Merge PRs snabbco#619 snabbco#620 snabbco#621 (LuaJIT extensions) int…

…o next

lib/luajit/src/jit/dump.lua

@@ -75,6 +75,9 @@ local bcline, disass
 -- Active flag, output file handle and dump mode.
 local active, out, dumpmode
 
+-- Information about traces that is remembered for future reference.
+local info = {}
+
 ------------------------------------------------------------------------------
 
 local symtabmt = { __index = false }
@@ -550,6 +553,7 @@ local function dump_trace(what, tr, func, pc, otr, oex)
     if dumpmode.m then dump_mcode(tr) end
   end
   if what == "start" then
+    info[tr] = { func = func, pc = pc, otr = otr, oex = oex }
     if dumpmode.H then out:write('<pre class="ljdump">\n') end
     out:write("---- TRACE ", tr, " ", what)
     if otr then out:write(" ", otr, "/", oex) end
@@ -701,6 +705,7 @@ end
 return {
   on = dumpon,
   off = dumpoff,
-  start = dumpon -- For -j command line option.
+  start = dumpon, -- For -j command line option.
+  info = info
 }
 

lib/luajit/src/jit/p.lua

@@ -36,6 +36,7 @@
 --   G  Produce raw output suitable for graphical tools (e.g. flame graphs).
 --   m<number> Minimum sample percentage to be shown. Default: 3.
 --   i<number> Sampling interval in milliseconds. Default: 10.
+--   S[<string>] Events source if performace events are enabled
 --
 ----------------------------------------------------------------------------
 
@@ -44,6 +45,8 @@ local jit = require("jit")
 assert(jit.version_num == 20100, "LuaJIT core/library version mismatch")
 local profile = require("jit.profile")
 local vmdef = require("jit.vmdef")
+local jutil = require("jit.util")
+local dump = require("jit.dump")
 local math = math
 local pairs, ipairs, tonumber, floor = pairs, ipairs, tonumber, math.floor
 local sort, format = table.sort, string.format
@@ -74,7 +77,38 @@ local function prof_cb(th, samples, vmmode)
   -- Collect keys for sample.
   if prof_states then
     if prof_states == "v" then
-      key_state = map_vmmode[vmmode] or vmmode
+      if map_vmmode[vmmode] then
+        key_state = map_vmmode[vmmode]
+      else
+         -- Sampling a trace: make an understandable one-line description.
+         local tr = tonumber(vmmode)
+         local info = jutil.traceinfo(tr)
+         local extra = dump.info[tr]
+         -- Show the parent of this trace (if this is a side trace)
+         local parent = ""
+         if extra and extra.otr and extra.oex then
+            parent = "("..extra.otr.."/"..extra.oex..")"
+         end
+         -- Show what the end of the trace links to (e.g. loop or other trace)
+         local lnk = ""
+         local link, ltype = info.link, info.linktype
+         if     link == tr or link == 0 then lnk = "->"..ltype
+         elseif ltype == "root"         then lnk = "->"..link
+         else                                lnk = "->"..link.." "..ltype end
+         -- Show the current zone (if zone profiling is enabled)
+         local z = ""
+         if zone and zone:get() then
+            z = (" %-16s"):format(zone:get())
+         end
+         -- Show the source location where the trace starts
+         local loc = ""
+         if extra and extra.func then
+            local fi = jutil.funcinfo(extra.func, extra.pc)
+            if fi.loc then loc = fi.loc end
+         end
+         local s = ("TRACE %3d %-8s %-10s%s %s"):format(vmmode, parent, lnk, z, loc)
+         key_state = map_vmmode[vmmode] or s
+      end
     else
       key_state = zone:get() or "(none)"
     end
@@ -243,15 +277,18 @@ end
 -- Start profiling.
 local function prof_start(mode)
   local interval = ""
-  mode = mode:gsub("i%d*", function(s) interval = s; return "" end)
+  mode = mode:gsub("i%d+", function(s) interval = s; return "" end)
   prof_min = 3
   mode = mode:gsub("m(%d+)", function(s) prof_min = tonumber(s); return "" end)
   prof_depth = 1
   mode = mode:gsub("%-?%d+", function(s) prof_depth = tonumber(s); return "" end)
+  local flavour = "S[vanilla]"
+  mode = mode:gsub("S%[.+%]", function(s) flavour = s; return "" end)
+
   local m = {}
   for c in mode:gmatch(".") do m[c] = c end
-  prof_states = m.z or m.v
-  if prof_states == "z" then zone = require("jit.zone") end
+  prof_states = m.v or m.z
+  if m.z == "z" then zone = require("jit.zone") end
   local scope = m.l or m.f or m.F or (prof_states and "" or "f")
   local flags = (m.p or "")
   prof_raw = m.r
@@ -285,7 +322,7 @@ local function prof_start(mode)
   prof_count1 = {}
   prof_count2 = {}
   prof_samples = 0
-  profile.start(scope:lower()..interval, prof_cb)
+  profile.start(scope:lower()..interval..flavour, prof_cb)
   prof_ud = newproxy(true)
   getmetatable(prof_ud).__gc = prof_finish
 end

