)]}' { "log": [ { "commit": "590e4d857153c5d4cf86052cdfd42cf9b0779841", "tree": "b8de0d7c163b81050f8c95f1b59ccbf05aed5d8f", "parents": [ "a200d8e44649de2cbb39de95f42ad4ef5dc8dc22" ], "author": { "name": "Anton Blanchard", "email": "anton@samba.org", "time": "Sun Jul 24 16:33:13 2011 +0000" }, "committer": { "name": "Benjamin Herrenschmidt", "email": "benh@kernel.crashing.org", "time": "Tue Sep 20 15:53:21 2011 +1000" }, "message": "sched: Allow SD_NODES_PER_DOMAIN to be overridden\n\nWe want to override the default value of SD_NODES_PER_DOMAIN on ppc64,\nso move it into linux/topology.h.\n\nSigned-off-by: Anton Blanchard \u003canton@samba.org\u003e\nAcked-by: Peter Zijlstra \u003ca.p.zijlstra@chello.nl\u003e\nSigned-off-by: Benjamin Herrenschmidt \u003cbenh@kernel.crashing.org\u003e\n" }, { "commit": "32e45ff43eaf5c17f5a82c9ad358d515622c2562", "tree": "be4984b6be4b3ffe259ef19efac7bb12f3b26c2a", "parents": [ "17441227f6258fc379c6ebfe21c3eec43b6f0de3" ], "author": { "name": "KOSAKI Motohiro", "email": "kosaki.motohiro@jp.fujitsu.com", "time": "Wed Jun 15 15:08:20 2011 -0700" }, "committer": { "name": "Linus Torvalds", "email": "torvalds@linux-foundation.org", "time": "Wed Jun 15 20:03:59 2011 -0700" }, "message": "mm: increase RECLAIM_DISTANCE to 30\n\nRecently, Robert Mueller reported (http://lkml.org/lkml/2010/9/12/236)\nthat zone_reclaim_mode doesn\u0027t work properly on his new NUMA server (Dual\nXeon E5520 + Intel S5520UR MB). He is using Cyrus IMAPd and it\u0027s built on\na very traditional single-process model.\n\n * a master process which reads config files and manages the other\n process\n * multiple imapd processes, one per connection\n * multiple pop3d processes, one per connection\n * multiple lmtpd processes, one per connection\n * periodical \"cleanup\" processes.\n\nThere are thousands of independent processes. The problem is, recent\nIntel motherboard turn on zone_reclaim_mode by default and traditional\nprefork model software don\u0027t work well on it. Unfortunatelly, such models\nare still typical even in the 21st century. We can\u0027t ignore them.\n\nThis patch raises the zone_reclaim_mode threshold to 30. 30 doesn\u0027t have\nany specific meaning. but 20 means that one-hop QPI/Hypertransport and\nsuch relatively cheap 2-4 socket machine are often used for traditional\nservers as above. The intention is that these machines don\u0027t use\nzone_reclaim_mode.\n\nNote: ia64 and Power have arch specific RECLAIM_DISTANCE definitions.\nThis patch doesn\u0027t change such high-end NUMA machine behavior.\n\nDave Hansen said:\n\n: I know specifically of pieces of x86 hardware that set the information\n: in the BIOS to \u002721\u0027 *specifically* so they\u0027ll get the zone_reclaim_mode\n: behavior which that implies.\n:\n: They\u0027ve done performance testing and run very large and scary benchmarks\n: to make sure that they _want_ this turned on. What this means for them\n: is that they\u0027ll probably be de-optimized, at least on newer versions of\n: the kernel.\n:\n: If you want to do this for particular systems, maybe _that_\u0027s what we\n: should do. Have a list of specific configurations that need the\n: defaults overridden either because they\u0027re buggy, or they have an\n: unusual hardware configuration not really reflected in the distance\n: table.\n\nAnd later said:\n\n: The original change in the hardware tables was for the benefit of a\n: benchmark. Said benchmark isn\u0027t going to get run on mainline until the\n: next batch of enterprise distros drops, at which point the hardware where\n: this was done will be irrelevant for the benchmark. I\u0027m sure any new\n: hardware will just set this distance to another yet arbitrary value to\n: make the kernel do what it wants. :)\n:\n: Also, when the hardware got _set_ to this initially, I complained. So, I\n: guess I\u0027m getting my way now, with this patch. I\u0027m cool with it.\n\nReported-by: Robert Mueller \u003crobm@fastmail.fm\u003e\nSigned-off-by: KOSAKI Motohiro \u003ckosaki.motohiro@jp.fujitsu.com\u003e\nAcked-by: Christoph Lameter \u003ccl@linux.com\u003e\nAcked-by: David Rientjes \u003crientjes@google.com\u003e\nReviewed-by: KAMEZAWA Hiroyuki \u003ckamezawa.hiroyu@jp.fujitsu.com\u003e\nCc: Benjamin Herrenschmidt \u003cbenh@kernel.crashing.org\u003e\nCc: \"Luck, Tony\" \u003ctony.luck@intel.com\u003e\nAcked-by: Dave Hansen \u003cdave@linux.vnet.ibm.com\u003e\nSigned-off-by: Andrew Morton \u003cakpm@linux-foundation.org\u003e\nSigned-off-by: Linus Torvalds \u003ctorvalds@linux-foundation.org\u003e\n" }, { "commit": "01a08546af311c065f34727787dd0cc8dc0c216f", "tree": "759d8898d8714b587985433a189070497081828f", "parents": [ "f269893c575167447cc9f6d1867e639fb5b6f0c5" ], "author": { "name": "Heiko Carstens", "email": "heiko.carstens@de.ibm.com", "time": "Tue Aug 31 10:28:16 2010 +0200" }, "committer": { "name": "Ingo Molnar", "email": "mingo@elte.hu", "time": "Thu Sep 09 20:41:20 2010 +0200" }, "message": "sched: Add book scheduling domain\n\nOn top of the SMT and MC scheduling domains this adds the BOOK scheduling\ndomain. This is useful for NUMA like machines which do not have an interface\nwhich tells which piece of memory is attached to which node or where the\nhardware performs striping.\n\nSigned-off-by: Heiko Carstens \u003cheiko.carstens@de.ibm.com\u003e\nSigned-off-by: Peter Zijlstra \u003ca.p.zijlstra@chello.nl\u003e\nLKML-Reference: \u003c20100831082844.253053798@de.ibm.com\u003e\nSigned-off-by: Ingo Molnar \u003cmingo@elte.hu\u003e\n" }, { "commit": "251060006003b79b788f8ce5a827ee5354a42910", "tree": "eb69ce45bd31ffc0609300afbe0cca2a2cbf2ab1", "parents": [ "b645bd1286f2fbcd2eb4ab3bed5884f63c42e363" ], "author": { "name": "Lee Schermerhorn", "email": "Lee.Schermerhorn@hp.com", "time": "Mon Aug 09 17:19:00 2010 -0700" }, "committer": { "name": "Linus Torvalds", "email": "torvalds@linux-foundation.org", "time": "Mon Aug 09 20:44:57 2010 -0700" }, "message": "topology: alternate fix for ia64 tiger_defconfig build breakage\n\nDefine stubs for the numa_*_id() generic percpu related functions for\nnon-NUMA configurations in \u003casm-generic/topology.h\u003e where the other\nnon-numa stubs live.\n\nFixes ia64 !NUMA build breakage -- e.g., tiger_defconfig\n\nBack out now unneeded \u0027#ifndef CONFIG_NUMA\u0027 guards from ia64 smpboot.c\n\nSigned-off-by: Lee Schermerhorn \u003clee.schermerhorn@hp.com\u003e\nTested-by: Tony Luck \u003ctony.luck@intel.com\u003e\nAcked-by: Tony Luck \u003ctony.luck@intel.com\u003e\nSigned-off-by: Andrew Morton \u003cakpm@linux-foundation.org\u003e\nSigned-off-by: Linus Torvalds \u003ctorvalds@linux-foundation.org\u003e\n" }, { "commit": "2ec57d448b2e8fcfba539a46701b43f14f037f17", "tree": "92f014a595860b3124816ba4bb1ec8751ea66543", "parents": [ "0b2e918aa99fe6c3b8f163aa323a275ad8577828" ], "author": { "name": "Michael Neuling", "email": "mikey@neuling.org", "time": "Tue Jun 29 12:02:01 2010 +1000" }, "committer": { "name": "Ingo Molnar", "email": "mingo@elte.hu", "time": "Tue Jun 29 10:44:29 2010 +0200" }, "message": "sched: Fix spelling of sibling\n\nNo logic changes, only spelling.\n\nSigned-off-by: Michael Neuling \u003cmikey@neuling.org\u003e\nCc: linuxppc-dev@ozlabs.org\nCc: David Howells \u003cdhowells@redhat.com\u003e\nCc: Peter Zijlstra \u003cpeterz@infradead.org\u003e\nLKML-Reference: \u003c15249.1277776921@neuling.org\u003e\nSigned-off-by: Ingo Molnar \u003cmingo@elte.hu\u003e\n" }, { "commit": "532cb4c401e225b084c14d6bd6a2f8ee561de2f1", "tree": "0ce57c2e21cd12ee05561ab2b9c4b66729da8e5a", "parents": [ "9d5efe05eb0c904545a28b19c18b949f23334de0" ], "author": { "name": "Michael Neuling", "email": "mikey@neuling.org", "time": "Tue Jun 08 14:57:02 2010 +1000" }, "committer": { "name": "Ingo Molnar", "email": "mingo@elte.hu", "time": "Wed Jun 09 10:34:55 2010 +0200" }, "message": "sched: Add asymmetric group packing option for sibling domain\n\nCheck to see if the group is packed in a sched doman.\n\nThis is primarily intended to used at the sibling level. Some cores\nlike POWER7 prefer to use lower numbered SMT threads. In the case of\nPOWER7, it can move to lower SMT modes only when higher threads are\nidle. When in lower SMT modes, the threads will perform better since\nthey share less core resources. Hence when we have idle threads, we\nwant them to be the higher ones.\n\nThis adds a hook into f_b_g() called check_asym_packing() to check the\npacking. This packing function is run on idle threads. It checks to\nsee if the busiest CPU in this domain (core in the P7 case) has a\nhigher CPU number than what where the packing function is being run\non. If it is, calculate the imbalance and return the higher busier\nthread as the busiest group to f_b_g(). Here we are assuming a lower\nCPU number will be equivalent to a lower SMT thread number.\n\nIt also creates a new SD_ASYM_PACKING flag to enable this feature at\nany scheduler domain level.\n\nIt also creates an arch hook to enable this feature at the sibling\nlevel. The default function doesn\u0027t enable this feature.\n\nBased heavily on patch from Peter Zijlstra.\nFixes from Srivatsa Vaddagiri.