Accelerator cores in low-power heterogeneous processors have on-chip local memories to enable parallel data access. The memory capacities of the local memories are very small. Therefore, the data should be transferred from the global memory to the local memories many times. These data transfers greatly increase the total processing time. Memory allocation technique to increase the data sharing ...

This paper introduces an approach for dense 3D reconstruc7 7 tion from unregistered Internet-scale photo collections with about 3 mil8 8 lion of images within the span of a day on a single PC (“cloudless”). Our 9 9 method advances image clustering, stereo, stereo fusion and structure 10 10 from motion to achieve high computational performance. We leverage 11 11 geometric and appearance constrai...

Massively parallel architectures such as the GPU are becoming increasingly important due to the recent proliferation of data. In this paper, we propose a key class of hybrid parallel graphlet algorithms that leverages multiple CPUs and GPUs simultaneously for computing k-vertex induced subgraph statistics (called graphlets). In addition to the hybrid multi-core CPU-GPU framework, we also invest...

The initial objective was to implement a 3D Gaussian beam model on multi-core, high-performance computing hardware in order to support mid and high frequency applications like reverberation modeling and acoustic channel simulation. As graphic processing unit or GPU hardware rapidly emerged as a compelling platform for such work, my objective became to adapt selected acoustic wave propagation mo...

SUMMARY When choosing a read mapper, one faces the trade off between speed and the ability to map reads in highly polymorphic regions. Here, we report NextGenMap, a fast and accurate read mapper, which reduces this dilemma. NextGenMap aligns reads reliably to a reference genome even when the sequence difference between target and reference genome is large, i.e. highly polymorphic genome. At the...

Solar potential estimation using LiDAR data is an efficient approach for finding suitable roofs for photovoltaic systems’ installations. As the amount of LiDAR data increases, the non-parallel methods take considerable time to accurately estimate the solar potential. Although supercomputing provides a possible solution, it is still too expensive and thus infeasible for general usage. Fortunatel...

Designing scalable transaction processing systems on modern multicore hardware has been a challenge for almost a decade. The typical characteristics of transaction processing workloads lead to a high degree of unbounded communication on multicores for conventional system designs. In this tutorial, we initially present a systematic way of eliminating scalability bottlenecks of a transaction proc...

Heterogeneous multi-core processors are at tracted by media processing applications due to their ca pability of drawing strengths of different cores to improve the overall performance. However, the unavailability of an efficient programing environment is a major problem. In this paper, we propose an FPGA-oriented heterogeneous multi-core processor, with GPU-like accelerator cores. We use a CUDA...

The era of programming with single processors has ended. Decades of prophesies have at last come true: programming with multiple processors has now entered the mainstream. Two forces have caused this transition to happen now. First, the Internet, which is a network of many loosely coupled processors. It had been gaining relevance for many years, but only recently has it achieved sufficient band...

We present in this paper Video++, a new framework targeting image and video applications running on multi-core processors. While offering a high expressive power, we show that it generates code running up to 32 times faster than the naive equivalents. Taking advantage of the new C++11/C++14 features, tools, we propose simple abstractions matching the performance of hand optimized code. This pap...