| Age | Commit message (Collapse) | Author | Files | Lines |
|---|
|
In general a read fills 4kb so filling the buffer is a 1 in 4096
operation, move it out of the io__get_char function to avoid some
checking overhead and to better hint the function is good to inline.
For perf's IO intensive internal (non-rigorous) benchmarks there's a
small improvement to kallsyms-parsing with a default build.
Before:
```
$ perf bench internals all
Computing performance of single threaded perf event synthesis by
synthesizing events on the perf process itself:
Average synthesis took: 146.322 usec (+- 0.305 usec)
Average num. events: 61.000 (+- 0.000)
Average time per event 2.399 usec
Average data synthesis took: 145.056 usec (+- 0.155 usec)
Average num. events: 329.000 (+- 0.000)
Average time per event 0.441 usec
Average kallsyms__parse took: 162.313 ms (+- 0.599 ms)
...
Computing performance of sysfs PMU event scan for 100 times
Average core PMU scanning took: 53.720 usec (+- 7.823 usec)
Average PMU scanning took: 375.145 usec (+- 23.974 usec)
```
After:
```
$ perf bench internals all
Computing performance of single threaded perf event synthesis by
synthesizing events on the perf process itself:
Average synthesis took: 127.829 usec (+- 0.079 usec)
Average num. events: 61.000 (+- 0.000)
Average time per event 2.096 usec
Average data synthesis took: 133.652 usec (+- 0.101 usec)
Average num. events: 327.000 (+- 0.000)
Average time per event 0.409 usec
Average kallsyms__parse took: 150.415 ms (+- 0.313 ms)
...
Computing performance of sysfs PMU event scan for 100 times
Average core PMU scanning took: 47.790 usec (+- 1.178 usec)
Average PMU scanning took: 376.945 usec (+- 23.683 usec)
```
Signed-off-by: Ian Rogers <[email protected]>
Signed-off-by: Namhyung Kim <[email protected]>
Link: https://lore.kernel.org/r/[email protected]
|
|
filename__read_str() has its own string reading code that allocates
memory before reading into it. The memory allocated is sized at BUFSIZ
that is 8kb. Most strings are short and so most of this 8kb is wasted.
Refactor io__getline(), as io__getdelim(), so that the newline character
can be configurable and ignored in the case of filename__read_str().
Code like build_caches_for_cpu() in perf's header.c will read many strings
and hold them in a data structure, in this case multiple strings per
cache level per CPU.
Using io.h's io__getline() avoids the wasted memory as strings are
temporarily read into a buffer on the stack before being copied to a
buffer that grows 128 bytes at a time and is never sized larger than the
string.
For a 16 hyperthread system the memory consumption of "perf record
true" is reduced by 180kb, primarily through saving memory when
reading the cache information.
Signed-off-by: Ian Rogers <[email protected]>
Acked-by: Namhyung Kim <[email protected]>
Cc: Adrian Hunter <[email protected]>
Cc: Alexander Shishkin <[email protected]>
Cc: Andi Kleen <[email protected]>
Cc: Athira Jajeev <[email protected]>
Cc: Changbin Du <[email protected]>
Cc: Colin Ian King <[email protected]>
Cc: Dmitrii Dolgov <[email protected]>
Cc: German Gomez <[email protected]>
Cc: Guilherme Amadio <[email protected]>
Cc: Huacai Chen <[email protected]>
Cc: Ingo Molnar <[email protected]>
Cc: James Clark <[email protected]>
Cc: Jiri Olsa <[email protected]>
Cc: K Prateek Nayak <[email protected]>
Cc: Kajol Jain <[email protected]>
Cc: Kan Liang <[email protected]>
Cc: Leo Yan <[email protected]>
Cc: Li Dong <[email protected]>
Cc: Liam Howlett <[email protected]>
Cc: Mark Rutland <[email protected]>
Cc: Masami Hiramatsu (Google) <[email protected]>
Cc: Miguel Ojeda <[email protected]>
Cc: Ming Wang <[email protected]>
Cc: Nick Terrell <[email protected]>
Cc: Paolo Bonzini <[email protected]>
Cc: Peter Zijlstra <[email protected]>
Cc: Ravi Bangoria <[email protected]>
Cc: Sandipan Das <[email protected]>
Cc: Sean Christopherson <[email protected]>
Cc: Steinar H. Gunderson <[email protected]>
Cc: Vincent Whitchurch <[email protected]>
Cc: Wenyu Liu <[email protected]>
Cc: Yang Jihong <[email protected]>
Link: https://lore.kernel.org/r/[email protected]
Signed-off-by: Arnaldo Carvalho de Melo <[email protected]>
|
|
There are functions using __u64, so we need to have the linux/types.h
header otherwise we'll break when its not included before api/io.h.
