Dan Hoeflinger

Results 127 comments of Dan Hoeflinger

I'm going to break this into two PRs. setting draft for now.

> 2\. What is a motivation to preserve the backend type names in the tests? I suppose it might be removed as non-necessary now. I too am confused by our...

For this to stop the copy from happening, changes are required for `__future` as well I believe. I think @julianmi may be working on this already, we have discussed it...

As written, it seems like the `parallel_find_or` case seems to be doing a `||` reduction using global atomic variables to combine the results from different work items. I would expect...

> Theoretically, it might be true. But, practically, access to a global atomic is very "expensive".. I agree. I'm not sure without running benchmarks what the best option is. I'll...

To the same goal of future performance improvement of merge sort: It seems worthwhile to investigate leaf size changes based on total input size to avoid the extra copy from...

Additionally, it would be good to re-investigate selection of leaf size and its effect on performance. I believe when this was looked at previously, the number of iterations of the...

Is there a reason there are 2 nested copies of `TestUtils::unique_kernel_name` here? Are they redundant? If we change the number from `0` to `1`, in your code, do both copies...

> In any case now `new_kernel_name` is implemented in this PR exactly as > > ```c++ > template > using __policy_kernel_name = typename ::std::decay_t::kernel_name; > ``` Yes, but in your...