Mikael Simberg
Mikael Simberg
**Describe the bug** With certain allocation patterns and sizes QuickPool never coalesces a backing allocation to be large enough, leading to unnecessary reallocations. **To Reproduce** Compile the following file with...
Use custom coalescing heuristic and introduce a number of new configuration options for memory pools
This is intended to replace https://github.com/eth-cscs/DLA-Future/pull/1143 as a more permanent solution. This introduces: - configuration options for controlling the "next" allocation size (i.e. all blocks after the first one) and...
Replaces #1083 (at least "temporarily"). This simply takes the patch we already use in upstream spack and proposes to make it semi-permanent, until we some day decide to revisit #1083....
This is the non-production ready debugging version that I used to print Umpire pool statistics for debugging. It needs cleanup to be ready for merging, but more importantly if we...
We currently have the command line option `--dlaf:print-config` which dumps the configuration used by DLA-Future. Is there any opposition to adding a corresponding `DLAF_PRINT_CONFIG` environment variable (or similar, but I...
Fixes #1258. Follow-up PR to https://github.com/pika-org/pika/pull/1260. Let's get #1260 merged first.
This needs a corresponding version addition to the spack package.
- [ ] don't allow failures on gh200 anymore
When running HIP CI with GCC version 14 or 13 we bump into an issue that is essentially identical to https://github.com/llvm/llvm-project/issues/69288. The issue seems to happen whenever `` is included...
In connection with #16 we've been changing includes back from `pika/modules/x.hpp` to more fine-grained modules. Once that is done we can remove the generated headers (and hopefully also slightly speed...