I noticed the SIZEOF intrinsic documented in pgicudaforug.pdf (11.4.) I’ve tried this out on non-CUDA code and for the derived types I’ve tested it seems to work like the F2008 storage_size intrinsic except the result is in bytes not bits. Several questions:
-
Will you be moving the documentation of this intrinsic into pgifortref.pdf and solidifying support for this intrinsic going forward? b) is there a guarantee of support for this host intrinsic going forward outside of the CUDA domain?
-
Why was the non-standard SIZEOF chosen instead of just getting a jump on the standard STORAGE_SIZE approach?
-
Will SIZEOF work on polymorphic types? b) is it supposed to be like STORAGE_SIZE / 8 (using dynamic type) or something more limited–say closer to the C_SIZEOF F2008 intrinsic that only operates on interoperable types?
-
Is there a list of all intrinsics supported in each release separate from pgifortref.pdf and pgicudaforug.pdf? I ask this because I almost missed this very useful host intrinsic because it’s not listed in the main fortran reference–I’m wondering what other useful intrinsics I’m missing out on ;)
Thanks,
Joe