autofdo incorrectly expects branch flags to include either mispred or
predicted. In fact mispred = predicted = 0 is valid and means the flags
are not supported, which they aren't by Intel PT.
To make autofdo work, add a config option which will cause Intel PT
decoder to set the mispred flag on all branches.
Below is an example of using Intel PT with autofdo. The example is
also added to the Intel PT documentation. It requires autofdo
(https://github.com/google/autofdo) and gcc version 5. The bubble
sort example is from the AutoFDO tutorial (https://gcc.gnu.org/wiki/AutoFDO/Tutorial)
amended to take the number of elements as a parameter.
$ gcc-5 -O3 sort.c -o sort_optimized
$ ./sort_optimized 30000
Bubble sorting array of 30000 elements
2254 ms
$ cat ~/.perfconfig
[intel-pt]
mispred-all
$ perf record -e intel_pt//u ./sort 3000
Bubble sorting array of 3000 elements
58 ms
[ perf record: Woken up 2 times to write data ]
[ perf record: Captured and wrote 3.939 MB perf.data ]
$ perf inject -i perf.data -o inj --itrace=i100usle --strip
$ ./create_gcov --binary=./sort --profile=inj --gcov=sort.gcov -gcov_version=1
$ gcc-5 -O3 -fauto-profile=sort.gcov sort.c -o sort_autofdo
$ ./sort_autofdo 30000
Bubble sorting array of 30000 elements
2155 ms
Note there is currently no advantage to using Intel PT instead of LBR,
but that may change in the future if greater use is made of the data.
Signed-off-by: Adrian Hunter <adrian.hunter@intel.com>
Cc: Jiri Olsa <jolsa@redhat.com>
Link: http://lkml.kernel.org/r/1443186956-18718-26-git-send-email-adrian.hunter@intel.com
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Add a new option --strip which is used with --itrace to strip out
non-synthesized events. This results in a perf.data file that is
simpler for external tools to parse. In particular, this can be used to
prepare a perf.data file for consumption by autofdo.
A subsequent patch makes a change to Intel PT also to enable use with
autofdo and gives an example of that use.
Signed-off-by: Adrian Hunter <adrian.hunter@intel.com>
Cc: Jiri Olsa <jolsa@redhat.com>
Link: http://lkml.kernel.org/r/1443186956-18718-25-git-send-email-adrian.hunter@intel.com
[ Made it use perf_evlist__remove() + perf_evsel__delete() ]
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
perf inject can process instruction traces (using the --itrace option)
which removes aux-related events and replaces them with the requested
synthesized events.
However there are still some leftovers, namely PERF_RECORD_ITRACE_START
events and the original evsel (selected event) e.g. intel_pt//
For the sake of completeness, remove them too.
Signed-off-by: Adrian Hunter <adrian.hunter@intel.com>
Cc: Jiri Olsa <jolsa@redhat.com>
Link: http://lkml.kernel.org/r/1443186956-18718-24-git-send-email-adrian.hunter@intel.com
[ Made it use perf_evlist__remove() + perf_evsel__delete() ]
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
perf script has a setting to set the maximum stack depth when processing
callchains. The setting defaults to the hard-coded maximum definition
PERF_MAX_STACK_DEPTH which is 127.
It is possible, when processing instruction traces, to synthesize
callchains. Synthesized callchains do not have the kernel size
limitation and are whatever size the user requests, although validation
presently prevents the user requested a value greater that 1024. The
default value is 16.
To allow for synthesized callchains, make the scripting_max_stack value
at least the same size as the synthesized callchain size.
Signed-off-by: Adrian Hunter <adrian.hunter@intel.com>
Cc: Jiri Olsa <jolsa@redhat.com>
Link: http://lkml.kernel.org/r/1443186956-18718-21-git-send-email-adrian.hunter@intel.com
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
perf report has an option (--max-stack) to set the maximum stack depth
when processing callchains. The option defaults to the hard-coded
maximum definition PERF_MAX_STACK_DEPTH which is 127. The intention of
the option is to allow the user to reduce the processing time by
reducing the amount of the callchain that is processed.
It is also possible, when processing instruction traces, to synthesize
callchains. Synthesized callchains do not have the kernel size
limitation and are whatever size the user requests, although validation
presently prevents the user requested a value greater that 1024. The
default value is 16.
To allow for synthesized callchains, make the max_stack value at least
the same size as the synthesized callchain size.