lib/luajit/src/lib_jit.c

@@ -555,7 +555,10 @@ static void jit_profile_callback(lua_State *L2, lua_State *L, int samples,
     setfuncV(L2, L2->top++, funcV(tv));
     setthreadV(L2, L2->top++, L);
     setintV(L2->top++, samples);
-    setstrV(L2, L2->top++, lj_str_new(L2, &vmst, 1));
+    if (vmstate >= 256)
+      setintV(L2->top++, vmstate-256);
+    else
+      setstrV(L2, L2->top++, lj_str_new(L2, &vmst, 1));
     status = lua_pcall(L2, 3, 0, 0);  /* callback(thread, samples, vmstate) */
     if (status) {
       if (G(L2)->panic) G(L2)->panic(L2);

lib/luajit/src/lj_profile.c

@@ -5,6 +5,7 @@
 
 #define lj_profile_c
 #define LUA_CORE
+#define _GNU_SOURCE 1
 
 #include "lj_obj.h"
 
@@ -29,6 +30,17 @@
 #define profile_lock(ps)	UNUSED(ps)
 #define profile_unlock(ps)	UNUSED(ps)
 
+#if 1
+#include <stdio.h>
+#include <unistd.h>
+#include <fcntl.h>
+#include <sys/syscall.h>
+#include <sys/ioctl.h>
+#include <linux/perf_event.h>
+#include <sys/prctl.h>
+#endif
+
+
 #elif LJ_PROFILE_PTHREAD
 
 #include <pthread.h>
@@ -62,6 +74,8 @@ typedef struct ProfileState {
   SBuf sb;			/* String buffer for stack dumps. */
   int interval;			/* Sample interval in milliseconds. */
   int samples;			/* Number of samples for next callback. */
+  char *flavour;		/* What generates profiling events. */
+  int perf_event_fd;		/* Performace event file descriptor */
   int vmstate;			/* VM state when profile timer triggered. */
 #if LJ_PROFILE_SIGPROF
   struct sigaction oldsa;	/* Previous SIGPROF state. */
@@ -155,7 +169,7 @@ static void profile_trigger(ProfileState *ps)
   mask = g->hookmask;
   if (!(mask & (HOOK_PROFILE|HOOK_VMEVENT))) {  /* Set profile hook. */
     int st = g->vmstate;
-    ps->vmstate = st >= 0 ? 'N' :
+    ps->vmstate = st >= 0 ? 256+st :
 		  st == ~LJ_VMST_INTERP ? 'I' :
 		  st == ~LJ_VMST_C ? 'C' :
 		  st == ~LJ_VMST_GC ? 'G' : 'J';
@@ -176,29 +190,178 @@ static void profile_signal(int sig)
   profile_trigger(&profile_state);
 }
 
