LuaJIT profiler extension to count loops of traces separately [draft]

luajit-profiler-separate-loops.patch

diff --git a/lib/luajit/src/jit/p.lua b/lib/luajit/src/jit/p.lua
index d894bb7..b619517 100644
--- a/lib/luajit/src/jit/p.lua
+++ b/lib/luajit/src/jit/p.lua
@@ -71,7 +71,8 @@ local map_vmmode = {
 }
 
 -- Profiler callback.
-local function prof_cb(th, samples, vmmode)
+local function prof_cb(th, samples, vmmode, ip)
+  print("samples", samples, "vmmode", vmmode, "ip", bit.tohex(ip))
   prof_samples = prof_samples + samples
   local key_stack, key_stack2, key_state
   -- Collect keys for sample.
@@ -106,7 +107,9 @@ local function prof_cb(th, samples, vmmode)
             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)
+         print("ip", bit.tohex(ip), "szmcode", info.szmcode, "mcode", bit.tohex(info.mcode), "mcloop", info.mcloop)
+         local loop = (ip >= info.mcode+info.mcloop) and ':LOOP' or '     '
+         local s = ("TRACE %3d%s %-8s %-10s%s %s"):format(vmmode, loop, parent, lnk, z, loc)
          key_state = map_vmmode[vmmode] or s
       end
     else
diff --git a/lib/luajit/src/lib_jit.c b/lib/luajit/src/lib_jit.c
index 9e0e01d..9f7c041 100644
--- a/lib/luajit/src/lib_jit.c
+++ b/lib/luajit/src/lib_jit.c
@@ -293,14 +293,23 @@ LJLIB_CF(jit_util_traceinfo)
   GCtrace *T = jit_checktrace(L);
   if (T) {
     GCtab *t;
+    //int gc = 0;
+    int i;
     lua_createtable(L, 0, 8);  /* Increment hash size if fields are added. */
     t = tabV(L->top-1);
     setintfield(L, t, "nins", (int32_t)T->nins - REF_BIAS - 1);
     setintfield(L, t, "nk", REF_BIAS - (int32_t)T->nk);
     setintfield(L, t, "link", T->link);
     setintfield(L, t, "nexit", T->nsnap);
+    setintfield(L, t, "szmcode", T->szmcode);
+    setintfield(L, t, "mcode", T->mcode);
+    setintfield(L, t, "mcloop", T->mcloop);
     setstrV(L, L->top++, lj_str_newz(L, jit_trlinkname[T->linktype]));
     lua_setfield(L, -2, "linktype");
+
+    for (i = 0; i > T->nins-1; i++) {
+      if ((IROp)T->ir[i].o == IR_GCSTEP) { printf("GC\n"); }
+    }
     /* There are many more fields. Add them only when needed. */
     return 1;
   }
@@ -543,7 +552,7 @@ static const char KEY_PROFILE_THREAD = 't';
 static const char KEY_PROFILE_FUNC = 'f';
 
 static void jit_profile_callback(lua_State *L2, lua_State *L, int samples,
-				 int vmstate)
+				 int vmstate, uint64_t ip)
 {
   TValue key;
   cTValue *tv;
@@ -559,7 +568,8 @@ static void jit_profile_callback(lua_State *L2, lua_State *L, int samples,
       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) */
+    setintV(L2->top++, ip);
+    status = lua_pcall(L2, 4, 0, 0);  /* callback(thread, samples, vmstate) */
     if (status) {
       if (G(L2)->panic) G(L2)->panic(L2);
       exit(EXIT_FAILURE);
diff --git a/lib/luajit/src/lj_profile.c b/lib/luajit/src/lj_profile.c
index f4d6fe1..d8c14fb 100644
--- a/lib/luajit/src/lj_profile.c
+++ b/lib/luajit/src/lj_profile.c
@@ -38,6 +38,7 @@
 #include <sys/ioctl.h>
 #include <linux/perf_event.h>
 #include <sys/prctl.h>
+#include <ucontext.h>
 #endif
 