Signed-off-by: Adrian Hunter <adrian.hunter@intel.com>
Cc: Jiri Olsa <jolsa@redhat.com>
Link: http://lkml.kernel.org/r/1443186956-18718-16-git-send-email-adrian.hunter@intel.com
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Add support for generating branch stack context for PT samples. The
decoder reports a configurable number of branches as branch context for
each sample. Internally it keeps track of them by using a simple sliding
window. We also flush the last branch buffer on each sample to avoid
overlapping intervals.
This is useful for:
- Reporting accurate basic block edge frequencies through the perf
report branch view
- Using with --branch-history to get the wider context of samples
- Other users of LBRs
Also the Documentation is updated.
Examples:
Record with Intel PT:
perf record -e intel_pt//u ls
Branch stacks are used by default if synthesized so:
perf report --itrace=ile
is the same as:
perf report --itrace=ile -b
Branch history can be requested also:
perf report --itrace=igle --branch-history
Based-on-patch-by: Andi Kleen <ak@linux.intel.com>
Signed-off-by: Adrian Hunter <adrian.hunter@intel.com>
Cc: Jiri Olsa <jolsa@redhat.com>
Link: http://lkml.kernel.org/r/1443186956-18718-15-git-send-email-adrian.hunter@intel.com
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
A non-synthesized event might not have a branch stack if branch stacks
have been synthesized (using itrace options).
An example of that is when Intel PT records sched_switch events for
decoding purposes. Those sched_switch events do not have branch stacks
even though the Intel PT decoder may be synthesizing other events that
do due to the itrace options.
Signed-off-by: Adrian Hunter <adrian.hunter@intel.com>
Cc: Jiri Olsa <jolsa@redhat.com>
Link: http://lkml.kernel.org/r/1443186956-18718-12-git-send-email-adrian.hunter@intel.com
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
perf report looks at event sample types to determine if branch stacks
have been sampled. Adjust the validation to know about instruction
tracing options.
This change allows the use of the -b option which otherwise would
complain with an error like:
Error:
Selected -b but no branch data. Did you call perf record without -b?
# To display the perf.data header info,
# please use --header/--header-only options.
#
Signed-off-by: Adrian Hunter <adrian.hunter@intel.com>
Cc: Jiri Olsa <jolsa@redhat.com>
Link: http://lkml.kernel.org/r/1443186956-18718-10-git-send-email-adrian.hunter@intel.com
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
TSC packets contain only 7 bytes of TSC. The 8th byte is assumed to
change so infrequently that its value can be inferred. However the
logic must cater for a 7 byte wraparound, which it does by adding 1 to
the top byte.
The existing code was doing that with a while loop even though the
addition should only need to be done once. That logic won't work (will
loop forever) if TSC wraps around at the 8th byte. Theoretically that
would take at least 10 years, unless something else went wrong.
And what else could go wrong. Well, if the chunks of trace data are
processed out of order, it will make it look like the 7-byte TSC has
gone backwards (i.e. wrapped). If that happens 256 times then stuck in
the while loop it will be.
Fix that by getting rid of the unnecessary while loop.
Signed-off-by: Adrian Hunter <adrian.hunter@intel.com>
Cc: Jiri Olsa <jolsa@redhat.com>
Link: http://lkml.kernel.org/r/1443186956-18718-4-git-send-email-adrian.hunter@intel.com
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Processing instruction tracing data (e.g. Intel PT) can synthesize
callchains e.g.
$ perf record -e intel_pt//u uname
$ perf report --stdio --itrace=ige
However perf report's callgraph option gets extra validation, so:
$ perf report --stdio --itrace=ige -gflat
Error:
Selected -g or --branch-history but no callchain data. Did
you call 'perf record' without -g?
# To display the perf.data header info,
# please use --header/--header-only options.
#
Fix the validation to know about instruction tracing options so
above command works.
A side-effect of the change is that the default option to
accumulate the callchain of child functions comes into force.
To get the previous behaviour the --no-children option can be
used e.g.
$ perf report --stdio --itrace=ige -gflat --no-children
Signed-off-by: Adrian Hunter <adrian.hunter@intel.com>
Cc: Jiri Olsa <jolsa@redhat.com>
Link: http://lkml.kernel.org/r/1443186956-18718-3-git-send-email-adrian.hunter@intel.com
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Instruction tracing options (i.e. --itrace) include an option for
sampling instructions at an arbitrary period. e.g.
--itrace=i10us
means make an 'instructions' sample for every 10us of trace.
Currently the logic does not distinguish between a period of
zero and no period being specified at all, so it gets treated
as the default period which is 100000. That doesn't really
make sense.
