Matching Items (21)
Filtering by
- All Subjects: Computer Engineering
- Genre: Masters Thesis
- Creators: Ogras, Umit Y.
- Creators: Seo, Jae-Sun
Description
Mobile platforms are becoming highly heterogeneous by combining a powerful multiprocessor system-on-chip (MpSoC) with numerous resources including display, memory, power management IC (PMIC), battery and wireless modems into a compact package. Furthermore, the MpSoC itself is a heterogeneous resource that integrates many processing elements such as CPU cores, GPU, video, image, and audio processors. As a result, optimization approaches targeting mobile computing needs to consider the platform at various levels of granularity.
Platform energy consumption and responsiveness are two major considerations for mobile systems since they determine the battery life and user satisfaction, respectively. In this work, the models for power consumption, response time, and energy consumption of heterogeneous mobile platforms are presented. Then, these models are used to optimize the energy consumption of baseline platforms under power, response time, and temperature constraints with and without introducing new resources. It is shown, the optimal design choices depend on dynamic power management algorithm, and adding new resources is more energy efficient than scaling existing resources alone. The framework is verified through actual experiments on Qualcomm Snapdragon 800 based tablet MDP/T. Furthermore, usage of the framework at both design and runtime optimization is also presented.
Platform energy consumption and responsiveness are two major considerations for mobile systems since they determine the battery life and user satisfaction, respectively. In this work, the models for power consumption, response time, and energy consumption of heterogeneous mobile platforms are presented. Then, these models are used to optimize the energy consumption of baseline platforms under power, response time, and temperature constraints with and without introducing new resources. It is shown, the optimal design choices depend on dynamic power management algorithm, and adding new resources is more energy efficient than scaling existing resources alone. The framework is verified through actual experiments on Qualcomm Snapdragon 800 based tablet MDP/T. Furthermore, usage of the framework at both design and runtime optimization is also presented.
ContributorsGupta, Ujjwala (Author) / Ogras, Umit Y. (Thesis advisor) / Ozev, Sule (Committee member) / Chakrabarti, Chaitali (Committee member) / Arizona State University (Publisher)
Created2014
Description
Heterogeneous multiprocessor systems-on-chip (MPSoCs) powering mobile platforms integrate multiple asymmetric CPU cores, a GPU, and many specialized processors. When the MPSoC operates close to its peak performance, power dissipation easily increases the temperature, hence adversely impacts reliability. Since using a fan is not a viable solution for hand-held devices, there is a strong need for dynamic thermal and power management (DTPM) algorithms that can regulate temperature with minimal performance impact. This abstract presents a DTPM algorithm based on a practical temperature prediction methodology using system identification. The DTPM algorithm dynamically computes a power budget using the predicted temperature, and controls the types and number of active processors as well as their frequencies. Experiments on an octa-core big.LITTLE processor and common Android apps demonstrate that the proposed technique predicts temperature within 3% accuracy, while the DTPM algorithm provides around 6x reduction in temperature variance, and as large as 16% reduction in total platform power compared to using a fan.
ContributorsSingla, Gaurav (Author) / Ogras, Umit Y. (Thesis advisor) / Bakkaloglu, Bertan (Committee member) / Unver, Ali (Committee member) / Arizona State University (Publisher)
Created2015
Description
Object tracking is an important topic in multimedia, particularly in applications such as teleconferencing, surveillance and human-computer interface. Its goal is to determine the position of objects in images continuously and reliably. The key steps involved in object tracking are foreground detection to detect moving objects, clustering to enable representation of an object by its centroid, and tracking the centroids to determine the motion parameters.
In this thesis, a low cost object tracking system is implemented on a hardware accelerator that is a warp based processor for SIMD/Vector style computations. First, the different foreground detection techniques are explored to figure out the best technique that involves the least number of computations without compromising on the performance. It is found that the Gaussian Mixture Model proposed by Zivkovic gives the best performance with respect to both accuracy and number of computations. Pixel level parallelization is applied to this algorithm and it is mapped onto the hardware accelerator.
Next, the different clustering algorithms are studied and it is found that while DBSCAN is highly accurate and robust to outliers, it is very computationally intensive. In contrast, K-means is computationally simple, but it requires that the number of means to be specified beforehand. So, a new clustering algorithm is proposed that uses a combination of both DBSCAN and K-means algorithm along with a diagnostic algorithm on K-means to estimate the right number of centroids. The proposed hybrid algorithm is shown to be faster than the DBSCAN algorithm by ~2.5x with minimal loss in accuracy. Also, the 1D Kalman filter is implemented assuming constant acceleration model. Since the computations involved in Kalman filter is just a set of recursive equations, the sequential model in itself exhibits good performance, thereby alleviating the need for parallelization. The tracking performance of the low cost implementation is evaluated against the sequential version. It is found that the proposed hybrid algorithm performs very close to the reference algorithm based on the DBSCAN algorithm.
