Felix19350 2nd submission. Further improved performance by improving the parsing logic #323
Conversation
|
@gunnarmorling rebased PR as per your request. Sorry about that! |
|
|
||
| public static Stream<AverageAggregatorTask> createStreamOf(List<MemorySegment> memorySegments) { | ||
| return memorySegments.stream().map(AverageAggregatorTask::new); | ||
| } | ||
|
|
||
| public Map<String, ResultRow> processChunk() { | ||
| final var result = new TreeMap<String, ResultRow>(); | ||
| final var measurements = new HashMap<Integer, ResultRow>(EXPECTED_MAX_NUM_CITIES); | ||
| final var cityNames = new HashMap<Integer, String>(EXPECTED_MAX_NUM_CITIES); |
There was a problem hiding this comment.
What's used as the key here? How does it prevent collisions between entries?
There was a problem hiding this comment.
Hmm good point, it doesn't really. Let me see what I can do in terms of proper perfect hashing :)
There was a problem hiding this comment.
@gunnarmorling addressed your comment above (check the equals method on the CityRef class) and squeezed a bit more performance out of this while (hopefully) maintaining a relatively readable and idiomatic solution.
I'm seeing some fluctuation in my machine, I was able to go as low as 15s in some runs. Curious to see how this behaves in the test environment.
…trings for city names are not allocated with each row.
|
00:26.500. It fluctuates quite a bit indeed. |
Further improved performance by improving the parsing logic so that strings for city names are not allocated with each row.
Removed JVM options to control heap memory as they are now unnecessary.
Check List:
./test.sh <username>shows no differences between expected and actual outputs)calculate_average_<username>.sh(make sure to match casing of your GH user name) and is executablecalculate_average_baseline.sh