+
+static int perf_event_open(struct perf_event_attr *attr,
+			   pid_t pid, int cpu, int group_fd,
+			   unsigned long flags)
+{
+  return syscall(SYS_perf_event_open, attr, pid, cpu, group_fd, flags);
+}
+
+
+static void register_prof_events(ProfileState *ps)
+{
+  struct flavour_t {
+    char *name; uint32_t type; uint64_t config;
+  };
+
+  static struct flavour_t flavours[] =
+      {
+	{ "sw-cpu-clock",
+	  PERF_TYPE_SOFTWARE, PERF_COUNT_SW_CPU_CLOCK },
+
+	{ "sw-context-switches",
+	  PERF_TYPE_SOFTWARE, PERF_COUNT_SW_CONTEXT_SWITCHES },
+
+	{ "sw-page-faults",
+	  PERF_TYPE_SOFTWARE, PERF_COUNT_SW_PAGE_FAULTS },
+
+	{ "sw-minor-page-faults",
+	  PERF_TYPE_SOFTWARE, PERF_COUNT_SW_PAGE_FAULTS_MIN },
+
+	{ "sw-major-page-faults",
+	  PERF_TYPE_SOFTWARE, PERF_COUNT_SW_PAGE_FAULTS_MAJ },
+
+	{ "branch-instructions",
+	  PERF_TYPE_HARDWARE, PERF_COUNT_HW_BRANCH_INSTRUCTIONS },
+
+	{ "cpu-cycles",
+	  PERF_TYPE_HARDWARE, PERF_COUNT_HW_CPU_CYCLES },
+
+	{ "instructions",
+	  PERF_TYPE_HARDWARE, PERF_COUNT_HW_INSTRUCTIONS },
+
+	{ "cache-references",
+	  PERF_TYPE_HARDWARE, PERF_COUNT_HW_CACHE_REFERENCES },
+
+	{ "cache-misses",
+	  PERF_TYPE_HARDWARE, PERF_COUNT_HW_CACHE_MISSES },
+
+	{ "branch-instructions",
+	  PERF_TYPE_HARDWARE, PERF_COUNT_HW_BRANCH_INSTRUCTIONS },
+
+	{ "branch-misses",
+	  PERF_TYPE_HARDWARE, PERF_COUNT_HW_BRANCH_MISSES },
+
+	{ "bus-cycles",
+	  PERF_TYPE_HARDWARE, PERF_COUNT_HW_BUS_CYCLES },
+
+	{ "stalled-cycles-frontend",
+	  PERF_TYPE_HARDWARE, PERF_COUNT_HW_STALLED_CYCLES_FRONTEND },
+
+	{ "stalled-cycles-backend",
+	  PERF_TYPE_HARDWARE, PERF_COUNT_HW_STALLED_CYCLES_BACKEND },
+
+	{ "cpu-cycles",
+	  PERF_TYPE_HARDWARE, PERF_COUNT_HW_REF_CPU_CYCLES },
+
+	{ 0, 0, 0 }
+  };
+
+
+  struct perf_event_attr attr = { };
+
+  memset(&attr, 0, sizeof(struct perf_event_attr));
+
+  const struct flavour_t *f;
+  for (f = flavours; f->name != 0; f++)
+    {
+      if (strcmp (ps->flavour, f->name) == 0)
+	{
+	  attr.type = f->type;
+	  attr.config = f->config;
+	  break;
+	}
+    }
+
+  if (strcmp (ps->flavour, "?") == 0)
+    {
+      const struct flavour_t *f;
+      fprintf (stderr, "I know: ");
+      for (f = flavours; f->name != 0; f++)
+	fprintf (stderr, "%s ", f->name);
+      fprintf(stderr, "\n");
+    }
+  else if (! f->name)
+    {
+      fprintf (stderr, "unknown profiling flavour `%s', S[?] to list\n", ps->flavour);
+    }
+
+  attr.size = sizeof(struct perf_event_attr);
+  attr.sample_type = PERF_SAMPLE_IP;
+  /* attr.read_format = PERF_FORMAT_GROUP | PERF_FORMAT_ID; */
+  attr.disabled=1;
+  attr.pinned=1;
+  attr.exclude_kernel=1;
+  attr.exclude_hv=1;
+
+  attr.sample_period = ps->interval;
+  /* attr.watermark=0; */
+  /* attr.wakeup_events=1; */
+
+  int fd = perf_event_open(&attr, 0, -1, -1, 0);
+  if (fd == -1)
+    {
+      printf ("! perf_event_open %m\n");
+    }
+
+  ps->perf_event_fd = fd;
+
+  fcntl(fd, F_SETFL, O_RDWR|O_NONBLOCK|O_ASYNC);
+  fcntl(fd, F_SETSIG, SIGPROF);
+  fcntl(fd, F_SETOWN, getpid());
+
+  ioctl(fd, PERF_EVENT_IOC_RESET, 0);
+
+  int err = ioctl(fd, PERF_EVENT_IOC_ENABLE, 0);
+  if (err != 0)
+    printf ("! perf_events enable\n");
+}
+
+
+
 /* Start profiling timer. */
 static void profile_timer_start(ProfileState *ps)
 {
-  int interval = ps->interval;
-  struct itimerval tm;
-  struct sigaction sa;
-  tm.it_value.tv_sec = tm.it_interval.tv_sec = interval / 1000;
-  tm.it_value.tv_usec = tm.it_interval.tv_usec = (interval % 1000) * 1000;
-  setitimer(ITIMER_PROF, &tm, NULL);
-  sa.sa_flags = SA_RESTART;
-  sa.sa_handler = profile_signal;
+  struct sigaction sa = {
+    .sa_flags = SA_RESTART,
+    .sa_handler = profile_signal
+  };
+
   sigemptyset(&sa.sa_mask);
   sigaction(SIGPROF, &sa, &ps->oldsa);
+
+  if (strcmp(ps->flavour, "vanilla") == 0)
+    {
+      int interval = ps->interval;
+      struct itimerval tm;
+      tm.it_value.tv_sec = tm.it_interval.tv_sec = interval / 1000;
+      tm.it_value.tv_usec = tm.it_interval.tv_usec = (interval % 1000) * 1000;
+      setitimer(ITIMER_PROF, &tm, NULL);
+    }
+  else
+    {
+      register_prof_events(ps);
+    }
 }
 