In this thesis, a low cost object tracking system is implemented on a hardware accelerator that is a warp based processor for SIMD/Vector style computations. First, the different foreground detection techniques are explored to figure out the best technique that involves the least number of computations without compromising on the performance. It is found that the Gaussian Mixture Model proposed by Zivkovic gives the best performance with respect to both accuracy and number of computations. Pixel level parallelization is applied to this algorithm and it is mapped onto the hardware accelerator.
Next, the different clustering algorithms are studied and it is found that while DBSCAN is highly accurate and robust to outliers, it is very computationally intensive. In contrast, K-means is computationally simple, but it requires that the number of means to be specified beforehand. So, a new clustering algorithm is proposed that uses a combination of both DBSCAN and K-means algorithm along with a diagnostic algorithm on K-means to estimate the right number of centroids. The proposed hybrid algorithm is shown to be faster than the DBSCAN algorithm by ~2.5x with minimal loss in accuracy. Also, the 1D Kalman filter is implemented assuming constant acceleration model. Since the computations involved in Kalman filter is just a set of recursive equations, the sequential model in itself exhibits good performance, thereby alleviating the need for parallelization. The tracking performance of the low cost implementation is evaluated against the sequential version. It is found that the proposed hybrid algorithm performs very close to the reference algorithm based on the DBSCAN algorithm.
ContributorsSasikumar, Asha (Author) / Chakrabarti, Chaitali (Thesis advisor) / Ogras, Umit Y. (Committee member) / Suppapola, Antonia Pappandreau (Committee member) / Arizona State University (Publisher)
Created2015
Description
Graph theory is a critical component of computer science and software engineering, with algorithms concerning graph traversal and comprehension powering much of the largest problems in both industry and research. Engineers and researchers often have an accurate view of their target graph, however they struggle to implement a correct, and efficient, search over that graph.
To facilitate rapid, correct, efficient, and intuitive development of graph based solutions we propose a new programming language construct - the search statement. Given a supra-root node, a procedure which determines the children of a given parent node, and optional definitions of the fail-fast acceptance or rejection of a solution, the search statement can conduct a search over any graph or network. Structurally, this statement is modelled after the common switch statement and is put into a largely imperative/procedural context to allow for immediate and intuitive development by most programmers. The Go programming language has been used as a foundation and proof-of-concept of the search statement. A Go compiler is provided which implements this construct.
To facilitate rapid, correct, efficient, and intuitive development of graph based solutions we propose a new programming language construct - the search statement. Given a supra-root node, a procedure which determines the children of a given parent node, and optional definitions of the fail-fast acceptance or rejection of a solution, the search statement can conduct a search over any graph or network. Structurally, this statement is modelled after the common switch statement and is put into a largely imperative/procedural context to allow for immediate and intuitive development by most programmers. The Go programming language has been used as a foundation and proof-of-concept of the search statement. A Go compiler is provided which implements this construct.
ContributorsHenderson, Christopher (Author) / Bansal, Ajay (Thesis advisor) / Lindquist, Timothy (Committee member) / Acuna, Ruben (Committee member) / Arizona State University (Publisher)
Created2018
Description
Deep neural networks (DNN) have shown tremendous success in various cognitive tasks, such as image classification, speech recognition, etc. However, their usage on resource-constrained edge devices has been limited due to high computation and large memory requirement.
To overcome these challenges, recent works have extensively investigated model compression techniques such as element-wise sparsity, structured sparsity and quantization. While most of these works have applied these compression techniques in isolation, there have been very few studies on application of quantization and structured sparsity together on a DNN model.
This thesis co-optimizes structured sparsity and quantization constraints on DNN models during training. Specifically, it obtains optimal setting of 2-bit weight and 2-bit activation coupled with 4X structured compression by performing combined exploration of quantization and structured compression settings. The optimal DNN model achieves 50X weight memory reduction compared to floating-point uncompressed DNN. This memory saving is significant since applying only structured sparsity constraints achieves 2X memory savings and only quantization constraints achieves 16X memory savings. The algorithm has been validated on both high and low capacity DNNs and on wide-sparse and deep-sparse DNN models. Experiments demonstrated that deep-sparse DNN outperforms shallow-dense DNN with varying level of memory savings depending on DNN precision and sparsity levels. This work further proposed a Pareto-optimal approach to systematically extract optimal DNN models from a huge set of sparse and dense DNN models. The resulting 11 optimal designs were further evaluated by considering overall DNN memory which includes activation memory and weight memory. It was found that there is only a small change in the memory footprint of the optimal designs corresponding to the low sparsity DNNs. However, activation memory cannot be ignored for high sparsity DNNs.
