Equinox | EPFL Graph Search

A solar equinox is a moment in time when the Sun crosses the Earth's equator, which is to say, appears directly above the equator, rather than north or south of the equator. On the day of the equinox, the Sun appears to rise "due east" and set "due west". This occurs twice each year, around 20 March and 23 September. More precisely, an equinox is traditionally defined as the time when the plane of Earth's equator passes through the geometric center of the Sun's disk. Equivalently, this is the moment when Earth's rotation axis is directly perpendicular to the Sun-Earth line, tilting neither toward nor away from the Sun. In modern times, since the Moon (and to a lesser extent the planets) causes Earth's orbit to vary slightly from a perfect ellipse, the equinox is officially defined by the Sun's more regular ecliptic longitude rather than by its declination. The instants of the equinoxes are currently defined to be when the apparent geocentric longitude of the Sun is 0° and 180°. The

Equinox: Training (for Free) on a Custom Inference Accelerator

Louis Coulon, Mario Paulo Drumond Lages De Oliveira, Babak Falsafi, Martin Jaggi, Arash Pourhabibi Zarandi, Ahmet Caner Yüzügüler

DNN inference accelerators executing online services exhibit low average loads because of service demand variability, leading to poor resource utilization. Unfortunately, reclaiming idle inference cycles is difficult as other workloads can not execute on a custom accelerator. With recent proposals for the use of fixed-point arithmetic in training, there are opportunities for training services to piggyback on inference accelerators. We make the observation that a key challenge in doing so is maintaining service-level latency constraints for inference. We show that relaxing latency constraints in an inference accelerator with ALU arrays that are batching-optimized achieves near-optimal throughput for a given area and power envelope while maintaining inference services' tail latency goals. We present Equinox, a custom inference accelerator designed to piggyback training. Equinox employs a uniform arithmetic encoding to accommodate inference and training and a priority hardware scheduler with adaptive batching that interleaves training during idle inference cycles. For a500𝜇𝑠 inference service time constraint, Equinox achieves 6.67× higher throughput than a latency-optimal inference accelerator. Despite not being optimized for training services, Equinox achieves up to 78% of the throughput of a dedicated training accelerator that saturates the available compute resources and DRAM bandwidth. Finally, Equinox’s controller logic incurs less than 1% power and area overhead, while the uniform encoding (to enable training) incurs 13% power and 4% area overhead compared to a fixed-point inference accelerator.

ACM2021

On advanced daylighting simulations and integrated performance assessment of complex fenestration systems for sunny climates

Chantal Basurto Dávila

The inclusion of daylight in buildings represents several benefits: its use not only signifies a reduction in the building energy consumption through the compensation of electric lighting, it also has positive effects in the execution of human activities. Through its spectral composition, it contributes to create a better interior atmosphere for visual comfort, leading to a better performance on working tasks. It also influences biological human cycles, which has an impact on human alertness, mood and well-being. Compared with electric light, its intensity and dynamic variations have a stimulating effect by providing a connection with the exterior environment. However, for buildings located at low latitudes (between 23°N and S), the inclusion of daylight implies the admission of sunrays, which affect the visual comfort and perception of the indoor environment by altering the interior luminance distribution increasing the risk of glare. An additional unfavourable effect is the increment of the interior thermal loads which represent a risk of overheating for the occupants. In those regions, common practices such as: the reduction of the window size, the use of tinted glazing and solar protection are usually applied to buildings in the search of overcoming such problems. However, the use of such strategies also implies a reduction of the admission of daylight inducing the use of electric light, which is incongruent in countries with large amounts of this natural resource. The use of Complex Fenestration Systems (CFS) may represent a solution given their characteristics: daylight redirection and direct solar rays protection. The latter would contribute to control the admission of solar gains while maintaining a good visual interior environment. However, the appropriate selection of CFS requires a careful evaluation of their features and performance regarding their suitability to the building location. This thesis explores the potentiality of using CFS to improve the interior daylight distribution in buildings located at low latitudes while maintaining a satisfactory visual and thermal interior environment for the occupants. In order to do this, the existing daylight situation of two office rooms located in the central-north of México (Zacatecas 22° 783' N., 102° 583' W, Altitude: 2543m) were monitored from 2011 to 2013. Illuminance and luminance were measured on periods of the year that are crucial for the interior luminous environment (summer and winter solstices together with spring equinox), in order to characterize the existing daylighting situation and to study the dispersion of results obtained using a virtual model which reproduces the features of both offices. The performance of five CFS was then tested using computer simulations in order to assess their suitability to the local sky conditions. For this, three main factors were taken into account: the improvement of the interior daylight distribution, the risk of glare and overheating. The effects of the CFS in a room regarding such conditions were simulated using RADIANCE and Energy Plus using BTDF data (Bidirectional Transmission Distribution Function that characterizes the CFS’s lighting transmission properties) assessed by the use of a gonio-photometer. The assessment was performed first during the winter and summer solstices as well as spring equinox and secondly then on annual basis. The latter was done in order to take into account the daylight variation characteristics of two locations with prevailing clear sky conditions. The results obtained allow determining the CFS that better contributes to a better interior luminous environment in each building without compromising the thermal and visual comfort of the occupants.

EPFL2014

Towards novel glazing with seasonal dynamics based on micro compound parabolic concentrators

Jing Gong, André Gabriel Kostro, Josef Andreas Schuler

Thermal loss and overheating caused by glazing is a key issue in reducing CO2 emission and energy consumption of buildings. A novel glazing based on micro compound parabolic concentrators (CPCs) is proposed. The glazing consists of a polymer layer with embedded micro CPCs, which is attached to a glass pane of glazing. Thanks to the geometry and the micrometric size of the embedded CPCs, the proposed novel glazing can reduce energy consumption in cooling, improve visual comfort and maintain clear view through glazing. In the present work, the potential benefits of such glazing are preliminary estimated for Dubai. The seasonal dynamics are investigated by analyzing the direct solar transmittance for the working hours on the spring equinox and the winter solstice. The improvement of visual comfort is studied based on the assessment of glare. First samples with micro CPCs are fabricated, and the high transparency is achieved. The optical characterization using goniophotometer confirms the seasonal dynamics of the novel glazing.

2018

On advanced daylighting simulations and integrated performance assessment of complex fenestration systems for sunny climates

Chantal Basurto Dávila

EPFL2014