Gantt chart

[aafshinfard]

an example run to test it:

Rscript -e 'rmarkdown::render("/projects/btl_scratch/aafshinfard/projects/physlr/port_code/sqcos/physlr/data/profile-physlr.Rmd", "html_document", "/projects/btl_scratch/aafshinfard/projects/physlr/port_code/sqcos/physlr/data/test.prof.html", knit_root_dir="/projects/btl_scratch/aafshinfard/projects/physlr/port_code/sqcos/physlr/data", params=list(path="/projects/btl_scratch/aafshinfard/projects/physlr/port_code/sqcos/physlr/data/"))'

you can change the path to any path you like to test it with other files.

[jwcodee]

Can you add a makefile command to generate the Gantt chart?

[aafshinfard]

Sure @jowong4, I'll just do it.

[aafshinfard]

Pufff I was struggling to make the makefile command work! and the problem was the formatting of spaces and newlines!

Now it works, try make {NAME}.prof.html path_profile={PATH} and use a name prefix you like and determine the path for *.time files. by defaults it is set to data path for physlr

[jwcodee]

Hmm, are the ram values correct? It doesn't quite look correct.

[jwcodee]

I would suggest making the current directory the default directory. You can probably do that by assining pwd to a variable and using that.

[aafshinfard]

@jowong4 The ram values I extract from the .time file, so in case of python with multiprocessing it's the ram usage for only a single process, I can simply fix that by multiplying the ram usage with number of threads if the language is python, but I think it's better not to scale it up so we can see other bars easily. also note that ram usage is converted into GB and also rounded (ceiling). So is there still any other problem you see with the reports?

[jwcodee]

Just change the green like we suggested in the meeting. Thanks!

[lcoombe]

@afshinfard - If you had different physlr runs in the same directory, would this script be able to distinguish the runs for the Gantt chart?

[aafshinfard]

@lcoombe In that situation, the code orders all the *.time files and if there are multiple files for a single step it will keep the one which is older. so if you have two different runs in a single directory it will report the first run. One simple way to profile different runs is to mv the files you want to a subdirectory and run this on that subdirectory.