\n\nSigned-off-by: Michael Neuling \u003cmikey@neuling.org\u003e\nSigned-off-by: Srivatsa Vaddagiri \u003cvatsa@linux.vnet.ibm.com\u003e\nSigned-off-by: Peter Zijlstra \u003ca.p.zijlstra@chello.nl\u003e\nCc: Arjan van de Ven \u003carjan@linux.intel.com\u003e\nCc: \"H. Peter Anvin\" \u003chpa@zytor.com\u003e\nCc: Thomas Gleixner \u003ctglx@linutronix.de\u003e\nLKML-Reference: \u003c20100608045702.2936CCC897@localhost.localdomain\u003e\nSigned-off-by: Ingo Molnar \u003cmingo@elte.hu\u003e\n" }, { "commit": "7aac789885512388a66d47280d7e7777ffba1e59", "tree": "af4ac98260268889a422dd264102d2f15d5c1983", "parents": [ "3bccd996276b108c138e8176793a26ecef54d573" ], "author": { "name": "Lee Schermerhorn", "email": "lee.schermerhorn@hp.com", "time": "Wed May 26 14:45:00 2010 -0700" }, "committer": { "name": "Linus Torvalds", "email": "torvalds@linux-foundation.org", "time": "Thu May 27 09:12:57 2010 -0700" }, "message": "numa: introduce numa_mem_id()- effective local memory node id\n\nIntroduce numa_mem_id(), based on generic percpu variable infrastructure\nto track \"nearest node with memory\" for archs that support memoryless\nnodes.\n\nDefine API in \u003clinux/topology.h\u003e when CONFIG_HAVE_MEMORYLESS_NODES\ndefined, else stubs. Architectures will define HAVE_MEMORYLESS_NODES\nif/when they support them.\n\nArchs can override definitions of:\n\nnuma_mem_id() - returns node number of \"local memory\" node\nset_numa_mem() - initialize [this cpus\u0027] per cpu variable \u0027numa_mem\u0027\ncpu_to_mem() - return numa_mem for specified cpu; may be used as lvalue\n\nGeneric initialization of \u0027numa_mem\u0027 occurs in __build_all_zonelists().\nThis will initialize the boot cpu at boot time, and all cpus on change of\nnuma_zonelist_order, or when node or memory hot-plug requires zonelist\nrebuild. Archs that support memoryless nodes will need to initialize\n\u0027numa_mem\u0027 for secondary cpus as they\u0027re brought on-line.\n\n[akpm@linux-foundation.org: fix build]\nSigned-off-by: Lee Schermerhorn \u003clee.schermerhorn@hp.com\u003e\nSigned-off-by: Christoph Lameter \u003ccl@linux-foundation.org\u003e\nCc: Tejun Heo \u003ctj@kernel.org\u003e\nCc: Mel Gorman \u003cmel@csn.ul.ie\u003e\nCc: Christoph Lameter \u003ccl@linux-foundation.org\u003e\nCc: Nick Piggin \u003cnpiggin@suse.de\u003e\nCc: David Rientjes \u003crientjes@google.com\u003e\nCc: Eric Whitney \u003ceric.whitney@hp.com\u003e\nCc: KAMEZAWA Hiroyuki \u003ckamezawa.hiroyu@jp.fujitsu.com\u003e\nCc: Ingo Molnar \u003cmingo@elte.hu\u003e\nCc: Thomas Gleixner \u003ctglx@linutronix.de\u003e\nCc: \"H. Peter Anvin\" \u003chpa@zytor.com\u003e\nCc: \"Luck, Tony\" \u003ctony.luck@intel.com\u003e\nCc: Pekka Enberg \u003cpenberg@cs.helsinki.fi\u003e\nCc: \u003clinux-arch@vger.kernel.org\u003e\nSigned-off-by: Andrew Morton \u003cakpm@linux-foundation.org\u003e\nSigned-off-by: Linus Torvalds \u003ctorvalds@linux-foundation.org\u003e\n" }, { "commit": "7281201922a0063fa60804ce39c277fc98142a47", "tree": "4bf089d077b1055e54bc1411dcc0db121d01d9fa", "parents": [ "866707fc2721df8fee637fcf0239628b9231f9ea" ], "author": { "name": "Lee Schermerhorn", "email": "lee.schermerhorn@hp.com", "time": "Wed May 26 14:44:56 2010 -0700" }, "committer": { "name": "Linus Torvalds", "email": "torvalds@linux-foundation.org", "time": "Thu May 27 09:12:57 2010 -0700" }, "message": "numa: add generic percpu var numa_node_id() implementation\n\nRework the generic version of the numa_node_id() function to use the new\ngeneric percpu variable infrastructure.\n\nGuard the new implementation with a new config option:\n\n CONFIG_USE_PERCPU_NUMA_NODE_ID.\n\nArchs which support this new implemention will default this option to \u0027y\u0027\nwhen NUMA is configured. This config option could be removed if/when all\narchs switch over to the generic percpu implementation of numa_node_id().\nArch support involves:\n\n 1) converting any existing per cpu variable implementations to use\n this implementation. x86_64 is an instance of such an arch.\n 2) archs that don\u0027t use a per cpu variable for numa_node_id() will\n need to initialize the new per cpu variable \"numa_node\" as cpus\n are brought on-line. ia64 is an example.\n 3) Defining USE_PERCPU_NUMA_NODE_ID in arch dependent Kconfig--e.g.,\n when NUMA is configured. This is required because I have\n retained the old implementation by default to allow archs to\n be modified incrementally, as desired.\n\nSubsequent patches will convert x86_64 and ia64 to use this implemenation.\n\nSigned-off-by: Lee Schermerhorn \u003clee.schermerhorn@hp.com\u003e\nCc: Tejun Heo \u003ctj@kernel.org\u003e\nCc: Mel Gorman \u003cmel@csn.ul.ie\u003e\nReviewed-by: Christoph Lameter \u003ccl@linux-foundation.org\u003e\nCc: Nick Piggin \u003cnpiggin@suse.de\u003e\nCc: David Rientjes \u003crientjes@google.com\u003e\nCc: Eric Whitney \u003ceric.whitney@hp.com\u003e\nCc: KAMEZAWA Hiroyuki \u003ckamezawa.hiroyu@jp.fujitsu.com\u003e\nCc: Ingo Molnar \u003cmingo@elte.hu\u003e\nCc: Thomas Gleixner \u003ctglx@linutronix.de\u003e\nCc: \"H. Peter Anvin\" \u003chpa@zytor.com\u003e\nCc: \"Luck, Tony\" \u003ctony.luck@intel.com\u003e\nCc: Pekka Enberg \u003cpenberg@cs.helsinki.fi\u003e\nCc: \u003clinux-arch@vger.kernel.org\u003e\nSigned-off-by: Andrew Morton \u003cakpm@linux-foundation.org\u003e\nSigned-off-by: Linus Torvalds \u003ctorvalds@linux-foundation.org\u003e\n" }, { "commit": "50b926e439620c469565e8be0f28be78f5fca1ce", "tree": "3eb4b4aa89ed7c6d61457cdf745c5d52f577671a", "parents": [ "24bc7347da73a9ed3383056c3d0f28c0e361621e" ], "author": { "name": "Mike Galbraith", "email": "efault@gmx.de", "time": "Mon Jan 04 14:44:56 2010 +0100" }, "committer": { "name": "Ingo Molnar", "email": "mingo@elte.hu", "time": "Thu Jan 21 13:39:03 2010 +0100" }, "message": "sched: Fix vmark regression on big machines\n\nSD_PREFER_SIBLING is set at the CPU domain level if power saving isn\u0027t\nenabled, leading to many cache misses on large machines as we traverse\nlooking for an idle shared cache to wake to. Change the enabler of\nselect_idle_sibling() to SD_SHARE_PKG_RESOURCES, and enable same at the\nsibling domain level.\n\nReported-by: Lin Ming \u003cming.m.lin@intel.com\u003e\nSigned-off-by: Mike Galbraith \u003cefault@gmx.de\u003e\nSigned-off-by: Peter Zijlstra \u003ca.p.zijlstra@chello.nl\u003e\nLKML-Reference: \u003c1262612696.15495.15.camel@marge.simson.net\u003e\nSigned-off-by: Ingo Molnar \u003cmingo@elte.hu\u003e\n" }, { "commit": "799e2205ec65e174f752b558c62a92c4752df313", "tree": "dc2438771be941b485f0da12a1ff177f0797e57b", "parents": [ "e17b38bf9e70d74f3739a600db75240078ac1407" ], "author": { "name": "Peter Zijlstra", "email": "a.p.zijlstra@chello.nl", "time": "Fri Oct 09 12:16:40 2009 +0200" }, "committer": { "name": "Ingo Molnar", "email": "mingo@elte.hu", "time": "Wed Oct 14 15:02:34 2009 +0200" }, "message": "sched: Disable SD_PREFER_LOCAL for MC/CPU domains\n\nYanmin reported that both tbench and hackbench were significantly\nhurt by trying to keep tasks local on these domains, esp on small\ncache machines.\n\nSo disable it in order to promote spreading outside of the cache\ndomains.\n\nReported-by: \"Zhang, Yanmin\" \u003cyanmin_zhang@linux.intel.com\u003e\nSigned-off-by: Peter Zijlstra \u003ca.p.zijlstra@chello.nl\u003e\nCC: Mike Galbraith \u003cefault@gmx.de\u003e\nLKML-Reference: \u003c1255083400.8802.15.camel@laptop\u003e\nSigned-off-by: Ingo Molnar \u003cmingo@elte.hu\u003e\n" }, { "commit": "6f401420e2822c24c36e6e1c657f6e7f7f777a93", "tree": "bd11d56f964190886245c383746a736076850711", "parents": [ "fe71a3c7dc8cfe0f239c04b4fc6501f4aa56aa0a" ], "author": { "name": "Rusty Russell", "email": "rusty@rustcorp.com.au", "time": "Thu Sep 24 09:34:40 2009 -0600" }, "committer": { "name": "Rusty Russell", "email": "rusty@rustcorp.com.au", "time": "Thu Sep 24 09:34:41 2009 +0930" }, "message": "cpumask: remove obsolete topology_core_siblings and topology_thread_siblings: core\n\nThere were replaced by topology_core_cpumask and topology_thread_cpumask.\n\nSigned-off-by: Rusty Russell \u003crusty@rustcorp.com.au\u003e\n" }, { "commit": "182a85f8a119c789610a9d464f4129ded9f3c107", "tree": "ffc36e5049895bf85543ba8d0225805e15fe8d92", "parents": [ "5a9b86f647a56862cdc0a1362bfb015ae921af7f" ], "author": { "name": "Peter Zijlstra", "email": "a.p.zijlstra@chello.nl", "time": "Wed Sep 16 13:24:49 2009 +0200" }, "committer": { "name": "Ingo Molnar", "email": "mingo@elte.hu", "time": "Wed Sep 16 16:44:33 2009 +0200" }, "message": "sched: Disable wakeup balancing\n\nSysbench thinks SD_BALANCE_WAKE is too agressive and kbuild doesn\u0027t\nreally mind too much, SD_BALANCE_NEWIDLE picks up most of the\nslack.