I have encountered an anomalous situation when reprojecting some specific S1 IW products using gpt and a graph.
It seems to be related to products corresponding to the end of a data-take, and probably particularly when the duration of data acquisition is longer than normal..
Processing time for any typical IW product is of the order of 60 seconds, sometimes out to 120 seconds.
The abnormal cases may take more than an hour, and a worst case I have noticed is more than 4 hours, i.e. more than 200 times typical processing time.
When I have noticed these excessively long execution-times I have been using my Windows 11 PC:
x64 system, 4 processor, 32 GB RAM.
An example is for product:
S1C_IW_GRDH_1SSH_20260124T105720_20260124T105756_006044_00C1F2_2426
I processed four consecutive products from the data-take. Each of the first three took about a minute to process, the fourth (above example) took more than an hour. The SAFE product for the first three had a size of about 820 MB, the fourth was 1172 MB. It had an acquisition duration of 36 seconds instead of the usual 25 seconds.
The next acquisition over the site occurred on 20260126, and all four products processed with typical execution times. In this case the fourth product of the sequence had a total size of 716 MB (shorter duration), less than the typical 820 MB.
Inspection of the performance during processing of the example product shows that gpt has typically 6 to 6.5 GB of RAM assigned, and uses a variable amount of CPU between 15% and 70%. The total for all other processes is less than 5% CPU.
There is typically 15% of RAM (32 GB) not committed or containing cached data not in use.
The rate of progress gets progressively slower as the progress bar indicates greater completed. only 70 completed after more than 35 minutes execution time. Total execution time was around 1 hr 45 m.
The content of the xml file for the relevant graph follows.
[file-name “flat.tiff” points at the tiff file of the IW product]