[X] The "/nb-scala/nb-scala 6.7v1/nb-scala-6.7v1.zip" file could not be found or is not available. Please select another file.

mib is an MPI-based I/O benchmark for clusters.


http://mibtest.sourceforge.net





Separate each tag with a space.

Release Date:

2006-07-31

Topic:

Operating System:

License:

Intended Audience:

User Interface:

Programming Language:

C

Registered:

2006-07-29

Ratings and Reviews

Be the first to post a text review of mib. Rate and review a project by clicking thumbs up or thumbs down in the right column.

Project Feed

  • File released: /mib/mib-1.9.6/mib-1.9.6.tar.gz

    posted 1255 days ago

  • mib mib-1.9.6 file released: mib-1.9.6.tar.gz

    2006-07-13 Andrew Uselton <uselton2@llnl.gov> * Tag 1.9.6 * Update all files with UCRL number in preparation for release. * Sanity check documentation. * Caught and fixed a bug in options handling. Send the correct argc and argv to MPI_Init if you plan on consuming command line parameters and options. It would segfault trying to get to argv[0] when sent argc = 0, which is wrong in a couple of ways. 2006-06-29 Andrew Uselton <uselton2@llnl.gov> * Tag 1.9.5 * Modify dlopen() code to be configurable, set up elan and infiniband variants. * Cast the "size_t count" parameter to Write() and Read() to be ssize_t. This prevents incorrect execution when the system call returns an error condition (-1) and misinterprets it as a very large number. There is also one in Snprintf. * Include "-Wextra" as a CCFLAGS flag to catch future instances of the above. 2006-04-26 Andrew Uselton <uselton2@llnl.gov> * Tag 1.9.4 * Add lwatch.py to the project, since lmtd actually is available as open source. Modify references saying otherwise in the documentations. 2006-04-20 Andrew Uselton <uselton2@llnl.gov> * Tag 1.9.3 * Fix really stupid, pernicious, and fundamental bug regarding use of opt->flags and verbosity references in macros. They appear in several places, especially base_report, before opts is initialized. 1) split base_report from flags test and introduce new conditional_report. 2) do not refer to opts in macros defined in options.h. Use opts in macros only locally to mib.c. 2006-04-3 Andrew Uselton <uselton2@llnl.gov> * Tag 1.9.2 * Take out the node averaging. The BGL code has entirely forked and node averaging is only an issue there. It is also less reasonable on non-BGL clusters because the tasks do not stay coordinated across a node like they do on BGL. * The -V option should include the architecture, not just the signon message. * regularize all the documentation so it all agrees. 2006-03-24 Andrew Uselton <uselton2@llnl.gov> * Tag 1.9.1 * Add the architecture to the version string 2006-03-17 Andrew Uselton <uselton2@llnl.gov> * Tag 1.9 * Allow for operation without MPI * Introduce random read option * Improve the mib.sh wrapper script * Make the target a required argument to avoid accidentally writing to NFS and causing disruption * Check for the file system type of the target and refuse to use NFS unless a "-F" (--force) flag is set. * Add a version option that prints the version string and exits. * Remove the "options" file code, use command line arguments only * Add a unit test "mib-test.sh" to the tools directory * Fix Mib_Init, it didn't return(mib) its result * When there is no SLURM or MPI we set the rank and base to the host's index. That means the read test's file name needs the base added to its modulo arithmatic. When we have arbitrary subcommunicators, if ever, that would also come up. * "Open" for writes should have O_CREAT, and "Open" for reads should not. Doing this makes the "Open" implementation straight forward. * If the statfs call fails, which it will if you gave it a bogus target directory, then you need to bail with an error message. I'd left out the "exit(1)". * Convert to using config.h, which is created by configure, to detect if MPI is present on the build host, and skip using it if so. * Set the correct include directory: /usr/lib/mpi/include in the Makefile.in 2006-03-01 Andrew Uselton <uselton2@llnl.gov> * Tag 1.8 * Dump the iterateions and pause options * Code factoring in the options settings * Consolidate boolean flags into fields in a single flag * Fixed ENOENT / ENONET bug :( * Get command line options (including long-opts) working and tested * Allow there to be a log_dir with no "options" file * Make a man page * Fix packaging 2006-02-15 Andrew Uselton <uselton2@llnl.gov> * Tag 1.7 * Introduce "new" and "remove" options to allow for managing files. * Introduce "use_ion_aves" option to suppress (by default) calculating averages accross IONs the way BGL does. * Fixed the "no averages" code path so it rightly preserves the whole profile from a run. * Promoted the res->transferred field to a double because as a 32 bit quantity it could overflow for really fast or long accumulations of transfers. 2006-01-15 Andrew Uselton <uselton2@llnl.gov> * Tag 1.6 * Introduce "profiles" option, and don't do profiles unless it is set * Adjust defaults for operation without parameters file * Reduce verbosity for default operation, summary line only * Clean up "lint" detected by icc compiler, missing "returns" pointer arithmetic, missing "include" files <stdlib> and <string> * init_timer() now uses MPI_COMM_WORLD and occurs outside the iteration loop. It also announces the program name and start time. * the fsync() seems to return errors that (probably) occurred earlier and were corrected. Work around, investigate, fix. 2005-12-07 Andrew Uselton <uselton2@llnl.gov> * Tag 1.5 * Discovered where the rogue ASSERT was, though not why. * Retrenched to only using FAIL and ASSERT and removed problematic error handling code * Documented, but did not implement, details of system call return codes * Timer intialization check 2005-12-01 Andrew Uselton <uselton2@llnl.gov> * Tag 1.4 * Bool enumeration * Optional read_only and write_only * Optional looping 2005-11-22 Andrew Uselton <uselton2@llnl.gov> * Tag 1.3 * Code refactoring 2005-11-15 Andrew Uselton <uselton2@llnl.gov> * Tag 1.2 * Progress meter 2005-11-01 Andrew Uselton <uselton2@llnl.gov> * tag 1.1 * Reorganize for single directory of output 2005-10-01 Andrew Uselton <uselton2@llnl.gov> * tag 1.0 * Working code

    posted 1255 days ago

  • Code committed

    Anonymous committed patchset 1 of module CVSROOT to the mib CVS repository, changing 11 files

    posted by nobody 1257 days ago

  • Forum thread added

    Anonymous created the Welcome to Open Discussion forum thread

    posted by nobody 1257 days ago

  • Forum thread added

    Anonymous created the Welcome to Developers forum thread

    posted by nobody 1257 days ago

  • Forum thread added

    Anonymous created the Welcome to Help forum thread

    posted by nobody 1257 days ago

  • Project Information Updated

    auselton registered the mib project

    posted by auselton 1257 days ago

Rate and Review

Be the first person to add a text review.

Would you recommend this project?






<

Related Projects

Thanks for your rating!

Would you also like to write a review?





Skip Review

Thanks for your review!

Get credit for your review by logging in via OpenID. Click your account provider:

No Thanks