Added
Hardware
- Extend and deploy documentation (#156)
- Add reduction support (sequential and parallel) with
floo_reduction_unitandfloo_alu; router exposes offload ports for functional units (DCA-compatible) (#163) - Add multiplane support in
floo_vc_arbiterto decouple AXI read and write wide channels into separate routers; add credit-based VC and preempt-valid arbitration modes (#153) - Extend
floo_nw_chimneyto support 2 input/output wide links for read/write stream decoupling (#153)
FlooGen
- Added support for custom/external templates (#155)
- Extend and deploy documentation (#156)
- Publish FlooGen on PyPI (#160)
- Add reduction configuration generation (#163)
Changed
Hardware
- Unified collective operation configuration (multicast, synchronization, reduction) into a single
CollectiveCfgstructure; individual enable parameters replaced by frontend macro-operation selection (collect_op_fe_cfg_t) (#163) - Use default remotes in Bender.yml manifest (#161)
- Bump dependencies (#161)
FlooGen
- Refactored CLI to use subcommands instead of flags (#155)