Visualizing surfaces is a fundamental technique in computer science and is frequently used across... more Visualizing surfaces is a fundamental technique in computer science and is frequently used across a wide range of fields such as computer graphics, biology, engineering, and scientific visualization. In many cases, visualizing an interface between boundaries can provide meaningful analysis or simplification of complex data. Some examples include physical simulation for animation, multimaterial mesh extraction in biophysiology, flow on airfoils in aeronautics, and integral surfaces. However, the quest for high-quality visualization, coupled with increasingly complex data, comes with a high computational cost. Therefore, new techniques are needed to solve surface visualization problems within a reasonable amount of time while also providing sophisticated visuals that are meaningful to scientists and engineers. In this dissertation, novel techniques are presented to facilitate surface visualization. First, a particle system for mesh extraction is parallelized on the graphics processing unit (GPU) with a red-black update scheme to achieve an order of magnitude speed-up over a central processing unit (CPU) implementation. Next, extending the red-black technique to multiple materials showed inefficiencies on the GPU. Therefore, we borrow the underlying data structure from the closest point method, the closest point embedding, and the particle system solver is switched to hierarchical octree-based approach on the GPU. Third, to demonstrate that the closest point embedding is a fast, flexible data structure for surface particles, it is adapted to unsteady surface flow visualization at near-interactive speeds. Finally, the closest point embedding is a three-dimensional dense structure that does not scale well. Therefore, we introduce a closest point sparse octree that allows the closest point embedding to scale to higher resolution. Further, we demonstrate unsteady line integral convolution using the closest point method.
2018 IEEE 38th International Conference on Distributed Computing Systems (ICDCS)
One of the core issues across computer and computational science today is adapting to, managing, ... more One of the core issues across computer and computational science today is adapting to, managing, and learning from the influx of "Big Data". In the commercial space, this problem has led to a huge investment in new technologies and capabilities that are well adapted to dealing with the sorts of human-generated logs, videos, texts, and other large-data artifacts that are processed and resulted in an explosion of useful platforms and languages (Hadoop, Spark, Pandas, etc.). However, translating this work from the enterprise space to the computational science and HPC community has proven somewhat difficult, in part because of some of the fundamental differences in type and scale of data and timescales surrounding its generation and use. We describe a forward-looking research and development plan which centers around the concept of making Input/Output (I/O) intelligent for users in the scientific community, whether they are accessing scalable storage or performing in situ workflow tasks. Much of our work is based on our experience with the Adaptable I/O System (ADIOS 1.X), and our next generation version of the software ADIOS 2.X [1].
This work compares the two major paradigms for doing in situ visualization: in-line, where the si... more This work compares the two major paradigms for doing in situ visualization: in-line, where the simulation and visualization share the same resources, and in-transit, where simulation and visualization are given dedicated resources. Our runs vary many parameters, including simulation cycle time, visualization frequency, and dedicated resources, to study how tradeoffs change over configuration. In particular, we consider simulations as large as 1,024 nodes (16,384 cores) and dedicated visualization resources with as many as 512 nodes (8,192 cores). We draw conclusions about when each paradigm is superior, such as in-line being superior when the simulation cycle time is very fast. Surprisingly, we also find that in-transit can minimize the total resources consumed for some configurations, since it can cause the visualization routines to require fewer overall resources when they run at lower concurrency. For example, one of our scenarios finds that allocating 25% more resources for visualization allows the simulation to run 61% faster than its in-line comparator. Finally, we explore various models for quantifying the cost for each paradigm, and consider transition points when one paradigm is superior to the other. Our contributions inform design decisions for simulation scientists when performing in situ visualization.
People are two qubits closer to useful quantum computers. That might not sound like much, but the... more People are two qubits closer to useful quantum computers. That might not sound like much, but the quantum computing arms race playing out could eventually deliver the biggest breakthrough in technology since the advent of the computer itself. In July, a quantum simulator with 51 quantum bits, or qubits, was announced. Now, a team led by Christopher Monroe at the Joint Quantum Institute in Maryland has gone two better. The latest 53-qubit device, made from a chain of 53 charged atoms, or ions, trapped in electric fields, has run scientific simulations that don't seem to be possible with everyday computers.
Google has come up with a strategy for demonstrating quantum supremacy, the claim that quantum co... more Google has come up with a strategy for demonstrating quantum supremacy, the claim that quantum computers can perform tasks that no current computers can. While it's widely assumed that we will eventually reach quantum supremacy, nobody has done it yet because current quantum computers can only run a few specialized algorithms. The plan is based on simulating coin flips. An ordinary computer does this by storing two numbers and choosing one of them at random for each flip. To simulate 50 coin tosses, it selects 50 times in a row. This is simple with regular coins, but if the coins behave like particles obeying the laws of quantum mechanics, things get more complicated
Half a century of research on getting computers to work even when errors pop up makes their moder... more Half a century of research on getting computers to work even when errors pop up makes their modern machines pretty reliable. Unfortunately, the laws of quantum mechanics render all that research useless for quantum computers, the sheer complexity of which leaves them prone to errors. Now, they finally have the first quantum program that detects data corruption. Two research groups--one from the University of Maryland and Georgia Tech and the other from IBM--have demonstrated the same quantum error-detecting program, albeit implemented with different hardware.
