--- - branch: MAIN date: Sat Aug 15 21:27:46 UTC 2015 files: - new: '1.16' old: '1.15' path: pkgsrc/parallel/hwloc/Makefile pathrev: pkgsrc/parallel/hwloc/Makefile@1.16 type: modified - new: '1.4' old: '1.3' path: pkgsrc/parallel/hwloc/PLIST pathrev: pkgsrc/parallel/hwloc/PLIST@1.4 type: modified - new: '1.7' old: '1.6' path: pkgsrc/parallel/hwloc/distinfo pathrev: pkgsrc/parallel/hwloc/distinfo@1.7 type: modified id: 20150815T212746Z.41a4a3ad45b80090df9b76be6eb4d45e95243138 log: "Update to hwloc 1.11.1\n>From Benedek Gergely on pkgsrc-users.\n\nVersion 1.11.1\n--------------\n* hwloc-annotate does not drop unavailable resources from the output anymore\n and those may be annotated as well.\n* Fix an overzealous assertion when intermediate Group object while Groups\n are totally ignored.\n* Fix a memory leak on failure to load a xml diff file.\n* Fix some segfaults when inputting an invalid synthetic description.\n\nVersion 1.11.0\n--------------\n* API\n + Socket objects are renamed into Package to align with the terminology\n \ used by processor vendors. The old HWLOC_OBJ_SOCKET type and \"Socket\"\n \ name are still supported for backward compatibility.\n + HWLOC_OBJ_NODE is replaced with HWLOC_OBJ_NUMANODE for clarification.\n HWLOC_OBJ_NODE is still supported for backward compatibility.\n \"Node\" and \"NUMANode\" strings are supported as in earlier releases.\n* Detection improvements\n + Add support for Intel Knights Landing Xeon Phi.\n Thanks to Grzegorz Andrejczuk and Lukasz Anaczkowski.\n + Add Vendor, Model, Revision, SerialNumber, Type and LinuxDeviceID\n \ info attributes to Block OS devices on Linux. Thanks to Vineet Pedaballe\n \ for the help.\n - Add --disable-libudev to avoid dependency on the libudev library.\n + Add \"MemoryDevice\" Misc objects with information about DIMMs, on Linux\n when privileged and when I/O is enabled.\n Thanks to Vineet Pedaballe for the help.\n + Add a PCISlot attribute to PCI devices on Linux when supported to\n identify the physical PCI slot where the board is plugged.\n + Add CPUStepping info attribute on x86 processors,\n thanks to Thomas Ræ\x97¦hl for the suggestion.\n \ + Ignore the device-tree on non-Power architectures to avoid buggy\n detection on ARM. Thanks to Orion Poplawski for reporting the issue.\n + Work-around buggy Xeon E5v3 BIOS reporting invalid PCI-NUMA affinity\n for the PCI links on the second processor.\n + Add support for CUDA compute capability 5.x, thanks Benjamin Worpitz.\n + Many fixes to the x86 backend\n - Add L1i and fix L2/L3 type on old AMD processors without topoext support.\n - Fix Intel CPU family and model numbers when basic family isn't 6 or 15.\n - Fix package IDs on recent AMD processors.\n - Fix misc issues due to incomplete APIC IDs on x2APIC processors.\n \ - Avoid buggy discovery on old SGI Altix UVs with non-unique APIC IDs.\n + Gather total machine memory on NetBSD.\n* Tools\n + lstopo\n - Collapse identical PCI devices unless --no-collapse is given.\n This avoids gigantic outputs when a PCI device contains dozens of\n identical virtual functions.\n - The ASCII art output is now called \"ascii\", for instance in\n \"lstopo -.ascii\".\n The former \"txt\" extension is retained for backward compatibility.\n \ - Automatically scales graphical box width to the inner text in Cairo,\n ASCII and Windows outputs.\n - Add --rect to lstopo to force rectangular layout even for NUMA nodes.\n - Add --restrict-flags to configure the behavior of --restrict.\n \ - Objects may have a \"Type\" info attribute to specify a better type name\n \ and display it in lstopo.\n - Really export all verbose information to the given output file.\n + hwloc-annotate\n - May now operate on all types of objects, including I/O.\n - May now insert Misc objects in the topology.\n \ - Do not drop instruction caches and I/O devices from the output anymore.\n \ + Fix lstopo path in hwloc-gather-topology after install.\n* Misc\n + Fix hwloc/cudart.h for machines with multiple PCI domains,\n thanks to Imre Kerr for reporting the problem.\n + Fix PCI Bridge-specific depth attribute.\n + Fix hwloc_bitmap_intersect() for two infinite bitmaps.\n + Fix some corner cases in the building of levels on large NUMA machines\n with non-uniform NUMA groups and I/Os.