Accelerating LLM Decoding with In-DRAM Parallelismwith Kevin Xu, Daichi Tokuda, Shinya Takamaeda-Yamazaki, Issei Sato1 versionW_jzfgdsppv1 · current
PuDGhost: Experimental Analysis of Computation Result Corruption in Processing-using-DRAM Operations on Real DRAM Chips and Implications for Future SystemsDaichi Tokuda, İsmail Emir Yuksel, Tatsuya Kubo et al.0 citations2606.19119v12026arXiv
PUDTune: Multi-Level Charging for High-Precision Calibration in Processing-Using-DRAMTatsuya Kubo, Daichi Tokuda, Lei Qu et al.2 citations2505.05266v12025arXiv
MVDRAM: Enabling GeMV Execution in Unmodified DRAM for Low-Bit LLM AccelerationTatsuya Kubo, Daichi Tokuda, Tomoya Nagatani et al.1 citation2503.23817v22025arXiv