\n\nOn a dual socket, quad core, dual thread nehalem system:\n\nsysbench (--num_threads\u003d16):\n\n SD_BALANCE_WAKE-: 13982 tx/s\n SD_BALANCE_WAKE+: 15688 tx/s\n\nkbuild (-j16):\n\n SD_BALANCE_WAKE-: 47.648295846 seconds time elapsed ( +- 0.312% )\n SD_BALANCE_WAKE+: 47.608607360 seconds time elapsed ( +- 0.026% )\n\n(same within noise)\n\nSigned-off-by: Peter Zijlstra \u003ca.p.zijlstra@chello.nl\u003e\nLKML-Reference: \u003cnew-submission\u003e\nSigned-off-by: Ingo Molnar \u003cmingo@elte.hu\u003e\n" }, { "commit": "59abf02644c45f1591e1374ee7bb45dc757fcb88", "tree": "ce9cb0a4a57873b9d7d75cce4d28308019367192", "parents": [ "e69b0f1b41c0e57bb1e29100b5810a5914efcb45" ], "author": { "name": "Peter Zijlstra", "email": "a.p.zijlstra@chello.nl", "time": "Wed Sep 16 08:28:30 2009 +0200" }, "committer": { "name": "Ingo Molnar", "email": "mingo@elte.hu", "time": "Wed Sep 16 08:42:40 2009 +0200" }, "message": "sched: Add SD_PREFER_LOCAL\n\nAnd turn it on for NUMA and MC domains. This improves\nlocality in balancing decisions by keeping up to\ncapacity amount of tasks local before looking for idle\nCPUs. (and twice the capacity if SD_POWERSAVINGS_BALANCE\nis set.)\n\nSigned-off-by: Peter Zijlstra \u003ca.p.zijlstra@chello.nl\u003e\nLKML-Reference: \u003cnew-submission\u003e\nSigned-off-by: Ingo Molnar \u003cmingo@elte.hu\u003e\n" }, { "commit": "b8a543ea5a5896830a9969bacfd047f9d15940b2", "tree": "45bae76596f2a51f3cc79d55c9c4e0da064a7a2c", "parents": [ "0ec9fab3d186d9cbb00c0f694d4a260d07c198d9" ], "author": { "name": "Peter Zijlstra", "email": "a.p.zijlstra@chello.nl", "time": "Tue Sep 15 15:22:03 2009 +0200" }, "committer": { "name": "Ingo Molnar", "email": "mingo@elte.hu", "time": "Tue Sep 15 16:51:23 2009 +0200" }, "message": "sched: Reduce forkexec_idx\n\nIf we\u0027re looking to place a new task, we might as well find the\nidlest position _now_, not 1 tick ago.\n\nSigned-off-by: Peter Zijlstra \u003ca.p.zijlstra@chello.nl\u003e\nLKML-Reference: \u003cnew-submission\u003e\nSigned-off-by: Ingo Molnar \u003cmingo@elte.hu\u003e\n" }, { "commit": "0ec9fab3d186d9cbb00c0f694d4a260d07c198d9", "tree": "07773edcece2dd82a63265e027793fe8b2231960", "parents": [ "6bd7821f905a8d6c471f0d6675f5cb7ea448d791" ], "author": { "name": "Mike Galbraith", "email": "efault@gmx.de", "time": "Tue Sep 15 15:07:03 2009 +0200" }, "committer": { "name": "Ingo Molnar", "email": "mingo@elte.hu", "time": "Tue Sep 15 16:51:16 2009 +0200" }, "message": "sched: Improve latencies and throughput\n\nMake the idle balancer more agressive, to improve a\nx264 encoding workload provided by Jason Garrett-Glaser:\n\n NEXT_BUDDY NO_LB_BIAS\n encoded 600 frames, 252.82 fps, 22096.60 kb/s\n encoded 600 frames, 250.69 fps, 22096.60 kb/s\n encoded 600 frames, 245.76 fps, 22096.60 kb/s\n\n NO_NEXT_BUDDY LB_BIAS\n encoded 600 frames, 344.44 fps, 22096.60 kb/s\n encoded 600 frames, 346.66 fps, 22096.60 kb/s\n encoded 600 frames, 352.59 fps, 22096.60 kb/s\n\n NO_NEXT_BUDDY NO_LB_BIAS\n encoded 600 frames, 425.75 fps, 22096.60 kb/s\n encoded 600 frames, 425.45 fps, 22096.60 kb/s\n encoded 600 frames, 422.49 fps, 22096.60 kb/s\n\nPeter pointed out that this is better done via newidle_idx,\nnot via LB_BIAS, newidle balancing should look for where\nthere is load _now_, not where there was load 2 ticks ago.\n\nWorst-case latencies are improved as well as no buddies\nmeans less vruntime spread. (as per prior lkml discussions)\n\nThis change improves kbuild-peak parallelism as well.\n\nReported-by: Jason Garrett-Glaser \u003cdarkshikari@gmail.com\u003e\nSigned-off-by: Mike Galbraith \u003cefault@gmx.de\u003e\nSigned-off-by: Peter Zijlstra \u003ca.p.zijlstra@chello.nl\u003e\nLKML-Reference: \u003c1253011667.9128.16.camel@marge.simson.net\u003e\nSigned-off-by: Ingo Molnar \u003cmingo@elte.hu\u003e\n" }, { "commit": "6bd7821f905a8d6c471f0d6675f5cb7ea448d791", "tree": "edf79a70823c3f6903d4a6a3406f36078a19199b", "parents": [ "78e7ed53c9f42f04f9401ada6f7047db60781676" ], "author": { "name": "Peter Zijlstra", "email": "a.p.zijlstra@chello.nl", "time": "Fri Sep 11 18:42:15 2009 +0200" }, "committer": { "name": "Ingo Molnar", "email": "mingo@elte.hu", "time": "Tue Sep 15 16:01:08 2009 +0200" }, "message": "sched: Fix some domain tunings\n\nCPU level should have WAKE_AFFINE, whereas ALLNODES is dubious.\n\nSigned-off-by: Peter Zijlstra \u003ca.p.zijlstra@chello.nl\u003e\nLKML-Reference: \u003cnew-submission\u003e\nSigned-off-by: Ingo Molnar \u003cmingo@elte.hu\u003e\n" }, { "commit": "78e7ed53c9f42f04f9401ada6f7047db60781676", "tree": "24f45333ce4479b27c96b425c7d09c080a26609f", "parents": [ "d7c33c4930f569caf6b2ece597432853c4151a45" ], "author": { "name": "Peter Zijlstra", "email": "a.p.zijlstra@chello.nl", "time": "Thu Sep 03 13:16:51 2009 +0200" }, "committer": { "name": "Ingo Molnar", "email": "mingo@elte.hu", "time": "Tue Sep 15 16:01:07 2009 +0200" }, "message": "sched: Tweak wake_idx\n\nWhen merging select_task_rq_fair() and sched_balance_self() we lost\nthe use of wake_idx, restore that and set them to 0 to make wake\nbalancing more aggressive.\n\nSigned-off-by: Peter Zijlstra \u003ca.p.zijlstra@chello.nl\u003e\nLKML-Reference: \u003cnew-submission\u003e\nSigned-off-by: Ingo Molnar \u003cmingo@elte.hu\u003e\n" }, { "commit": "c88d5910890ad35af283344417891344604f0438", "tree": "4e2025d569c3e03a7ec5163f0a9bc159114ee14e", "parents": [ "e9c8431185d6c406887190519f6dbdd112641686" ], "author": { "name": "Peter Zijlstra", "email": "a.p.zijlstra@chello.nl", "time": "Thu Sep 10 13:50:02 2009 +0200" }, "committer": { "name": "Ingo Molnar", "email": "mingo@elte.hu", "time": "Tue Sep 15 16:01:05 2009 +0200" }, "message": "sched: Merge select_task_rq_fair() and sched_balance_self()\n\nThe problem with wake_idle() is that is doesn\u0027t respect things like\ncpu_power, which means it doesn\u0027t deal well with SMT nor the recent\nRT interaction.\n\nTo cure this, it needs to do what sched_balance_self() does, which\nleads to the possibility of merging select_task_rq_fair() and\nsched_balance_self().\n\nModify sched_balance_self() to:\n\n - update_shares() when walking up the domain tree,\n (it only called it for the top domain, but it should\n have done this anyway), which allows us to remove\n this ugly bit from try_to_wake_up().\n\n - do wake_affine() on the smallest domain that contains\n both this (the waking) and the prev (the wakee) cpu for\n WAKE invocations.\n\nThen use the top-down balance steps it had to replace wake_idle().\n\nThis leads to the dissapearance of SD_WAKE_BALANCE and\nSD_WAKE_IDLE_FAR, with SD_WAKE_IDLE replaced with SD_BALANCE_WAKE.\n\nSD_WAKE_AFFINE needs SD_BALANCE_WAKE to be effective.\n\nTouch all topology bits to replace the old with new SD flags --\nplatforms might need re-tuning, enabling SD_BALANCE_WAKE\nconditionally on a NUMA distance seems like a good additional\nfeature, magny-core and small nehalem systems would want this\nenabled, systems with slow interconnects would not.\n\nSigned-off-by: Peter Zijlstra \u003ca.p.zijlstra@chello.nl\u003e\nLKML-Reference: \u003cnew-submission\u003e\nSigned-off-by: Ingo Molnar \u003cmingo@elte.hu\u003e\n" }, { "commit": "a8fae3ec5f118dc92517dcbed3ecf69ddb641d0f", "tree": "e01715204f2a6da7277cb4d397449df9fcf0ea6b", "parents": [ "71a29aa7b600595d0ef373ea605ac656876d1f2f" ], "author": { "name": "Peter Zijlstra", "email": "a.p.zijlstra@chello.nl", "time": "Mon Sep 07 18:32:32 2009 +0200" }, "committer": { "name": "Ingo Molnar", "email": "mingo@elte.hu", "time": "Mon Sep 07 22:00:17 2009 +0200" }, "message": "sched: enable SD_WAKE_IDLE\n\nNow that SD_WAKE_IDLE doesn\u0027t make pipe-test suck anymore,\nenable it by default for MC, CPU and NUMA domains.\n\nSigned-off-by: Peter Zijlstra \u003ca.p.zijlstra@chello.nl\u003e\nLKML-Reference: \u003cnew-submission\u003e\nSigned-off-by: Ingo Molnar \u003cmingo@elte.hu\u003e\n" }, { "commit": "840a0653100dbde599ae8ddf83fa214dfa5fd1aa", "tree": "273ffb82cb6d6f48c35d94b68e54874451956001", "parents": [ "47734f89be0614b5acbd6a532390f9c72f019648" ], "author": { "name": "Ingo Molnar", "email": "mingo@elte.hu", "time": "Fri Sep 04 11:32:54 2009 +0200" }, "committer": { "name": "Ingo Molnar", "email": "mingo@elte.hu", "time": "Fri Sep 04 11:52:54 2009 +0200" }, "message": "sched: Turn on SD_BALANCE_NEWIDLE\n\nStart the re-tuning of the balancer by turning on newidle.\n\nIt improves hackbench performance and parallelism on a 4x4 box.\nThe \"perf stat --repeat 10\" measurements give us:\n\n domain0 domain1\n .......................................\n -SD_BALANCE_NEWIDLE -SD_BALANCE_NEWIDLE:\n 2041.273208 task-clock-msecs # 9.354 CPUs ( +- 0.363% )\n\n +SD_BALANCE_NEWIDLE -SD_BALANCE_NEWIDLE:\n 2086.326925 task-clock-msecs # 11.934 CPUs ( +- 0.301% )\n\n +SD_BALANCE_NEWIDLE +SD_BALANCE_NEWIDLE:\n 2115.289791 task-clock-msecs # 12.158 CPUs ( +- 0.263% )\n\nAcked-by: Peter Zijlstra \u003ca.p.zijlstra@chello.nl\u003e\nCc: Andreas Herrmann \u003candreas.herrmann3@amd.com\u003e\nCc: Andreas Herrmann \u003candreas.herrmann3@amd.com\u003e\nCc: Gautham R Shenoy \u003cego@in.ibm.com\u003e\nCc: Balbir Singh \u003cbalbir@in.ibm.com\u003e\nLKML-Reference: \u003cnew-submission\u003e\nSigned-off-by: Ingo Molnar \u003cmingo@elte.hu\u003e\n" }, { "commit": "47734f89be0614b5acbd6a532390f9c72f019648", "tree": "48c7a765e90927ffed0158518157650c9b09dda8", "parents": [ "d7ea17a76916e456fcc78e45142c66f7fb875e3d" ], "author": { "name": "Ingo Molnar", "email": "mingo@elte.hu", "time": "Fri Sep 04 11:21:24 2009 +0200" }, "committer": { "name": "Ingo Molnar", "email": "mingo@elte.hu", "time": "Fri Sep 04 11:52:53 2009 +0200" }, "message": "sched: Clean up topology.h\n\nRe-organize the flag settings so that it\u0027s visible at a glance\nwhich sched-domains flags are set and which not.