Skip to main content
Code Review

Return to Answer

added 1 characters in body
Source Link
konijn
konijn
  • 34.4k
  • 5
  • 71
  • 267

Algorithms

I would also like to add couple observation on top of previous answers. I was slightly confused if you were really comparing your merge sort with the Arrays.sort(int[])?? The Arrays.sort(int[]) uses a different algorithm. Tuned quick sort according to a documentation Arrays doc

So I have a feeling that you might be comparing two different algorithms.

PerfomancePerformance tests

Another observation that I found was that I would suggest to adjust the performance test. Firstly, for performance testing you should be using System.nanos(). System.currentMillis() dont have a milisecond precision, depends on a OS it could even have a 16ms interval of incrementing.

Try to run the tests for a different size of the array. I notice you used always 100 items. I would suggest to try more such as 1K, 10K, 100K, 1M... This gives you a overview if the 10% speed up is always there or if, for example, with 1M you only get 1%. The reason why this is good to test is if you find that the speed up decreases with higher numbers then there is some constant overhead that is not connected to a number of items in the array. Which you might want to know. On the other hand if you know that your algorithm will be run for arrays of size around hundred this is a relevant performance test that gives you relevant results.

Algorithms

I would also like to add couple observation on top of previous answers. I was slightly confused if you were really comparing your merge sort with the Arrays.sort(int[])?? The Arrays.sort(int[]) uses a different algorithm. Tuned quick sort according to a documentation Arrays doc

So I have a feeling that you might be comparing two different algorithms.

Perfomance tests

Another observation that I found was that I would suggest to adjust the performance test. Firstly, for performance testing you should be using System.nanos(). System.currentMillis() dont have a milisecond precision, depends on a OS it could even have a 16ms interval of incrementing.

Try to run the tests for a different size of the array. I notice you used always 100 items. I would suggest to try more such as 1K, 10K, 100K, 1M... This gives you a overview if the 10% speed up is always there or if, for example, with 1M you only get 1%. The reason why this is good to test is if you find that the speed up decreases with higher numbers then there is some constant overhead that is not connected to a number of items in the array. Which you might want to know. On the other hand if you know that your algorithm will be run for arrays of size around hundred this is a relevant performance test that gives you relevant results.

Algorithms

I would also like to add couple observation on top of previous answers. I was slightly confused if you were really comparing your merge sort with the Arrays.sort(int[])?? The Arrays.sort(int[]) uses a different algorithm. Tuned quick sort according to a documentation Arrays doc

So I have a feeling that you might be comparing two different algorithms.

Performance tests

Another observation that I found was that I would suggest to adjust the performance test. Firstly, for performance testing you should be using System.nanos(). System.currentMillis() dont have a milisecond precision, depends on a OS it could even have a 16ms interval of incrementing.

Try to run the tests for a different size of the array. I notice you used always 100 items. I would suggest to try more such as 1K, 10K, 100K, 1M... This gives you a overview if the 10% speed up is always there or if, for example, with 1M you only get 1%. The reason why this is good to test is if you find that the speed up decreases with higher numbers then there is some constant overhead that is not connected to a number of items in the array. Which you might want to know. On the other hand if you know that your algorithm will be run for arrays of size around hundred this is a relevant performance test that gives you relevant results.

Source Link
Vojta
Vojta
  • 656
  • 3
  • 10

Algorithms

I would also like to add couple observation on top of previous answers. I was slightly confused if you were really comparing your merge sort with the Arrays.sort(int[])?? The Arrays.sort(int[]) uses a different algorithm. Tuned quick sort according to a documentation Arrays doc

So I have a feeling that you might be comparing two different algorithms.

Perfomance tests

Another observation that I found was that I would suggest to adjust the performance test. Firstly, for performance testing you should be using System.nanos(). System.currentMillis() dont have a milisecond precision, depends on a OS it could even have a 16ms interval of incrementing.

Try to run the tests for a different size of the array. I notice you used always 100 items. I would suggest to try more such as 1K, 10K, 100K, 1M... This gives you a overview if the 10% speed up is always there or if, for example, with 1M you only get 1%. The reason why this is good to test is if you find that the speed up decreases with higher numbers then there is some constant overhead that is not connected to a number of items in the array. Which you might want to know. On the other hand if you know that your algorithm will be run for arrays of size around hundred this is a relevant performance test that gives you relevant results.