulated the intercept from my fitted GDM model

mod <- 0.209

• test par_index function

par_index(x, pa.gdm.values, mod, n = NULL, ncores = -1, filename = "test") # this ran for 4.5 hours and hadn't finished.

---

My second trial (1 layer raster; 1 protected area):

• memory require to process raster
  mem_info()

Memory (GB)
check threshold : 1 (memmin)
available : 19.77
allowed (50%) : 9.89
needed (n=1) : 17.08
proc in memory : FALSE
nr chunks : 4

mod <- 0.209 -- intercept from fitted model

• test par_index function

par_index(x, pa.gdm.values, mod, n = NULL, ncores = 4, filename = "test") # this ran for 2.5 hours and hadn't finished

Based on your tests, is there anything that you'd do differently? Thanks very much