- May 20, ‘24:
1) Working on the changes Toby suggested in #7. (7 hours)
2) Working on a blog page for lessons on code improvements learned by following Toby’s guide. (2 hours)
- May 21, ‘24:
1) Continuining work on the points above. (6 hours)
2) Hosted and led zoom meetings with GSoC students (data.table projects) apart from Doris. (2 hours)
- May 22, ‘24:
1) Discussed improvements to make and future work with Toby and then talked with Doris regarding code issues and progress. (3 hours)
2) Introduced new arguments to benchmarking functions (number of times the benchmark is to be run + an option to toggle detailed messaging on/off) + small improvements such as not exporting runBenchmarks() and using namespace in the examples for what Toby suggested, changes to the plot code to make share colors between line and points of the same speedup category. (6 hours)
- May 23, ‘24:
1) Simplified code for runBenchmarks and findOptimalThreadCount, making the former return the entire microbenchmark summary and the later properly parse the required values (min, max, median). Got them and the print method to work easily but the plot method threw a column incosistency error:
Error in rbindlist(l, use.names, fill, idcol) :
Item 2 has 6 columns, inconsistent with item 1 which has 4 columns. To fill missing columns use fill=TRUE.
All the five data.tables to be combined had the 6 columns but it turns out (after some debugging) that the types are not right, and that that additional code was needed, so I made the plot function’s data.table combination code more robust to accodmodate that. (7 hours)
- May 24, ‘24:
1) Tried several different approaches to separate the legend from the plot as I want 5 values in my plot (one point and one line with the same color, one point and one line with another color, and finally a line with yet another color) but only 3 showing up in my legend, for which I need a custom legend. Used libraries such as grid and patchwork as well, but only partially successful since the grobs have issues when merged and the legends repeat for each subplot/expr. Finally got it done (went for a simpler solution, granted that shape was not reasonable enough to include). (7 hours)