Search

In nature, microbes use extracellular electron transfer (EET) to recover noble metals. Most attention has been paid to the biorecovery process occurring intracellularly and on the cell surface. In this work, we report that Pd nanorods could be biosynthesized by Enterobacter cloacae SgZ-5T in the extracellular space. This bacterium possesses both a direct EET pathway through membrane redox systems and an indirect EET pathway via the self-secreted electron carrier hydroquinone (HQ). When exposed to Pd(II), the bacteria adjusted their metabolic pathway and membrane-bound proteins to secrete riboflavin (RF). However, no HQ was detected in the supernatant in presence of Pd(II). No significant change was observed through metabolomic analysis regarding the abundance of HQ in presence of Pd(II) compared to Pd(II)-free supernatant. Similar results were also obtained through transcriptomic analysis of YqjG gene encoding glutathionyl-HQ reductase synthase. X-ray photoelectron spectroscopic evidence indicated that HQ may adsorb to the surface of Pd nanorods. Moreover, the gene encoding RF synthase (ribE) was up-regulated in the present of Pd(II), suggesting that this bioreduction process induced RF synthase, which had been shown in previous results. The UV–vis spectroscopy data demonstrated that the Pd(II) reduction rate was enhanced by 5%, 5.5% and 30% by the addition of 3.33 μM HQ, 3.33 μM RF and the both, respectively. All these results revealed that the bi-mediators secreted by bacteria were beneficial for biorecovery of Pd. This work is of significance for understanding metal biorecovery processes and natural biogeochemical processes. Keywords: Extracellular electron transfer, Enterobacter cloacae, Palladium, Bi-mediator, Biorecovery

With the proliferation of deep learning and the increasing pressure to deploy inference applications at the edge, many AI chip makers integrate the open source NVIDIA Deep Learning Accelerator (NVDLA) design in their AI solutions. Lack of open source compiler support and having only limited configurability support in the software stacks erect a barrier for developers to freely explore the NVDLA design space at system level. This paper presents an ONNC-based software development platform that includes the first open source compiler for NVDLA-based designs, a virtual platform with various CPU models as well as configurable NVDLA models, and auxiliary tools for debugging. The platform is tightly coupled with the hardware design tradeoffs and provides extendibility for compiler optimization, more CPU types, and more NVDLA hardware configurations. It lifts many restrictions of software development for those who like to leverage the NVDLA design in inference applications.

This paper presents ONNC (Open Neural Network Compiler), a retargetable compilation framework designed to connect ONNX (Open Neural Network Exchange) models to proprietary deep learning accelerators (DLAs). The intermediate representations (IRs) of ONNC have one-to-one mapping to ONNX IRs, thus making porting ONNC to proprietary DLAs much simpler than other compilation frameworks such as TVM and Glow especially for hardware with coarse-grained operators that are not part of the generic IRs in the LLVM backend. ONNC also has a flexible pass manager designed to support compiler optimizations at all levels. A docker image of ONNC bundled with a Vanilla backend is released with this paper to enable fast porting to new hardware targets. To illustrate how an ONNC-based toolkit guides our research and development in DLA design, we present a case study on compiler optimizations for activation memory consumption. The study shows that the Best-Fit algorithm with a proposed heuristic and a reordering scheme may act as a near-optimal strategy, getting the memory consumption close to the ideal lower bound in 11 of 12 models from the ONNX model zoo. To our best knowledge, ONNC is the first open source compilation framework that is specially designed to support the ONNX-based models for both commercial and research projects for deep learning applications.

Anger and related emotions have been identified as triggers in substance use. Forgiveness therapy (FT) targets anger, anxiety, and depression as foci of treatment. Fourteen patients with substance dependence from a local residential treatment facility were randomly assigned to and completed either 12 approximately twice-weekly sessions of individual FT or 12 approximately twice-weekly sessions of an alternative individual treatment based on routine drug and alcohol therapy topics. Participants who completed FT had significantly more improvement in total and trait anger, depression, total and trait anxiety, self-esteem, forgiveness, and vulnerability to drug use than did the alternative treatment group. Most benefits of FT remained significant at 4-month follow-up. These results support FT as an efficacious newly developed model for residential drug rehabilitation.

In this paper we address the severe performance gap caused by high processor clock rates and slow DRAM accesses. We show that even with an aggressive, next-generation memory system using four Direct Rambus channels and an integrated one-megabyte level-two cache, a processor still spends over half of its time stalling for L2 misses. Large cache blocks can improve performance, but only when coupled with wide memory channels. DRAM address mappings also affect performance significantly. We evaluate an aggressive prefetch unit integrated with the L2 cache and memory, controllers. By issuing prefetches only when the Rambus channels are idle, prioritizing them to maximize DRAM row buffer hits, and giving them low replacement priority, we achieve a 43% speedup across 10 of the 26 SPEC2000 benchmarks, without degrading performance an the others. With eight Rambus channels, these ten benchmarks improve to within 10% of the performance of a perfect L2 cache.