Search

The rapidly growing compute demands of AI necessitate the creation of new computing architectures and approaches. Tenstorrent designed its architecture (embodied in Grayskull and Wormhole devices) to tackle this challenge via two fundamental and synergistic approaches. The first is via compute-on-packets fabric that is built from ground up for massive scaleout. The second is the ability to execute dynamic computation, built into the compiler, runtime software and hardware architecture. By combining these approaches, TensTorrent will enable continued scaling of AI workloads.

Because of the limitations of current staging systems and evolving definitions, there are limited data on oligometastatic non-small cell lung cancer (NSCLC) epidemiology. The purpose of this study was to evaluate metastatic disease burden and the incidence of oligometastatic disease using recent clinical trial eligibility criteria.A cohort of patients with metastatic NSCLC, diagnosed from 2016 to 2019, were randomly sampled from a curated tumor registry. Definitions for oligometastatic disease were obtained from relevant clinical trials. The Stanford Cancer Institute Research Database was used to identify baseline patient factors, systemic and local therapy, extent and location of metastatic lesions, and survival outcomes.Among 120 patients presenting with metastatic NSCLC, the majority had de novo metastatic disease (75%) with a median of 4 metastatic lesions involving 3 organ systems. Of these, 37.5% would have been eligible for at least 1 oligometastatic trial, with 28.3% meeting criteria for the MD Anderson Cancer Center trial, 20.0% for NRG-LU002, 6.7% for SINDAS, and 16.7% for SABR-COMET. By adding malignant pleural effusions and early progression as exclusionary criteria, only 54.1% of patients with ≤3 synchronous metastases were eligible for consideration of local therapy. Early progression on systemic therapy was associated with worse survival (10.0 vs 42.4 months; P.001), whereas presence of malignant pleural effusions was not. Of those tumors identified as oligometastatic, 44.4% received local therapy and 28.9% underwent ablative therapy to all sites. There was a trend toward greater overall survival (44.4 vs 24.9 months; P = .055) and progression-free survival (8.0 vs 5.4 months; P = .06) in patients meeting eligibility for at least 1 oligometastatic trial.Around 48% of patients with metastatic NSCLC had ≤3 metastases at presentation and 28% met clinical trial criteria for oligometastatic disease. Future research is needed to better define the oligometastatic state and identify patients most likely to benefit from local therapy.

Traditional management of metastatic sarcoma primarily relies on systemic therapy, with surgery often used for tumor control. We analyzed the rates of recurrence, overall survival, and treatment complications in patients undergoing either surgical resection or stereotactic body radiation therapy (SBRT) for metastatic sarcoma of the bone and/or soft tissue.The records of patients with metastatic sarcoma between 2009 and 2020 were reviewed. Local recurrence (LR) was defined as tumor growth or recurrence at the tumor site. Cumulative LR incidence was analyzed accounting for the competing risk of death, and groups were compared using the Gray test. Overall survival (OS) was assessed using the Kaplan-Meier method and log-rank test. Hazard ratios were determined using the Cox proportional hazards model.A total of 525 metastatic lesions in 217 patients were analyzed. The mean age of patients was 57 years (range, 4-88 years). The lung was the predominant site treated (50%), followed by intra-abdominal (13%) and soft tissue (11%). Two-year cumulative incidences of LR for surgery and SBRT were 14.8% (95% confidence interval [CI], 11.6%-18.5%) and 1.7% (95% CI, 0.1%-8.2%), respectively (P = .003). Local recurrence occurred in 72 (16.5%) of 437 tumors treated with surgery and 2 (2.3%) of 88 tumors treated with SBRT. The adjusted hazard ratio for LR of lesions treated surgically was 11.5 (P = .026) when controlling for tumor size and tumor site. Median OS was 29.8 months (95% CI, 25.6-40.9 months). There were 47 surgical complications among a total of 275 procedures (18%). Of 58 radiation treatment courses, radiation-related toxic effects were reported during the treatment of 7 lesions (12%), and none were higher than grade 2.We observed excellent local control among patients selected for treatment with SBRT for metastatic sarcoma, with no evidence of an increase in LR after SBRT compared with surgical management. Further investigation is necessary to better define the most appropriate local control strategies for metastatic sarcoma.

Hybrid configurable logic block architectures for field-programmable gate arrays that contain a mixture of lookup tables and hardened multiplexers are evaluated toward the goal of higher logic density and area reduction. Multiple hybrid configurable logic block architectures, both nonfracturable and fracturable with varying MUX:LUT logic element ratios are evaluated across two benchmark suites (VTR and CHStone) using a custom tool flow consisting of LegUp-HLS, Odin-II front-end synthesis, ABC logic synthesis and technology mapping, and VPR for packing, placement, routing, and architecture exploration. Technology mapping optimizations that target the proposed architectures are also implemented within ABC. Experimentally, we show that for nonfracturable architectures, without any mapper optimizations, we naturally save up to $\sim 8$ % area postplace and route; both accounting for complex logic block and routing area while maintaining mapping depth. With architecture-aware technology mapper optimizations in ABC, additional area is saved, post-place-and-route. For fracturable architectures, experiments show that only marginal gains are seen after place-and-route up to $\sim 2$ %. For both nonfracturable and fracturable architectures, we see minimal impact on timing performance for the architectures with best area-efficiency.

