)]}'
{
  "commit": "f1dd9c379cac7d5a76259e7dffcd5f8edc697d17",
  "tree": "ea1870f5720d842ae8968922226a424f3cb28726",
  "parents": [
    "22626216c46f2ec86287e75ea86dd9ac3df54265"
  ],
  "author": {
    "name": "Zhang Yanmin",
    "email": "yanmin.zhang@intel.com",
    "time": "Wed Mar 12 22:52:37 2008 -0700"
  },
  "committer": {
    "name": "David S. Miller",
    "email": "davem@davemloft.net",
    "time": "Wed Mar 12 22:52:37 2008 -0700"
  },
  "message": "[NET]: Fix tbench regression in 2.6.25-rc1\n\nComparing with kernel 2.6.24, tbench result has regression with\n2.6.25-rc1.\n\n1) On 2 quad-core processor stoakley: 4%.\n2) On 4 quad-core processor tigerton: more than 30%.\n\nbisect located below patch.\n\nb4ce92775c2e7ff9cf79cca4e0a19c8c5fd6287b is first bad commit\ncommit b4ce92775c2e7ff9cf79cca4e0a19c8c5fd6287b\nAuthor: Herbert Xu \u003cherbert@gondor.apana.org.au\u003e\nDate:   Tue Nov 13 21:33:32 2007 -0800\n\n    [IPV6]: Move nfheader_len into rt6_info\n\n    The dst member nfheader_len is only used by IPv6.  It\u0027s also currently\n    creating a rather ugly alignment hole in struct dst.  Therefore this patch\n    moves it from there into struct rt6_info.\n\nAbove patch changes the cache line alignment, especially member\n__refcnt. I did a testing by adding 2 unsigned long pading before\nlastuse, so the 3 members, lastuse/__refcnt/__use, are moved to next\ncache line. The performance is recovered.\n\nI created a patch to rearrange the members in struct dst_entry.\n\nWith Eric and Valdis Kletnieks\u0027s suggestion, I made finer arrangement.\n\n1) Move tclassid under ops in case CONFIG_NET_CLS_ROUTE\u003dy. So\n   sizeof(dst_entry)\u003d200 no matter if CONFIG_NET_CLS_ROUTE\u003dy/n. I\n   tested many patches on my 16-core tigerton by moving tclassid to\n   different place. It looks like tclassid could also have impact on\n   performance.  If moving tclassid before metrics, or just don\u0027t move\n   tclassid, the performance isn\u0027t good. So I move it behind metrics.\n\n2) Add comments before __refcnt.\n\nOn 16-core tigerton:\n\nIf CONFIG_NET_CLS_ROUTE\u003dy, the result with below patch is about 18%\nbetter than the one without the patch;\n\nIf CONFIG_NET_CLS_ROUTE\u003dn, the result with below patch is about 30%\nbetter than the one without the patch.\n\nWith 32bit 2.6.25-rc1 on 8-core stoakley, the new patch doesn\u0027t\nintroduce regression.\n\nThank Eric, Valdis, and David!\n\nSigned-off-by: Zhang Yanmin \u003cyanmin.zhang@intel.com\u003e\nAcked-by: Eric Dumazet \u003cdada1@cosmosbay.com\u003e\nSigned-off-by: David S. Miller \u003cdavem@davemloft.net\u003e\n",
  "tree_diff": [
    {
      "type": "modify",
      "old_id": "e3ac7d0fc4e1807e62d1246b74aefa5d9ae9f23a",
      "old_mode": 33188,
      "old_path": "include/net/dst.h",
      "new_id": "ae13370e8484528a8501b0dbf4a8091ff5dd7e95",
      "new_mode": 33188,
      "new_path": "include/net/dst.h"
    }
  ]
}