FM / profiling

/usr/share/systemtap/examples/profiling/

/usr/share/systemtap/examples/profiling

Name	Size	Mode	Actions
container_check.meta	1064	0644	edit dl rm
container_check.stp	6323	0755	edit dl rm
errno.meta	456	0644	edit dl rm
errno.stp	984	0755	edit dl rm
fileline-profile.meta	783	0644	edit dl rm
fileline-profile.stp	1218	0755	edit dl rm
fntimes.meta	512	0644	edit dl rm
fntimes.stp	870	0755	edit dl rm
fntimes.txt	1375	0644	edit dl rm
functioncallcount.meta	649	0644	edit dl rm
functioncallcount.stp	490	0755	edit dl rm
functioncallcount.tcl	399	0644	edit dl rm
graphcall.stp	534	0755	edit dl rm
ioctl_handler.meta	1209	0644	edit dl rm
ioctl_handler.stp	811	0755	edit dl rm
latencytap.meta	877	0644	edit dl rm
latencytap.stp	12755	0755	edit dl rm
latencytap.tcl	140	0644	edit dl rm
latencytap.txt	2104	0644	edit dl rm
linetimes.meta	688	0644	edit dl rm
linetimes.stp	1608	0755	edit dl rm
linetimes.txt	13936	0644	edit dl rm
perf.meta	578	0644	edit dl rm
perf.stp	1936	0755	edit dl rm
perf.tcl	166	0644	edit dl rm
periodic.meta	895	0644	edit dl rm
periodic.stp	2659	0755	edit dl rm
pf2.meta	420	0644	edit dl rm
pf2.stp	370	0755	edit dl rm
pf2.txt	406	0644	edit dl rm
pf3.meta	439	0644	edit dl rm
pf3.stp	992	0755	edit dl rm
pf3.txt	489	0644	edit dl rm
pf4.meta	471	0644	edit dl rm
pf4.stp	952	0755	edit dl rm
pf4.tcl	79	0644	edit dl rm
pf4.txt	754	0644	edit dl rm
sched_switch.meta	780	0644	edit dl rm
sched_switch.stp	2127	0755	edit dl rm
syscallerrorsbypid.meta	1067	0644	edit dl rm
syscallerrorsbypid.stp	950	0755	edit dl rm
syscalllatency.meta	1072	0644	edit dl rm
syscalllatency.stp	989	0755	edit dl rm
syscallsbypid.meta	1020	0644	edit dl rm
syscallsbypid.stp	785	0755	edit dl rm
syscallsrw.meta	489	0644	edit dl rm
syscallsrw.stp	375	0755	edit dl rm
thread-times.meta	384	0644	edit dl rm
thread-times.stp	1241	0755	edit dl rm
thread-times.txt	1009	0644	edit dl rm
timeout.meta	728	0644	edit dl rm
timeout.stp	2775	0755	edit dl rm
topsys.meta	446	0644	edit dl rm
topsys.stp	745	0755	edit dl rm
ucalls.meta	1065	0644	edit dl rm
ucalls.stp	3544	0755	edit dl rm

Edit: /usr/share/systemtap/examples/profiling/perf.stp (1936B)

#!/usr/bin/stap
global cycles_per_insn
global branch_per_insn
global cacheref_per_insn
global insns
global cycles
global branches
global cacherefs
global insn

probe perf.hw.instructions.process("/usr/bin/find").counter("find_insns") {} 
probe perf.hw.cpu_cycles.process("/usr/bin/find").counter("find_cycles") {} 
probe perf.hw.branch_instructions.process("/usr/bin/find").counter("find_branches") {} 
probe perf.hw.cache_references.process("/usr/bin/find").counter("find_cache_refs") {}

probe process("/usr/bin/find").function("visit")  # generally .inline'd, thus the need for debuginfo
{
 insn["find_insns"] = @perf("find_insns")
 insns <<< (insn["find_insns"])
 insn["find_cycles"] = @perf("find_cycles")
 cycles <<< insn["find_cycles"]
 insn["find_branches"] = @perf("find_branches")
 branches <<< insn["find_branches"]
 insn["find_cache_refs"] = @perf("find_cache_refs")
 cacherefs <<< insn["find_cache_refs"]
}

probe process("/usr/bin/find").function("visit").return !,
      process("/usr/bin/find").statement("visit@ftsfind.c+13").nearest  # in lieu of .inline.return
{
    dividend = (@perf("find_cycles") - insn["find_cycles"])
    divisor =  (@perf("find_insns") - insn["find_insns"])
    q = (divisor > 0 ? dividend / divisor : 0)
    if (q > 0)
	cycles_per_insn <<< q

dividend = (@perf("find_branches") - insn["find_branches"])
    q = (divisor > 0 ? dividend / divisor : 0)
    if (q > 0)
	branch_per_insn <<< q

dividend = (@perf("find_cycles") - insn["find_cycles"])
    q = (divisor > 0 ? dividend / divisor : 0)
    if (q > 0)
	cacheref_per_insn <<< q
}

probe end
{
 if (@count(cycles_per_insn)) {
   printf ("Cycles per Insn\n\n")
   print (@hist_log(cycles_per_insn))
 }
 if (@count(branch_per_insn)) {
   printf ("\nBranches per Insn\n\n")
   print (@hist_log(branch_per_insn))
 }
 if (@count(cacheref_per_insn)) {
   printf ("Cache Refs per Insn\n\n")
   print (@hist_log(cacheref_per_insn))
 }
}

#!/usr/bin/stap global cycles_per_insn global branch_per_insn global cacheref_per_insn global insns global cycles global branches global cacherefs global insn probe perf.hw.instructions.process("/usr/bin/find").counter("find_insns") {} probe perf.hw.cpu_cycles.process("/usr/bin/find").counter("find_cycles") {} probe perf.hw.branch_instructions.process("/usr/bin/find").counter("find_branches") {} probe perf.hw.cache_references.process("/usr/bin/find").counter("find_cache_refs") {} probe process("/usr/bin/find").function("visit") # generally .inline'd, thus the need for debuginfo { insn["find_insns"] = @perf("find_insns") insns <<< (insn["find_insns"]) insn["find_cycles"] = @perf("find_cycles") cycles <<< insn["find_cycles"] insn["find_branches"] = @perf("find_branches") branches <<< insn["find_branches"] insn["find_cache_refs"] = @perf("find_cache_refs") cacherefs <<< insn["find_cache_refs"] } probe process("/usr/bin/find").function("visit").return !, process("/usr/bin/find").statement("visit@ftsfind.c+13").nearest # in lieu of .inline.return { dividend = (@perf("find_cycles") - insn["find_cycles"]) divisor = (@perf("find_insns") - insn["find_insns"]) q = (divisor > 0 ? dividend / divisor : 0) if (q > 0) cycles_per_insn <<< q dividend = (@perf("find_branches") - insn["find_branches"]) q = (divisor > 0 ? dividend / divisor : 0) if (q > 0) branch_per_insn <<< q dividend = (@perf("find_cycles") - insn["find_cycles"]) q = (divisor > 0 ? dividend / divisor : 0) if (q > 0) cacheref_per_insn <<< q } probe end { if (@count(cycles_per_insn)) { printf ("Cycles per Insn\n\n") print (@hist_log(cycles_per_insn)) } if (@count(branch_per_insn)) { printf ("\nBranches per Insn\n\n") print (@hist_log(branch_per_insn)) } if (@count(cacheref_per_insn)) { printf ("Cache Refs per Insn\n\n") print (@hist_log(cacheref_per_insn)) } }