\n\nWith the new balancer code we\u0027ll need to re-tune these details\nanyway, so make it cleaner to make fewer mistakes down the\nroad ;-)\n\nCc: Peter Zijlstra \u003ca.p.zijlstra@chello.nl\u003e\nCc: Andreas Herrmann \u003candreas.herrmann3@amd.com\u003e\nCc: Andreas Herrmann \u003candreas.herrmann3@amd.com\u003e\nCc: Gautham R Shenoy \u003cego@in.ibm.com\u003e\nCc: Balbir Singh \u003cbalbir@in.ibm.com\u003e\nLKML-Reference: \u003cnew-submission\u003e\nSigned-off-by: Ingo Molnar \u003cmingo@elte.hu\u003e\n" }, { "commit": "a52bfd73589eaf88d9c95ad2c1de0b38a6b27972", "tree": "33cee609a46624525abf1643d71d0fbc79e25f87", "parents": [ "cc9fba7d7672fa3ed58d9d9ecb6c45b1351c29a6" ], "author": { "name": "Peter Zijlstra", "email": "a.p.zijlstra@chello.nl", "time": "Tue Sep 01 10:34:35 2009 +0200" }, "committer": { "name": "Ingo Molnar", "email": "mingo@elte.hu", "time": "Fri Sep 04 10:09:54 2009 +0200" }, "message": "sched: Add smt_gain\n\nThe idea is that multi-threading a core yields more work\ncapacity than a single thread, provide a way to express a\nstatic gain for threads.\n\nSigned-off-by: Peter Zijlstra \u003ca.p.zijlstra@chello.nl\u003e\nTested-by: Andreas Herrmann \u003candreas.herrmann3@amd.com\u003e\nAcked-by: Andreas Herrmann \u003candreas.herrmann3@amd.com\u003e\nAcked-by: Gautham R Shenoy \u003cego@in.ibm.com\u003e\nCc: Balbir Singh \u003cbalbir@in.ibm.com\u003e\nLKML-Reference: \u003c20090901083826.073345955@chello.nl\u003e\nSigned-off-by: Ingo Molnar \u003cmingo@elte.hu\u003e\n" }, { "commit": "082edb7bf443eb8eda15b482d16ad9dd8137ad24", "tree": "167d8c2ca193af9161aded5f368f300981c59535", "parents": [ "0b966252d9e5d95ec2d11e63d7e55b42913aa5b7" ], "author": { "name": "Rusty Russell", "email": "rusty@rustcorp.com.au", "time": "Fri Mar 13 23:43:37 2009 +1030" }, "committer": { "name": "Ingo Molnar", "email": "mingo@elte.hu", "time": "Fri Mar 13 14:35:31 2009 +0100" }, "message": "numa, cpumask: move numa_node_id default implementation to topology.h\n\nImpact: cleanup, potential bugfix\n\nNot sure what changed to expose this, but clearly that numa_node_id()\ndoesn\u0027t belong in mmzone.h (the inline in gfp.h is probably overkill, too).\n\nIn file included from include/linux/topology.h:34,\n from arch/x86/mm/numa.c:2:\n/home/rusty/patches-cpumask/linux-2.6/arch/x86/include/asm/topology.h:64:1: warning: \"numa_node_id\" redefined\nIn file included from include/linux/topology.h:32,\n from arch/x86/mm/numa.c:2:\ninclude/linux/mmzone.h:770:1: warning: this is the location of the previous definition\n\nSigned-off-by: Rusty Russell \u003crusty@rustcorp.com.au\u003e\nCc: Mike Travis \u003ctravis@sgi.com\u003e\nLKML-Reference: \u003c200903132343.37661.rusty@rustcorp.com.au\u003e\nSigned-off-by: Ingo Molnar \u003cmingo@elte.hu\u003e\n" }, { "commit": "a70f730282019f487aa33a84e5ac9a5e89c5abd0", "tree": "e6891ec5db5383c6f39617d0cc9671e1a0d1a988", "parents": [ "c69fc56de1df5769f2ec69c915c7ad5afe63804c" ], "author": { "name": "Rusty Russell", "email": "rusty@rustcorp.com.au", "time": "Fri Mar 13 14:49:46 2009 +1030" }, "committer": { "name": "Rusty Russell", "email": "rusty@rustcorp.com.au", "time": "Fri Mar 13 14:49:46 2009 +1030" }, "message": "cpumask: replace node_to_cpumask with cpumask_of_node.\n\nImpact: cleanup\n\nnode_to_cpumask (and the blecherous node_to_cpumask_ptr which\ncontained a declaration) are replaced now everyone implements\ncpumask_of_node.\n\nSigned-off-by: Rusty Russell \u003crusty@rustcorp.com.au\u003e\n" }, { "commit": "fbd59a8d1f7cf325fdb6828659f1fb76631e87b3", "tree": "ccb7bbf70ef6e2e08e8ae83964b09fbdcf74808a", "parents": [ "7f7ace0cda64c99599c23785f8979a072e118058" ], "author": { "name": "Rusty Russell", "email": "rusty@rustcorp.com.au", "time": "Sat Jan 10 21:58:08 2009 -0800" }, "committer": { "name": "Ingo Molnar", "email": "mingo@elte.hu", "time": "Sun Jan 11 19:12:49 2009 +0100" }, "message": "cpumask: Use topology_core_cpumask()/topology_thread_cpumask()\n\nImpact: reduce stack usage, use new cpumask API.\n\nThis actually uses topology_core_cpumask() and\ntopology_thread_cpumask(), removing the only users of\ntopology_core_siblings() and topology_thread_siblings()\n\nSigned-off-by: Rusty Russell \u003crusty@rustcorp.com.au\u003e\nSigned-off-by: Mike Travis \u003ctravis@sgi.com\u003e\nCc: linux-net-drivers@solarflare.com\n" }, { "commit": "100fdaee70ebf5f31b9451fbc01300c627091328", "tree": "81493242ff99c8ba52163da5b44ffc51929622cf", "parents": [ "ad273b32e482cdef306eac32b28d97f513a022f4" ], "author": { "name": "Vaidyanathan Srinivasan", "email": "svaidy@linux.vnet.ibm.com", "time": "Thu Dec 18 23:26:47 2008 +0530" }, "committer": { "name": "Ingo Molnar", "email": "mingo@elte.hu", "time": "Fri Dec 19 09:21:55 2008 +0100" }, "message": "sched: add SD_BALANCE_NEWIDLE at MC and CPU level for sched_mc\u003e0\n\nImpact: change task balancing to save power more agressively\n\nAdd SD_BALANCE_NEWIDLE flag at MC level and CPU level\nif sched_mc is set. This helps power savings and\nwill not affect performance when sched_mc\u003d0\n\nIngo and Mike Galbraith have optimised the SD flags by\nremoving SD_BALANCE_NEWIDLE at MC and CPU level. This\nhelps performance but hurts power savings since this\nslows down task consolidation by reducing the number\nof times load_balance is run.\n\n sched: fine-tune SD_MC_INIT\n commit 14800984706bf6936bbec5187f736e928be5c218\n Author: Mike Galbraith \u003cefault@gmx.de\u003e\n Date: Fri Nov 7 15:26:50 2008 +0100\n\n sched: re-tune balancing -- revert\n commit 9fcd18c9e63e325dbd2b4c726623f760788d5aa8\n Author: Ingo Molnar \u003cmingo@elte.hu\u003e\n Date: Wed Nov 5 16:52:08 2008 +0100\n\nThis patch selectively enables SD_BALANCE_NEWIDLE flag\nonly when sched_mc is set to 1 or 2. This helps power savings\nby task consolidation and also does not hurt performance at\nsched_mc\u003d0 where all power saving optimisations are turned off.\n\nSigned-off-by: Vaidyanathan Srinivasan \u003csvaidy@linux.vnet.ibm.com\u003e\nAcked-by: Balbir Singh \u003cbalbir@linux.vnet.ibm.com\u003e\nAcked-by: Peter Zijlstra \u003ca.p.zijlstra@chello.nl\u003e\nSigned-off-by: Ingo Molnar \u003cmingo@elte.hu\u003e\n" }, { "commit": "716707b29906e1d8d190defe3d646610b097a861", "tree": "02ddc0300b74d19bc93cc00d033b72e3e279dbf7", "parents": [ "3b11ce7f542e415c90267b4482d4611410b468e6" ], "author": { "name": "Vaidyanathan Srinivasan", "email": "svaidy@linux.vnet.ibm.com", "time": "Thu Dec 18 23:26:02 2008 +0530" }, "committer": { "name": "Ingo Molnar", "email": "mingo@elte.hu", "time": "Fri Dec 19 09:21:45 2008 +0100" }, "message": "sched: convert BALANCE_FOR_xx_POWER to inline functions\n\nImpact: cleanup\n\nBALANCE_FOR_MC_POWER and similar macros defined in sched.h are\nnot constants and have various condition checks and significant\namount of code that is not suitable to be contain in a macro.\nAlso there could be side effects on the expressions passed to\nsome of them like test_sd_parent().\n\nThis patch converts all complex macros related to power savings\nbalance to inline functions.\n\nSigned-off-by: Vaidyanathan Srinivasan \u003csvaidy@linux.vnet.ibm.com\u003e\nAcked-by: Balbir Singh \u003cbalbir@linux.vnet.ibm.com\u003e\nAcked-by: Peter Zijlstra \u003ca.p.zijlstra@chello.nl\u003e\nSigned-off-by: Ingo Molnar \u003cmingo@elte.hu\u003e\n" }, { "commit": "ee79d1bdb6a10499e53f80b1e8d14110215178ba", "tree": "e9273d48cadb0b2ba77865e3405ad5e3e9664ff9", "parents": [ "c1dfdc7597d051b09555d4ae2acb90403e238746" ], "author": { "name": "Heiko Carstens", "email": "heiko.carstens@de.ibm.com", "time": "Tue Dec 09 18:49:50 2008 +0100" }, "committer": { "name": "Ingo Molnar", "email": "mingo@elte.hu", "time": "Fri Dec 12 13:47:21 2008 +0100" }, "message": "sched: let arch_update_cpu_topology indicate if topology changed\n\nChange arch_update_cpu_topology so it returns 1 if the cpu topology changed\nand 0 if it didn\u0027t change. This will be useful for the next patch which adds\na call to this function in partition_sched_domains.\n\nSigned-off-by: Heiko Carstens \u003cheiko.carstens@de.ibm.com\u003e\nSigned-off-by: Ingo Molnar \u003cmingo@elte.hu\u003e\n" }, { "commit": "52c642f33b14bfa1b00ef2b68296effb34a573f3", "tree": "508a49591138e91733ad228ae6a367d91b318b39", "parents": [ "14800984706bf6936bbec5187f736e928be5c218" ], "author": { "name": "Ingo Molnar", "email": "mingo@elte.hu", "time": "Fri Nov 07 16:09:23 2008 +0100" }, "committer": { "name": "Ingo Molnar", "email": "mingo@elte.hu", "time": "Fri Nov 07 16:09:23 2008 +0100" }, "message": "sched: fine-tune SD_SIBLING_INIT\n\nfine-tune the HT sched-domains parameters as well.