To overcome these challenges, recent works have extensively investigated model compression techniques such as element-wise sparsity, structured sparsity and quantization. While most of these works have applied these compression techniques in isolation, there have been very few studies on application of quantization and structured sparsity together on a DNN model.
This thesis co-optimizes structured sparsity and quantization constraints on DNN models during training. Specifically, it obtains optimal setting of 2-bit weight and 2-bit activation coupled with 4X structured compression by performing combined exploration of quantization and structured compression settings. The optimal DNN model achieves 50X weight memory reduction compared to floating-point uncompressed DNN. This memory saving is significant since applying only structured sparsity constraints achieves 2X memory savings and only quantization constraints achieves 16X memory savings. The algorithm has been validated on both high and low capacity DNNs and on wide-sparse and deep-sparse DNN models. Experiments demonstrated that deep-sparse DNN outperforms shallow-dense DNN with varying level of memory savings depending on DNN precision and sparsity levels. This work further proposed a Pareto-optimal approach to systematically extract optimal DNN models from a huge set of sparse and dense DNN models. The resulting 11 optimal designs were further evaluated by considering overall DNN memory which includes activation memory and weight memory. It was found that there is only a small change in the memory footprint of the optimal designs corresponding to the low sparsity DNNs. However, activation memory cannot be ignored for high sparsity DNNs.
ContributorsSrivastava, Gaurav (Author) / Seo, Jae-Sun (Thesis advisor) / Chakrabarti, Chaitali (Committee member) / Berisha, Visar (Committee member) / Arizona State University (Publisher)
Created2018
Description
Hardware implementation of deep neural networks is earning significant importance nowadays. Deep neural networks are mathematical models that use learning algorithms inspired by the brain. Numerous deep learning algorithms such as multi-layer perceptrons (MLP) have demonstrated human-level recognition accuracy in image and speech classification tasks. Multiple layers of processing elements called neurons with several connections between them called synapses are used to build these networks. Hence, it involves operations that exhibit a high level of parallelism making it computationally and memory intensive. Constrained by computing resources and memory, most of the applications require a neural network which utilizes less energy. Energy efficient implementation of these computationally intense algorithms on neuromorphic hardware demands a lot of architectural optimizations. One of these optimizations would be the reduction in the network size using compression and several studies investigated compression by introducing element-wise or row-/column-/block-wise sparsity via pruning and regularization. Additionally, numerous recent works have concentrated on reducing the precision of activations and weights with some reducing to a single bit. However, combining various sparsity structures with binarized or very-low-precision (2-3 bit) neural networks have not been comprehensively explored. Output activations in these deep neural network algorithms are habitually non-binary making it difficult to exploit sparsity. On the other hand, biologically realistic models like spiking neural networks (SNN) closely mimic the operations in biological nervous systems and explore new avenues for brain-like cognitive computing. These networks deal with binary spikes, and they can exploit the input-dependent sparsity or redundancy to dynamically scale the amount of computation in turn leading to energy-efficient hardware implementation. This work discusses configurable spiking neuromorphic architecture that supports multiple hidden layers exploiting hardware reuse. It also presents design techniques for minimum-area/-energy DNN hardware with minimal degradation in accuracy. Area, performance and energy results of these DNN and SNN hardware is reported for the MNIST dataset. The Neuromorphic hardware designed for SNN algorithm in 28nm CMOS demonstrates high classification accuracy (>98% on MNIST) and low energy (51.4 - 773 (nJ) per classification). The optimized DNN hardware designed in 40nm CMOS that combines 8X structured compression and 3-bit weight precision showed 98.4% accuracy at 33 (nJ) per classification.
