Closed deanrockit closed 5 years ago
@deanrockit nice work!
In the future, please create one pull request (PR) per issue and reference the issue in the pull request. That way, we can more easily keep track of what issue each PR resolves. This comment applies to me as well.
I can review items (1) - (4) , @hzfmer can you please review (5) ?
@deanrockit can you please create an issue and PR in https://github.com/SCECcode/awp-benchmarks that contains the code you used to test this patch?
@ooreilly Sure! I should have done this more often, instead of pushing all the fixes at once like what I did this time. I'll submit one PR for one issue at a time for the next one. One thing I wasn't sure is I didn't know I should whether merge my change into fix_source_recv_dm or the master. I used the master as the base when creating this pull request. Is that okay?
@deanrockit wonderful! thanks.
@deanrockit merging into master is fine for now.
@ooreilly @deanrockit Revision (5) looks good to me.
Te-yang and I discussed on this issue and it's nice that Te-yang implemented it now.
The revision helps avoid zero-value outputs if using previous output approach. Maybe Te-yang can elaborate this in detail in https://github.com/SCECcode/awp-benchmarks
(1) Fix source/receiver placement in multiple-block configurations
(2) Fix a bug in source_find_grid_number to make the length of z1 array consistent with Nz of each block. Modify the lower and upper bounds of each layer to allow sources to be placed in the overlap zone
(3) Add special treatments to source_init_common to take care of sources placed in the overlap zone
(4) Remove FORCE_HIGH_Q block from the pmcl3d.c
(5) Modify velbuff function in kernel.cu to so that sfc output automatically writes out the first non-zero velocity grid with or w/o FOLLOWBATHY option