\n\nOn a HT capable box, this increases lat_ctx performance from 23.87\nusecs to 1.49 usecs:\n\n # before\n\n $ ./lat_ctx -s 0 2\n\n \"size\u003d0k ovr\u003d1.89\n 2 23.87\n\n # after\n\n $ ./lat_ctx -s 0 2\n\n \"size\u003d0k ovr\u003d1.84\n 2 1.49\n\nSigned-off-by: Ingo Molnar \u003cmingo@elte.hu\u003e\n" }, { "commit": "14800984706bf6936bbec5187f736e928be5c218", "tree": "09b462191420a5652601a30711211693fb0e3239", "parents": [ "ca3273f9646694e0419cfb9d6c12deb1c9aff27c" ], "author": { "name": "Mike Galbraith", "email": "efault@gmx.de", "time": "Fri Nov 07 15:26:50 2008 +0100" }, "committer": { "name": "Ingo Molnar", "email": "mingo@elte.hu", "time": "Fri Nov 07 15:35:11 2008 +0100" }, "message": "sched: fine-tune SD_MC_INIT\n\nTune SD_MC_INIT the same way as SD_CPU_INIT:\nunset SD_BALANCE_NEWIDLE, and set SD_WAKE_BALANCE.\n\nThis improves vmark by 5%:\n\nvmark 132102 125968 125497 messages/sec avg 127855.66 .984\nvmark 139404 131719 131272 messages/sec avg 134131.66 1.033\n\nSigned-off-by: Mike Galbraith \u003cefault@gmx.de\u003e\nAcked-by: Peter Zijlstra \u003ca.p.zijlstra@chello.nl\u003e\nSigned-off-by: Ingo Molnar \u003cmingo@elte.hu\u003e\n\n # *DOCUMENTATION*\n" }, { "commit": "9fcd18c9e63e325dbd2b4c726623f760788d5aa8", "tree": "fcc7f3c4b17ff66b62feb72194bc14560d948980", "parents": [ "02479099c286894644f8e96c6bbb535ab64662fd" ], "author": { "name": "Ingo Molnar", "email": "mingo@elte.hu", "time": "Wed Nov 05 16:52:08 2008 +0100" }, "committer": { "name": "Ingo Molnar", "email": "mingo@elte.hu", "time": "Wed Nov 05 18:04:38 2008 +0100" }, "message": "sched: re-tune balancing\n\nImpact: improve wakeup affinity on NUMA systems, tweak SMP systems\n\nGiven the fixes+tweaks to the wakeup-buddy code, re-tweak the domain\nbalancing defaults on NUMA and SMP systems.\n\nTurn on SD_WAKE_AFFINE which was off on x86 NUMA - there\u0027s no reason\nwhy we would not want to have wakeup affinity across nodes as well.\n(we already do this in the standard NUMA template.)\n\nlat_ctx on a NUMA box is particularly happy about this change:\n\nbefore:\n\n | phoenix:~/l\u003e ./lat_ctx -s 0 2\n | \"size\u003d0k ovr\u003d2.60\n | 2 5.70\n\nafter:\n\n | phoenix:~/l\u003e ./lat_ctx -s 0 2\n | \"size\u003d0k ovr\u003d2.65\n | 2 2.07\n\na 2.75x speedup.\n\npipe-test is similarly happy about it too:\n\n | phoenix:~/sched-tests\u003e ./pipe-test\n | 18.26 usecs/loop.\n | 14.70 usecs/loop.\n | 14.38 usecs/loop.\n | 10.55 usecs/loop. # +WAKE_AFFINE on domain0+domain1\n | 8.63 usecs/loop.\n | 8.59 usecs/loop.\n | 9.03 usecs/loop.\n | 8.94 usecs/loop.\n | 8.96 usecs/loop.\n | 8.63 usecs/loop.\n\nAlso:\n\n - disable SD_BALANCE_NEWIDLE on NUMA and SMP domains (keep it for siblings)\n - enable SD_WAKE_BALANCE on SMP domains\n\nSysbench+postgresql improves all around the board, quite significantly:\n\n .28-rc3-11474e2c .28-rc3-11474e2c-tune\n-------------------------------------------------\n 1: 571 688 +17.08%\n 2: 1236 1206 -2.55%\n 4: 2381 2642 +9.89%\n 8: 4958 5164 +3.99%\n 16: 9580 9574 -0.07%\n 32: 7128 8118 +12.20%\n 64: 7342 8266 +11.18%\n 128: 7342 8064 +8.95%\n 256: 7519 7884 +4.62%\n 512: 7350 7731 +4.93%\n-------------------------------------------------\n SUM: 55412 59341 +6.62%\n\nSo it\u0027s a win both for the runup portion, the peak area and the tail.\n\nSigned-off-by: Ingo Molnar \u003cmingo@elte.hu\u003e\n" }, { "commit": "c50cbb05a05cf1f9ca3592272eff053c847727d8", "tree": "e64a380282a16a9a593c07df8d50b729c825ccc0", "parents": [ "aab2545fdd6641b76af0ae96456c4ca9d1e50dad" ], "author": { "name": "Ben Hutchings", "email": "bhutchings@solarflare.com", "time": "Wed Jun 04 21:47:29 2008 -0700" }, "committer": { "name": "Ingo Molnar", "email": "mingo@elte.hu", "time": "Fri Jun 13 10:09:46 2008 +0200" }, "message": "cpu topology: always define CPU topology information\n\nThis can result in an empty topology directory in sysfs, and requires\nin-kernel users to protect all uses with #ifdef - see\n\u003chttp://marc.info/?l\u003dlinux-netdev\u0026m\u003d120639033904472\u0026w\u003d2\u003e.\n\nThe documentation of CPU topology specifies what the defaults should be if\nonly partial information is available from the hardware. So we can\nprovide these defaults as a fallback.\n\nThis patch:\n\n- Adds default definitions of the 4 topology macros to \u003clinux/topology.h\u003e\n- Changes drivers/base/topology.c to use the topology macros unconditionally\n and to cope with definitions that aren\u0027t lvalues\n- Updates documentation accordingly\n\n[ From: Andrew Morton \u003cakpm@linux-foundation.org\u003e\n - fold now-duplicated code\n - fix layout\n]\n\nSigned-off-by: Ben Hutchings \u003cbhutchings@solarflare.com\u003e\nCc: Vegard Nossum \u003cvegard.nossum@gmail.com\u003e\nCc: Nick Piggin \u003cnickpiggin@yahoo.com.au\u003e\nCc: Chandra Seetharaman \u003csekharan@us.ibm.com\u003e\nCc: Suresh Siddha \u003csuresh.b.siddha@intel.com\u003e\nCc: Mike Travis \u003ctravis@sgi.com\u003e\nCc: Christoph Lameter \u003cclameter@sgi.com\u003e\nCc: John Hawkes \u003chawkes@sgi.com\u003e\nCc: Zhang, Yanmin \u003cyanmin.zhang@intel.com\u003e\nSigned-off-by: Andrew Morton \u003cakpm@linux-foundation.org\u003e\nSigned-off-by: Ingo Molnar \u003cmingo@elte.hu\u003e\n" }, { "commit": "ea3f01f8afd3bc5daff915cc4ea5cc5ea9e7d427", "tree": "e2330451fc9d6cc02840c31be3e4956db7974a76", "parents": [ "b3137bc8e77962a8e3b4dfdc1bcfd38e437bd278" ], "author": { "name": "Ingo Molnar", "email": "mingo@elte.hu", "time": "Thu May 29 14:32:23 2008 +0200" }, "committer": { "name": "Ingo Molnar", "email": "mingo@elte.hu", "time": "Thu May 29 14:46:30 2008 +0200" }, "message": "sched: re-tune NUMA topologies\n\nimprove the sysbench ramp-up phase and its peak throughput on\na 16way NUMA box, by turning on WAKE_AFFINE:\n\n tip/sched tip/sched+wake-affine\n-------------------------------------------------\n 1: 700 830 +15.65%\n 2: 1465 1391 -5.28%\n 4: 3017 3105 +2.81%\n 8: 5100 6021 +15.30%\n 16: 10725 10745 +0.19%\n 32: 10135 10150 +0.16%\n 64: 9338 9240 -1.06%\n 128: 8599 8252 -4.21%\n 256: 8475 8144 -4.07%\n-------------------------------------------------\n SUM: 57558 57882 +0.56%\n\nthis change also improves lat_ctx from 6.69 usecs to 1.11 usec:\n\n $ ./lat_ctx -s 0 2\n \"size\u003d0k ovr\u003d1.19\n 2 1.11\n\n $ ./lat_ctx -s 0 2\n \"size\u003d0k ovr\u003d1.22\n 2 6.69\n\nin sysbench it\u0027s an overall win with some weakness at the lots-of-clients\nside. That happens because we now under-balance this workload\na bit. To counter that effect, turn on NEWIDLE:\n\n wake-idle wake-idle+newidle\n -------------------------------------------------\n 1: 830 834 +0.43%\n 2: 1391 1401 +0.65%\n 4: 3105 3091 -0.43%\n 8: 6021 6046 +0.42%\n 16: 10745 10736 -0.08%\n 32: 10150 10206 +0.55%\n 64: 9240 9533 +3.08%\n 128: 8252 8355 +1.24%\n 256: 8144 8384 +2.87%\n -------------------------------------------------\n SUM: 57882 58591 +1.21%\n\nas a bonus this not only improves the many-clients case but\nalso improves the (more important) rampup phase.\n\nsysbench is a workload that quickly breaks down if the\nscheduler over-balances, so since it showed an improvement\nunder NEWIDLE this change is definitely good.\n" }, { "commit": "7c16ec585c558960a508ccf9a08fcb9ed49b3754", "tree": "cca2b12203a10944d7095a07df7292421f578dc9", "parents": [ "c5f59f0833df945eef7ff35f3dc6ba61c5f293dd" ], "author": { "name": "Mike Travis", "email": "travis@sgi.com", "time": "Fri Apr 04 18:11:11 2008 -0700" }, "committer": { "name": "Ingo Molnar", "email": "mingo@elte.hu", "time": "Sat Apr 19 19:44:59 2008 +0200" }, "message": "cpumask: reduce stack usage in SD_x_INIT initializers\n\n * Remove empty cpumask_t (and all non-zero/non-null) variables\n in SD_*_INIT macros. Use memset(0) to clear. Also, don\u0027t\n inline the initializer functions to save on stack space in\n build_sched_domains().\n\n * Merge change to include/linux/topology.h that uses the new\n node_to_cpumask_ptr function in the nr_cpus_node macro into\n this patch.\n\nDepends on:\n\t[mm-patch]: asm-generic-add-node_to_cpumask_ptr-macro.patch\n\t[sched-devel]: sched: add new set_cpus_allowed_ptr function\n\nCc: H. Peter Anvin \u003chpa@zytor.com\u003e\nSigned-off-by: Mike Travis \u003ctravis@sgi.com\u003e\nSigned-off-by: Ingo Molnar \u003cmingo@elte.hu\u003e\n" }, { "commit": "22e52b072dd87faa9b2559fe89d4e8f2370f81ca", "tree": "79e71d0aad3c9ea9f326977c3a3ff1dceaec81b5", "parents": [ "9aefd0abd8610e8f3bb097debf3afb73f8b7b210" ], "author": { "name": "Heiko Carstens", "email": "heiko.carstens@de.ibm.com", "time": "Wed Mar 12 18:31:59 2008 +0100" }, "committer": { "name": "Ingo Molnar", "email": "mingo@elte.hu", "time": "Fri Mar 21 16:43:48 2008 +0100" }, "message": "sched: add arch_update_cpu_topology hook.\n\nWill be called each time the scheduling domains are rebuild.\nNeeded for architectures that don\u0027t have a static cpu topology.\n\nSigned-off-by: Heiko Carstens \u003cheiko.carstens@de.ibm.com\u003e\nSigned-off-by: Martin Schwidefsky \u003cschwidefsky@de.ibm.com\u003e\nSigned-off-by: Ingo Molnar \u003cmingo@elte.hu\u003e\n" }, { "commit": "33b0c4217dcd67b788318c3192a2912b530e4eef", "tree": "bd428acf9edb2d37c28e22d0ae8c772d5f9a3e6e", "parents": [ "74e3cd7f480ae1888b7cd196bf8125a1d3bfee05" ], "author": { "name": "Ingo Molnar", "email": "mingo@elte.hu", "time": "Sun Mar 16 11:14:30 2008 +0100" }, "committer": { "name": "Ingo Molnar", "email": "mingo@elte.hu", "time": "Wed Mar 19 04:27:53 2008 +0100" }, "message": "sched: tune multi-core idle balancing\n\nWAKE_IDLE is too agressive on multi-core CPUs with the new\nwake-affine code, keep it on for SMT/HT balancing alone\n(where there\u0027s no cache affinity at all between logical CPUs).