Hello everyone, how are you doing?
Recently a potential client contacted us needing to run seismic inversion on a 25GiB angle stack dataset (3x25GiB seismic + 3x25GiB elastic property trends).
The problem is that our plugin was storing the inverted traces on a few SeisTrcBuf objects. I've managed to implement a pooling mechanism to consume the buffer and delete the traces as I sent them to SeisTrcWriter objects. However, SeisTrcWriter is keeping a copy of the traces inside it (trcblock_) and still taking up a lot of memory. How can I dump the written traces from SeisTrcWriter? Is there another way to write humongous cubes using less memory as the data become available to write? I need to write it without caching the whole cube in memory, since the inversion is producing 150GiB of data.
Thanks!