ContributorsKolala Venkataramanaiah, Shreyas (Author) / Seo, Jae-Sun (Thesis advisor) / Chakrabarti, Chaitali (Committee member) / Cao, Yu (Committee member) / Arizona State University (Publisher)
Created2018
Description
The reduced availability of 3He is a motivation for developing alternative neutron detectors. 6Li-enriched CLYC (Cs2LiYCl6), a scintillator, is a promising candidate to replace 3He. The neutron and gamma ray signals from CLYC have different shapes due to the slower decay of neutron pulses. Some of the well-known pulse shape discrimination techniques are charge comparison method, pulse gradient method and frequency gradient method. In the work presented here, we have applied a normalized cross correlation (NCC) approach to real neutron and gamma ray pulses produced by exposing CLYC scintillators to a mixed radiation environment generated by 137Cs, 22Na, 57Co and 252Cf/AmBe at different event rates. The cross correlation analysis produces distinctive results for measured neutron pulses and gamma ray pulses when they are cross correlated with reference neutron and/or gamma templates. NCC produces good separation between neutron and gamma rays at low (< 100 kHz) to mid event rate (< 200 kHz). However, the separation disappears at high event rate (> 200 kHz) because of pileup, noise and baseline shift. This is also confirmed by observing the pulse shape discrimination (PSD) plots and figure of merit (FOM) of NCC. FOM is close to 3, which is good, for low event rate but rolls off significantly along with the increase in the event rate and reaches 1 at high event rate. Future efforts are required to reduce the noise by using better hardware system, remove pileup and detect the NCC shapes of neutron and gamma rays using advanced techniques.
ContributorsChandhran, Premkumar (Author) / Holbert, Keith E. (Thesis advisor) / Spanias, Andreas (Committee member) / Ogras, Umit Y. (Committee member) / Arizona State University (Publisher)
Created2015
Description
Historically, wireless communication devices have been developed to process one specific waveform. In contrast, a modern cellular phone supports multiple waveforms corresponding to LTE, WCDMA(3G) and 2G standards. The selection of the network is controlled by software running on a general purpose processor, not by the user. Now, instead of selecting from a set of complete radios as in software controlled radio, what if the software could select the building blocks based on the user needs. This is the new software-defined flexible radio which would enable users to construct wireless systems that fit their needs, rather than forcing to use from a small set of pre-existing protocols.
To develop and implement flexible protocols, a flexible hardware very similar to a Software Defined Radio (SDR) is required. In this thesis, the Intel T2200 board is chosen as the SDR platform. It is a heterogeneous platform with ARM, CEVA DSP and several accelerators. A wide range of protocols is mapped onto this platform and their performance evaluated. These include two OFDM based protocols (WiFi-Lite-A, WiFi-Lite-B), one DFT-spread OFDM based protocol (SCFDM-Lite) and one single carrier based protocol (SC-Lite). The transmitter and receiver blocks of the different protocols are first mapped on ARM in the T2200 board. The timing results show that IFFT, FFT, and Viterbi decoder blocks take most of the transmitter and receiver execution time and so in the next step these are mapped onto CEVA DSP. Mapping onto CEVA DSP resulted in significant execution time savings. The savings for WiFi-Lite-A were 60%, for WiFi-Lite-B were 64%, and for SCFDM-Lite were 71.5%. No savings are reported for SC-Lite since it was not mapped onto CEVA DSP.
Significant reduction in execution time is achieved for WiFi-Lite-A and WiFi-Lite-B protocols by implementing the entire transmitter and receiver chains on CEVA DSP. For instance, for WiFi-Lite-A, the savings were as large as 90%. Such huge savings are because the entire transmitter or receiver chain are implemented on CEVA and the timing overhead due to ARM-CEVA communication is completely eliminated. Finally, over-the-air testing was done for WiFi-Lite-A and WiFi-Lite-B protocols. Data was sent over the air using one Intel T2200 WBS board and received using another Intel T2200 WBS board. The received frames were decoded with no errors, thereby validating the over-the-air-communications.
To develop and implement flexible protocols, a flexible hardware very similar to a Software Defined Radio (SDR) is required. In this thesis, the Intel T2200 board is chosen as the SDR platform. It is a heterogeneous platform with ARM, CEVA DSP and several accelerators. A wide range of protocols is mapped onto this platform and their performance evaluated. These include two OFDM based protocols (WiFi-Lite-A, WiFi-Lite-B), one DFT-spread OFDM based protocol (SCFDM-Lite) and one single carrier based protocol (SC-Lite). The transmitter and receiver blocks of the different protocols are first mapped on ARM in the T2200 board. The timing results show that IFFT, FFT, and Viterbi decoder blocks take most of the transmitter and receiver execution time and so in the next step these are mapped onto CEVA DSP. Mapping onto CEVA DSP resulted in significant execution time savings. The savings for WiFi-Lite-A were 60%, for WiFi-Lite-B were 64%, and for SCFDM-Lite were 71.5%. No savings are reported for SC-Lite since it was not mapped onto CEVA DSP.
