Closed jinisusan closed 3 years ago
Is there a specific reason, why you reference TR9 and not 5.1? I don't recall any change which would impact this patch.
The spec says in 5.5.9.1 ompd_enumerate_icvs
:
The ICVs num-procs-var, thread-num-var, final-task-var, implicit-task-var and team-size-var must also be available with an ompd- prefix.
The intention was, that we didn't want to remove the old names without deprecation (and Debuggers @jdelsign might already match for the old names?). Looking at your patch, we missed the change ompd-implicit-var
-> implicit-task-var
. This name change would not be addressed by above sentence.
To comply with the spec, you should add back the ompd- variants to the FOREACH_OMPD_ICV
like:
macro (ompd_num_procs_var, "ompd-num-procs-var", ompd_scope_address_space, 0) \
In ompd_get_icv_from_scope
/ ompd_get_icv_string_from_scope
you can just duplicate the cases and reuse the function call:
case ompd_icv_num_procs_var:
case ompd_icv_ompd_num_procs_var:
return ompd_get_num_procs((ompd_address_space_handle_t*)handle, icv_value);
I guess for compatibility with 5.0 we should have ompd-implicit-var
and for 5.1 compliance ompd-implicit-task-var
.
Thanks for pointing this out, @jprotze ! I have made the necessary changes. (includes a change to include ompd-final-var also -- like ompd-implicit-var). There is no specific reason to mention TR9. (I assumed that was the convention -- since I saw other TR references in a couple of places).
There is no specific reason to mention TR9. (I assumed that was the convention -- since I saw other TR references in a couple of places).
We typically referenced TR documents, when the TR was the latest released document, to make clear that the change relates to something newer than the latest released spec version.
lgtm now, thanks!
Thank you !
Changes to incorporate the new ICV names as per TR9 (sans the ompd prefix), some cleanup and changes to support the ompd retrieval of the tool-verbose-init-var ICV.