Technical Reports from UMIACS

Permanent URI for this collection

http://hdl.handle.net/1903/7

Browse

Recent Submissions

Now showing 1 - 5 of 950
  • Thumbnail Image
    Item
    Pipelined CPU-GPU Scheduling for Caches
    (2021-03-23) Gerzhoy, Daniel; Yeung, Donald
    Heterogeneous microprocessors integrate a CPU and GPU with a shared cache hierarchy on the same chip, affording low-overhead communication between the CPU and GPU's cores. Often times, large array data structures are communicated from the CPU to the GPU and back. While the on-chip cache hierarchy can support such CPU-GPU producer-consumer sharing, this almost never happens due to poor temporal reuse. Because the data structures can be quite large, by the time the consumer reads the data, it has been evicted from cache even though the producer had brought it on-chip when it originally wrote the data. As a result, the CPU-GPU communication happens through main memory instead of the cache, hurting performance and energy. This paper exploits the on-chip caches in a heterogeneous microprocessor to improve CPU-GPU communication efficiency. We divide streaming computations executed by the CPU and GPU that exhibit producer-consumer sharing into chunks, and overlap the execution of CPU chunks with GPU chunks in a software pipeline. To enforce data dependences, the producer executes one chunk ahead of the consumer at all times. We also propose a low-overhead synchronization mechanism in which the CPU directly controls thread-block scheduling in the GPU to maintain the producer's "run-ahead distance" relative to the consumer. By adjusting the chunk size or run-ahead distance, we can make the CPU-GPU working set fit in the last-level cache, thus permitting the producer-consumer sharing to occur through the LLC. We show through simulation that our technique reduces the number of DRAM accesses by 30.4%, improves performance by 26.8%, and lowers memory system energy by 27.4% averaged across 7 benchmarks.
  • Thumbnail Image
    Item
    Nervous system maps on the C. elegans genome
    (2020-09-28) Cherniak, Christopher; Mokhtarzada, Zekeria; Rodriguez-Esteban, Raul
    This project begins from a synoptic point of view, focusing upon the large-scale (global) landscape of the genome. This is along the lines of combinatorial network optimization in computational complexity theory [1]. Our research program here in turn originated along parallel lines in computational neuroanatomy [2,3,4,5]. Rather than mapping body structure onto the genome, the present report focuses upon statistically significant mappings of the Caenorhabditis elegans nervous system onto its genome. Via published datasets, evidence is derived for a "wormunculus", on the model of a homunculus representation, but on the C. elegans genome. The main method of testing somatic-genomic position-correlations here is via public genome databases, with r^2 analyses and p evaluations. These findings appear to yield some of the basic structural and functional organization of invertebrate nucleus and chromosome architecture. The design rationale for somatic maps on the genome in turn may be efficient interconnections. A next question this study raises: How do these various somatic maps mesh (interrelate, interact) with each other?
  • Thumbnail Image
    Item
    Design and Evaluation of Monolithic Computers Implemented Using Crossbar ReRAM
    (2019-07-16) Jagasivamani, Meenatchi; Walden, Candace; Singh, Devesh; Li, Shang; Kang, Luyi; Asnaashari, Mehdi; Dubois, Sylvain; Jacob, Bruce; Yeung, Donald
    A monolithic computer is an emerging architecture in which a multicore CPU and a high-capacity main memory system are all integrated in a single die. We believe such architectures will be possible in the near future due to nonvolatile memory technology, such as the resistive random access memory, or ReRAM, from Crossbar Incorporated. Crossbar's ReRAM can be fabricated in a standard CMOS logic process, allowing it to be integrated into a CPU's die. The ReRAM cells are manufactured in between metal wires and do not employ per-cell access transistors, leaving the bulk of the base silicon area vacant. This means that a CPU can be monolithically integrated directly underneath the ReRAM memory, allowing the cores to have massively parallel access to the main memory. This paper presents the characteristics of Crossbar's ReRAM technology, informing architects on how ReRAM can enable monolithic computers. Then, it develops a CPU and memory system architecture around those characteristics, especially to exploit the unprecedented memory-level parallelism. The architecture employs a tiled CPU, and incorporates memory controllers into every compute tile that support a variable access granularity to enable high scalability. Lastly, the paper conducts an experimental evaluation of monolithic computers on graph kernels and streaming computations. Our results show that compared to a DRAM-based tiled CPU, a monolithic computer achieves 4.7x higher performance on the graph kernels, and achieves roughly parity on the streaming computations. Given a future 7nm technology node, a monolithic computer could outperform the conventional system by 66% for the streaming computations.
  • Thumbnail Image
    Item
    Exploiting Multi-Loop Parallelism on Heterogeneous Microprocessors
    (2016-11-10) Zuzak, Michael; Yeung, Donald
    Heterogeneous microprocessors integrate CPUs and GPUs on the same chip, providing fast CPU-GPU communication and enabling cores to compute on data "in place." These advantages will permit integrated GPUs to exploit a smaller unit of parallelism. But one challenge will be exposing sufficient parallelism to keep all of the on-chip compute resources fully utilized. In this paper, we argue that integrated CPU-GPU chips should exploit parallelism from multiple loops simultaneously. One example of this is nested parallelism in which one or more inner SIMD loops are nested underneath a parallel outer (non- SIMD) loop. By scheduling the parallel outer loop on multiple CPU cores, multiple dynamic instances of the inner SIMD loops can be scheduled on the GPU cores. This boosts GPU utilization and parallelizes the non-SIMD code. Our preliminary results show exploiting such multi-loop parallelism provides a 3.12x performance gain over exploiting parallelism from individual loops one at a time.
  • Thumbnail Image
    Item
    Body Maps on Human Chromosomes
    (2015-11-08) Cherniak, Christopher; Rodriguez-Esteban, Raul
    An exploration of the hypothesis that human genes are organized somatotopically: For each autosomal chromosome, its tissue-specific genes tend to have relative positions on the chromosome that mirror corresponding positions of the tissues in the body. In addition, there appears to be a division of labor: Such a homunculus representation on a chromosome holds significantly for either the anteroposterior or the dorsoventral body axis. In turn, anteroposterior and dorsoventral chromosomes tend to occupy separate zones in the spermcell nucleus. One functional rationale of such largescale organization is for efficient interconnections in the genome.