)]}'
{
  "commit": "cd881a6b22902b356cacf8fd2e4e895871068eec",
  "tree": "e26ecdf5ef148b34b1b827ce4da083194c022984",
  "parents": [
    "74a0b5762713a26496db72eac34fbbed46f20fce"
  ],
  "author": {
    "name": "Andy Whitcroft",
    "email": "apw@shadowen.org",
    "time": "Tue Oct 16 01:24:10 2007 -0700"
  },
  "committer": {
    "name": "Linus Torvalds",
    "email": "torvalds@woody.linux-foundation.org",
    "time": "Tue Oct 16 09:42:51 2007 -0700"
  },
  "message": "sparsemem: clean up spelling error in comments\n\nSPARSEMEM is a pretty nice framework that unifies quite a bit of code over all\nthe arches.  It would be great if it could be the default so that we can get\nrid of various forms of DISCONTIG and other variations on memory maps.  So far\nwhat has hindered this are the additional lookups that SPARSEMEM introduces\nfor virt_to_page and page_address.  This goes so far that the code to do this\nhas to be kept in a separate function and cannot be used inline.\n\nThis patch introduces a virtual memmap mode for SPARSEMEM, in which the memmap\nis mapped into a virtually contigious area, only the active sections are\nphysically backed.  This allows virt_to_page page_address and cohorts become\nsimple shift/add operations.  No page flag fields, no table lookups, nothing\ninvolving memory is required.\n\nThe two key operations pfn_to_page and page_to_page become:\n\n   #define __pfn_to_page(pfn)      (vmemmap + (pfn))\n   #define __page_to_pfn(page)     ((page) - vmemmap)\n\nBy having a virtual mapping for the memmap we allow simple access without\nwasting physical memory.  As kernel memory is typically already mapped 1:1\nthis introduces no additional overhead.  The virtual mapping must be big\nenough to allow a struct page to be allocated and mapped for all valid\nphysical pages.  This vill make a virtual memmap difficult to use on 32 bit\nplatforms that support 36 address bits.\n\nHowever, if there is enough virtual space available and the arch already maps\nits 1-1 kernel space using TLBs (f.e.  true of IA64 and x86_64) then this\ntechnique makes SPARSEMEM lookups even more efficient than CONFIG_FLATMEM.\nFLATMEM needs to read the contents of the mem_map variable to get the start of\nthe memmap and then add the offset to the required entry.  vmemmap is a\nconstant to which we can simply add the offset.\n\nThis patch has the potential to allow us to make SPARSMEM the default (and\neven the only) option for most systems.  It should be optimal on UP, SMP and\nNUMA on most platforms.  Then we may even be able to remove the other memory\nmodels: FLATMEM, DISCONTIG etc.\n\nThe current aim is to bring a common virtually mapped mem_map to all\narchitectures.  This should facilitate the removal of the bespoke\nimplementations from the architectures.  This also brings performance\nimprovements for most architecture making sparsmem vmemmap the more desirable\nmemory model.  The ultimate aim of this work is to expand sparsemem support to\nencompass all the features of the other memory models.  This could allow us to\ndrop support for and remove the other models in the longer term.\n\nBelow are some comparitive kernbench numbers for various architectures,\ncomparing default memory model against SPARSEMEM VMEMMAP.  All but ia64 show\nmarginal improvement; we expect the ia64 figures to be sorted out when the\nlarger mapping support returns.\n\nx86-64 non-NUMA\n             Base    VMEMAP    % change (-ve good)\nUser        85.07     84.84    -0.26\nSystem      34.32     33.84    -1.39\nTotal      119.38    118.68    -0.59\n\nia64\n             Base    VMEMAP    % change (-ve good)\nUser      1016.41   1016.93    0.05\nSystem      50.83     51.02    0.36\nTotal     1067.25   1067.95    0.07\n\nx86-64 NUMA\n             Base   VMEMAP    % change (-ve good)\nUser        30.77   431.73     0.22\nSystem      45.39    43.98    -3.11\nTotal      476.17   475.71    -0.10\n\nppc64\n             Base   VMEMAP    % change (-ve good)\nUser       488.77   488.35    -0.09\nSystem      56.92    56.37    -0.97\nTotal      545.69   544.72    -0.18\n\nBelow are some AIM bencharks on IA64 and x86-64 (thank Bob).  