scubbo/blogcontent
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

Mention further work in the Zig speedtest experiment

Browse Source
This commit is contained in:
Jack Jackson 2024-12-28 15:33:04 -08:00
parent 493e5da5f4
commit 495a36f3dc
1 changed files with 1 additions and 1 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

2
blog/content/posts/zig-zig-zag-as-fast-as-you-can.md
View File

@ -206,7 +206,7 @@ Summary of non-pointer-based times: test.Summary{ .max = 278000, .mean = 175750
You'll note that the variables `TIMES_TO_RUN_A_SINGLE_TEST` and `NUMBER_OF_TESTS_TO_RUN` are super low, so these data are hardly statistically sound. I started out with `1000` and `50` - but experienced a Segmentation Fault (`aborting due to recursive panic`, with a code pointer into `lib/std/array_list.zig`), that I was unable to debug. Even values as low as `10` and `5` caused this issue. I hope I can figure this out to run a more scientific experiment. You'll note that the variables `TIMES_TO_RUN_A_SINGLE_TEST` and `NUMBER_OF_TESTS_TO_RUN` are super low, so these data are hardly statistically sound. I started out with `1000` and `50` - but experienced a Segmentation Fault (`aborting due to recursive panic`, with a code pointer into `lib/std/array_list.zig`), that I was unable to debug. Even values as low as `10` and `5` caused this issue. I hope I can figure this out to run a more scientific experiment.
Still, even with these low counts - it's surprising for the pointer-based approach to be pretty consistently _slower_. I suspect I'm doing something wrong in my test cases, since the explanation I was given seems intuitively sensible - "_finding the value-location_" twice is always going to be slower than finding it once. Still, even with these low counts - it's surprising for the pointer-based approach to be pretty consistently _slower_. I suspect I'm doing something wrong in my test cases, since the explanation I was given seems intuitively sensible - "_finding the value-location_" twice is always going to be slower than finding it once. I wonder if it's possible that some other part of the test-setup (say, random number generation) dominates the time-spent, and so I'm not actually getting an accurate comparison of pointer-based vs. non-pointer-based interaction with Map Values. I'll try some further experiments in that direction.
[^advent-of-code]: you can see my solutions [here](https://gitea.scubbo.org/scubbo/advent-of-code-2024) - though, since I'd written zero lines of Zig before these challenges, and I've mostly been focused on achieving solutions quickly rather than optimally or maintainably, please don't judge me on Code Quality! 😆 [^advent-of-code]: you can see my solutions [here](https://gitea.scubbo.org/scubbo/advent-of-code-2024) - though, since I'd written zero lines of Zig before these challenges, and I've mostly been focused on achieving solutions quickly rather than optimally or maintainably, please don't judge me on Code Quality! 😆