This expository thesis contains a study of four interpolation theorems, the requisite background ... more This expository thesis contains a study of four interpolation theorems, the requisite background material, and a few applications. The materials introduced in the first three sections of Chapter 1 are used to motivate and prove the Riesz-Thorin interpolation theorem and its extension by Stein, both of which are presented in the fourth section. Chapter 2 revolves around Calderón's complex method of interpolation and the interpolation theorem of Fefferman and Stein, with the material in between providing the necessary examples and tools. The two theorems are then applied to a brief study of linear partial differential equations, Sobolev spaces, and Fourier integral operators, presented in the last section of the second chapter.
This article addresses the question of whether operational efficiency is recognized and rewarded ... more This article addresses the question of whether operational efficiency is recognized and rewarded by the private funders that support nonprofit organizations in fields ranging from education to social service to arts and beyond. Looking at the administrative efficiency and fundraising results of a large sample of nonprofit organizations over an 11 year period, we find that nonprofits that position themselves as cost efficient-reporting low administrative to total expense ratios-fared no better over time than less efficient appearing organizations in the market for individuals, foundations, and corporate contributions. From this analysis, we suggest that economizing may not always be the best strategy in the nonprofit sector.
Proceedings of the 5th Annual Workshop on General Purpose Processing with Graphics Processing Units - GPGPU-5, 2012
Extracting isosurfaces represented as high quality meshes from threedimensional scalar fields is ... more Extracting isosurfaces represented as high quality meshes from threedimensional scalar fields is needed for many important applications, particularly visualization and numerical simulations. One recent advance for extracting high quality meshes for isosurface computation is based on a dynamic particle system. Unfortunately, this state-of-the-art particle placement technique requires a significant amount of time to produce a satisfactory mesh. To address this issue, we study the parallelism property of the particle placement and make use of CUDA, a parallel programming technique on the GPU, to significantly improve the performance of particle placement. This paper describes the curvature dependent sampling method used to extract high quality meshes and describes its implementation using CUDA on the GPU.
Visualizing surfaces is a fundamental technique in computer science and is frequently used across... more Visualizing surfaces is a fundamental technique in computer science and is frequently used across a wide range of fields such as computer graphics, biology, engineering, and scientific visualization. In many cases, visualizing an interface between boundaries can provide meaningful analysis or simplification of complex data. Some examples include physical simulation for animation, multimaterial mesh extraction in biophysiology, flow on airfoils in aeronautics, and integral surfaces. However, the quest for high-quality visualization, coupled with increasingly complex data, comes with a high computational cost. Therefore, new techniques are needed to solve surface visualization problems within a reasonable amount of time while also providing sophisticated visuals that are meaningful to scientists and engineers. In this dissertation, novel techniques are presented to facilitate surface visualization. First, a particle system for mesh extraction is parallelized on the graphics processing unit (GPU) with a red-black update scheme to achieve an order of magnitude speed-up over a central processing unit (CPU) implementation. Next, extending the red-black technique to multiple materials showed inefficiencies on the GPU. Therefore, we borrow the underlying data structure from the closest point method, the closest point embedding, and the particle system solver is switched to hierarchical octree-based approach on the GPU. Third, to demonstrate that the closest point embedding is a fast, flexible data structure for surface particles, it is adapted to unsteady surface flow visualization at near-interactive speeds. Finally, the closest point embedding is a three-dimensional dense structure that does not scale well. Therefore, we introduce a closest point sparse octree that allows the closest point embedding to scale to higher resolution. Further, we demonstrate unsteady line integral convolution using the closest point method.
2018 IEEE 38th International Conference on Distributed Computing Systems (ICDCS)
One of the core issues across computer and computational science today is adapting to, managing, ... more One of the core issues across computer and computational science today is adapting to, managing, and learning from the influx of "Big Data". In the commercial space, this problem has led to a huge investment in new technologies and capabilities that are well adapted to dealing with the sorts of human-generated logs, videos, texts, and other large-data artifacts that are processed and resulted in an explosion of useful platforms and languages (Hadoop, Spark, Pandas, etc.). However, translating this work from the enterprise space to the computational science and HPC community has proven somewhat difficult, in part because of some of the fundamental differences in type and scale of data and timescales surrounding its generation and use. We describe a forward-looking research and development plan which centers around the concept of making Input/Output (I/O) intelligent for users in the scientific community, whether they are accessing scalable storage or performing in situ workflow tasks. Much of our work is based on our experience with the Adaptable I/O System (ADIOS 1.X), and our next generation version of the software ADIOS 2.X [1].