+
+
 /* Stop profiling timer. */
 static void profile_timer_stop(ProfileState *ps)
 {
-  struct itimerval tm;
-  tm.it_value.tv_sec = tm.it_interval.tv_sec = 0;
-  tm.it_value.tv_usec = tm.it_interval.tv_usec = 0;
-  setitimer(ITIMER_PROF, &tm, NULL);
-  sigaction(SIGPROF, &ps->oldsa, NULL);
+  if (ps->perf_event_fd)
+    {
+      ioctl(ps->perf_event_fd, PERF_EVENT_IOC_DISABLE, 0);
+    }
+  else
+    {
+      struct itimerval tm;
+      tm.it_value.tv_sec = tm.it_interval.tv_sec = 0;
+      tm.it_value.tv_usec = tm.it_interval.tv_usec = 0;
+      setitimer(ITIMER_PROF, &tm, NULL);
+      sigaction(SIGPROF, &ps->oldsa, NULL);
+    }
 }
 
 #elif LJ_PROFILE_PTHREAD
@@ -300,6 +463,8 @@ LUA_API void luaJIT_profile_start(lua_State *L, const char *mode,
 {
   ProfileState *ps = &profile_state;
   int interval = LJ_PROFILE_INTERVAL_DEFAULT;
+  char *flavour;
+
   while (*mode) {
     int m = *mode++;
     switch (m) {
@@ -315,6 +480,13 @@ LUA_API void luaJIT_profile_start(lua_State *L, const char *mode,
       lj_trace_flushall(L);
       break;
 #endif
+    case 'S':
+      {
+	int k;
+	if (sscanf (mode, "[%m[^]]]%n", &flavour, &k) > 0)
+	  mode += k;
+      }
+
     default:  /* Ignore unknown mode chars. */
       break;
     }
@@ -328,6 +500,7 @@ LUA_API void luaJIT_profile_start(lua_State *L, const char *mode,
   ps->cb = cb;
   ps->data = data;
   ps->samples = 0;
+  ps->flavour = flavour;
   lj_buf_init(L, &ps->sb);
   profile_timer_start(ps);
 }