New directions and developments in mobile GPU design

During this decade, pixels have become mobile. Cell phones, PDAs, handheld gaming consoles and other similar devices start to have color displays by standard and color displays are hungry for high-quality graphics. QVGA and VGA display resolutions are common, requiring dedicated hardware for graphics acceleration. Color displays and open platforms also invite games and other applications, which build on the availability of robust graphics. Handheld graphics acceleration is close to its desktop and games console counterparts Ű with content running on an embedded version of OpenGL, the OpenGL ES 2.0, vertex and pixel shaders are a requirement. Floating-point accuracy, lots of texture surfaces, plenty of performance Ű handheld pixels are of good quality and there are lots of them. Handheld gaming drives the handheld 3D graphics performance, but unlike on desktops, vector graphics hardware acceleration will become an even widely spread requirement on new handheld platforms. Applications such as the device's main graphical user interface and interactive maps are driving these requirements. In addition to performance, a strong driver for vector graphics on handhelds is image quality.The first handheld devices, including cell phones, with dedicated 3D graphics accelerators have already hit the market. By 2010, a large number of new cell phones and PDAs will be enabled with hardware vector- and 3D graphics acceleration. The volume of graphics acceleration enabled silicon chips shipping for handheld devices is expected to be significantly higher than for desktop PCs and gaming consoles. This creates a lucrative platform for game and application developers who want to develop handheld content with high-quality graphics.As there are numerous different handheld devices, the industry is fighting against fragmentation Ű widely adopted platforms must be created to enable universal content development across a wide range of platforms and end devices Ű the platform race is already on.All in all, the industry is busy creating all the essential components to bring high-quality programmable pixels to handheld devices. Content developers are already up-to speed to provide winning content for these devices. All in all, the future of handheld pixels looks rosy!

In this chapter, you will tackle two critical subjects in today’s mobile graphics development: portability or multiplatform support and the new features of Open GL ES 2.0.KeywordsGame EngineVoid MainPublic VoidEfficient GraphicConfiguration ChooserThese keywords were added by machine and not by the authors. This process is experimental and the keywords may be updated as the learning algorithm improves.

Current radio frequency (RF) sensors at the Edge lack the computational resources to support practical, in-situ training for intelligent spectrum monitoring. This is true for sensor data classification in general. We propose a solution via Deep Delay Loop Reservoir Computing (DLR), a processing architecture that supports general machine learning algorithms on compact mobile devices by leveraging delay-loop reservoir computing in combination with innovative electro-optical hardware. With both digital and photonic realizations of our design of the loops, DLR delivers reductions in form factor, hardware complexity and latency, compared to the State-of-the-Art (SoA). The main impact of the reservoir is to project the input data into a higher dimensional space of reservoir state vectors in order to linearly separate the input classes. Once the classes are well separated, traditionally complex, power-hungry classification models are no longer needed for the learning process. Yet, even with simple classifiers based on Ridge regression (RR), the complexity grows at least quadratically with the input size. Hence, the hardware reduction required for training on compact devices is in contradiction with the large dimension of state vectors. DLR employs a RR-based classifier to exceed the SoA accuracy, while further reducing power consumption by leveraging the architecture of parallel (split) loops. We present DLR architectures composed of multiple smaller loops whose state vectors are linearly combined to create a lower dimensional input into Ridge regression. We demonstrate the advantages of using DLR for two distinct applications: RF Specific Emitter Identification (SEI) for IoT authentication, and wireless protocol recognition for IoT situational awareness.

Data transfer overhead between computing cores and memory hierarchy has been a persistent issue for von Neumann architectures and the problem has only become more challenging with the emergence of manycore systems. A conceptually powerful approach to mitigate this overhead is to bring the computation closer to data, known as Near Data Computing (NDC). Recently, NDC has been investigated in different flavors for CPU-based multicores, while the GPU domain has received little attention. In this paper, we present a novel NDC solution for GPU architectures with the objective of minimizing on-chip data transfer between the computing cores and Last-Level Cache (LLC). To achieve this, we first identify frequently occurring Load-Compute-Store instruction chains in GPU applications. These chains, when offloaded to a compute unit closer to where the data resides, can significantly reduce data movement. We develop two offloading techniques, called LLC-Compute and Omni-Compute. The first technique, LLC-Compute, augments the LLCs with computational hardware for handling the computation offloaded to them. The second technique (Omni-Compute) employs simple bookkeeping hardware to enable GPU cores to compute instructions offloaded by other GPU cores. Our experimental evaluations on nine GPGPU workloads indicate that the LLC-Compute technique provides, on an average, 19% performance improvement (IPC), 11% performance/watt improvement, and 29% reduction in on-chip data movement compared to the baseline GPU design. The Omni-Compute design boosts these benefits to 31%, 16% and 44%, respectively.