 
@@ -77,6 +78,7 @@ typedef struct ProfileState {
   char *flavour;		/* What generates profiling events. */
   int perf_event_fd;		/* Performace event file descriptor */
   int vmstate;			/* VM state when profile timer triggered. */
+  uint64_t ip;                  /* CPU Instruction Pointer when triggered. */
 #if LJ_PROFILE_SIGPROF
   struct sigaction oldsa;	/* Previous SIGPROF state. */
 #elif LJ_PROFILE_PTHREAD
@@ -144,13 +146,15 @@ void LJ_FASTCALL lj_profile_interpreter(lua_State *L)
   uint8_t mask;
   profile_lock(ps);
   mask = (g->hookmask & ~HOOK_PROFILE);
+  printf("lj_profile_interpreter\n");
   if (!(mask & HOOK_VMEVENT)) {
     int samples = ps->samples;
     ps->samples = 0;
     g->hookmask = HOOK_VMEVENT;
     lj_dispatch_update(g);
     profile_unlock(ps);
-    ps->cb(ps->data, L, samples, ps->vmstate);  /* Invoke user callback. */
+    printf("Calling cb with ps->ip %p\n", (void*)ps->ip);
+    ps->cb(ps->data, L, samples, ps->vmstate, ps->ip);  /* Invoke user callback. */
     profile_lock(ps);
     mask |= (g->hookmask & HOOK_PROFILE);
   }
@@ -160,12 +164,13 @@ void LJ_FASTCALL lj_profile_interpreter(lua_State *L)
 }
 
 /* Trigger profile hook. Asynchronous call from OS-specific profile timer. */
-static void profile_trigger(ProfileState *ps)
+static void profile_trigger(ProfileState *ps, uint64_t ip)
 {
   global_State *g = ps->g;
   uint8_t mask;
   profile_lock(ps);
   ps->samples++;  /* Always increment number of samples. */
+  ps->ip = ip;
   mask = g->hookmask;
   if (!(mask & (HOOK_PROFILE|HOOK_VMEVENT))) {  /* Set profile hook. */
     int st = g->vmstate;
@@ -176,6 +181,13 @@ static void profile_trigger(ProfileState *ps)
     g->hookmask = (mask | HOOK_PROFILE);
     lj_dispatch_update(g);
   }
+  /*
+  else {
+    int samples = ps->samples;
+    ps->samples = 0;
+    ps->cb(ps->data, g, samples, ps->vmstate, ps->ip);
+  }
+*/
   profile_unlock(ps);
 }
 
@@ -184,13 +196,13 @@ static void profile_trigger(ProfileState *ps)
 #if LJ_PROFILE_SIGPROF
 
 /* SIGPROF handler. */
-static void profile_signal(int sig)
+static void profile_signal(int sig, siginfo_t *info, void *data)
 {
   UNUSED(sig);
-  profile_trigger(&profile_state);
+  uint64_t ip = (uint64_t)((ucontext_t*)data)->uc_mcontext.gregs[REG_RIP];
+  profile_trigger(&profile_state, ip);
 }
 
-
 static int perf_event_open(struct perf_event_attr *attr,
 			   pid_t pid, int cpu, int group_fd,
 			   unsigned long flags)
@@ -324,8 +336,8 @@ static void register_prof_events(ProfileState *ps)
 static void profile_timer_start(ProfileState *ps)
 {
   struct sigaction sa = {
-    .sa_flags = SA_RESTART,
-    .sa_handler = profile_signal
+    .sa_flags = SA_RESTART|SA_SIGINFO,
+    .sa_sigaction = profile_signal
   };
 
   sigemptyset(&sa.sa_mask);
diff --git a/lib/luajit/src/luajit.h b/lib/luajit/src/luajit.h
index 9604185..fe67cd4 100644
--- a/lib/luajit/src/luajit.h
+++ b/lib/luajit/src/luajit.h
@@ -29,6 +29,7 @@
 #define _LUAJIT_H
 
 #include "lua.h"
+#include <stdint.h>
 
 #define LUAJIT_VERSION		"LuaJIT 2.1.0-beta1"
 #define LUAJIT_VERSION_NUM	20100  /* Version 2.1.0 = 02.01.00. */
@@ -66,7 +67,8 @@ LUA_API int luaJIT_setmode(lua_State *L, int idx, int mode);
 
 /* Low-overhead profiling API. */
 typedef void (*luaJIT_profile_callback)(void *data, lua_State *L,
-					int samples, int vmstate);
+					int samples, int vmstate,
+                                        uint64_t ip);
 LUA_API void luaJIT_profile_start(lua_State *L, const char *mode,
 				  luaJIT_profile_callback cb, void *data);
 LUA_API void luaJIT_profile_stop(lua_State *L);