[ubuntu/partner/trusty-proposed] ti-openmpi 1.0.0.15-0ubuntu1 (Accepted)

[dann frazier]

ti-openmpi (1.0.0.15-0ubuntu1) trusty; urgency=medium

  [ TI Keystone PPA ]
  * SRIO user space library used inside SRIO BTL
  * Updated module dependency on CMEM
  * Fix for HLINK (mpm-transport) crash w/o BTLs specified
  * Makefile.am modified to workaround gcc4.6 dynamic linking issue; two pass
    linking for edma3 and edma3rm
  * Upgrade to new MPM-transport, support 6.25 Gbps with hyperlink
  * Fix uninitialized variable
  * Reduce MPI SRIO memory use to fit into reduced CMEM block
  * Remove sysroot from CFLAGS for srio module (causing err in stdint.h
    inclusion)
  * Modified debian control file with Linux headers keystone version
  * sysroot removed from CPPFLAGS in Makefile-in, in srio folder
  * Fixed linux kernel header version extraction, in debian rules file
  * Update debian control file for libedma3, libmpmtransport and lnx headers
    deps
  * MPI version, and libmptransport name typo fixed
  * Code cleanup
  * Update dependency on mpmtransport, cmem, and cmem-mod-dkms
  * Use upstream version for dependencies in debian/control
  * Restore documentation of opal/libltdl copyright/license in
    debian/copyright.
  * Fix for 45th cartridge
  * All transfers in BTL HLINK are (temporarily) made EDMA driven
  * Bump up priority on NREAD requests (so they don't get timed out
    competing with packet forwarding data traffic)
  * Topology/Routing changes to avoid loops in a cluster (because,
    there'll be a LSU lock up with an SRIO loop)
  * Packet forwarding fixes to exclude a node's own id, and to have
    no packet forwarding entries for end point nodes
  * Integrated PDSP firmware based SRIO routing
  * TX side throttling added only during starvation of tx free pool.
  * Manual cache operation restored
  * Debug load for IMB 4 node test
  * Uninitizalized variable fixed in ORTE hostfile.c. It was causing MPMD to
    fail intermittently, with probability increased with number of nodes
    involved

Date: Wed, 10 Dec 2014 15:00:04 -0700
Changed-By: dann frazier <dann.frazier at canonical.com>
Maintainer: TI Keystone PPA <ti-keystone-ppa at list.ti.com>
https://launchpad.net/ubuntu/+source/ti-openmpi/1.0.0.15-0ubuntu1
-------------- next part --------------
Format: 1.8
Date: Wed, 10 Dec 2014 15:00:04 -0700
Source: ti-openmpi
Binary: ti-openmpi
Architecture: source
Version: 1.0.0.15-0ubuntu1
Distribution: trusty
Urgency: medium
Maintainer: TI Keystone PPA <ti-keystone-ppa at list.ti.com>
Changed-By: dann frazier <dann.frazier at canonical.com>
Description:
 ti-openmpi - openmpi patched to use for TI Keystone
Changes:
 ti-openmpi (1.0.0.15-0ubuntu1) trusty; urgency=medium
 .
   [ TI Keystone PPA ]
   * SRIO user space library used inside SRIO BTL
   * Updated module dependency on CMEM
   * Fix for HLINK (mpm-transport) crash w/o BTLs specified
   * Makefile.am modified to workaround gcc4.6 dynamic linking issue; two pass
     linking for edma3 and edma3rm
   * Upgrade to new MPM-transport, support 6.25 Gbps with hyperlink
   * Fix uninitialized variable
   * Reduce MPI SRIO memory use to fit into reduced CMEM block
   * Remove sysroot from CFLAGS for srio module (causing err in stdint.h
     inclusion)
   * Modified debian control file with Linux headers keystone version
   * sysroot removed from CPPFLAGS in Makefile-in, in srio folder
   * Fixed linux kernel header version extraction, in debian rules file
   * Update debian control file for libedma3, libmpmtransport and lnx headers
     deps
   * MPI version, and libmptransport name typo fixed
   * Code cleanup
   * Update dependency on mpmtransport, cmem, and cmem-mod-dkms
   * Use upstream version for dependencies in debian/control
   * Restore documentation of opal/libltdl copyright/license in
     debian/copyright.
   * Fix for 45th cartridge
   * All transfers in BTL HLINK are (temporarily) made EDMA driven
   * Bump up priority on NREAD requests (so they don't get timed out
     competing with packet forwarding data traffic)
   * Topology/Routing changes to avoid loops in a cluster (because,
     there'll be a LSU lock up with an SRIO loop)
   * Packet forwarding fixes to exclude a node's own id, and to have
     no packet forwarding entries for end point nodes
   * Integrated PDSP firmware based SRIO routing
   * TX side throttling added only during starvation of tx free pool.
   * Manual cache operation restored
   * Debug load for IMB 4 node test
   * Uninitizalized variable fixed in ORTE hostfile.c. It was causing MPMD to
     fail intermittently, with probability increased with number of nodes
     involved
Checksums-Sha1:
 3a4f510e887ad9ad0016200600056daaa85f0d07 1626 ti-openmpi_1.0.0.15-0ubuntu1.dsc
 a8a49b74ff19e6c24ed612ed84326609c5dfe3e8 18284139 ti-openmpi_1.0.0.15-0ubuntu1.tar.gz
Checksums-Sha256:
 6177aa7df814814d65d7afec623c8736425224ef800a41662deeb0c4ae6946aa 1626 ti-openmpi_1.0.0.15-0ubuntu1.dsc
 66d22c6880f7c3a5d69b9771bc37464ae223a4cf105cb7af6667ec71ec7b7e66 18284139 ti-openmpi_1.0.0.15-0ubuntu1.tar.gz
Files:
 bdc301ce1a13dde03da2407badea6bc8 1626 partner/net extra ti-openmpi_1.0.0.15-0ubuntu1.dsc
 610861f44043801d8d6b1612c6b089b2 18284139 partner/net extra ti-openmpi_1.0.0.15-0ubuntu1.tar.gz