I have taken 3 s-1 SLC bursts and then applied some processing to that using snappy. If I don’t write the output, it takes around 3 sec for the entire processing but when I’m writing the product into tif format it takes more than 20 mins. Can reduce the writing time?
The Graph Processing Facility used by gpt doesn’t actually do calculations until the result is needed, so no actual processing is done if you don’t use the result, e.g., to write a file.
Hi @gnwiii
I checked the final output before writing it, the processing was done.
My processing flow includes debursts and terrain correction, you can see both of them were performed.
Memory management between Python and Java is tricky. Your plot has to allocate memory in Python (for band_data). GPT processing uses memory controlled by Java, so you end up storing the same data in two places. How long doe the processing take if you omit the Python plot? It could be useful to watch memory and CPU usage. My GPT batch workflows (done years ago, when memory was often a limiting resource) have always saved the results with separate processing for visualizations.