)]}'
{
  "commit": "1a076689cda8a1d623dcda170b2dc2b476cc6f1a",
  "tree": "4c71eb88be4803eb8fd3b60884e8e9b2bc496b5e",
  "parents": [
    "281922a1d4f59bdebbe78c1d9f4c50a967eb6cff"
  ],
  "author": {
    "name": "Kim Phillips",
    "email": "kim.phillips@freescale.com",
    "time": "Fri Jun 22 19:48:53 2012 -0500"
  },
  "committer": {
    "name": "Herbert Xu",
    "email": "herbert@gondor.apana.org.au",
    "time": "Wed Jun 27 14:42:06 2012 +0800"
  },
  "message": "crypto: caam - disable IRQ coalescing by default\n\nIt has been observed that in zero-loss benchmarks, when a\nslow traffic rate is being tested, the IRQ timer coalescing\nparameter was set too high, and the ethernet controller\nwould start dropping packets because the job ring back half\nwouldn\u0027t be executed in time before the ethernet controller\nwould fill its buffers, thereby significantly reducing the\nzero-loss performance figures.\n\nEmpirical testing has shown that the best zero-loss performance\nis achieved when IRQ coalescing is set to minimum values and/or\nturned off, since apparently the job ring driver already implements\nan adequately-performing general-purpose IRQ mitigation strategy\nin software.\n\nWhilst we could go with minimal count (2-8) and timing settings\n(192-256), we prefer just turning h/w coalescing altogether off\nto minimize setkey latency (due to split key generation), and\nfor consistent cross-SoC performance (the SEC vs. core clock\nratio changes).\n\nSigned-off-by: Kim Phillips \u003ckim.phillips@freescale.com\u003e\nSigned-off-by: Herbert Xu \u003cherbert@gondor.apana.org.au\u003e\n",
  "tree_diff": [
    {
      "type": "modify",
      "old_id": "502a7f3cca5f8dd9012dcdfa0616ceaa3d208d73",
      "old_mode": 33188,
      "old_path": "drivers/crypto/caam/Kconfig",
      "new_id": "65c7668614ab4faf7880b9f34d9cbfc23351ce68",
      "new_mode": 33188,
      "new_path": "drivers/crypto/caam/Kconfig"
    }
  ]
}