Low operating temperature micro solid oxide fuel cell (micro-SOFC) systems which utilize liquid fuels are attractive for high energy efficiency and density. Long time operation is possible for micro-SOFCs by refueling and it is not restricted by battery charge. Hence, it is expected to alternative power source for compact mobile electronic devices which consume a large amount of power. In previous, diverse types of micro-SOFCs focused on its performance was reported and a maximum power density of 1 W/cm2 at 500oC was achieved. However, they have many problems of durability and repeatability for practical use. Additionally, it is necessary to consider how to heat up micro-SOFC cells to operating temperature, compatibility with other electronic circuits and thermally self-sustained condition. In this study, micro-SOFC system was proposed in order to adopt the micro-SOFCs into a power source of compact mobile electric devices. We designed the micro-SOFC package with thermal insulation and heat exchanger. Its insulation performance was experimentally confirmed and power generation test with micro heater in the micro-SOFC package was performed. Micro-SOFC packages require thermal insulation to keep the surface temperature below about 85oC, which is the maximum operating temperature of other electronic circuits. We applied vacuum insulation widely used for thermal insulation in MEMS devices to the package. The heat of the exhaust gas was recovered by the liquid fuel through heat exchanger. From the calculation of temperature distribution by the finite element method, surface temperature of the package was kept from 57 to 88oC when operating temperature is 500oC. As long as the liquid is evaporated, temperature of the heat exchanger is kept constant. Therefore, heat conduction from high temperature glass flow channels to vacuum insulator was prevented and low surface temperature was achieved. Applying micro-SOFC system as independent power sources, thermally self-sustained condition is required, i.e. micro-SOFC packages should keep operating temperature more than 300oC only by the generated heat from reaction of fuel cells during power generation. For thermally self-sustained condition, the amount of heat generated during power generation needs to exceed that of heat loss from surface of the package and transferred heat by exhaust gas. We calculated thermally self-sustained condition of a micro-SOFC system with 4 W output fueled by methanol and water that has reformer and heat exchanger from exhaust gas. From the result, generated heat from reaction of fuel cells was 4.8W, heat loss from surface was 1.5 W and transferred heat by exhaust gas was 4.6 W at 500oC. The result confirmed that transferred heat by exhaust gas was dominated in this system. Hence, recovering the heat of exhaust gas with the gas-liquid heat exchanger was essential to achieve thermally self-sustained condition. Micro-SOFC cells and micro heater were fabricated in general MEMS technology. Power generation test and electrochemical impedance spectroscopy of micro-SOFC (Pt-Pd | 15 mol% Y-doped BaZrO3 | Pt-Pd) by micro heater were measured in the temperature range of 200-330oC. A maximum power density of 1.12 mW/cm2 and open circuit voltage of 1.09 V at 330oC were obtained using wet H2 as fuel and air as the oxidant. As a result of impedance measurement, electrode reaction resistance was very high in cell total resistance. Therefore, the improvement of electrode will reduce cell resistance to a level for actual utility. We actually fabricated the micro-SOFC package installed vacuum structure and heat exchanger to evaluate cooling performance of the insulation package and performed power generation test in the package. In the case of heat exchanger filled with 50 vol% ethanol solution, surface temperature was 89.6oC when micro heater temperature was 500oC. In accordance with the result, it was possible to keep surface temperature of the package sufficiently low with heat exchanger. As a result of power generation test of micro-SOFC inside the package, a maximum power density of 1.27 mW/cm2 and open circuit voltage of 0.94 V at 300oC were obtained using wet 25% or 50% H2 as fuel and air as the oxidant. Additionally, surface temperature of the package was kept under 80oC. Therefore I achieved power generation in compact micro-SOFC system with thermal insulation. The micro-SOFC package was designed for applying micro-SOFCs to power source of mobile electronic devices in this study. As a result of calculation, the package installed vacuum insulation and heat exchanger could achieve thermally self-sustained condition and low surface temperature. Power generation test of micro-SOFC inside the prototype package was achieved and the package kept low surface temperature. Hence, this study presented that applicability of micro-SOFC system to power source for compact mobile electronic devices.

While the potential of mobile information visualisation is widely recognized, there is still relatively little research in this area and few practical guidelines for the design of mobile information visualisation interfaces. Indeed, it would appear that there is still a general feeling in the interface design community that mobile visualisation should be limited to simple operations and small scale data. Information visualisation research has concentrated thus far on desktop PCs and larger displays while interfaces for more compact mobile device have been neglected. This is in spite of the increasing popularity and widespread use of smart-phones and other new mobile technologies. In this paper we address this issue by developing a set of low-level interface design guidelines for mobile information visualisation development. This is done by considering a basic set of interactions and relating these to mobile device limitations. Our results suggest that the mindful application of existing information visualisation techniques can overcome many mobile device limitations and that proper implementation of interaction mechanisms and animated view transitions are key to effective mobile information visualisation. This is illustrated with case studies looking at a coordinated map and timeline interface for geo-temporal data, a distorted scatter-plot, and a space filling hierarchy view.

