Improve spullara's solution (performs 25% better)

[yemreinci]

Copied spullara's solution and made a few improvements. Performs ~25% better on my machine.

• Class Name: CalculateAverage_yemreinci
• Machine specs: 16 cores, 64GB RAM
• JVM: 21.0.1-graalce

Improvements:

• Calculate the hashcode while reading the data, instead of later in the hashmap implementation. This is expected to increase instruction level parallelism as CPU can work on the math while waiting for data from memory/cache.
• Convert the number parsing while loop into a few if statements. While loop with a switch-case inside is likely not so great for the branch predictor.

Benchmarks

$ for i in {1..5}; do  ./calculate_average_spullara.sh > /tmp/spullara.txt; done

real    0m4.096s
user    0m59.311s
sys     0m0.750s

real    0m4.117s
user    0m59.624s
sys     0m0.818s

real    0m4.160s
user    1m0.216s
sys     0m0.659s

real    0m4.119s
user    1m0.066s
sys     0m0.703s

real    0m4.059s
user    0m58.870s
sys     0m0.601s

$ for i in {1..5}; do  ./calculate_average_yemreinci.sh > /tmp/yemreinci.txt; done

real    0m3.195s
user    0m45.039s
sys     0m0.840s

real    0m3.071s
user    0m43.438s
sys     0m0.731s

real    0m3.184s
user    0m44.765s
sys     0m0.718s

real    0m3.075s
user    0m43.489s
sys     0m0.759s

real    0m3.093s
user    0m43.659s
sys     0m0.692s

[spullara]

I added changes similar to yours. Very good!

[gunnarmorling]

Does this pass the tests: ./test.sh yemreinci?

[gunnarmorling]

00:13.463, nice. Thx for participating!