Coarse-grained reconfigurable architectures (CGRAs) have gained traction as a potential solution to implement accelerators for compute-intensive kernels, particularly in domains requiring hardware programmability. Architecture and CAD for CGRAs are tightly intertwined, with many prior works having combined architectures and tools. In this work, we present an architecture-agnostic integer linear programming (ILP) approach for CGRA mapping, integrated within an open-source CGRA architecture evaluation framework. The mapper accepts an application and an architecture description as input and can generate an optimal mapping, if indeed mapping is feasible. An experimental study demonstrates its effectiveness over a range of CGRA architectures.

We describe an open-source software framework,CGRA-ME, for the modeling and exploration of coarse-grained reconfigurable architectures (CGRAs). CGRAs are programmable hardware devices having large ALU-like logic blocks, and datapath bus-style inter-connect. CGRAs are positioned between fine-grained FPGAs and standard-cell ASICs on the spectrum of programmability - they are less flexible than FPGAs, yet are more flexible than ASICs. With CGRA-ME, an architect can describe a CGRA architecture in an XML-based language. The framework also allows the architect to map benchmarks onto the architecture and provides automatic generation of Verilog RTL for the modeled architecture. This allows the architect to simulate for verification purposes, and perform synthesis to either an ASIC or FPGA-overlay implementation of the CGRA, assessing performance, area, and power consumption. In an experimental study, we use CGRA-ME to model, map benchmarks onto, and evaluate several variants of a widely known CGRA, considering both standard-cell and FPGA-overlay physical realizations of the CGRA.

Coarse-grained reconfigurable arrays (CGRAs) are a style of programmable logic device situated between FPGAs and custom ASICs on the spectrum of programmability, performance, power and cost. CGRAs have been proposed by both academia and industry; however, prior works have been mainly self-contained without broad architectural exploration and comparisons with competing CGRAs. We present CGRA-ME - a unified CGRA framework that encompasses generic architecture description, architecture modelling, application mapping, and physical implementation. Within this framework, we discuss our architecture description language CGRA-ADL, a generic LLVM-based simulated annealing mapper, and a standard cell flow for physical implementation. An architecture exploration case study is presented, highlighting the capabilities of CGRA-ME by exploring a variety of architectures with varying functionality, interconnect, array size, and execution contexts through the mapping of application benchmarks and the production of standard cell designs.

Aircraft structures have varying stiffness levels making them flexible. Consequently, this elastic property becomes increasingly important at high speeds affecting the flight dynamics of the aircraft. In high speed aircraft such as the F/A-18, elastic structural properties must be accounted for to ensure confidence in predicted flight dynamics in order to avoid adverse aeroelastic phenomena throughout flight. Data from the F/A-18 Active Aeroelastic Wing (AAW) program was used to create aeroservoelastic (ASE) models at varying flight conditions. This paper focuses on the integration process of these models into the NASA Dryden F/A-18 digital simulator. The ASE models were discretized and added in parallel with the traditional rigid body response model to ensure minimal disruption to the existing operating framework of the digital simulator. An interpolation scheme was used to construct ASE models within the known flight condition models. Data was processed through the state space ASE models to compute the elastic effects during flight. Total flight dynamics from the simulation were analyzed and showed expected behavior for the combined elastic and rigid body components in flight. The model integration process will be discussed along with an overview of the results of the combined rigid body and elastic flight dynamics output from the simulator.

In this paper, we present a holistic methodology and framework that uniquely integrates the aerodynamics reduced order model (ROM), structural dynamics model, and active structural control into a single platform to enable integrated, ultra-fast ROMs for efficient Aeroelastic (AE) and Aeroservoelastic (ASE) analysis, and controller design. The aerodynamic ROM (AeroROM) was developed based on the system identification approach to establish the mapping relationship between the structural displacement and the Generalized Aerodynamic Force, and hence, could be used to accurately predict aerodynamic response at various flight conditions. The AeroROM was then coupled with a rigid-body dynamics model by virtue of appropriate sensors, actuators, and control laws to form an integrated ASE ROM for closed-loop simulation and analysis. Several case studies using 2 DOF rigid body (pitch-plunge motion) and 3 DOF (pitch-plunge-flap motion) airfoil in conjunction with control laws and at different flight conditions were performed. The case studies demonstrate that the integrated ROM can be used for accurate AE and ASE analysis. The ROM results show excellent agreement with high-fidelity CFD-CSD simulations (

This paper presents a case for a hybrid configurable logic block that contains a mixture of LUTs and hardened multiplexers towards the goal of higher logic density and area reduction. Technology mapping optimizations, called MuxMap, that target the proposed architecture are implemented using a modified version of the mapper in the ABC logic synthesis tool. VPR is used to model the new hybrid configurable logic block and verify post place and route implementation. Multiple hybrid configurable logic block architectures with varying MUX:LUT ratios are evaluated across three benchmark suites with both Quartus II and Odin-II front-end RTL synthesis tools. Experimentally, we show that without any mapper optimizations we naturally save ~4% area post place and route and with MuxMap optimizations in ABC yielding ~6% area reduction post place and route while maintaining mapping depth, overall configurable logic block count, and routing demand.

The development of the CubeSat standard, a picosatellite standard, has become a tool that encourages engineering collaboration, trains students with real-world satellite experience, and provides technology advancement in the aerospace industry. The Poly-Picosatellite Orbital Deployer (P-POD), in conjuction with the CubeSat standard, plays a key role in providing access to space for CubeSats. Developing satellites at the CubeSat level highlight the increasing opportunities for access to space while yielding quicker development times. The upcoming launches demonstrate the growing interest of universities, companies, and government organizations to develop CubeSats to perform valuable scientific experiments and missions. The educational benefits of CubeSat development are emphasized by providing an ideal training ground for future scientists and engineers. I. Background