Significant reduction in execution time is achieved for WiFi-Lite-A and WiFi-Lite-B protocols by implementing the entire transmitter and receiver chains on CEVA DSP. For instance, for WiFi-Lite-A, the savings were as large as 90%. Such huge savings are because the entire transmitter or receiver chain are implemented on CEVA and the timing overhead due to ARM-CEVA communication is completely eliminated. Finally, over-the-air testing was done for WiFi-Lite-A and WiFi-Lite-B protocols. Data was sent over the air using one Intel T2200 WBS board and received using another Intel T2200 WBS board. The received frames were decoded with no errors, thereby validating the over-the-air-communications.
ContributorsChagari, Vamsi Reddy (Author) / Chakrabarti, Chaitali (Thesis advisor) / Lee, Hyunseok (Committee member) / Ogras, Umit Y. (Committee member) / Arizona State University (Publisher)
Created2016
Description
Coarse-grained Reconfigurable Arrays (CGRAs) are promising accelerators capable
of accelerating even non-parallel loops and loops with low trip-counts. One challenge
in compiling for CGRAs is to manage both recurring and nonrecurring variables in
the register file (RF) of the CGRA. Although prior works have managed recurring
variables via rotating RF, they access the nonrecurring variables through either a
global RF or from a constant memory. The former does not scale well, and the latter
degrades the mapping quality. This work proposes a hardware-software codesign
approach in order to manage all the variables in a local nonrotating RF. Hardware
provides modulo addition based indexing mechanism to enable correct addressing
of recurring variables in a nonrotating RF. The compiler determines the number of
registers required for each recurring variable and configures the boundary between the
registers used for recurring and nonrecurring variables. The compiler also pre-loads
the read-only variables and constants into the local registers in the prologue of the
schedule. Synthesis and place-and-route results of the previous and the proposed RF
design show that proposed solution achieves 17% better cycle time. Experiments of
mapping several important and performance-critical loops collected from MiBench
show proposed approach improves performance (through better mapping) by 18%,
compared to using constant memory.
of accelerating even non-parallel loops and loops with low trip-counts. One challenge
in compiling for CGRAs is to manage both recurring and nonrecurring variables in
the register file (RF) of the CGRA. Although prior works have managed recurring
variables via rotating RF, they access the nonrecurring variables through either a
global RF or from a constant memory. The former does not scale well, and the latter
degrades the mapping quality. This work proposes a hardware-software codesign
approach in order to manage all the variables in a local nonrotating RF. Hardware
provides modulo addition based indexing mechanism to enable correct addressing
of recurring variables in a nonrotating RF. The compiler determines the number of
registers required for each recurring variable and configures the boundary between the
registers used for recurring and nonrecurring variables. The compiler also pre-loads
the read-only variables and constants into the local registers in the prologue of the
schedule. Synthesis and place-and-route results of the previous and the proposed RF
design show that proposed solution achieves 17% better cycle time. Experiments of
mapping several important and performance-critical loops collected from MiBench
show proposed approach improves performance (through better mapping) by 18%,
compared to using constant memory.
ContributorsDave, Shail (Author) / Shrivastava, Aviral (Thesis advisor) / Ren, Fengbo (Committee member) / Ogras, Umit Y. (Committee member) / Arizona State University (Publisher)
Created2016
Description
Integrated circuits must be energy efficient. This efficiency affects all aspects of chip design, from the battery life of embedded devices to thermal heating on high performance servers. As technology scaling slows, future generations of transistors will lack the energy efficiency gains as it has had in previous generations. Therefore, other sources of energy efficiency will be much more important. Many computations have the potential to be executed for extreme energy efficiency but are not instigated because the platforms they run on are not optimized for efficient execution. ASICs improve energy efficiency by reducing flexibility and leveraging the properties of a specific computation. However, ASICs are fixed in function and therefore have incredible opportunity cost. FPGAs offer a reconfigurable solution but are 25x less energy efficient than ASIC implementation. Spatially programmable architectures (SPAs) are similar in design and structure to ASICs and FPGAs but are able bridge the ASIC-FPGA energy efficiency gap by trading flexibility for efficiency. However, SPAs are difficult to program because they do not share the same programming model as normal architectures that execute in time. This work addresses compiler challenges for coarse grained, locally interconnected SPA for domain efficiency (SPADE). A novel SPADE topology, called the wave pipeline, is introduced that is designed for the image signal processing domain that is both efficient and simple to compile to. A compiler for the wave pipeline is created that solves for maximum energy and area efficiency using low complexity, greedy methods. The wave pipeline topology and compiler allow for us to investigate and experiment with image signal processing applications to prove the feasibility of SPADE compilers.
ContributorsMackay, Curtis (Author) / Brunhaver, John (Thesis advisor) / Karam, Lina J (Committee member) / Seo, Jae-Sun (Committee member) / Arizona State University (Publisher)
Created2016