In our experiments, both values are set to 48 GB (out of 64 GB available on the machine), resulting in a fixed heap size. SABER_JVM_MS and SABER_JVM_MX set the minimum and maximum Java heap size (in GB) of a Saber application. The variable is used to construct the Java class path when running Saber appications from the command line. The default location on Linux is $HOME/.m2. MVN_REPOSITORY is the local folder where Saber's Java dependencies are downloaded and stored automatically by Maven. The nf fileĮdit scripts/nf and revisit Saber's default configuration parameters if necessary. Figures that have not been reproduced are replotted using the original results from the paper. Unless the -recompile-paper (or -r) flag is set to false, the script uses results from any reproduced experiments to replot figures (say, Figure 11(b) and Figure 14 from the example above). Reproduces Figure 11(b) and Figure 14 from the paper. The prepare-software.sh script will guide you through the installation and compilation process of our system. Support host page-locked memory mapping: Yesĭevice supports Unified Addressing (UVA): Yesĭevice PCI Domain ID / Bus ID / location ID: 0 / 4 / 0Ĭ) Experimentation Info Installing the system Max dimension size of a grid size (x,y,z): (2147483647, 65535, 65535)Ĭoncurrent copy and kernel execution: Yes with 2 copy engine(s) Max dimension size of a thread block (x,y,z): (1024, 1024, 64) Maximum number of threads per block: 1024 Maximum number of threads per multiprocessor: 2048 Total number of registers available per block: 65536 Total amount of shared memory per block: 49152 bytes Total amount of constant memory: 65536 bytes Maximum Layered 1D Texture Size, (num) layers 1D=(16384), 2048 layers (12) Multiprocessors, (192) CUDA Cores/MP: 2304 CUDA Cores Total amount of global memory: 8123 MBytes (8517124096 bytes)
JAVA VER DRIVER
The GPU has two copy engines, allowing for concurrent copy and kernel execution.ĬUDA Driver Version / Runtime Version 8.0 / 7.0ĬUDA Capability Major/Minor version number: 3.5 The GPU used in our experiments is an NVIDIA Quadro K5200 with 2,304 CUDA cores and 8,123 MB of RAM, attached to the host via PCIe 3.0 x16. Our machine has a total of 64 GB of DDR4 SDRAM, spread equally across 8 CPU DIMM slots. The L1d (L1i), L2, and 元 cache sizes are 32 KB, 256 KB, and 20 MB, respectively. Our machine has 2 Intel Xeon CPU E5-2640 v3 2.60GHz. We performed our experiments on an HP Z840 Workstation. The OpenCL headers are located under /usr/local/cuda/include. In our installation, it is located under /usr/local/cuda/lib64/. Saber requires libOpenCL.so to be present on the OS library path ( LD_LIBRARY_PATH). These are used to replot figures from any reproduced results. Optional commands and their packages are epstopdf (texlive-font-utils), and gnuplot (gnuplot). The prepare-software.sh script uses APT to handle the installation of these dependencies.Įssential commands and their corresponding Ubuntu packages are: javac and java (openjdk-7-jdk), mvn (maven), make (make), gcc (gcc), and python (python). The system has been mainly tested on Ubuntu Linux 14.04. The GPU OpenCL operators are compiled just-in-time, when an operator is instantiated by a Saber application. The CPU and GPU query operators have been implemented in Java and OpenCL, respectively.
JAVA VER CODE
The C code is compiled and packaged using GNU make (3.81) and gcc (4.8.4). The Java code is compiled and packaged using Apache Maven (3.3.1) and the Java SDK (1.7.0.79). Saber has been implemented in Java and C.
JAVA VER DOWNLOAD
The Saber source code is available to download from: Owner of artifact grants ACM permission to serve the artifact to users of the ACM Digital Library. Wolf, Paolo Costa, and Peter Pietzuch Permissions SABER: Window-based hybrid stream processing for heterogeneous architectures AuthorsĪlexandros Koliousis, Matthias Weildich, Raul Castro Fernandez, Alexander L. Readme for reproducibility submission of paper ID 195 Paper title
0 kommentar(er)
