CuAsmRL uses deep reinforcement learning to reorder GPU SASS instructions and reports average 9% throughput improvement on six LLM kernels on an NVIDIA A100.
Title resolution pending
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.AR 1years
2025 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
CuAsmRL: Optimizing GPU SASS Schedules via Deep Reinforcement Learning
CuAsmRL uses deep reinforcement learning to reorder GPU SASS instructions and reports average 9% throughput improvement on six LLM kernels on an NVIDIA A100.