Fix Performance of 'by' Operations when verbose=TRUE

[joshhwuu]

Closes #6286

This PR changes calls to C's clock() to internal wallclock() in dogroups.c. Apparently calling clock() repeatedly on non-Windows platforms used to cause a significant overhead, leading to significantly longer processing times. I found this SO thread that discusses some of the differences between C's clock() on different platforms.

• Add a performance test

[github-actions]

Generated via commit fc0c1e7

Download link for the artifact containing the test results: ↓ atime-results.zip

Time taken to finish the standard R installation steps: 12 minutes and 1 seconds

Time taken to run atime::atime_pkg on the tests: 3 minutes and 51 seconds

[joshhwuu]

@tdhock @Anirban166 I'm trying to run the suite of atime tests locally so I can make sure everything is working before I add a new test, however I'm having trouble with running it based on instructions in https://github.com/Rdatatable/data.table/wiki/Performance-testing

I tried to use atime::atime_pkg(getwd(),".ci"), which gives me a host of C level errors associated with the first commit listed in the test list:

ERROR: compilation failed for package ‘data.table.b1b1832b0d2d4032b46477d9fe6efb15006664f4’
* removing ‘/home/joshhwuu/R/x86_64-pc-linux-gnu-library/4.4/data.table.b1b1832b0d2d4032b46477d9fe6efb15006664f4’
Error in atime_versions_install(Package, normalizePath(pkg.path), new.Package.vec,  : 
  '/usr/lib/R/bin/R' CMD INSTALL -l '/home/joshhwuu/R/x86_64-pc-linux-gnu-library/4.4' /tmp/RtmpyGd6uW/file1d8c26f36015d/data.table.b1b1832b0d2d4032b46477d9fe6efb15006664f4 returned error status code 1

Not sure why this is happening, can anyone chime in?

[Anirban166]

I created a performance test for this - Result

Could probably do with a better/succinct name for the title (open to suggestions)

I'm not using a 'Fast' label (as displayed in the latest plot) since it would only exist after this PR is merged onto master here and is available to be installed using the commit SHA associated with it (after this is merged, I can create a follow-up PR for the test). Doesn't seem to be the case where this was previously fast so 'Before' and related labels won't fit. I initially tested 1.15.4 vs 1.15.99 for checking on this - no change (I assume you switched operating systems in between the process of testing other versions yesterday?) so likely not a regression.

[tdhock]

Not sure why this is happening, can anyone chime in?

please file an issue with additional details (OS, full input/output, traceback) https://github.com/tdhock/atime/issues

[joshhwuu]

no change (I assume you switched operating systems in between the process of testing other versions yesterday?) so likely not a regression.

Yup, I confirmed this here