The optically transparent antenna technology has been gaining attention in consumer electronic industry owing to its potential to release the antenna volume constraints in compact mobile devices. Earlier studies have been heavily focused on the transparent antenna integration to the display for smartphone applications, for example, to improve its spatial coverage of 5G mm-wave at the forehead side. To our knowledge, the application of the transparent antenna technology to smart glasses and augmented reality glasses, especially considering the constraints from the coexistence with optical features in the lens stack-up, has remained unexplored. Thus, for the first time, we suggest a feasible existence solution between antennas and the RF lossy layers in the lens stack-up, to reduce the negative impacts on antenna performances. A slot loop antenna formed by metal-mesh based transparent antenna film and metallic glass frame has been investigated. Experimental results of such an antenna on human head phantom shows a total efficiency better than −4.5 dB in 2.4 GHz band. Even though it is demonstrated through a 2.4 GHz antenna example, the same principle can be applied to any other sub-6 GHz antenna applications, i.e., LTE, WiFi, GNSS, etc.

The continued drive for more compact and lightweight handheld mobile devices has forcibly pushed the electronics assembly industry to look for novel packaging and assembly technologies. One of the newest advances in recent years is for semiconductors to be stacked, one on top of the other, in a single package. This die stacking allows system designers to take advantage of the often more readily available “Z” axis of the cubic area while saving on the valuable “X” and “Y” square dimensional space on PCB layouts.

In this paper, a low complexity microwave based direction finding (DF) system is presented. The proposed system consists of single six-port (SP) circuit integrated with a reconfigurable multiple-input-multiple-output (MIMO) antenna system. The proposed design is a complete integrated solution with DF capabilities for cognitive radio platforms. The design is suitable to be used in compact wireless handheld and mobile communication devices. The SP circuit covers a wide frequency band (1.68∼2.25 GHz) while the reconfigurable MIMO antenna system cover several frequency bands between 0.7 GHz to 3 GHz. The fabricated integrated system achieves ±16° accuracy in its direction of arrival estimates.

Color-changing materials, such as photochromic pigments, allow objects to have reprogrammable multicolor surface images. Existing systems that reprogram these images are based on projectors and LEDs, each with advantages and limitations in device portability and image resolution. In this paper, we present ChromoLCD, a surface reprogrammer that uses a liquid crystal display (LCD) to achieve a compact handheld device without sacrificing image resolution. ChromoLCD consists of an LCD panel with a custom backlight containing R,G,B and UV LEDs, forming high-resolution light patterns with the required wavelengths. The compact form factor of ChromoLCD enables on-the-fly reprogramming of everyday surfaces. Our technical evaluation shows that ChromoLCD achieves a resolution of 25 ppi, which is 8 times better than the prior work. We demonstrate ChromoLCD with three applications, including the stamping of reprogrammable AR markers on a kitchen counter, on-the-fly designs on personal accessories, and reference pictures on a whiteboard.

As GPUs scale their low-precision matrix math throughput to boost deep learning (DL) performance, they upset the balance between math throughput and memory system capabilities. We demonstrate that a converged GPU design trying to address diverging architectural requirements between FP32 (or larger)-based HPC and FP16 (or smaller)-based DL workloads results in sub-optimal configurations for either of the application domains. We argue that a C omposable O n- PA ckage GPU (COPA-GPU) architecture to provide domain-specialized GPU products is the most practical solution to these diverging requirements. A COPA-GPU leverages multi-chip-module disaggregation to support maximal design reuse, along with memory system specialization per application domain. We show how a COPA-GPU enables DL-specialized products by modular augmentation of the baseline GPU architecture with up to 4× higher off-die bandwidth, 32× larger on-package cache, and 2.3× higher DRAM bandwidth and capacity, while conveniently supporting scaled-down HPC-oriented designs. This work explores the microarchitectural design necessary to enable composable GPUs and evaluates the benefits composability can provide to HPC, DL training, and DL inference. We show that when compared to a converged GPU design, a DL-optimized COPA-GPU featuring a combination of 16× larger cache capacity and 1.6× higher DRAM bandwidth scales per-GPU training and inference performance by 31% and 35%, respectively, and reduces the number of GPU instances by 50% in scale-out training scenarios.

We propose Conslide, a new Thai input method for compact mobile devices. To address the complexity of Thai language, our design adopts drawing approach for text input. This results in a smaller keyboard (more area for applications). With appropriate layout, users can input words without lifting finger from the screen, hence the better efficiency. Empirical result shows that the average movement distance and minimal movement time required for each Thai word are comparable to those of the quasi-Kedmanee keyboards such as Droidsans Thai keyboard.

SneakLeak+: Large-scale klepto apps analysis