This work compares the two major paradigms for doing in situ visualization: in-line, where the si... more This work compares the two major paradigms for doing in situ visualization: in-line, where the simulation and visualization share the same resources, and in-transit, where simulation and visualization are given dedicated resources. Our runs vary many parameters, including simulation cycle time, visualization frequency, and dedicated resources, to study how tradeoffs change over configuration. In particular, we consider simulations as large as 1,024 nodes (16,384 cores) and dedicated visualization resources with as many as 512 nodes (8,192 cores). We draw conclusions about when each paradigm is superior, such as in-line being superior when the simulation cycle time is very fast. Surprisingly, we also find that in-transit can minimize the total resources consumed for some configurations, since it can cause the visualization routines to require fewer overall resources when they run at lower concurrency. For example, one of our scenarios finds that allocating 25% more resources for visualization allows the simulation to run 61% faster than its in-line comparator. Finally, we explore various models for quantifying the cost for each paradigm, and consider transition points when one paradigm is superior to the other. Our contributions inform design decisions for simulation scientists when performing in situ visualization.
People are two qubits closer to useful quantum computers. That might not sound like much, but the... more People are two qubits closer to useful quantum computers. That might not sound like much, but the quantum computing arms race playing out could eventually deliver the biggest breakthrough in technology since the advent of the computer itself. In July, a quantum simulator with 51 quantum bits, or qubits, was announced. Now, a team led by Christopher Monroe at the Joint Quantum Institute in Maryland has gone two better. The latest 53-qubit device, made from a chain of 53 charged atoms, or ions, trapped in electric fields, has run scientific simulations that don't seem to be possible with everyday computers.
Google has come up with a strategy for demonstrating quantum supremacy, the claim that quantum co... more Google has come up with a strategy for demonstrating quantum supremacy, the claim that quantum computers can perform tasks that no current computers can. While it's widely assumed that we will eventually reach quantum supremacy, nobody has done it yet because current quantum computers can only run a few specialized algorithms. The plan is based on simulating coin flips. An ordinary computer does this by storing two numbers and choosing one of them at random for each flip. To simulate 50 coin tosses, it selects 50 times in a row. This is simple with regular coins, but if the coins behave like particles obeying the laws of quantum mechanics, things get more complicated
Half a century of research on getting computers to work even when errors pop up makes their moder... more Half a century of research on getting computers to work even when errors pop up makes their modern machines pretty reliable. Unfortunately, the laws of quantum mechanics render all that research useless for quantum computers, the sheer complexity of which leaves them prone to errors. Now, they finally have the first quantum program that detects data corruption. Two research groups--one from the University of Maryland and Georgia Tech and the other from IBM--have demonstrated the same quantum error-detecting program, albeit implemented with different hardware.
This expository thesis contains a study of four interpolation theorems, the requisite background ... more This expository thesis contains a study of four interpolation theorems, the requisite background material, and a few applications. The materials introduced in the first three sections of Chapter 1 are used to motivate and prove the Riesz-Thorin interpolation theorem and its extension by Stein, both of which are presented in the fourth section. Chapter 2 revolves around Calderón's complex method of interpolation and the interpolation theorem of Fefferman and Stein, with the material in between providing the necessary examples and tools. The two theorems are then applied to a brief study of linear partial differential equations, Sobolev spaces, and Fourier integral operators, presented in the last section of the second chapter.
This article addresses the question of whether operational efficiency is recognized and rewarded ... more This article addresses the question of whether operational efficiency is recognized and rewarded by the private funders that support nonprofit organizations in fields ranging from education to social service to arts and beyond. Looking at the administrative efficiency and fundraising results of a large sample of nonprofit organizations over an 11 year period, we find that nonprofits that position themselves as cost efficient-reporting low administrative to total expense ratios-fared no better over time than less efficient appearing organizations in the market for individuals, foundations, and corporate contributions. From this analysis, we suggest that economizing may not always be the best strategy in the nonprofit sector.
Proceedings of the 5th Annual Workshop on General Purpose Processing with Graphics Processing Units - GPGPU-5, 2012
Extracting isosurfaces represented as high quality meshes from threedimensional scalar fields is ... more Extracting isosurfaces represented as high quality meshes from threedimensional scalar fields is needed for many important applications, particularly visualization and numerical simulations. One recent advance for extracting high quality meshes for isosurface computation is based on a dynamic particle system. Unfortunately, this state-of-the-art particle placement technique requires a significant amount of time to produce a satisfactory mesh. To address this issue, we study the parallelism property of the particle placement and make use of CUDA, a parallel programming technique on the GPU, to significantly improve the performance of particle placement. This paper describes the curvature dependent sampling method used to extract high quality meshes and describes its implementation using CUDA on the GPU.
Uploads
Papers by Mark Kim