lkml.org 
[lkml]   [2021]   [Jan]   [16]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
 
Messages in this thread
/
From
Date
SubjectRe: [PATCH v5] pgo: add clang's Profile Guided Optimization infrastructure
On Sat, Jan 16, 2021 at 9:39 AM Sedat Dilek <sedat.dilek@gmail.com> wrote:
> On Sat, Jan 16, 2021 at 10:44 AM 'Bill Wendling' via Clang Built Linux
> <clang-built-linux@googlegroups.com> wrote:
> >
> > From: Sami Tolvanen <samitolvanen@google.com>
> >
> > Enable the use of clang's Profile-Guided Optimization[1]. To generate a
> > profile, the kernel is instrumented with PGO counters, a representative
> > workload is run, and the raw profile data is collected from
> > /sys/kernel/debug/pgo/profraw.
> >
> > The raw profile data must be processed by clang's "llvm-profdata" tool
> > before it can be used during recompilation:
> >
> > $ cp /sys/kernel/debug/pgo/profraw vmlinux.profraw
> > $ llvm-profdata merge --output=vmlinux.profdata vmlinux.profraw
> >
> > Multiple raw profiles may be merged during this step.
> >
> > The data can now be used by the compiler:
> >
> > $ make LLVM=1 KCFLAGS=-fprofile-use=vmlinux.profdata ...
> >
> > This initial submission is restricted to x86, as that's the platform we
> > know works. This restriction can be lifted once other platforms have
> > been verified to work with PGO.
> >
> > Note that this method of profiling the kernel is clang-native, unlike
> > the clang support in kernel/gcov.
> >
> > [1] https://clang.llvm.org/docs/UsersManual.html#profile-guided-optimization
> >
> > Signed-off-by: Sami Tolvanen <samitolvanen@google.com>
> > Co-developed-by: Bill Wendling <morbo@google.com>
> > Signed-off-by: Bill Wendling <morbo@google.com>
> > ---
> > v2: - Added "__llvm_profile_instrument_memop" based on Nathan Chancellor's
> > testing.
> > - Corrected documentation, re PGO flags when using LTO, based on Fangrui
> > Song's comments.
> > v3: - Added change log section based on Sedat Dilek's comments.
> > v4: - Remove non-x86 Makfile changes and se "hweight64" instead of using our
> > own popcount implementation, based on Nick Desaulniers's comment.
> > v5: - Correct padding calculation, discovered by Nathan Chancellor.
> > ---
> > Documentation/dev-tools/index.rst | 1 +
> > Documentation/dev-tools/pgo.rst | 127 +++++++++
> > MAINTAINERS | 9 +
> > Makefile | 3 +
> > arch/Kconfig | 1 +
> > arch/x86/Kconfig | 1 +
> > arch/x86/boot/Makefile | 1 +
> > arch/x86/boot/compressed/Makefile | 1 +
> > arch/x86/crypto/Makefile | 2 +
> > arch/x86/entry/vdso/Makefile | 1 +
> > arch/x86/kernel/vmlinux.lds.S | 2 +
> > arch/x86/platform/efi/Makefile | 1 +
> > arch/x86/purgatory/Makefile | 1 +
> > arch/x86/realmode/rm/Makefile | 1 +
> > arch/x86/um/vdso/Makefile | 1 +
> > drivers/firmware/efi/libstub/Makefile | 1 +
> > include/asm-generic/vmlinux.lds.h | 44 +++
> > kernel/Makefile | 1 +
> > kernel/pgo/Kconfig | 35 +++
> > kernel/pgo/Makefile | 5 +
> > kernel/pgo/fs.c | 382 ++++++++++++++++++++++++++
> > kernel/pgo/instrument.c | 185 +++++++++++++
> > kernel/pgo/pgo.h | 206 ++++++++++++++
> > scripts/Makefile.lib | 10 +
> > 24 files changed, 1022 insertions(+)
> > create mode 100644 Documentation/dev-tools/pgo.rst
> > create mode 100644 kernel/pgo/Kconfig
> > create mode 100644 kernel/pgo/Makefile
> > create mode 100644 kernel/pgo/fs.c
> > create mode 100644 kernel/pgo/instrument.c
> > create mode 100644 kernel/pgo/pgo.h
> >
> > diff --git a/Documentation/dev-tools/index.rst b/Documentation/dev-tools/index.rst
> > index f7809c7b1ba9e..8d6418e858062 100644
> > --- a/Documentation/dev-tools/index.rst
> > +++ b/Documentation/dev-tools/index.rst
> > @@ -26,6 +26,7 @@ whole; patches welcome!
> > kgdb
> > kselftest
> > kunit/index
> > + pgo
> >
> >
> > .. only:: subproject and html
> > diff --git a/Documentation/dev-tools/pgo.rst b/Documentation/dev-tools/pgo.rst
> > new file mode 100644
> > index 0000000000000..b7f11d8405b73
> > --- /dev/null
> > +++ b/Documentation/dev-tools/pgo.rst
> > @@ -0,0 +1,127 @@
> > +.. SPDX-License-Identifier: GPL-2.0
> > +
> > +===============================
> > +Using PGO with the Linux kernel
> > +===============================
> > +
> > +Clang's profiling kernel support (PGO_) enables profiling of the Linux kernel
> > +when building with Clang. The profiling data is exported via the ``pgo``
> > +debugfs directory.
> > +
> > +.. _PGO: https://clang.llvm.org/docs/UsersManual.html#profile-guided-optimization
> > +
> > +
> > +Preparation
> > +===========
> > +
> > +Configure the kernel with:
> > +
> > +.. code-block:: make
> > +
> > + CONFIG_DEBUG_FS=y
> > + CONFIG_PGO_CLANG=y
> > +
> > +Note that kernels compiled with profiling flags will be significantly larger
> > +and run slower.
> > +
> > +Profiling data will only become accessible once debugfs has been mounted:
> > +
> > +.. code-block:: sh
> > +
> > + mount -t debugfs none /sys/kernel/debug
> > +
> > +
> > +Customization
> > +=============
> > +
> > +You can enable or disable profiling for individual file and directories by
> > +adding a line similar to the following to the respective kernel Makefile:
> > +
> > +- For a single file (e.g. main.o)
> > +
> > + .. code-block:: make
> > +
> > + PGO_PROFILE_main.o := y
> > +
> > +- For all files in one directory
> > +
> > + .. code-block:: make
> > +
> > + PGO_PROFILE := y
> > +
> > +To exclude files from being profiled use
> > +
> > + .. code-block:: make
> > +
> > + PGO_PROFILE_main.o := n
> > +
> > +and
> > +
> > + .. code-block:: make
> > +
> > + PGO_PROFILE := n
> > +
> > +Only files which are linked to the main kernel image or are compiled as kernel
> > +modules are supported by this mechanism.
> > +
> > +
> > +Files
> > +=====
> > +
> > +The PGO kernel support creates the following files in debugfs:
> > +
> > +``/sys/kernel/debug/pgo``
> > + Parent directory for all PGO-related files.
> > +
> > +``/sys/kernel/debug/pgo/reset``
> > + Global reset file: resets all coverage data to zero when written to.
> > +
> > +``/sys/kernel/debug/profraw``
> > + The raw PGO data that must be processed with ``llvm_profdata``.
> > +
> > +
> > +Workflow
> > +========
> > +
> > +The PGO kernel can be run on the host or test machines. The data though should
> > +be analyzed with Clang's tools from the same Clang version as the kernel was
> > +compiled. Clang's tolerant of version skew, but it's easier to use the same
> > +Clang version.
> > +
> > +The profiling data is useful for optimizing the kernel, analyzing coverage,
> > +etc. Clang offers tools to perform these tasks.
> > +
> > +Here is an example workflow for profiling an instrumented kernel with PGO and
> > +using the result to optimize the kernel:
> > +
> > +1) Install the kernel on the TEST machine.
> > +
> > +2) Reset the data counters right before running the load tests
> > +
> > + .. code-block:: sh
> > +
> > + $ echo 1 > /sys/kernel/debug/pgo/reset
> > +
>
> I do not get this...
>
> # mount | grep debugfs
> debugfs on /sys/kernel/debug type debugfs (rw,nosuid,nodev,noexec,relatime)
>
> After the load-test...?
>
> echo 0 > /sys/kernel/debug/pgo/reset
>
Writing anything to /sys/kernel/debug/pgo/reset will cause it to reset
the profiling counters. I picked 1 (one) semi-randomly, but it could
be any number, letter, your favorite short story, etc. You don't want
to reset it before collecting the profiling data from your load tests
though.

