An official website of the United States government
The rapidly increasing size of deep-learning models has renewed interest in alternatives to digital-electronic computers as a means to dramatically reduce the energy cost of running state-of-the-art neural networks. Optical matrix-vector multipliers are best suited to performing computations with very large operands, which suggests that large Transformer models could be a good target for them. In this paper, we investigate---through a combination of simulations and experiments on prototype optical hardware---the feasibility and potential energy benefits of running Transformer models on future optical accelerators that perform matrix-vector multiplication. We use simulations, with noise models validated by small-scale optical experiments, to show that optical accelerators for matrix-vector multiplication should be able to accurately run a typical Transformer architecture model for language processing. We demonstrate that optical accelerators can achieve the same (or better) perplexity as digital-electronic processors at 8-bit precision, provided that the optical hardware uses sufficiently many photons per inference, which translates directly to a requirement on optical energy per inference. We studied numerically how the requirement on optical energy per inference changes as a function of the Transformer width $$d$$ and found that the optical energy per multiply--accumulate (MAC) scales approximately as $$\frac{1}{d}$$, giving an asymptotic advantage over digital systems. We also analyze the total system energy costs for optical accelerators running Transformers, including both optical and electronic costs, as a function of model size. We predict that well-engineered, large-scale optical hardware should be able to achieve a $$100 \times$$ energy-efficiency advantage over current digital-electronic processors in running some of the largest current Transformer models, and if both the models and the optical hardware are scaled to the quadrillion-parameter regime, optical accelerators could have a $$>8,000\times$$ energy-efficiency advantage. Under plausible assumptions about future improvements to electronics and Transformer quantization techniques (5× cheaper memory access, double the digital--analog conversion efficiency, and 4-bit precision), we estimate that the energy advantage for optical processors versus electronic processors operating at 300~fJ/MAC could grow to $$>100,000\times$$.
more »
« less
All optical NOR gate via tunnel-junction transistor lasers for high speed optical logic processors
Tunnel-junction transistor lasers (TJ-TLs) is a critical element to form a universal electro-optical NOR gate and an optical bistable latch which can be developed into a compact chip-level solution for optical logic processors operating at GHz speed.
more »
« less
- Award ID(s):
- 1640196
- PAR ID:
- 10064926
- Date Published:
- Journal Name:
- 2018 International Symposium on VLSI Technology, Systems and Application (VLSI-TSA)
- Format(s):
- Medium: X
- Sponsoring Org:
- National Science Foundation
More Like this
-
-
null (Ed.)We present a hybrid optical-electrical analog deep learning (DL) accelerator, the first work to use incoherent optical signals for DL workloads. Incoherent optical designs are more attractive than coherent ones as the former can be more easily realized in practice. However, a significant challenge in analog DL accelerators, where multiply-accumulate operations are dominant, is that there is no known solution to perform accumulation using incoherent optical signals. We overcome this challenge by devising a hybrid approach: accumulation is done in the electrical domain, while multiplication is performed in the optical domain. The key technology enabler of our design is the transistor laser, which performs electrical-to-optical and optical-to-electrical conversions efficiently to tightly integrate electrical and optical devices into compact circuits. As such, our design fully realizes the ultra high-speed and high-energy-efficiency advantages of analog and optical computing. Our evaluation results using the MNIST benchmark show that our design achieves 2214× and 65× improvements in latency and energy, respectively, compared to a state-of-the-art memristor-based analog design.more » « less
-
Broken spatial and time reversal symmetries in materials often give rise to new emergent phenomena in the interaction between light and matter. The combination of chirality and broken time reversal symmetry in a magnetic field leads to magneto–chiral phenomena, such as the nonreciprocity of transmission. Here, we construct a terahertz hybrid metamaterial that combines the natural optical activity of a chiral metallic gammadion bilayer and the magneto-optical activity of semiconductor indium antimonide in a magnetic field. We report a resonant magneto–chiral effect that leads to polarization-independent nonreciprocal optical transmittance. Furthermore, we discover a magneto-optical Faraday effect that is resonantly controlled by the natural optical activity of the chiral gammadion bilayer. Unlike optical activity due to chirality, the novel Faraday effect is odd under time reversal. Both phenomena are activated by a modest magnetic field, which may open doors for their potential applications in polarization-independent optical isolation and highly efficient polarization control at terahertz frequencies.more » « less
-
null (Ed.)Abstract Optical bottle beams can be used to trap atoms and small low-index particles. We introduce a figure of merit (FoM) for optical bottle beams, specifically in the context of optical traps, and use it to compare optical bottle-beam traps obtained by three different methods. Using this FoM and an optimization algorithm, we identified the optical bottle-beam traps based on a Gaussian beam illuminating a metasurface that are superior in terms of power efficiency than existing approaches. We numerically demonstrate a silicon metasurface for creating an optical bottle-beam trap.more » « less
- Free Publicly Accessible Full Text
-
Accepted Manuscript1.0
- Conference Paper:
- https://doi.org/10.1109/VLSI-TSA.2018.8403847