\n + Improve the performance of object insertion by cpuset for large\n topologies.\n + Prefix verbose XML import errors with the source name.\n + Improve pkg-config checks and error messages.\n + Fix excluding after a component with an argument in the HWLOC_COMPONENTS\n environment variable.\n* Documentation\n + Fix the recommended way in documentation and examples to allocate memory\n on some node, it should use HWLOC_MEMBIND_BIND.\n Thanks to Nicolas Bouzat for reporting the issue.\n + Add a \"Miscellaneous objects\" section in the documentation.\n \ + Add a FAQ entry \"What happens to my topology if I disable symmetric\n multithreading, hyper-threading, etc. ?\" to the documentation.\n\nVersion 1.10.1\n--------------\n* Actually remove disallowed NUMA nodes from nodesets when the whole-system\n flag isn't enabled.\n* Fix the gathering of PCI domains. Thanks to James Custer for reporting\n the issue and providing a patch.\n* Fix the merging of identical parent and child in presence of Misc objects.\n Thanks to Dave Love for reporting the issue.\n* Fix some misordering of children when merging with ignore_keep_structure()\n \ in partially allowed topologies.\n* Fix an overzealous assertion in the debug code when running on a single-PU\n host with I/O. Thanks to Thomas Van Doren for reporting the issue.\n* Don't forget to setup NUMA node object nodesets in x86 backend (for BSDs)\n and OSF/Tru64 backend.\n* Fix cpuid-x86 build error with gcc -O3 on x86-32. Thanks to Thomas Van Doren\n for reporting the issue.\n* Fix support for future very large caches in the x86 backend.\n* Fix vendor/device names for SR-IOV PCI devices on Linux.\n* Fix an unlikely crash in case of buggy hierarchical distance matrix.\n* Fix PU os_index on some AIX releases. Thanks to Hendryk Bockelmann and\n Erik Schnetter for helping debugging.\n* Fix hwloc_bitmap_isincluded() in case of infinite sets.\n* Change hwloc-ls.desktop into a lstopo.desktop and only install it if\n lstopo is built with Cairo/X11 support. It cannot work with a non-graphical\n lstopo or hwloc-ls.\n* Add support for the renaming of Socket into Package in future releases.\n* Add support for the replacement of HWLOC_OBJ_NODE with HWLOC_OBJ_NUMANODE\n in future releases.\n* Clarify the documentation of distance matrices in hwloc.h and in the manpage\n of the hwloc-distances. Thanks to Dave Love for the suggestion.\n* Improve some error messages by displaying more information about the\n hwloc library in use.\n* Document how to deal with the ABI break when upgrading to the upcoming 2.0\n See \"How do I handle ABI breaks and API upgrades ?\" in the FAQ.\n\nVersion 1.10.0\n--------------\n* API\n \ + Add hwloc_topology_export_synthetic() to export a topology to a\n synthetic string without using lstopo. See the Synthetic topologies\n section in the documentation.\n + Add hwloc_topology_set/get_userdata() to let the application save\n a private pointer in the topology whenever it needs a way to find\n \ its own object corresponding to a topology.\n + Add hwloc_get_numanode_obj_by_os_index() and document that this function\n as well as hwloc_get_pu_obj_by_os_index() are good at converting\n nodesets and cpusets into objects.\n + hwloc_distrib() does not ignore any objects anymore when there are\n too many of them. They get merged with others instead.\n Thanks to Tim Creech for reporting the issue.\n* Tools\n + hwloc-bind --get now executes the command after displaying\n \ the binding instead of ignoring the command entirely.\n Thanks to John Donners for the suggestion.\n + Clarify that memory sizes shown in lstopo are local by default\n unless specified (total memory added in the root object).\n* Synthetic topologies\n + Synthetic topology descriptions may now specify attributes such as\n memory sizes and OS indexes. See the Synthetic topologies section\n \ in the documentation.\n + lstopo now exports in this fully-detailed format by default.\n The new option --export-synthetic-flags may be used to revert\n \ back the old format.\n* Documentation\n + Add the doc/examples/ subdirectory with several real-life examples,\n including the already existing hwloc-hello.C for basics.\n Thanks to Rob Aulwes for the suggestion.