> > +3) Run the load tests.
> > +
> > +4) Collect the raw profile data
> > +
> > + .. code-block:: sh
> > +
> > + $ cp -a /sys/kernel/debug/pgo/profraw /tmp/vmlinux.profraw
> > +
>
> This is only 4,9M small and seen from the date 5mins before I run the
> echo-1 line.
>
> # ll /sys/kernel/debug/pgo
> insgesamt 0
> drwxr-xr-x 2 root root 0 16. Jan 17:29 .
> drwx------ 41 root root 0 16. Jan 17:29 ..
> -rw------- 1 root root 0 16. Jan 17:29 profraw
> --w------- 1 root root 0 16. Jan 18:19 reset
>
> # cp -a /sys/kernel/debug/pgo/profraw /tmp/vmlinux.profraw
>
> # ll /tmp/vmlinux.profraw
> -rw------- 1 root root 4,9M 16. Jan 17:29 /tmp/vmlinux.profraw
>
> For me there was no prof-data collected from my defconfig kernel-build.
>
The /sys/kernel/debug/pgo/profraw file is read-only. Nothing writes to
it, not even the kernel. All it does is serialize the profiling
counters from a memory location in the kernel into a format that
LLVM's tools can understand.

> > +5) (Optional) Download the raw profile data to the HOST machine.
> > +
> > +6) Process the raw profile data
> > +
> > + .. code-block:: sh
> > +
> > + $ llvm-profdata merge --output=vmlinux.profdata vmlinux.profraw
> > +
>
> Is that executed in /path/to/linux/git?
>
The llvm-profdata tool is not in the linux source tree. You need to
grab it from a clang distribution (or built from clang's git repo).

> > + Note that multiple raw profile data files can be merged during this step.
> > +
> > +7) Rebuild the kernel using the profile data (PGO disabled)
> > +
> > + .. code-block:: sh
> > +
> > + $ make LLVM=1 KCFLAGS=-fprofile-use=vmlinux.profdata ...
>
> How big is vmlinux.profdata (make defconfig)?
>
I don't have numbers for this, but from what you listed here, it's ~5M
in size. The size is proportional to the number of counters
instrumented in the kernel.

> Do I need to do a full defconfig build or can I stop the build after
> let me say 10mins?
>
You should do a full rebuild. Make sure that PGO is disabled during the rebuild.

-bw

\
 
 \ /
  Last update: 2021-01-16 21:25    [W:0.597 / U:0.096 seconds]
©2003-2020 Jasper Spaans|hosted at Digital Ocean and TransIP|Read the blog|Advertise on this site