The seems\npretty much flat as you would expect.\n\nia64 results 2 cpu non-numa 4Gb SCSI disk\n\nBenchmark\tVersion\tMachine\tRun Date\nAIM Multiuser Benchmark - Suite VII\t\"1.1\"\textreme\tJun  1 07:17:24 2007\n\nTasks\tJobs/Min\tJTI\tReal\tCPU\tJobs/sec/task\n1\t98.9\t\t100\t58.9\t1.3\t1.6482\n101\t5547.1\t\t95\t106.0\t79.4\t0.9154\n201\t6377.7\t\t95\t183.4\t158.3\t0.5288\n301\t6932.2\t\t95\t252.7\t237.3\t0.3838\n401\t7075.8\t\t93\t329.8\t316.7\t0.2941\n501\t7235.6\t\t94\t403.0\t396.2\t0.2407\n600\t7387.5\t\t94\t472.7\t475.0\t0.2052\n\nBenchmark\tVersion\tMachine\tRun Date\nAIM Multiuser Benchmark - Suite VII\t\"1.1\"\tvmemmap\tJun  1 09:59:04 2007\n\nTasks\tJobs/Min\tJTI\tReal\tCPU\tJobs/sec/task\n1\t99.1\t\t100\t58.8\t1.2\t1.6509\n101\t5480.9\t\t95\t107.2\t79.2\t0.9044\n201\t6490.3\t\t95\t180.2\t157.8\t0.5382\n301\t6886.6\t\t94\t254.4\t236.8\t0.3813\n401\t7078.2\t\t94\t329.7\t316.0\t0.2942\n501\t7250.3\t\t95\t402.2\t395.4\t0.2412\n600\t7399.1\t\t94\t471.9\t473.9\t0.2055\n\nopen power 710 2 cpu, 4 Gb, SCSI and configured physically\n\nBenchmark\tVersion\tMachine\tRun Date\nAIM Multiuser Benchmark - Suite VII\t\"1.1\"\textreme\tMay 29 15:42:53 2007\n\nTasks\tJobs/Min\tJTI\tReal\tCPU\tJobs/sec/task\n1\t25.7\t\t100\t226.3\t4.3\t0.4286\n101\t1096.0\t\t97\t536.4\t199.8\t0.1809\n201\t1236.4\t\t96\t946.1\t389.1\t0.1025\n301\t1280.5\t\t96\t1368.0\t582.3\t0.0709\n401\t1270.2\t\t95\t1837.4\t771.0\t0.0528\n501\t1251.4\t\t96\t2330.1\t955.9\t0.0416\n601\t1252.6\t\t96\t2792.4\t1139.2\t0.0347\n701\t1245.2\t\t96\t3276.5\t1334.6\t0.0296\n918\t1229.5\t\t96\t4345.4\t1728.7\t0.0223\n\nBenchmark\tVersion\tMachine\tRun Date\nAIM Multiuser Benchmark - Suite VII\t\"1.1\"\tvmemmap\tMay 30 07:28:26 2007\n\nTasks\tJobs/Min\tJTI\tReal\tCPU\tJobs/sec/task\n1\t25.6\t\t100\t226.9\t4.3\t0.4275\n101\t1049.3\t\t97\t560.2\t198.1\t0.1731\n201\t1199.1\t\t97\t975.6\t390.7\t0.0994\n301\t1261.7\t\t96\t1388.5\t591.5\t0.0699\n401\t1256.1\t\t96\t1858.1\t771.9\t0.0522\n501\t1220.1\t\t96\t2389.7\t955.3\t0.0406\n601\t1224.6\t\t96\t2856.3\t1133.4\t0.0340\n701\t1252.0\t\t96\t3258.7\t1314.1\t0.0298\n915\t1232.8\t\t96\t4319.7\t1704.0\t0.0225\n\namd64 2 2-core, 4Gb and SATA\n\nBenchmark\tVersion\tMachine\tRun Date\nAIM Multiuser Benchmark - Suite VII\t\"1.1\"\textreme\tJun  2 03:59:48 2007\n\nTasks\tJobs/Min\tJTI\tReal\tCPU\tJobs/sec/task\n1\t13.0\t\t100\t446.4\t2.1\t0.2173\n101\t533.4\t\t97\t1102.0\t110.2\t0.0880\n201\t578.3\t\t97\t2022.8\t220.8\t0.0480\n301\t583.8\t\t97\t3000.6\t332.3\t0.0323\n401\t580.5\t\t97\t4020.1\t442.2\t0.0241\n501\t574.8\t\t98\t5072.8\t558.8\t0.0191\n600\t566.5\t\t98\t6163.8\t671.0\t0.0157\n\nBenchmark\tVersion\tMachine\tRun Date\nAIM Multiuser Benchmark - Suite VII\t\"1.1\"\tvmemmap\tJun  3 04:19:31 2007\n\nTasks\tJobs/Min\tJTI\tReal\tCPU\tJobs/sec/task\n1\t13.0\t\t100\t447.8\t2.0\t0.2166\n101\t536.5\t\t97\t1095.6\t109.7\t0.0885\n201\t567.7\t\t97\t2060.5\t219.3\t0.0471\n301\t582.1\t\t96\t3009.4\t330.2\t0.0322\n401\t578.2\t\t96\t4036.4\t442.4\t0.0240\n501\t585.1\t\t98\t4983.2\t555.1\t0.0195\n600\t565.5\t\t98\t6175.2\t660.6\t0.0157\n\nThis patch:\n\nFix some spelling errors.\n\nSigned-off-by: Christoph Lameter \u003cclameter@sgi.com\u003e\nSigned-off-by: Andy Whitcroft \u003capw@shadowen.org\u003e\nAcked-by: Mel Gorman \u003cmel@csn.ul.ie\u003e\nCc: \"Luck, Tony\" \u003ctony.luck@intel.com\u003e\nCc: Andi Kleen \u003cak@suse.de\u003e\nCc: \"David S. Miller\" \u003cdavem@davemloft.net\u003e\nCc: Paul Mackerras \u003cpaulus@samba.org\u003e\nCc: Benjamin Herrenschmidt \u003cbenh@kernel.crashing.org\u003e\nCc: KAMEZAWA Hiroyuki \u003ckamezawa.hiroyu@jp.fujitsu.com\u003e\nSigned-off-by: Andrew Morton \u003cakpm@linux-foundation.org\u003e\nSigned-off-by: Linus Torvalds \u003ctorvalds@linux-foundation.org\u003e\n",
  "tree_diff": [
    {
      "type": "modify",
      "old_id": "239f5a720d38fbe781751d2e3cb422cf49509d7d",
      "old_mode": 33188,
      "old_path": "mm/sparse.c",
      "new_id": "e8f36e4796d0e51ce9bc7b62b78204eddd0c57c3",
      "new_mode": 33188,
      "new_path": "mm/sparse.c"
    }
  ]
}