\n + Improve the documentation of CPU and memory binding in the API.\n + Add a FAQ entry about operating system errors, especially on AMD\n platforms with buggy cache information.\n + Add a FAQ entry about loading many topologies in a single program.\n* Misc\n + Work around buggy Linux kernels reporting 2 sockets instead\n 1 socket with 2 NUMA nodes for each Xeon E5 v3 (Haswell) processor.\n + pciutils/libpci support is now removed since libpciaccess works\n well and there's also a Linux-specific PCI backend. For the record,\n pciutils was GPL and therefore disabled by default since v1.6.2.\n + Add --disable-cpuid configure flag to work around buggy processor\n \ simulators reporting invalid CPUID information.\n Thanks for Andrew Friedley for reporting the issue.\n + Fix a racy use of libltdl when manipulating multiple topologies in\n different threads.\n Thanks to Andra Hugo for reporting the issue and testing patches.\n + Fix some build failures in private/misc.h.\n \ Thanks to Pavan Balaji and Ralph Castain for the reports.\n + Fix failures to detect X11/Xutil.h on some Solaris platforms.\n Thanks to Siegmar Gross for reporting the failure.\n + The plugin ABI has changed, this release will not load plugins\n built against previous hwloc releases.\n\nVersion 1.9.1\n-------------\n* Fix a crash when the PCI locality is invalid. Attach to the root object\n instead. Thanks to Nicolas Denoyelle for reporting the issue.\n* Fix -f in lstopo manpage. Thanks to Jirka Hladky for reporting the issue.\n* Fix hwloc_obj_type_sscanf() and others when strncasecmp() is not properly\n available. Thanks to Nick Papior Andersen for reporting the problem.\n* Mark Linux file descriptors as close-on-exec to avoid leaks on exec.\n* Fix some minor memory leaks.\n\nVersion 1.9.0\n-------------\n* API\n + Add hwloc_obj_type_sscanf() to extend hwloc_obj_type_of_string() with\n \ type-specific attributes such as Cache/Group depth and Cache type.\n hwloc_obj_type_of_string() is moved to hwloc/deprecated.h.\n + Add hwloc_linux_get_tid_last_cpu_location() for retrieving the\n last CPU where a Linux thread given by TID ran.\n + Add hwloc_distrib() to extend the old hwloc_distribute[v]() functions.\n hwloc_distribute[v]() is moved to hwloc/deprecated.h.\n + Don't mix total and local memory when displaying verbose object attributes\n with hwloc_obj_attr_snprintf() or in lstopo.\n* Backends\n + Add CPUVendor, CPUModelNumber and CPUFamilyNumber info attributes for\n x86, ia64 and Xeon Phi sockets on Linux, to extend the x86-specific\n \ support added in v1.8.1. Requested by Ralph Castain.\n + Add many CPU- and Platform-related info attributes on ARM and POWER\n platforms, in the Machine and Socket objects.\n + Add CUDA info attributes describing the number of multiprocessors and\n cores and the size of the global, shared and L2 cache memories in CUDA\n \ OS devices.\n + Add OpenCL info attributes describing the number of compute units and\n the global memory size in OpenCL OS devices.\n + The synthetic backend now accepts extended types such as L2Cache, L1i or\n Group3. lstopo also exports synthetic strings using these extended types.\n* Tools\n + lstopo\n \ - Do not overwrite output files by default anymore.\n Pass -f or --force to enforce it.\n - Display OpenCL, CUDA and Xeon Phi numbers of cores and memory sizes\n in the graphical output.\n - Fix export to stdout when specifying a Cairo-based output type\n with --of.\n + hwloc-ps\n - Add -e or --get-last-cpu-location to report where processes/threads\n run instead of where they are bound.\n \ - Report locations as likely-more-useful objects such as Cores or Sockets\n \ instead of Caches when possible.\n + hwloc-bind\n - Fix failure on Windows when not using --pid.\n - Add -e as a synonym to --get-last-cpu-location.\n \ + hwloc-distrib\n - Add --reverse to distribute using last objects first and singlify\n into last bits first. Thanks to Jirka Hladky for the suggestion.\n \ + hwloc-info\n - Report unified caches when looking for data or instruction cache\n ancestor objects.\n* Misc\n + Add experimental Visual Studio support under contrib/windows.\n Thanks to Eloi Gaudry for his help and for providing the first draft.