\n\nSigned-off-by: Ingo Molnar \u003cmingo@elte.hu\u003e\n" }, { "commit": "32525d022ad52a5c14e80e130260431e16e294b6", "tree": "1a4addc080a63c57d817320ae917a7425f10b7ad", "parents": [ "52d853431e8d9dc17ba94792123a3fe2bc039831" ], "author": { "name": "Ingo Molnar", "email": "mingo@elte.hu", "time": "Fri Jan 25 21:08:20 2008 +0100" }, "committer": { "name": "Ingo Molnar", "email": "mingo@elte.hu", "time": "Fri Jan 25 21:08:20 2008 +0100" }, "message": "sched: whitespace cleanups in topology.h\n\nwhitespace cleanups in topology.h.\n\nSigned-off-by: Ingo Molnar \u003cmingo@elte.hu\u003e\n" }, { "commit": "52d853431e8d9dc17ba94792123a3fe2bc039831", "tree": "996455f11e380eda8a37b91de0ca9245f610a35f", "parents": [ "b913176917399e92e6f741672038c73d7ce93be5" ], "author": { "name": "Ingo Molnar", "email": "mingo@elte.hu", "time": "Fri Jan 25 21:08:20 2008 +0100" }, "committer": { "name": "Ingo Molnar", "email": "mingo@elte.hu", "time": "Fri Jan 25 21:08:20 2008 +0100" }, "message": "sched: reactivate fork balancing\n\nreactivate fork balancing.\n\nSigned-off-by: Ingo Molnar \u003cmingo@elte.hu\u003e\n" }, { "commit": "7a6c6bcee029a978f866511d6e41dbc7301fde4c", "tree": "228cc1a31096d09fca917009833dfbf213ec251c", "parents": [ "95dbb421d12fdd9796ed153853daf3679809274f" ], "author": { "name": "Ingo Molnar", "email": "mingo@elte.hu", "time": "Mon Oct 15 17:00:19 2007 +0200" }, "committer": { "name": "Ingo Molnar", "email": "mingo@elte.hu", "time": "Mon Oct 15 17:00:19 2007 +0200" }, "message": "sched: enable wake-idle on CONFIG_SCHED_MC\u003dy\n\nmost multicore CPUs today have shared L2 caches, so tune things so\nthat the spreading amongst cores is more aggressive.\n\nSigned-off-by: Ingo Molnar \u003cmingo@elte.hu\u003e\n" }, { "commit": "95dbb421d12fdd9796ed153853daf3679809274f", "tree": "75ef5102937a206b687bfa3948850921033e700b", "parents": [ "6bc1665ba71de0f207391b01b187b21b2619c15c" ], "author": { "name": "Ingo Molnar", "email": "mingo@elte.hu", "time": "Mon Oct 15 17:00:19 2007 +0200" }, "committer": { "name": "Ingo Molnar", "email": "mingo@elte.hu", "time": "Mon Oct 15 17:00:19 2007 +0200" }, "message": "sched: reintroduce topology.h tunings\n\nreintroduce the 2.6.22 topology.h tunings again - they result in\nslightly better balancing.\n\nSigned-off-by: Ingo Molnar \u003cmingo@elte.hu\u003e\n" }, { "commit": "362a7016637648c6aefc98b706298baedfaa1543", "tree": "cb3f434fb71d29b247686b6dcfc76864cea4f57a", "parents": [ "1ed4395035a6791ebbbf618429a58ab9c207cc83" ], "author": { "name": "Ingo Molnar", "email": "mingo@elte.hu", "time": "Thu Aug 02 17:41:40 2007 +0200" }, "committer": { "name": "Ingo Molnar", "email": "mingo@elte.hu", "time": "Thu Aug 02 17:41:40 2007 +0200" }, "message": "[PATCH] sched: remove cache_hot_time\n\nremove the last unused remains of cache_hot_time.\n\nSigned-off-by: Ingo Molnar \u003cmingo@elte.hu\u003e\n" }, { "commit": "f787a50306680c187cf2896a8017937c1bf6dc7e", "tree": "1dbc2b37c53759bde6e227e6502ad5555388f305", "parents": [ "4bd77321a833077c5c9ac7b9d284e261e4a8906e" ], "author": { "name": "Ingo Molnar", "email": "mingo@elte.hu", "time": "Wed Jul 11 21:21:47 2007 +0200" }, "committer": { "name": "Linus Torvalds", "email": "torvalds@woody.linux-foundation.org", "time": "Fri Jul 13 10:11:52 2007 -0700" }, "message": "[PATCH] sched: small topology.h cleanup\n\ntrivial cleanup: LOCAL_DISTANCE and REMOTE_DISTANCE are only used in\ntopology.h and inside an #ifndef section - limit their existence to\nthat #ifndef.\n\nSigned-off-by: Ingo Molnar \u003cmingo@elte.hu\u003e\nSigned-off-by: Linus Torvalds \u003ctorvalds@linux-foundation.org\u003e\n" }, { "commit": "9c4801cebc2add1fe514bc8eb201b16372eee11a", "tree": "dc3987da0147c118ae1b62e9bbd6a1cfa20eca12", "parents": [ "0fec171cdbd7763ef86cbaccb91f3708de6a9003" ], "author": { "name": "Ingo Molnar", "email": "mingo@elte.hu", "time": "Mon Jul 09 18:52:01 2007 +0200" }, "committer": { "name": "Ingo Molnar", "email": "mingo@elte.hu", "time": "Mon Jul 09 18:52:01 2007 +0200" }, "message": "sched: more agressive idle balancing\n\nthe Linux scheduler is starving a number of workloads. So default\nto more agressive idle-balancing. This hurts lmbench context-switching\nnumbers (which was the main reason we sucked at idle-balancing for\nsuch a long time) but the lmbench numbers are fine once the system is\nminimally utilized.\n\nSigned-off-by: Ingo Molnar \u003cmingo@elte.hu\u003e\n" }, { "commit": "69f7c0a1be84b10a81b6edcce2dbee0cdec26eba", "tree": "a6d4988fda72595ea71ba7e2b4ac11f91fde0159", "parents": [ "759b9775c25f5e69aaea8a75c3914019e2dc5539" ], "author": { "name": "Con Kolivas", "email": "kernel@kolivas.org", "time": "Mon Mar 05 00:30:29 2007 -0800" }, "committer": { "name": "Linus Torvalds", "email": "torvalds@woody.linux-foundation.org", "time": "Mon Mar 05 07:57:51 2007 -0800" }, "message": "[PATCH] sched: remove SMT nice\n\nRemove the SMT-nice feature which idles sibling cpus on SMT cpus to\nfacilitiate nice working properly where cpu power is shared. The idling of\ncpus in the presence of runnable tasks is considered too fragile, easy to\nbreak with outside code, and the complexity of managing this system if an\narchitecture comes along with many logical cores sharing cpu power will be\nunworkable.\n\nRemove the associated per_cpu_gain variable in sched_domains used only by\nthis code.\n\nAlso:\n\n The reason is that with dynticks enabled, this code breaks without yet\n further tweaks so dynticks brought on the rapid demise of this code. So\n either we tweak this code or kill it off entirely. It was Ingo\u0027s preference\n to kill it off. Either way this needs to happen for 2.6.21 since dynticks\n has gone in.\n\nSigned-off-by: Con Kolivas \u003ckernel@kolivas.org\u003e\nAcked-by: Ingo Molnar \u003cmingo@elte.hu\u003e\nCc: Nick Piggin \u003cnickpiggin@yahoo.com.au\u003e\nSigned-off-by: Andrew Morton \u003cakpm@linux-foundation.org\u003e\nSigned-off-by: Linus Torvalds \u003ctorvalds@linux-foundation.org\u003e\n" }, { "commit": "08c183f31bdbb709f177f6d3110d5f288ea33933", "tree": "be7b84c07f3b0bf29473bad2b7b788fa189f948e", "parents": [ "1bd77f2da58e9cdd1f159217887343dadd9af417" ], "author": { "name": "Christoph Lameter", "email": "clameter@sgi.com", "time": "Sun Dec 10 02:20:29 2006 -0800" }, "committer": { "name": "Linus Torvalds", "email": "torvalds@woody.osdl.org", "time": "Sun Dec 10 09:55:43 2006 -0800" }, "message": "[PATCH] sched: add option to serialize load balancing\n\nLarge sched domains can be very expensive to scan. Add an option SD_SERIALIZE\nto the sched domain flags. If that flag is set then we make sure that no\nother such domain is being balanced.\n\n[akpm@osdl.org: build fix]\nSigned-off-by: Christoph Lameter \u003cclameter@sgi.com\u003e\nCc: Peter Williams \u003cpwil3058@bigpond.net.au\u003e\nCc: Nick Piggin \u003cnickpiggin@yahoo.com.au\u003e\nCc: Christoph Lameter \u003cclameter@sgi.com\u003e\nCc: \"Siddha, Suresh B\" \u003csuresh.b.siddha@intel.com\u003e\nCc: \"Chen, Kenneth W\" \u003ckenneth.w.chen@intel.com\u003e\nAcked-by: Ingo Molnar \u003cmingo@elte.hu\u003e\nCc: KAMEZAWA Hiroyuki \u003ckamezawa.hiroyu@jp.fujitsu.com\u003e\nSigned-off-by: Andrew Morton \u003cakpm@osdl.org\u003e\nSigned-off-by: Linus Torvalds \u003ctorvalds@osdl.org\u003e\n" }, { "commit": "ac7d550499e225efb51a53d0b00667f26b93bdff", "tree": "d1d210cefa91e732ab4ecccf19451bb5d1fbf05c", "parents": [ "054b9108e01ef27e2e6b32b4226abb6024626f06" ], "author": { "name": "Siddha, Suresh B", "email": "suresh.b.siddha@intel.com", "time": "Sun Dec 10 02:20:12 2006 -0800" }, "committer": { "name": "Linus Torvalds", "email": "torvalds@woody.osdl.org", "time": "Sun Dec 10 09:55:42 2006 -0800" }, "message": "[PATCH] sched domain: increase the SMT busy rebalance interval\n\nWith SMT, if the logical processor is busy, load balance happens for every\n8msec(min)-16msec(max). There is no need to do this often, as this is just\nfor fairness(to maintain uniform runqueue lengths) and default time slice\nanyhow is 100msec.\n\nAppended patch increases this interval to 64msec(min)-128msec(max) when the\nlogical processor is busy.\n\nSigned-off-by: Suresh Siddha \u003csuresh.b.siddha@intel.com\u003e\nCc: Nick Piggin \u003cnickpiggin@yahoo.com.au\u003e\nAcked-by: Ingo Molnar \u003cmingo@elte.hu\u003e\nSigned-off-by: Andrew Morton \u003cakpm@osdl.org\u003e\nSigned-off-by: Linus Torvalds \u003ctorvalds@osdl.org\u003e\n" }, { "commit": "89c4710ee9bbbefe6a4d469d9f36266a92c275c5", "tree": "f84fe28e48bbda210f01f22ae0065f7ed1fcc5e1", "parents": [ "1a84887080dc15f048db7c3a643e98f1435790d6" ], "author": { "name": "Siddha, Suresh B", "email": "suresh.b.siddha@intel.com", "time": "Tue Oct 03 01:14:09 2006 -0700" }, "committer": { "name": "Linus Torvalds", "email": "torvalds@g5.osdl.org", "time": "Tue Oct 03 08:04:06 2006 -0700" }, "message": "[PATCH] sched: cleanup sched_group cpu_power setup\n\nUp to now sched group\u0027s cpu_power for each sched domain is initialized\nindependently. This made the setup code ugly as the new sched domains are\ngetting added.