Dmitriy

Just putting this StackOverflow answer here for future reference. It's CUDA-specific, but I'm sure it's relevant for HIP as well. [How to have atomic load in CUDA](https://stackoverflow.com/questions/32341081/how-to-have-atomic-load-in-cuda).

Probably related to this issue, `monolith`-saved pages don't render [Nicky Case](https://ncase.me/)'s explorable explanations correctly. Just as an example: https://ncase.me/trust/

Ah, the `-b` flag does the trick, at least when online; without it, the page didn't work even online. I didn't know about it. Too bad that fully offline is...

Actually, fully offline seems to just work for that link, unless I'm hitting some browser cache.

And `-j` does the trick with substack. Thanks!

@tpeterka FYI

@daboehme This is quite a show-stopper for me. Would it be difficult to add something like `reset_time()`? Is this compatible with how services are designed now?

@daboehme Another question in the same line of understanding CalQL. I want to print out information about some MPI routines. I can do it for one specific routine at a...

@daboehme Thanks. I saw those docs, and I saw the `CALI_MPI_WHITELIST` option. I still think it would be great to have OR to be able to drill down into the...

Well, one hack would be to borrow angle brackets from unicode. U+3008 and U+3009,〈〉. From my cursory reading, they are allowed in URIs, but not in C++ identifiers. (This could...