Fixes: e95770af4c4a280f ("tools api: Add a lightweight buffered reading api")
Reviewed-by: Ian Rogers <[email protected]>
Cc: Adrian Hunter <[email protected]>
Cc: Jiri Olsa <[email protected]>
Cc: Namhyung Kim <[email protected]>
Link: https://lore.kernel.org/lkml/[email protected]
Signed-off-by: Arnaldo Carvalho de Melo <[email protected]>
|
|
io__getline will free the line on error but it doesn't clear the out
argument. This may lead to the line being freed twice, like in
tools/perf/util/srcline.c as detected by clang-tidy.
Signed-off-by: Ian Rogers <[email protected]>
Acked-by: Namhyung Kim <[email protected]>
Cc: Ravi Bangoria <[email protected]>
Cc: Nick Desaulniers <[email protected]>
Cc: Yang Jihong <[email protected]>
Cc: Huacai Chen <[email protected]>
Cc: Nathan Chancellor <[email protected]>
Cc: Kan Liang <[email protected]>
Cc: [email protected]
Cc: Ming Wang <[email protected]>
Cc: Tom Rix <[email protected]>
Cc: [email protected]
Link: https://lore.kernel.org/r/[email protected]
Signed-off-by: Namhyung Kim <[email protected]>
|
|
In situations like reading from a pipe it can be useful to have a
timeout so that the caller doesn't block indefinitely. Implement a
simple one based on poll.
Signed-off-by: Ian Rogers <[email protected]>
Cc: Adrian Hunter <[email protected]>
Cc: Alexander Shishkin <[email protected]>
Cc: Jiri Olsa <[email protected]>
Cc: Mark Rutland <[email protected]>
Cc: Namhyung Kim <[email protected]>
Cc: Peter Zijlstra <[email protected]>
Cc: Yang Jihong <[email protected]>
Link: http://lore.kernel.org/lkml/[email protected]
Signed-off-by: Arnaldo Carvalho de Melo <[email protected]>
|
|
Reads a line to allocated memory up to a newline following the getline
API.
Committer notes:
It also adds this new function to the 'api io' 'perf test' entry:
$ perf test "api io"
64: Test api io : Ok
$
Signed-off-by: Ian Rogers <[email protected]>
Tested-by: Arnaldo Carvalho de Melo <[email protected]>
Acked-by: Namhyung Kim <[email protected]>
Cc: Adrian Hunter <[email protected]>
Cc: Alexander Shishkin <[email protected]>
Cc: Ingo Molnar <[email protected]>
Cc: Jiri Olsa <[email protected]>
Cc: Mark Rutland <[email protected]>
Cc: Nathan Chancellor <[email protected]>
Cc: Nick Desaulniers <[email protected]>
Cc: Peter Zijlstra <[email protected]>
Cc: Tom Rix <[email protected]>
Cc: [email protected]
Link: https://lore.kernel.org/r/[email protected]
Signed-off-by: Arnaldo Carvalho de Melo <[email protected]>
|
|
'perf record' will call kallsyms__parse 4 times during startup and
process megabytes of data. This changes kallsyms__parse to use the io
library rather than fgets to improve performance of the user code by
over 8%.