\n\nMake the sched group cpu_power setup code generic, by using domain child\nfield and new domain flag in sched_domain. For most of the sched\ndomains(except NUMA), sched group\u0027s cpu_power is now computed generically\nusing the domain properties of itself and of the child domain.\n\nsched groups in NUMA domains are setup little differently and hence they\ndon\u0027t use this generic mechanism.\n\nSigned-off-by: Suresh Siddha \u003csuresh.b.siddha@intel.com\u003e\nAcked-by: Ingo Molnar \u003cmingo@elte.hu\u003e\nAcked-by: Nick Piggin \u003cnickpiggin@yahoo.com.au\u003e\nCc: Paul Jackson \u003cpj@sgi.com\u003e\nSigned-off-by: Andrew Morton \u003cakpm@osdl.org\u003e\nSigned-off-by: Linus Torvalds \u003ctorvalds@osdl.org\u003e\n" }, { "commit": "1a84887080dc15f048db7c3a643e98f1435790d6", "tree": "7cd335fee247c0b60f8562c82806b49435b5fb9d", "parents": [ "74732646431a1bb7e23e6b564127a8881cfef900" ], "author": { "name": "Siddha, Suresh B", "email": "suresh.b.siddha@intel.com", "time": "Tue Oct 03 01:14:08 2006 -0700" }, "committer": { "name": "Linus Torvalds", "email": "torvalds@g5.osdl.org", "time": "Tue Oct 03 08:04:06 2006 -0700" }, "message": "[PATCH] sched: introduce child field in sched_domain\n\nIntroduce the child field in sched_domain struct and use it in\nsched_balance_self().\n\nWe will also use this field in cleaning up the sched group cpu_power\nsetup(done in a different patch) code.\n\nSigned-off-by: Suresh Siddha \u003csuresh.b.siddha@intel.com\u003e\nAcked-by: Ingo Molnar \u003cmingo@elte.hu\u003e\nAcked-by: Nick Piggin \u003cnickpiggin@yahoo.com.au\u003e\nCc: Paul Jackson \u003cpj@sgi.com\u003e\nSigned-off-by: Andrew Morton \u003cakpm@osdl.org\u003e\nSigned-off-by: Linus Torvalds \u003ctorvalds@osdl.org\u003e\n" }, { "commit": "5c45bf279d378d436ce45825c0f136696c7b6109", "tree": "80e2fcf4866b84fccb787562e1a83b16f4bc8850", "parents": [ "369381694ddcf03f1de403501c8b97099b5109ec" ], "author": { "name": "Siddha, Suresh B", "email": "suresh.b.siddha@intel.com", "time": "Tue Jun 27 02:54:42 2006 -0700" }, "committer": { "name": "Linus Torvalds", "email": "torvalds@g5.osdl.org", "time": "Tue Jun 27 17:32:45 2006 -0700" }, "message": "[PATCH] sched: mc/smt power savings sched policy\n\nsysfs entries \u0027sched_mc_power_savings\u0027 and \u0027sched_smt_power_savings\u0027 in\n/sys/devices/system/cpu/ control the MC/SMT power savings policy for the\nscheduler.\n\nBased on the values (1-enable, 0-disable) for these controls, sched groups\ncpu power will be determined for different domains. When power savings\npolicy is enabled and under light load conditions, scheduler will minimize\nthe physical packages/cpu cores carrying the load and thus conserving\npower(with a perf impact based on the workload characteristics... see OLS\n2005 CMP kernel scheduler paper for more details..)\n\nSigned-off-by: Suresh Siddha \u003csuresh.b.siddha@intel.com\u003e\nCc: Ingo Molnar \u003cmingo@elte.hu\u003e\nCc: Nick Piggin \u003cnickpiggin@yahoo.com.au\u003e\nCc: Con Kolivas \u003ckernel@kolivas.org\u003e\nCc: \"Chen, Kenneth W\" \u003ckenneth.w.chen@intel.com\u003e\nCc: \"David S. Miller\" \u003cdavem@davemloft.net\u003e\nSigned-off-by: Andrew Morton \u003cakpm@osdl.org\u003e\nSigned-off-by: Linus Torvalds \u003ctorvalds@osdl.org\u003e\n" }, { "commit": "1e9f28fa1eb9773bf65bae08288c6a0a38eef4a7", "tree": "ccfa4927ebc7a8f663f9ac9e7789a713a33253ff", "parents": [ "77e4bfbcf071f795b54862455dce8902b3fc29c2" ], "author": { "name": "Siddha, Suresh B", "email": "suresh.b.siddha@intel.com", "time": "Mon Mar 27 01:15:22 2006 -0800" }, "committer": { "name": "Linus Torvalds", "email": "torvalds@g5.osdl.org", "time": "Mon Mar 27 08:44:43 2006 -0800" }, "message": "[PATCH] sched: new sched domain for representing multi-core\n\nAdd a new sched domain for representing multi-core with shared caches\nbetween cores. Consider a dual package system, each package containing two\ncores and with last level cache shared between cores with in a package. If\nthere are two runnable processes, with this appended patch those two\nprocesses will be scheduled on different packages.\n\nOn such systems, with this patch we have observed 8% perf improvement with\nspecJBB(2 warehouse) benchmark and 35% improvement with CFP2000 rate(with 2\nusers).\n\nThis new domain will come into play only on multi-core systems with shared\ncaches. On other systems, this sched domain will be removed by domain\ndegeneration code. This new domain can be also used for implementing power\nsavings policy (see OLS 2005 CMP kernel scheduler paper for more details..\nI will post another patch for power savings policy soon)\n\nMost of the arch/* file changes are for cpu_coregroup_map() implementation.\n\nSigned-off-by: Suresh Siddha \u003csuresh.b.siddha@intel.com\u003e\nCc: Ingo Molnar \u003cmingo@elte.hu\u003e\nSigned-off-by: Andrew Morton \u003cakpm@osdl.org\u003e\nSigned-off-by: Linus Torvalds \u003ctorvalds@osdl.org\u003e\n" }, { "commit": "9eeff2395e3cfd05c9b2e6074ff943a34b0c5c21", "tree": "20160098ec6ed8738cfecfc5f81181ad22b44e60", "parents": [ "f1fd1067ece574ab56e4a70878b9a5a1ed4c3c42" ], "author": { "name": "Christoph Lameter", "email": "clameter@sgi.com", "time": "Wed Jan 18 17:42:31 2006 -0800" }, "committer": { "name": "Linus Torvalds", "email": "torvalds@g5.osdl.org", "time": "Wed Jan 18 19:20:17 2006 -0800" }, "message": "[PATCH] Zone reclaim: Reclaim logic\n\nSome bits for zone reclaim exists in 2.6.15 but they are not usable. This\npatch fixes them up, removes unused code and makes zone reclaim usable.\n\nZone reclaim allows the reclaiming of pages from a zone if the number of\nfree pages falls below the watermarks even if other zones still have enough\npages available. Zone reclaim is of particular importance for NUMA\nmachines. It can be more beneficial to reclaim a page than taking the\nperformance penalties that come with allocating a page on a remote zone.\n\nZone reclaim is enabled if the maximum distance to another node is higher\nthan RECLAIM_DISTANCE, which may be defined by an arch. By default\nRECLAIM_DISTANCE is 20. 20 is the distance to another node in the same\ncomponent (enclosure or motherboard) on IA64. The meaning of the NUMA\ndistance information seems to vary by arch.\n\nIf zone reclaim is not successful then no further reclaim attempts will\noccur for a certain time period (ZONE_RECLAIM_INTERVAL).\n\nThis patch was discussed before. See\n\nhttp://marc.theaimsgroup.com/?l\u003dlinux-kernel\u0026m\u003d113519961504207\u0026w\u003d2\nhttp://marc.theaimsgroup.com/?l\u003dlinux-kernel\u0026m\u003d113408418232531\u0026w\u003d2\nhttp://marc.theaimsgroup.com/?l\u003dlinux-kernel\u0026m\u003d113389027420032\u0026w\u003d2\nhttp://marc.theaimsgroup.com/?l\u003dlinux-kernel\u0026m\u003d113380938612205\u0026w\u003d2\n\nSigned-off-by: Christoph Lameter \u003cclameter@sgi.com\u003e\nSigned-off-by: Andrew Morton \u003cakpm@osdl.org\u003e\nSigned-off-by: Linus Torvalds \u003ctorvalds@osdl.org\u003e\n" }, { "commit": "198e2f181163233b379dc7ce8a6d7516b84042e7", "tree": "cc4067ca1c81034ba8d214b7ff4c39f2f5be66ee", "parents": [ "4dc7a0bbeb6882ad665e588e82fabe5bb4645f2f" ], "author": { "name": "akpm@osdl.org", "email": "akpm@osdl.org", "time": "Thu Jan 12 01:05:30 2006 -0800" }, "committer": { "name": "Linus Torvalds", "email": "torvalds@g5.osdl.org", "time": "Thu Jan 12 09:08:50 2006 -0800" }, "message": "[PATCH] scheduler cache-hot-autodetect\n\n\r)\n\nFrom: Ingo Molnar \u003cmingo@elte.hu\u003e\n\nThis is the latest version of the scheduler cache-hot-auto-tune patch.\n\nThe first problem was that detection time scaled with O(N^2), which is\nunacceptable on larger SMP and NUMA systems. To solve this:\n\n- I\u0027ve added a \u0027domain distance\u0027 function, which is used to cache\n measurement results. Each distance is only measured once. This means\n that e.g. on NUMA distances of 0, 1 and 2 might be measured, on HT\n distances 0 and 1, and on SMP distance 0 is measured. The code walks\n the domain tree to determine the distance, so it automatically follows\n whatever hierarchy an architecture sets up. This cuts down on the boot\n time significantly and removes the O(N^2) limit. The only assumption\n is that migration costs can be expressed as a function of domain\n distance - this covers the overwhelming majority of existing systems,\n and is a good guess even for more assymetric systems.\n\n [ People hacking systems that have assymetries that break this\n assumption (e.g. different CPU speeds) should experiment a bit with\n the cpu_distance() function. Adding a -\u003emigration_distance factor to\n the domain structure would be one possible solution - but lets first\n see the problem systems, if they exist at all. Lets not overdesign. ]\n\nAnother problem was that only a single cache-size was used for measuring\nthe cost of migration, and most architectures didnt set that variable\nup. Furthermore, a single cache-size does not fit NUMA hierarchies with\nL3 caches and does not fit HT setups, where different CPUs will often\nhave different \u0027effective cache sizes\u0027. To solve this problem:\n\n- Instead of relying on a single cache-size provided by the platform and\n sticking to it, the code now auto-detects the \u0027effective migration\n cost\u0027 between two measured CPUs, via iterating through a wide range of\n cachesizes. The code searches for the maximum migration cost, which\n occurs when the working set of the test-workload falls just below the\n \u0027effective cache size\u0027. I.e. real-life optimized search is done for\n the maximum migration cost, between two real CPUs.