Twill automatically derives optimal software pipelining and warp specialization schedules for Tensor Core GPUs

Researchers have introduced Twill, a system that formulates software pipelining (SWP) and warp specialization (WS) as a joint optimization problem solvable by off-the-shelf constraint solvers. This approach automates the derivation of optimal execution schedules for iterative programs on GPU architectures.

• Twill is heuristic-free and guaranteed to produce optimal schedules for a large class of iterative programs.
• The system can rediscover and prove optimality of SWP and WS schedules manually developed by experts for Flash Attention.
• Validation was performed on both NVIDIA Hopper and Blackwell GPU architectures.

By replacing brittle compilation heuristics and human intuition with holistic optimization, Twill ensures maximal utilization of fixed-function units and general-purpose cores in complex modern GPUs.