Received-SPF: pass (google.com: best guess record for domain of linux-kernel-owner@vger.kernel.org designates 209.132.180.67 as permitted sender) client-ip=209.132.180.67;
MIME-Version: 1.0
References: <20200128021145.36774-1-palmerdabbelt@google.com>
In-Reply-To: <20200128021145.36774-1-palmerdabbelt@google.com>
From:   =?UTF-8?B?QmrDtnJuIFTDtnBlbA==?= <bjorn.topel@gmail.com>
Date:   Tue, 4 Feb 2020 20:30:23 +0100
Message-ID: <CAJ+HfNh2csyH2xZtGFXW1zwBEW4+bo_E60PWPydJkB6zZTVx3A@mail.gmail.com>
Subject: Re: arm64: bpf: Elide some moves to a0 after calls
To:     Palmer Dabbelt <palmerdabbelt@google.com>
Cc:     Daniel Borkmann <daniel@iogearbox.net>,
        Alexei Starovoitov <ast@kernel.org>, zlim.lnx@gmail.com,
        catalin.marinas@arm.com, will@kernel.org,
        Martin KaFai Lau <kafai@fb.com>,
        Song Liu <songliubraving@fb.com>, Yonghong Song <yhs@fb.com>,
        Andrii Nakryiko <andriin@fb.com>,
        Shuah Khan <shuah@kernel.org>, Netdev <netdev@vger.kernel.org>,
        bpf <bpf@vger.kernel.org>, linux-arm-kernel@lists.infradead.org,
        LKML <linux-kernel@vger.kernel.org>,
        linux-kselftest@vger.kernel.org,
        clang-built-linux@googlegroups.com, kernel-team@android.com
Content-Type: text/plain; charset="UTF-8"
Content-Transfer-Encoding: quoted-printable
Sender: linux-kernel-owner@vger.kernel.org
Precedence: bulk

On Tue, 28 Jan 2020 at 03:14, Palmer Dabbelt <palmerdabbelt@google.com> wro=
te:
>
> There's four patches here, but only one of them actually does anything.  =
The
> first patch fixes a BPF selftests build failure on my machine and has alr=
eady
> been sent to the list separately.  The next three are just staged such th=
at
> there are some patches that avoid changing any functionality pulled out f=
rom
> the whole point of those refactorings, with two cleanups and then the ide=
a.
>
> Maybe this is an odd thing to say in a cover letter, but I'm not actually=
 sure
> this patch set is a good idea.  The issue of extra moves after calls came=
 up as
> I was reviewing some unrelated performance optimizations to the RISC-V BP=
F JIT.
> I figured I'd take a whack at performing the optimization in the context =
of the
> arm64 port just to get a breath of fresh air, and I'm not convinced I lik=
e the
> results.
>
> That said, I think I would accept something like this for the RISC-V port
> because we're already doing a multi-pass optimization for shrinking funct=
ion
> addresses so it's not as much extra complexity over there.  If we do that=
 we
> should probably start puling some of this code into the shared BPF compil=
er,
> but we're also opening the doors to more complicated BPF JIT optimization=
s.
> Given that the BPF JIT appears to have been designed explicitly to be
> simple/fast as opposed to perform complex optimization, I'm not sure this=
 is a
> sane way to move forward.
>

Obviously I can only speak for myself and the RISC-V JIT, but given
that we already have opened the door for more advanced translations
(branch relaxation e.g.), I think that this makes sense. At the same
time we don't want to go all JVM on the JITs. :-P

> I figured I'd send the patch set out as more of a question than anything =
else.
> Specifically:
>
> * How should I go about measuring the performance of these sort of
>   optimizations?  I'd like to balance the time it takes to run the JIT wi=
th the
>   time spent executing the program, but I don't have any feel for what re=
al BPF
>   programs look like or have any benchmark suite to run.  Is there someth=
ing
>   out there this should be benchmarked against?  (I'd also like to know t=
hat to
>   run those benchmarks on the RISC-V port.)

If you run the selftests 'test_progs' with -v it'll measure/print the
execution time of the programs. I'd say *most* BPF program invokes a
helper (via call). It would be interesting to see, for say the
selftests, how often the optimization can be performed.

> * Is this the sort of thing that makes sense in a BPF JIT?  I guess I've =
just
>   realized I turned "review this patch" into a way bigger rabbit hole tha=
n I
>   really want to go down...
>

I'd say 'yes'. My hunch, and the workloads I've seen, BPF programs are
usually loaded, and then resident for a long time. So, the JIT time is
not super critical. The FB/Cilium folks can definitely provide a
better sample point, than my hunch. ;-)


Bj=C3=B6rn