\n\n This, amongst other things, has the positive effect hat if e.g. two\n CPUs share a L2/L3 cache, a different (and accurate) migration cost\n will be found than between two CPUs on the same system that dont share\n any caches.\n\n(The reliable measurement of migration costs is tricky - see the source\nfor details.)\n\nFurthermore i\u0027ve added various boot-time options to override/tune\nmigration behavior.\n\nFirstly, there\u0027s a blanket override for autodetection:\n\n\tmigration_cost\u003d1000,2000,3000\n\nwill override the depth 0/1/2 values with 1msec/2msec/3msec values.\n\nSecondly, there\u0027s a global factor that can be used to increase (or\ndecrease) the autodetected values:\n\n\tmigration_factor\u003d120\n\nwill increase the autodetected values by 20%. This option is useful to\ntune things in a workload-dependent way - e.g. if a workload is\ncache-insensitive then CPU utilization can be maximized by specifying\nmigration_factor\u003d0.\n\nI\u0027ve tested the autodetection code quite extensively on x86, on 3\nP3/Xeon/2MB, and the autodetected values look pretty good:\n\nDual Celeron (128K L2 cache):\n\n ---------------------\n migration cost matrix (max_cache_size: 131072, cpu: 467 MHz):\n ---------------------\n [00] [01]\n [00]: - 1.7(1)\n [01]: 1.7(1) -\n ---------------------\n cacheflush times [2]: 0.0 (0) 1.7 (1784008)\n ---------------------\n\nHere the slow memory subsystem dominates system performance, and even\nthough caches are small, the migration cost is 1.7 msecs.\n\nDual HT P4 (512K L2 cache):\n\n ---------------------\n migration cost matrix (max_cache_size: 524288, cpu: 2379 MHz):\n ---------------------\n [00] [01] [02] [03]\n [00]: - 0.4(1) 0.0(0) 0.4(1)\n [01]: 0.4(1) - 0.4(1) 0.0(0)\n [02]: 0.0(0) 0.4(1) - 0.4(1)\n [03]: 0.4(1) 0.0(0) 0.4(1) -\n ---------------------\n cacheflush times [2]: 0.0 (33900) 0.4 (448514)\n ---------------------\n\nHere it can be seen that there is no migration cost between two HT\nsiblings (CPU#0/2 and CPU#1/3 are separate physical CPUs). A fast memory\nsystem makes inter-physical-CPU migration pretty cheap: 0.4 msecs.\n\n8-way P3/Xeon [2MB L2 cache]:\n\n ---------------------\n migration cost matrix (max_cache_size: 2097152, cpu: 700 MHz):\n ---------------------\n [00] [01] [02] [03] [04] [05] [06] [07]\n [00]: - 19.2(1) 19.2(1) 19.2(1) 19.2(1) 19.2(1) 19.2(1) 19.2(1)\n [01]: 19.2(1) - 19.2(1) 19.2(1) 19.2(1) 19.2(1) 19.2(1) 19.2(1)\n [02]: 19.2(1) 19.2(1) - 19.2(1) 19.2(1) 19.2(1) 19.2(1) 19.2(1)\n [03]: 19.2(1) 19.2(1) 19.2(1) - 19.2(1) 19.2(1) 19.2(1) 19.2(1)\n [04]: 19.2(1) 19.2(1) 19.2(1) 19.2(1) - 19.2(1) 19.2(1) 19.2(1)\n [05]: 19.2(1) 19.2(1) 19.2(1) 19.2(1) 19.2(1) - 19.2(1) 19.2(1)\n [06]: 19.2(1) 19.2(1) 19.2(1) 19.2(1) 19.2(1) 19.2(1) - 19.2(1)\n [07]: 19.2(1) 19.2(1) 19.2(1) 19.2(1) 19.2(1) 19.2(1) 19.2(1) -\n ---------------------\n cacheflush times [2]: 0.0 (0) 19.2 (19281756)\n ---------------------\n\nThis one has huge caches and a relatively slow memory subsystem - so the\nmigration cost is 19 msecs.\n\nSigned-off-by: Ingo Molnar \u003cmingo@elte.hu\u003e\nSigned-off-by: Ashok Raj \u003cashok.raj@intel.com\u003e\nSigned-off-by: Ken Chen \u003ckenneth.w.chen@intel.com\u003e\nCc: \u003cwilder@us.ibm.com\u003e\nSigned-off-by: John Hawkes \u003chawkes@sgi.com\u003e\nSigned-off-by: Andrew Morton \u003cakpm@osdl.org\u003e\nSigned-off-by: Linus Torvalds \u003ctorvalds@osdl.org\u003e\n" }, { "commit": "9c1cfda20a508b181bdda8c0045f7c0c333880a5", "tree": "eaa5b7ef7407316c36def26169574d0e37b1e60a", "parents": [ "ef08e3b4981aebf2ba9bd7025ef7210e8eec07ce" ], "author": { "name": "John Hawkes", "email": "hawkes@sgi.com", "time": "Tue Sep 06 15:18:14 2005 -0700" }, "committer": { "name": "Linus Torvalds", "email": "torvalds@g5.osdl.org", "time": "Wed Sep 07 16:57:40 2005 -0700" }, "message": "[PATCH] cpusets: Move the ia64 domain setup code to the generic code\n\nSigned-off-by: John Hawkes \u003chawkes@sgi.com\u003e\nSigned-off-by: Ingo Molnar \u003cmingo@elte.hu\u003e\nSigned-off-by: Andrew Morton \u003cakpm@osdl.org\u003e\nSigned-off-by: Linus Torvalds \u003ctorvalds@osdl.org\u003e\n" }, { "commit": "687f1661d302bc70ce906594a6d3f615ef075a50", "tree": "2895a027851322c2badcd40acf6b871698f2962b", "parents": [ "68767a0ae428801649d510d9a65bb71feed44dd1" ], "author": { "name": "Nick Piggin", "email": "nickpiggin@yahoo.com.au", "time": "Sat Jun 25 14:57:21 2005 -0700" }, "committer": { "name": "Linus Torvalds", "email": "torvalds@ppc970.osdl.org", "time": "Sat Jun 25 16:24:42 2005 -0700" }, "message": "[PATCH] sched: sched tuning\n\nDo some basic initial tuning.\n\nSigned-off-by: Nick Piggin \u003cnickpiggin@yahoo.com.au\u003e\nSigned-off-by: Andrew Morton \u003cakpm@osdl.org\u003e\nSigned-off-by: Linus Torvalds \u003ctorvalds@osdl.org\u003e\n" }, { "commit": "147cbb4bbe991452698f0772d8292f22825710ba", "tree": "cb86550d7e440e7dfbe22b0af6d2cfc991cb76cf", "parents": [ "cafb20c1f9976a70d633bb1e1c8c24eab00e4e80" ], "author": { "name": "Nick Piggin", "email": "nickpiggin@yahoo.com.au", "time": "Sat Jun 25 14:57:19 2005 -0700" }, "committer": { "name": "Linus Torvalds", "email": "torvalds@ppc970.osdl.org", "time": "Sat Jun 25 16:24:42 2005 -0700" }, "message": "[PATCH] sched: balance on fork\n\nReimplement the balance on exec balancing to be sched-domains aware. Use this\nto also do balance on fork balancing. Make x86_64 do balance on fork over the\nNUMA domain.\n\nThe problem that the non sched domains aware blancing became apparent on dual\ncore, multi socket opterons. What we want is for the new tasks to be sent to\na different socket, but more often than not, we would first load up our\nsibling core, or fill two cores of a single remote socket before selecting a\nnew one.\n\nThis gives large improvements to STREAM on such systems.\n\nSigned-off-by: Nick Piggin \u003cnickpiggin@yahoo.com.au\u003e\nSigned-off-by: Andrew Morton \u003cakpm@osdl.org\u003e\nSigned-off-by: Linus Torvalds \u003ctorvalds@osdl.org\u003e\n" }, { "commit": "cafb20c1f9976a70d633bb1e1c8c24eab00e4e80", "tree": "7ff8e6060990889992d51d542abde31cf0d7a1ef", "parents": [ "a3f21bce1fefdf92a4d1705e888d390b10f3ac6f" ], "author": { "name": "Nick Piggin", "email": "nickpiggin@yahoo.com.au", "time": "Sat Jun 25 14:57:17 2005 -0700" }, "committer": { "name": "Linus Torvalds", "email": "torvalds@ppc970.osdl.org", "time": "Sat Jun 25 16:24:42 2005 -0700" }, "message": "[PATCH] sched: no aggressive idle balancing\n\nRemove the very aggressive idle stuff that has recently gone into 2.6 - it is\ngoing against the direction we are trying to go. Hopefully we can regain\nperformance through other methods.\n\nSigned-off-by: Nick Piggin \u003cnickpiggin@yahoo.com.au\u003e\nSigned-off-by: Andrew Morton \u003cakpm@osdl.org\u003e\nSigned-off-by: Linus Torvalds \u003ctorvalds@osdl.org\u003e\n" }, { "commit": "7897986bad8f6cd50d6149345aca7f6480f49464", "tree": "10a5e08e004ae685aaab6823a3774803455b7704", "parents": [ "99b61ccf0bf0e9a85823d39a5db6a1519caeb13d" ], "author": { "name": "Nick Piggin", "email": "nickpiggin@yahoo.com.au", "time": "Sat Jun 25 14:57:13 2005 -0700" }, "committer": { "name": "Linus Torvalds", "email": "torvalds@ppc970.osdl.org", "time": "Sat Jun 25 16:24:41 2005 -0700" }, "message": "[PATCH] sched: balance timers\n\nDo CPU load averaging over a number of different intervals. Allow each\ninterval to be chosen by sending a parameter to source_load and target_load.\n0 is instantaneous, idx \u003e 0 returns a decaying average with the most recent\nsample weighted at 2^(idx-1). To a maximum of 3 (could be easily increased).\n\nSo generally a higher number will result in more conservative balancing.\n\nSigned-off-by: Nick Piggin \u003cnickpiggin@yahoo.com.au\u003e\nSigned-off-by: Andrew Morton \u003cakpm@osdl.org\u003e\nSigned-off-by: Linus Torvalds \u003ctorvalds@osdl.org\u003e\n" }, { "commit": "1da177e4c3f41524e886b7f1b8a0c1fc7321cac2", "tree": "0bba044c4ce775e45a88a51686b5d9f90697ea9d", "parents": [], "author": { "name": "Linus Torvalds", "email": "torvalds@ppc970.osdl.org", "time": "Sat Apr 16 15:20:36 2005 -0700" }, "committer": { "name": "Linus Torvalds", "email": "torvalds@ppc970.osdl.org", "time": "Sat Apr 16 15:20:36 2005 -0700" }, "message": "Linux-2.6.12-rc2\n\nInitial git repository build. I\u0027m not bothering with the full history,\neven though we have it. We can create a separate \"historical\" git\narchive of that later if we want to, and in the meantime it\u0027s about\n3.2GB when imported into git - space that would just make the early\ngit days unnecessarily complicated, when we don\u0027t have a lot of good\ninfrastructure for it.\n\nLet it rip!\n" } ] }