Get Started
Research questionHow can analog compute-in-memory attention perform softmax without costly analog-to-digital conversion?Analog compute-in-memory attention produces scores as voltages, but softmax requires exponentiation and normalization. Converting those scores to digital form adds overhead before the attention weights can be computed.
AI
Inference Optimization
Technology
Latest papersRecent research connected to this question, newest first.Compute-in-Memory Attention: A Time-Domain Analog Softmax Circuit with RC-Tunable TemperatureThe evidence concerns a 128-element time-domain analog softmax circuit in GlobalFoundries 22-nm FDSOI that operates on compute-in-memory score voltages. It is supported by transistor-level and post-layout simulations, variation and mismatch analysis, and incorporation of extracted characteristics into a MemTorch hardware-aware Transformer model.research paper · Sep 2, 2026
Related questions
How can FP4 attention exploit Blackwell tensor cores when softmax overhead dominates?How can edge analog in-memory inference reduce energy while certifying errors from heterogeneous, imperfect accelerators?How can associative-memory capacity be compared across Hopfield and attention-like models without conflating assumptions?How can attention-head contributions be measured in prompt-injection classifiers across circuit and output scales?
Home
Topics
Search
Library