| lists.openwall.net | lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC | |
|
Open Source and information security mailing list archives
| ||
|
Message-ID: <CAP-5=fXME3cjNK-P9qk+kY0dA1Xkwvz6Su=99nY7CgcKrQTb2w@mail.gmail.com>
Date: Thu, 23 May 2024 21:47:02 -0700
From: Ian Rogers <irogers@...gle.com>
To: Namhyung Kim <namhyung@...nel.org>
Cc: Arnaldo Carvalho de Melo <acme@...hat.com>, linux-kernel@...r.kernel.org,
linux-perf-users@...r.kernel.org
Subject: Re: [PATCH v1] tools api io: Move filling the io buffer to its own function
On Thu, May 23, 2024 at 4:25 PM Namhyung Kim <namhyung@...nel.org> wrote:
>
> On Sun, May 19, 2024 at 11:17 AM Ian Rogers <irogers@...gle.com> wrote:
> >
> > 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
> > near 8% improvement to kallsyms-parsing with a default build.
>
> Oh, is it just from removing the io->eof check? Otherwise I don't
> see any difference.
I was hoping that by moving the code out-of-line then the hot part of
the function could be inlined into things like reading the hex
character. I didn't see that, presumably there are too many callers
and so that made the inliner think sharing would be best even though
the hot code is a compare, pointer dereference and an increment. I
tried forcing inlining but it didn't seem to win over just having the
code out-of-line. The eof check should be very well predicted. The
out-of-line code was branched over forward, which should be 1
mispredict but again not a huge deal. I didn't do a more thorough
analysis as I still prefer to have the cold code out-of-line.
Thanks,
Ian
> Thanks,
> Namhyung
>
> >
> > 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 <irogers@...gle.com>
> > ---
> > tools/lib/api/io.h | 69 +++++++++++++++++++++++++---------------------
> > 1 file changed, 38 insertions(+), 31 deletions(-)
> >
> > diff --git a/tools/lib/api/io.h b/tools/lib/api/io.h
> > index 84adf8102018..d3eb04d1bc89 100644
> > --- a/tools/lib/api/io.h
> > +++ b/tools/lib/api/io.h
> > @@ -43,48 +43,55 @@ static inline void io__init(struct io *io, int fd,
> > io->eof = false;
> > }
> >
> > -/* Reads one character from the "io" file with similar semantics to fgetc. */
> > -static inline int io__get_char(struct io *io)
> > +/* Read from fd filling the buffer. Called when io->data == io->end. */
> > +static inline int io__fill_buffer(struct io *io)
> > {
> > - char *ptr = io->data;
> > + ssize_t n;
> >
> > if (io->eof)
> > return -1;
> >
> > - if (ptr == io->end) {
> > - ssize_t n;
> > -
> > - if (io->timeout_ms != 0) {
> > - struct pollfd pfds[] = {
> > - {
> > - .fd = io->fd,
> > - .events = POLLIN,
> > - },
> > - };
> > -
> > - n = poll(pfds, 1, io->timeout_ms);
> > - if (n == 0)
> > - errno = ETIMEDOUT;
> > - if (n > 0 && !(pfds[0].revents & POLLIN)) {
> > - errno = EIO;
> > - n = -1;
> > - }
> > - if (n <= 0) {
> > - io->eof = true;
> > - return -1;
> > - }
> > + if (io->timeout_ms != 0) {
> > + struct pollfd pfds[] = {
> > + {
> > + .fd = io->fd,
> > + .events = POLLIN,
> > + },
> > + };
> > +
> > + n = poll(pfds, 1, io->timeout_ms);
> > + if (n == 0)
> > + errno = ETIMEDOUT;
> > + if (n > 0 && !(pfds[0].revents & POLLIN)) {
> > + errno = EIO;
> > + n = -1;
> > }
> > - n = read(io->fd, io->buf, io->buf_len);
> > -
> > if (n <= 0) {
> > io->eof = true;
> > return -1;
> > }
> > - ptr = &io->buf[0];
> > - io->end = &io->buf[n];
> > }
> > - io->data = ptr + 1;
> > - return *ptr;
> > + n = read(io->fd, io->buf, io->buf_len);
> > +
> > + if (n <= 0) {
> > + io->eof = true;
> > + return -1;
> > + }
> > + io->data = &io->buf[0];
> > + io->end = &io->buf[n];
> > + return 0;
> > +}
> > +
> > +/* Reads one character from the "io" file with similar semantics to fgetc. */
> > +static inline int io__get_char(struct io *io)
> > +{
> > + if (io->data == io->end) {
> > + int ret = io__fill_buffer(io);
> > +
> > + if (ret)
> > + return ret;
> > + }
> > + return *io->data++;
> > }
> >
> > /* Read a hexadecimal value with no 0x prefix into the out argument hex. If the
> > --
> > 2.45.0.rc1.225.g2a3ae87e7f-goog
> >
Powered by blists - more mailing lists