A brand new technical paper titled “Reaching the Efficiency of All-Financial institution In-DRAM PIM With Normal Reminiscence Interface: Reminiscence-Computation Decoupling” was revealed by researchers at Korea College.
“This paper proposed the memory-computation decoupled PIM structure to supply the efficiency corresponding to the all-bank PIM whereas preserving the usual DRAM interface, i.e., DRAM instructions, energy funds, timing constraints, and so forth. For reaching our objective, we launched two PIM execution phases: reminiscence and computation. The reminiscence part follows the per-bank PIM execution methodology. The computation part broadcasts the bank-shared operands to all banks, thus making their PIM engines carry out the computation concurrently with out accessing the financial institution knowledge arrays,” states the paper.
Discover the technical paper here. Printed August 2022.
Y. Paik, C. H. Kim, W. J. Lee and S. W. Kim, “Reaching the Efficiency of All-Financial institution In-DRAM PIM With Normal Reminiscence Interface: Reminiscence-Computation Decoupling,” in IEEE Entry, vol. 10, pp. 93256-93272, 2022, doi: 10.1109/ACCESS.2022.3203051. Inventive commons license here.