Fix it so that zero period is accepted and treated as meaning
"as often as possible".
In the case of Intel PT that is the same as a period of 1 and
a unit of 'instructions' (i.e. --itrace=i1i).
Signed-off-by: Adrian Hunter <adrian.hunter@intel.com>
Cc: Jiri Olsa <jolsa@redhat.com>
Link: http://lkml.kernel.org/r/1443186956-18718-2-git-send-email-adrian.hunter@intel.com
[ Add a few lines describing this in the Documentation/intel-pt.txt file ]
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
For dependency tracking we currently use targets that fall out of the
gcc -MD command. We store this info in the .cmd file and include as
makefile during the build.
This format put object as target and all the c and header files as
dependencies, like:
util/abspath.o: util/abspath.c /usr/include/stdc-predef.h util/cache.h \
/usr/include/bits/wordsize.h /usr/include/gnu/stubs.h \
...
If any of those dependency header files (krava.h below) is removed the
build fails on:
make[1]: *** No rule to make target 'krava.h', needed by 'inc.o'. Stop.
This patch adds fixdep helper, that is used by kbuild to alter the shape
of the object dependencies like:
source_util/abspath.o := util/abspath.c
deps_util/abspath.o := \
/usr/include/stdc-predef.h \
util/cache.h \
...
util/abspath.o: $(deps_util/abspath.o)
$(deps_util/abspath.o):
With this format the header removal won't make the build fail, because
it'll be picked up by the last empty target defined for each header.
As previously mentioned the fixdep tool is taken from kbuild. It's not
complete backport, only the part that alters the standard dependency
info was taken, the part that adds the CONFIG_* dependency logic will be
probably taken later on.
Signed-off-by: Jiri Olsa <jolsa@kernel.org>
Cc: David Ahern <dsahern@gmail.com>
Cc: Kai Germaschewski <kai.germaschewski@gmx.de>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Link: http://lkml.kernel.org/r/1443004442-32660-4-git-send-email-jolsa@kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Storing the actual tracing path mountpoint to display correct
error message hint ('Hint:' line). The error hint rediscovers
mountpoints, but it could be different from what we actually
used in tracing path.
Before we'd display debugfs mount even though tracefs was used:
$ perf record -e sched:sched_krava ls
event syntax error: 'sched:sched_krava'
\___ can't access trace events
Error: No permissions to read /sys/kernel/debug/tracing/events/sched/sched_krava
Hint: Try 'sudo mount -o remount,mode=755 /sys/kernel/debug'
...
After this change, correct mountpoint is displayed:
$ perf record -e sched:sched_krava ls
event syntax error: 'sched:sched_krava'
\___ can't access trace events
Error: No permissions to read /sys/kernel/debug/tracing/events/sched/sched_krava
Hint: Try 'sudo mount -o remount,mode=755 /sys/kernel/debug/tracing'
...
Signed-off-by: Jiri Olsa <jolsa@kernel.org>
Cc: David Ahern <dsahern@gmail.com>
Cc: Matt Fleming <matt@codeblueprint.co.uk>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Raphael Beamonte <raphael.beamonte@gmail.com>
Link: http://lkml.kernel.org/r/1442674027-19427-1-git-send-email-jolsa@kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Pull MIPS fixes from Ralf Baechle:
- Properly setup irq handling for ATH79 platforms
- Fix bootmem mapstart calculation for contiguous maps
- Handle little endian and older CPUs correct in BPF
- Fix console for Fulong 2E systems
- Handle FTLB correctly on R6 CPUs
- Fixes for CM, GIC and MAAR support code
* 'upstream' of git://git.linux-mips.org/pub/scm/ralf/upstream-linus:
MIPS: Initialise MAARs on secondary CPUs
MIPS: print MAAR configuration during boot
MIPS: mm: compile maar_init unconditionally
irqchip: mips-gic: Fix pending & mask reads for MIPS64 with 32b GIC.
irqchip: mips-gic: Convert CPU numbers to VP IDs.
MIPS: CM: Provide a function to map from CPU to VP ID.
