)]}'
{
  "commit": "9c2b957db1772ebf942ae7a9346b14eba6c8ca66",
  "tree": "0dbb83e57260ea7fc0dc421f214d5f1b26262005",
  "parents": [
    "0bbfcaff9b2a69c71a95e6902253487ab30cb498",
    "bea95c152dee1791dd02cbc708afbb115bb00f9a"
  ],
  "author": {
    "name": "Linus Torvalds",
    "email": "torvalds@linux-foundation.org",
    "time": "Tue Mar 20 10:29:15 2012 -0700"
  },
  "committer": {
    "name": "Linus Torvalds",
    "email": "torvalds@linux-foundation.org",
    "time": "Tue Mar 20 10:29:15 2012 -0700"
  },
  "message": "Merge branch \u0027perf-core-for-linus\u0027 of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip\n\nPull perf events changes for v3.4 from Ingo Molnar:\n\n - New \"hardware based branch profiling\" feature both on the kernel and\n   the tooling side, on CPUs that support it.  (modern x86 Intel CPUs\n   with the \u0027LBR\u0027 hardware feature currently.)\n\n   This new feature is basically a sophisticated \u0027magnifying glass\u0027 for\n   branch execution - something that is pretty difficult to extract from\n   regular, function histogram centric profiles.\n\n   The simplest mode is activated via \u0027perf record -b\u0027, and the result\n   looks like this in perf report:\n\n\t$ perf record -b any_call,u -e cycles:u branchy\n\n\t$ perf report -b --sort\u003dsymbol\n\t    52.34%  [.] main                   [.] f1\n\t    24.04%  [.] f1                     [.] f3\n\t    23.60%  [.] f1                     [.] f2\n\t     0.01%  [k] _IO_new_file_xsputn    [k] _IO_file_overflow\n\t     0.01%  [k] _IO_vfprintf_internal  [k] _IO_new_file_xsputn\n\t     0.01%  [k] _IO_vfprintf_internal  [k] strchrnul\n\t     0.01%  [k] __printf               [k] _IO_vfprintf_internal\n\t     0.01%  [k] main                   [k] __printf\n\n   This output shows from/to branch columns and shows the highest\n   percentage (from,to) jump combinations - i.e.  the most likely taken\n   branches in the system.  \"branches\" can also include function calls\n   and any other synchronous and asynchronous transitions of the\n   instruction pointer that are not \u0027next instruction\u0027 - such as system\n   calls, traps, interrupts, etc.\n\n   This feature comes with (hopefully intuitive) flat ascii and TUI\n   support in perf report.\n\n - Various \u0027perf annotate\u0027 visual improvements for us assembly junkies.\n   It will now recognize function calls in the TUI and by hitting enter\n   you can follow the call (recursively) and back, amongst other\n   improvements.\n\n - Multiple threads/processes recording support in perf record, perf\n   stat, perf top - which is activated via a comma-list of PIDs:\n\n\tperf top -p 21483,21485\n\tperf stat -p 21483,21485 -ddd\n\tperf record -p 21483,21485\n\n - Support for per UID views, via the --uid paramter to perf top, perf\n   report, etc.  For example \u0027perf top --uid mingo\u0027 will only show the\n   tasks that I am running, excluding other users, root, etc.\n\n - Jump label restructurings and improvements - this includes the\n   factoring out of the (hopefully much clearer) include/linux/static_key.h\n   generic facility:\n\n\tstruct static_key key \u003d STATIC_KEY_INIT_FALSE;\n\n\t...\n\n\tif (static_key_false(\u0026key))\n\t        do unlikely code\n\telse\n\t        do likely code\n\n\t...\n\tstatic_key_slow_inc();\n\t...\n\tstatic_key_slow_inc();\n\t...\n\n   The static_key_false() branch will be generated into the code with as\n   little impact to the likely code path as possible.  the\n   static_key_slow_*() APIs flip the branch via live kernel code patching.\n\n   This facility can now be used more widely within the kernel to\n   micro-optimize hot branches whose likelihood matches the static-key\n   usage and fast/slow cost patterns.\n\n - SW function tracer improvements: perf support and filtering support.\n\n - Various hardenings of the perf.data ABI, to make older perf.data\u0027s\n   smoother on newer tool versions, to make new features integrate more\n   smoothly, to support cross-endian recording/analyzing workflows\n   better, etc.\n\n - Restructuring of the kprobes code, the splitting out of \u0027optprobes\u0027,\n   and a corner case bugfix.\n\n - Allow the tracing of kernel console output (printk).\n\n - Improvements/fixes to user-space RDPMC support, allowing user-space\n   self-profiling code to extract PMU counts without performing any\n   system calls, while playing nice with the kernel side.\n\n - \u0027perf bench\u0027 improvements\n\n - ... and lots of internal restructurings, cleanups and fixes that made\n   these features possible.  And, as usual this list is incomplete as\n   there were also lots of other improvements\n\n* \u0027perf-core-for-linus\u0027 of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip: (120 commits)\n  perf report: Fix annotate double quit issue in branch view mode\n  perf report: Remove duplicate annotate choice in branch view mode\n  perf/x86: Prettify pmu config literals\n  perf report: Enable TUI in branch view mode\n  perf report: Auto-detect branch stack sampling mode\n  perf record: Add HEADER_BRANCH_STACK tag\n  perf record: Provide default branch stack sampling mode option\n  perf tools: Make perf able to read files from older ABIs\n  perf tools: Fix ABI compatibility bug in print_event_desc()\n  perf tools: Enable reading of perf.data files from different ABI rev\n  perf: Add ABI reference sizes\n  perf report: Add support for taken branch sampling\n  perf record: Add support for sampling taken branch\n  perf tools: Add code to support PERF_SAMPLE_BRANCH_STACK\n  x86/kprobes: Split out optprobe related code to kprobes-opt.c\n  x86/kprobes: Fix a bug which can modify kernel code permanently\n  x86/kprobes: Fix instruction recovery on optimized path\n  perf: Add callback to flush branch_stack on context switch\n  perf: Disable PERF_SAMPLE_BRANCH_* when not supported\n  perf/x86: Add LBR software filter support for Intel CPUs\n  ...\n",
  "tree_diff": []
}