Minor improvement to @spullara

[hundredwatt]

I forked @spullara's file and made a minor improvement that reduces execution time by 2-3%.

This reduced executive time from ~9.5 to ~9.3s in hyperfine with 5 runs

• Class Name: CalculateAverage_hundredwatt
• Machine specs: 8 cores, 32GB RAM
• JVM: 21.0.1-graalce

Improvement:

• Hard-code processing of the first character of the city name (which can never be ; according to the input value ranges in the README), commit: 0ffa9bcc6944df4db2e0177e09640861631bc813

I ran hyperfine twice with benchmark order reversed the second time and saw the same results.

Hyperfine command:

hyperfine --warmup 1 --runs 5 ' java $JAVA_OPTS --class-path target/average-1.0.0-SNAPSHOT.jar dev.morling.onebrc.CalculateAverage_yemreinci' ' java $JAVA_OPTS --class-path target/average-1.0.0-SNAPSHOT.jar dev.morling.onebrc.CalculateAverage_hundredwatt'

(Note: hostname of this machine is fedora-s-8vcpu-16gb-amd-sfo3-01, but I resized it to have 32gb of RAM)

[hundredwatt]

Rebased my fork with latest main and ensured test.sh passes

[hundredwatt]

The original results excited me because the difference was small, but this change always had the min time. Not true after rebasing.

Seems like it may be a wash for the purposes of this challenge, curious what others think:

[gunnarmorling]

I am on the fence tbh. As per the rules, there should be significant improvements when basing on existing impls. Hard to draw a line of course, but to me this one feels below the threshold, in particular as it seems not consistently faster? I'm inclined to close this one.