MIPS: Fix FTLB detection for R6
MIPS: cpu-features: Add cpu_has_ftlb
MIPS: ATH79: Add irq chip ar7240-misc-intc
MIPS: ATH79: Set missing irq ack handler for ar7100-misc-intc irq chip
MIPS: BPF: Fix build on pre-R2 little endian CPUs
MIPS: BPF: Avoid unreachable code on little endian
MIPS: bootmem: Fix mapstart calculation for contiguous maps
MIPS: Fix console output for Fulong2e system
Pull perf fixes from Thomas Gleixner:
"Another pile of fixes for perf:
- Plug overflows and races in the core code
- Sanitize the flow of the perf syscall so we error out before
handling the more complex and hard to undo setups
- Improve and fix Broadwell and Skylake hardware support
- Revert a fix which broke what it tried to fix in perf tools
- A couple of smaller fixes in various places of perf tools"
* 'perf-urgent-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip:
perf tools: Fix copying of /proc/kcore
perf intel-pt: Remove no_force_psb from documentation
perf probe: Use existing routine to look for a kernel module by dso->short_name
perf/x86: Change test_aperfmperf() and test_intel() to static
tools lib traceevent: Fix string handling in heterogeneous arch environments
perf record: Avoid infinite loop at buildid processing with no samples
perf: Fix races in computing the header sizes
perf: Fix u16 overflows
perf: Restructure perf syscall point of no return
perf/x86/intel: Fix Skylake FRONTEND MSR extrareg mask
perf/x86/intel/pebs: Add PEBS frontend profiling for Skylake
perf/x86/intel: Make the CYCLE_ACTIVITY.* constraint on Broadwell more specific
perf tools: Bool functions shouldn't return -1
tools build: Add test for presence of __get_cpuid() gcc builtin
tools build: Add test for presence of numa_num_possible_cpus() in libnuma
Revert "perf symbols: Fix mismatched declarations for elf_getphdrnum"
perf stat: Fix per-pkg event reporting bug
Pull scheduler fix from Thomas Gleixner:
"A single bug fix for the scheduler to prevent dequeueing of the idle
task when setting the cpus allowed mask"
* 'sched-urgent-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip:
sched: Fix crash trying to dequeue/enqueue the idle thread
Pull locking fix from Thomas Gleixner:
"A single bugfix for lockdep to preserve the pinning counter when
rebuilding the lock stack"
* 'locking-urgent-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip:
locking/lockdep: Fix hlock->pin_count reset on lock stack rebuilds
gic_handle_shared_int reads the GIC interrupt pending & mask registers
directly into a bitmap, which is defined as an array of unsigned longs.
The GIC pending registers may be 32 bits wide if the CM is older than
CM3, regardless of the bit width of the CPU, but for MIPS64 kernels
the unsigned longs in the bitmap will be 64 bits wide. In this case we
need to perform 2 x 32 bit reads per 64 bit unsigned long in order to
avoid missing interrupts.
Signed-off-by: Paul Burton <paul.burton@imgtec.com>
Acked-by: Thomas Gleixner <tglx@linutronix.de>
Cc: linux-mips@linux-mips.org
Cc: Marc Zyngier <marc.zyngier@arm.com>
Cc: Jason Cooper <jason@lakedaemon.net>
Cc: linux-kernel@vger.kernel.org
Patchwork: https://patchwork.linux-mips.org/patch/11213/
Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
The VP ID of a given CPU may not match up with the CPU number used by
Linux. For example, if the width of the VP part of the VP ID is wider
than log2(number of VPs per core) and the system has multiple cores then
this will be the case. Alternatively, if a pre-r6 system implements the
MT ASE with multiple VPEs per core and Linux is built without support
for the MT ASE then the numbers won't match up either. Provide a
function to convert from CPU number to VP ID.
Signed-off-by: Paul Burton <paul.burton@imgtec.com>
Acked-by: Thomas Gleixner <tglx@linutronix.de>
Cc: linux-mips@linux-mips.org
Cc: linux-kernel@vger.kernel.org
Cc: James Hogan <james.hogan@imgtec.com>
Cc: Markos Chandras <markos.chandras@imgtec.com>
Patchwork: https://patchwork.linux-mips.org/patch/11211/
Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
Pull x86 fixes from Thomas Gleixner:
"Two bugfixes from Andy addressing at least some of the subtle NMI
related wreckage which has been reported by Sasha Levin"
* 'x86-urgent-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip:
x86/nmi/64: Fix a paravirt stack-clobbering bug in the NMI code
x86/paravirt: Replace the paravirt nop with a bona fide empty function
Pull irq fix from Thomass Gleixner:
"A bugfix for the atmel aic5 irq chip driver which caches the wrong
data and thereby breaking resume"
* 'irq-urgent-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip:
irqchip/atmel-aic5: Use per chip mask caches in mask/unmask()