\n + Fix some overzealous assertions and warnings about the ordering of\n objects on a level with respect to cpusets. The ordering is only\n guaranteed for complete cpusets (based on the first bit in sets).\n + Fix some memory leaks when importing xml diffs and when exporting a\n \"too complex\" entry.\n\nVersion 1.8.1\n-------------\n* Fix the cpuid code on Windows 64bits so that the x86 backend gets\n enabled as expected and can populate CPU information.\n Thanks to Robin Scher for reporting the problem.\n* Add CPUVendor/CPUModelNumber/CPUFamilyNumber attributes when running\n on x86 architecture. Thanks to Ralph Castain for the suggestion.\n* Work around buggy BIOS reporting duplicate NUMA nodes on Linux.\n \ Thanks to Jeff Becker for reporting the problem and testing the patch.\n* Add a name to the lstopo graphical window. Thanks to Michael Prokop\n for reporting the issue.\n\nVersion 1.8.0\n-------------\n* New components\n + Add the \"linuxpci\" component that always works on Linux even when\n libpciaccess and libpci aren't available (and even with a modified\n file-system root). By default the old \"pci\" component runs first\n because \"linuxpci\" lacks device names (obj->name is always NULL).\n* API\n + Add the topology difference API in hwloc/diff.h for manipulating\n many similar topologies.\n + Add hwloc_topology_dup() for duplicating an entire topology.\n + hwloc.h and hwloc/helper.h have been reorganized to clarify the\n documentation sections. The actual inline code has moved out of hwloc.h\n \ into the new hwloc/inlines.h.\n + Deprecated functions are now in hwloc/deprecated.h, and not in the\n official documentation anymore.\n* Tools\n + Add hwloc-diff and hwloc-patch tools together with the new diff API.\n + Add hwloc-compress-dir to (de)compress an entire directory of XML files\n using hwloc-diff and hwloc-patch.\n \ + Object colors in the graphical output of lstopo may be changed by adding\n \ a \"lstopoStyle\" info attribute. See CUSTOM COLORS in the lstopo(1) manpage\n \ for details. Thanks to Jirka Hladky for discussing the idea.\n + hwloc-gather-topology may now gather I/O-related files on Linux when\n --io is given. Only the linuxpci component supports discovering I/O\n objects from these extended tarballs.\n \ + hwloc-annotate now supports --ri to remove/replace info attributes with\n \ a given name.\n + hwloc-info supports \"root\" and \"all\" special locations for dumping\n information about the root object.\n + lstopo now supports --append-legend to append custom lines of text\n to the legend in the graphical output. Thanks to Jirka Hladky for\n discussing the idea.\n + hwloc-calc and friends have a more robust parsing of locations given\n on the command-line and they report useful error messages about it.\n + Add --whole-system to hwloc-bind, hwloc-calc, hwloc-distances and\n hwloc-distrib, and add --restrict to hwloc-bind for uniformity among\n tools.\n* Misc\n + Calling hwloc_topology_load() or hwloc_topology_set_*() on an already\n loaded topology now returns an error (deprecated since release 1.6.1).\n + Fix the initialisation of cpusets and nodesets in Group objects added\n \ when inserting PCI hostbridges.\n + Never merge Group objects that were added explicitly by the user with\n hwloc_custom_insert_group_object_by_parent().\n \ + Add a sanity check during dynamic plugin loading to prevent some\n crashes when hwloc is dynamically loaded by another plugin mechanisms.\n + Add --with-hwloc-plugins-path to specify the install/load directories\n of plugins.\n + Add the MICSerialNumber info attribute to the root object when running\n hwloc inside a Xeon Phi to match the same attribute in the MIC OS device\n when running in the host.\n\nVersion 1.7.2\n-------------\n* Do not create invalid block OS devices on very old Linux kernel such\n as RHEL4 2.6.9.\n* Fix PCI subvendor/device IDs.\n* Fix the management of Misc objects inserted by parent.\n Thanks to Jirka Hladky for reporting the problem.\n* Add a PortState into attribute to OpenFabrics OS devices.\n* Add a MICSerialNumber info attribute to Xeon PHI/MIC OS devices.\n* Improve verbose error messages when failing to load from XML.\n" module: pkgsrc subject: 'CVS commit: pkgsrc/parallel/hwloc' unixtime: '1439674066' user: asau