notesum.ai
Published at November 8Palermo: Improving the Performance of Oblivious Memory using Protocol-Hardware Co-Design
cs.CR
cs.AR
Released Date: November 8, 2024
Authors: Haojie Ye1, Yuchen Xia1, Yuhan Chen1, Kuan-Yu Chen1, Yichao Yuan1, Shuwen Deng2, Baris Kasikci3, Trevor Mudge1, Nishil Talati1
Aff.: 1University of Michigan, USA; 2Tsinghua University, China; 3University of Washington, USA

| Category | Workload Name | Description |
| SPEC17 | mcf (mcf) | Route planning computation [3] |
| lbm (lbm) | Fluid dynamics computation [3] | |
| Graph | PageRank (pr) | Score ranking [2] on Livejournal [1] |
| Motif Mining (motif) | Graph mining [31] on Wikipedia [26] | |
| DL | DLRM(MemBound) (rm1) | Meta RM [19, 18] on Criteo 1T [22] |
| DLRM(Balanced) (rm2) | Alibaba RM [54] on DBLP [44] | |
| Large Lang. Model (llm) | GPT-2 [37] on OpenORCA [27] | |
| KV | Redis (redis) | Redis KV access [23] |
| Streaming (stm) | Streaming memory access | |
| Random (rand) | Random memory access |