Before:
Running 'internals/kallsyms-parse' benchmark:
Average kallsyms__parse took: 103.988 ms (+- 0.203 ms)
After:
Running 'internals/kallsyms-parse' benchmark:
Average kallsyms__parse took: 95.571 ms (+- 0.006 ms)
For a workload like:
$ perf record /bin/true
Run under 'perf record -e cycles:u -g' the time goes from:
Before
30.10% 1.67% perf perf [.] kallsyms__parse
After
25.55% 20.04% perf perf [.] kallsyms__parse
So a little under 5% of the start-up time is removed. A lot of what
remains is on the kernel side, but caching kallsyms within perf would at
least impact memory footprint.
Committer notes:
The internal/kallsyms-parse bench is run using:
[root@five ~]# perf bench internals kallsyms-parse
# Running 'internals/kallsyms-parse' benchmark:
Average kallsyms__parse took: 80.381 ms (+- 0.115 ms)
[root@five ~]#
And this pre-existing test uses these routines to parse kallsyms and
then compare with the info obtained from the matching ELF symtab:
[root@five ~]# perf test vmlinux
1: vmlinux symtab matches kallsyms : Ok
[root@five ~]#
Also we can't remove hex2u64() in this patch as this breaks the build:
/usr/bin/ld: /tmp/build/perf/perf-in.o: in function `modules__parse':
/home/acme/git/perf/tools/perf/util/symbol.c:607: undefined reference to `hex2u64'
/usr/bin/ld: /home/acme/git/perf/tools/perf/util/symbol.c:607: undefined reference to `hex2u64'
/usr/bin/ld: /tmp/build/perf/perf-in.o: in function `dso__load_perf_map':
/home/acme/git/perf/tools/perf/util/symbol.c:1477: undefined reference to `hex2u64'
/usr/bin/ld: /home/acme/git/perf/tools/perf/util/symbol.c:1483: undefined reference to `hex2u64'
collect2: error: ld returned 1 exit status
Leave it there, move it in the next patch.
Signed-off-by: Ian Rogers <[email protected]>
Tested-by: Arnaldo Carvalho de Melo <[email protected]>
Cc: Alexander Shishkin <[email protected]>
Cc: Jiri Olsa <[email protected]>
Cc: Mark Rutland <[email protected]>
Cc: Namhyung Kim <[email protected]>
Cc: Peter Zijlstra <[email protected]>
Cc: Stephane Eranian <[email protected]>
Cc: Thomas Gleixner <[email protected]>
Link: http://lore.kernel.org/lkml/[email protected]
Signed-off-by: Arnaldo Carvalho de Melo <[email protected]>
|
|
The synthesize benchmark shows the majority of execution time going to
fgets and sscanf, necessary to parse /proc/pid/maps. Add a new buffered
reading library that will be used to replace these calls in a follow-up
CL. Add tests for the library to perf test.
Committer tests:
$ perf test api
63: Test api io : Ok
$
Signed-off-by: Ian Rogers <[email protected]>
Acked-by: Jiri Olsa <[email protected]>
Acked-by: Namhyung Kim <[email protected]>
Tested-by: Arnaldo Carvalho de Melo <[email protected]>
Cc: Alexander Shishkin <[email protected]>
Cc: Andrey Zhizhikin <[email protected]>
Cc: Jiri Olsa <[email protected]>
Cc: Kan Liang <[email protected]>
Cc: Kefeng Wang <[email protected]>
Cc: Mark Rutland <[email protected]>
Cc: Peter Zijlstra <[email protected]>
Cc: Petr Mladek <[email protected]>
Cc: Stephane Eranian <[email protected]>
Cc: Thomas Gleixner <[email protected]>
Link: http://lore.kernel.org/lkml/[email protected]
Signed-off-by: Arnaldo Carvalho de Melo <[email protected]>
|