Saved in:
| Main Authors: | , , , |
|---|---|
| Format: | Preprint |
| Published: |
2026
|
| Subjects: | |
| Online Access: | https://arxiv.org/abs/2604.03829 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1866915916303826944 |
|---|---|
| author | Odemuyiwa, Toluwanimi O. Owens, John D. Emer, Joel S. Pellauer, Michael |
| author_facet | Odemuyiwa, Toluwanimi O. Owens, John D. Emer, Joel S. Pellauer, Michael |
| contents | Mamba is an emerging, complex workload with various short-range and long-range dependencies, nonlinearities, and elementwise computations that are unable to run at near-peak speeds on modern hardware. Specifically, Mamba's complex dependency graph makes fusion across its full operator cascade difficult, leaving substantial inter-operator memory traffic on the table. To address these challenges, we propose Mambalaya, a novel reconfigurable accelerator that leverages fusion to overcome the limitations of Mamba. We use the recently proposed cascade-of-Einsums abstraction to characterize Mamba's full computational structure, then apply the extended Einsum framework to systematically explore inter-Einsum fusion opportunities. This principled approach yields a series of fusion mappings that reduce off-chip inter-Einsum traffic. These mappings are supported by the underlying Mambalaya architecture. Mambalaya achieves a layer performance speedup of 4.9$\times$ for prefill and 1.9$\times$ for generation over MARCA. In prefill-dominated scenarios, it achieves up to 1.5$\times$ over a recent fine-grained, memory-aware fusion accelerator for Mamba. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2604_03829 |
| institution | arXiv |
| publishDate | 2026 |
| record_format | arxiv |
| spellingShingle | Mambalaya: Einsum-Based Fusion Optimizations on State-Space Models Odemuyiwa, Toluwanimi O. Owens, John D. Emer, Joel S. Pellauer, Michael Hardware Architecture Mamba is an emerging, complex workload with various short-range and long-range dependencies, nonlinearities, and elementwise computations that are unable to run at near-peak speeds on modern hardware. Specifically, Mamba's complex dependency graph makes fusion across its full operator cascade difficult, leaving substantial inter-operator memory traffic on the table. To address these challenges, we propose Mambalaya, a novel reconfigurable accelerator that leverages fusion to overcome the limitations of Mamba. We use the recently proposed cascade-of-Einsums abstraction to characterize Mamba's full computational structure, then apply the extended Einsum framework to systematically explore inter-Einsum fusion opportunities. This principled approach yields a series of fusion mappings that reduce off-chip inter-Einsum traffic. These mappings are supported by the underlying Mambalaya architecture. Mambalaya achieves a layer performance speedup of 4.9$\times$ for prefill and 1.9$\times$ for generation over MARCA. In prefill-dominated scenarios, it achieves up to 1.5$\times$ over a recent fine-grained, memory-aware fusion accelerator for Mamba. |
| title | Mambalaya: Einsum-Based Fusion Optimizations on State-Space Models |
| topic | Hardware Architecture |
| url | https://arxiv.org/abs/2604.03829 |