Mathias Rav
Mathias Rav
Right now we use buffers and mutexes in tp::parallel. Is it faster/slower/better/worse to use a lock-free queue than buffers and mutexes? There is a [lock-free queue under the BSD license](http://moodycamel.com/blog/2014/a-fast-general-purpose-lock-free-queue-for-c++)...
When we introduced data structures to pipelining, we started calling set_available_memory twice on each node: a) Once before freezing data structures, and b) once after freezing data structures. However, the...
We have two nodes that apply a function to the input and push to the destination: `map_t` in `map.h` and `lambda_t` in `std_glue.h`.
Jakob and I think it would be better to have phase names (with priorities) and node names (without priorities) than the current setup where we only have node names with...
The running time of merge sort is determined by the number of runs generated in run generation and the merge fanout d. If the number of runs is in the...
Somehow, the automatic add_push_destination does not happen to the stuff inside a fork().
It would be nice to have a serialization stack.
It is not possible to maintain the order in parallel in the general case when the output items do not correspond one-to-one to the input items. The current implementation fails...