sudhirtumati implementation

[sudhirtumati]

Check List:

• [x] Tests pass (./test.sh <username> shows no differences between expected and actual outputs)

• [x] All formatting changes by the build are committed

• [x] Your launch script is named calculate_average_<username>.sh (make sure to match casing of your GH user name) and is executable

• [x] Output matches that of calculate_average_baseline.sh

• [x] For new entries, or after substantial changes: When implementing custom hash structures, please point to where you deal with hash collisions (line number)

• Execution time: ~45 sec

• Execution time of reference implementation: ~4m 15s

My PC configuration:

• WSL2
• Intel Core i7-9750H
• 32GB RAM (24 GB allocated to WSL)
• 6 cores with 2 threads per code (8 / 12 threads allocated to WSL)

[gunnarmorling]

Produces incorrect output for the 1B file:

Validating calculate_average_sudhirtumati.sh -- measurements_1B.txt
6c6
< Adelaide;-34.2;17.2;67.1
---
> Adelaide;-34.2;17.3;67.1
31c31
< Baghdad;-30.2;22.7;70.2
---
> Baghdad;-30.2;22.8;70.2
142c142
< Hamilton;-34.7;18.0;73.1
---
> Hamilton;-34.7;18.1;73.1
191c191
< La Ceiba;-22.5;26.1;75.2
---
> La Ceiba;-22.5;26.2;75.2

FAILURE: ./test.sh sudhirtumati measurements_1B.txt failed

[sudhirtumati]

./test.sh sudhirtumati

Using java version 21.0.2-open in this shell.
Validating calculate_average_sudhirtumati.sh -- src/test/resources/samples/measurements-1.txt
Validating calculate_average_sudhirtumati.sh -- src/test/resources/samples/measurements-10.txt
Validating calculate_average_sudhirtumati.sh -- src/test/resources/samples/measurements-10000-unique-keys.txt
Validating calculate_average_sudhirtumati.sh -- src/test/resources/samples/measurements-2.txt
Validating calculate_average_sudhirtumati.sh -- src/test/resources/samples/measurements-20.txt
Validating calculate_average_sudhirtumati.sh -- src/test/resources/samples/measurements-3.txt
Validating calculate_average_sudhirtumati.sh -- src/test/resources/samples/measurements-boundaries.txt
Validating calculate_average_sudhirtumati.sh -- src/test/resources/samples/measurements-complex-utf8.txt
Validating calculate_average_sudhirtumati.sh -- src/test/resources/samples/measurements-dot.txt
Validating calculate_average_sudhirtumati.sh -- src/test/resources/samples/measurements-rounding.txt
Validating calculate_average_sudhirtumati.sh -- src/test/resources/samples/measurements-short.txt
Validating calculate_average_sudhirtumati.sh -- src/test/resources/samples/measurements-shortest.txt

This is the output of test execution on my environment. I do not see measurements_1B.txt file under `src/main/resources/samples' directory.

Could you please make measurements_1B.txt available for me to reproduce the issue?

[gunnarmorling]

That file is 13 G, so it's a bit hard to share. You can create it yourself using create_measurements.sh (for the standard eval file) and create_measurements3.sh (for the 10K key set file). Your implementation must show the same output for those as the base line (or compare to output of the top of the leaderboard who are known to be correct and will complete much faster).

[sudhirtumati]

Noted @gunnarmorling

Unfortunately, I couldn't reproduce the error in my environment. The output generated by my implementation is identical with the baseline. I am sure there must be an issue with the implementation as the same code is not working in your environment. I suspect thread contention might be resulting in incorrect map updates.

Modified my implementation to reduce thread contention to the extent possible. Total processing time (with 1B rows) also came down from ~40 seconds to ~30 seconds on my personal laptop.

Test suite execution is successful. In addition to the test suite, I ran tests with multiple files with different row counts (.5m, 1m, 10m, 50, 100, 500m, 1B) and found that the results are identical to the baseline output.

[gunnarmorling]

Hey, can you please rebase this to latest main and squash everything into one commit? There should be no unrelated commits in this PR.

[sudhirtumati]

Done. Please check

[gunnarmorling]

00:25.064.