Speed improvements
The following tables present the average results of the benchmark ran to compare the performance of GaiaXPy v2.1.0 and GaiaXPy v2.0.1 (its immediate predecessor).
The benchmark, which consisted on running each of the tools over input files in different formats, was repeated 20 times to compute the average times. All of the files contained the same 5,000 sources.
Environment details
Processor: Intel(R) Core(TM) i7-8565U CPU @ 1.80GHz x 8
Memory: 16GB
Operating system: Ubuntu 22.04.2 LTS 64-bit
Disk model: WDC PC SN730 SDBQNTY-512G-1001
Package dependencies: astropy==5.3, numpy==1.25.0, pandas==2.0.2, scipy==1.10.1.
Input reading and output writing
The tables below show the average time it takes GaiaXPy to read a file of 5,000 rows in different formats, and also the time it takes each tool to write the corresponding output.
GaiaXPy v.2.0.1 (old version)
Format/Function |
Input reading |
Output writing |
||
|---|---|---|---|---|
Calibrate |
Convert |
Generate |
||
CSV |
5.30s |
4.16s |
11.50s |
0.06s |
ECSV |
5.21s |
4.32s |
12.72s |
0.09s |
FITS |
3.14s |
0.05s |
0.30s |
0.04s |
XML binary* |
36.55s |
9.75s |
31.44s |
0.43s |
XML plain |
30.16s |
9.94s |
31.73s |
0.49s |
GaiaXPy v.2.1.0 (new version)
Format/Function |
Input reading |
Output writing |
||
|---|---|---|---|---|
Calibrate |
Convert |
Generate |
||
CSV |
6.00s |
4.29s |
12.11s |
0.11s |
ECSV |
5.91s |
4.35s |
12.42s |
0.10s |
FITS |
3.51s |
0.18s |
0.27s |
0.05s |
XML binary* |
34.09s |
7.76s |
25.28s |
0.58s |
XML plain |
28.90s |
7.84s |
25.25s |
0.54s |
*GaiaXPy XML output is always plain.
As it can be observed, the input reading time is longer for certain formats in v2.1.0 compared to the previous version, and the same applies to output writing.
The variations in input reading time arise from the fact that the parsing of some matrices, which was previously performed during processing, is now executed during reading. This introduces a slight overhead in input reading but greatly accelerates the processing.
Regarding the output writing, the files written by GaiaXPy v2.1.0 are similar but not identical to the ones generated by its predecessor, as mentioned in Release Notes.
Reading and processing times
The tables below show the time taken to run each of the tools over input files of 5,000 rows in different formats. It includes both the reading and processing steps, but it excludes the time used to save the output to a file.
GaiaXPy v.2.0.1 (old version)
Format/Function |
Calibrate |
Convert |
Generate (w/o error_corr) | Generate (w/ error_corr) |
||
|---|---|---|---|---|---|
CSV |
35.34s |
60.02s |
6.62s |
13.17s |
|
ECSV |
35.57s |
59.79s |
6.57s |
13.04s |
|
FITS |
33.37s |
57.62s |
4.48s |
10.96s |
|
XML binary |
66.46s |
90.52s |
37.91s |
44.42s |
|
XML plain |
60.08s |
84.00s |
31.59s |
38.23s |
|
GaiaXPy v.2.1.0 (new version)
Format/Function |
Calibrate |
Convert |
Generate (w/o error_corr) | Generate (w/ error_corr) |
||
|---|---|---|---|---|---|
CSV |
9.36s |
12.10s |
6.58s |
8.81s |
|
ECSV |
9.29s |
12.02s |
6.48s |
8.78s |
|
FITS |
6.97s |
9.51s |
4.08s |
6.37s |
|
XML binary |
37.27s |
40.00s |
34.49s |
36.82s |
|
XML plain |
32.14s |
34.89s |
29.51s |
31.76s |
|
Speed up factors
Format/Function |
Calibrate |
Convert |
Generate (w/o error_corr) | Generate (w/ error_corr) |
||
|---|---|---|---|---|---|
CSV |
3.78 |
4.96 |
1.01 |
1.49 |
|
ECSV |
3.83 |
4.97 |
1.01 |
1.49 |
|
FITS |
4.79 |
6.06 |
1.10 |
1.72 |
|
XML binary |
1.78 |
2.26 |
1.10 |
1.21 |
|
XML plain |
1.87 |
2.41 |
1.07 |
1.20 |
|
In general, the speed improvements when the input file is either XML binary or plain are smaller. This is likely due to the fact that most of the changes in the code were implemented in the processing step and not in the reading one. Reading XML is slow in general, and most of the execution time corresponds to reading these files. GaiaXPy uses Astropy to read these files, which may or may not improve its reading speeds in the future.
Processing times
The tables below show the results when only the processing step of each tool is considered (i.e. times for reading input and writing output are excluded).
GaiaXPy v.2.0.1 (old version)
Format/Function |
Calibrate |
Convert |
Generate (w/o error_corr) | Generate (w/ error_corr) |
||
|---|---|---|---|---|---|
CSV |
30.04s |
54.72s |
1.32s |
7.87s |
|
ECSV |
30.36s |
54.58s |
1.36s |
7.83s |
|
FITS |
30.23s |
54.48s |
1.34s |
7.82s |
|
XML binary |
29.91s |
53.97s |
1.36s |
7.87s |
|
XML plain |
29.92s |
53.84s |
1.43s |
8.07s |
|
GaiaXPy v.2.1.0 (new version)
Format/Function |
Calibrate |
Convert |
Generate (w/o error_corr) | Generate (w/ error_corr) |
||
|---|---|---|---|---|---|
CSV |
3.36s |
6.10s |
0.58s |
2.81s |
|
ECSV |
3.38s |
6.11s |
0.57s |
2.87s |
|
FITS |
3.46s |
6.00s |
0.57s |
2.86s |
|
XML binary |
3.18s |
5.91s |
0.40s |
2.73s |
|
XML plain |
3.24s |
5.99s |
0.61s |
2.86s |
|
Speed up factors
Format/Function |
Calibrate |
Convert |
Generate (w/o error_corr) |
Generate (w/ error_corr) |
|---|---|---|---|---|
CSV |
8.94 |
8.97 |
2.28 |
2.80 |
ECSV |
8.98 |
8.93 |
2.39 |
2.73 |
FITS |
8.74 |
9.08 |
2.35 |
2.73 |
XML binary |
9.41 |
9.13 |
3.40 |
2.88 |
XML plain |
9.23 |
8.99 |
2.34 |
2.82 |