Message-ID: <CA+EESO5P1P4_Mb_7q0E9Y9uv6f9wK4kTALqCOKsc36k+E4p-5Q@mail.gmail.com>
Date:   Thu, 1 Oct 2020 23:39:53 -0700
From:   Lokesh Gidra <lokeshgidra@...gle.com>
To:     "Kirill A. Shutemov" <kirill.shutemov@...ux.intel.com>
Cc:     Kalesh Singh <kaleshsingh@...gle.com>,
        Suren Baghdasaryan <surenb@...gle.com>,
        Minchan Kim <minchan@...gle.com>,
        Joel Fernandes <joelaf@...gle.com>,
        "Cc: Android Kernel" <kernel-team@...roid.com>,
        Catalin Marinas <catalin.marinas@....com>,
        Will Deacon <will@...nel.org>,
        Thomas Gleixner <tglx@...utronix.de>,
        Ingo Molnar <mingo@...hat.com>, Borislav Petkov <bp@...en8.de>,
        "the arch/x86 maintainers" <x86@...nel.org>,
        "H. Peter Anvin" <hpa@...or.com>,
        Andrew Morton <akpm@...ux-foundation.org>,
        Shuah Khan <shuah@...nel.org>,
        "Aneesh Kumar K.V" <aneesh.kumar@...ux.ibm.com>,
        Kees Cook <keescook@...omium.org>,
        Peter Zijlstra <peterz@...radead.org>,
        Sami Tolvanen <samitolvanen@...gle.com>,
        Masahiro Yamada <masahiroy@...nel.org>,
        Arnd Bergmann <arnd@...db.de>,
        Frederic Weisbecker <frederic@...nel.org>,
        Krzysztof Kozlowski <krzk@...nel.org>,
        Hassan Naveed <hnaveed@...ecomp.com>,
        Christian Brauner <christian.brauner@...ntu.com>,
        Mark Rutland <mark.rutland@....com>,
        Mike Rapoport <rppt@...nel.org>, Gavin Shan <gshan@...hat.com>,
        Zhenyu Ye <yezhenyu2@...wei.com>, Jia He <justin.he@....com>,
        John Hubbard <jhubbard@...dia.com>,
        William Kucharski <william.kucharski@...cle.com>,
        Sandipan Das <sandipan@...ux.ibm.com>,
        Ralph Campbell <rcampbell@...dia.com>,
        Mina Almasry <almasrymina@...gle.com>,
        Ram Pai <linuxram@...ibm.com>,
        Dave Hansen <dave.hansen@...el.com>,
        Kamalesh Babulal <kamalesh@...ux.vnet.ibm.com>,
        Masami Hiramatsu <mhiramat@...nel.org>,
        Brian Geffon <bgeffon@...gle.com>,
        SeongJae Park <sjpark@...zon.de>,
        linux-kernel <linux-kernel@...r.kernel.org>,
        "moderated list:ARM64 PORT (AARCH64 ARCHITECTURE)" 
        <linux-arm-kernel@...ts.infradead.org>,
        "open list:MEMORY MANAGEMENT" <linux-mm@...ck.org>,
        "open list:KERNEL SELFTEST FRAMEWORK" 
        <linux-kselftest@...r.kernel.org>
Subject: Re: [PATCH 0/5] Speed up mremap on large regions

On Thu, Oct 1, 2020 at 10:36 PM Kirill A. Shutemov
<kirill.shutemov@...ux.intel.com> wrote:
>
> On Thu, Oct 01, 2020 at 05:09:02PM -0700, Lokesh Gidra wrote:
> > On Thu, Oct 1, 2020 at 9:00 AM Kalesh Singh <kaleshsingh@...gle.com> wrote:
> > >
> > > On Thu, Oct 1, 2020 at 8:27 AM Kirill A. Shutemov
> > > <kirill.shutemov@...ux.intel.com> wrote:
> > > >
> > > > On Wed, Sep 30, 2020 at 03:42:17PM -0700, Lokesh Gidra wrote:
> > > > > On Wed, Sep 30, 2020 at 3:32 PM Kirill A. Shutemov
> > > > > <kirill.shutemov@...ux.intel.com> wrote:
> > > > > >
> > > > > > On Wed, Sep 30, 2020 at 10:21:17PM +0000, Kalesh Singh wrote:
> > > > > > > mremap time can be optimized by moving entries at the PMD/PUD level if
> > > > > > > the source and destination addresses are PMD/PUD-aligned and
> > > > > > > PMD/PUD-sized. Enable moving at the PMD and PUD levels on arm64 and
> > > > > > > x86. Other architectures where this type of move is supported and known to
> > > > > > > be safe can also opt-in to these optimizations by enabling HAVE_MOVE_PMD
> > > > > > > and HAVE_MOVE_PUD.
> > > > > > >
> > > > > > > Observed Performance Improvements for remapping a PUD-aligned 1GB-sized
> > > > > > > region on x86 and arm64:
> > > > > > >
> > > > > > >     - HAVE_MOVE_PMD is already enabled on x86 : N/A
> > > > > > >     - Enabling HAVE_MOVE_PUD on x86   : ~13x speed up
> > > > > > >
> > > > > > >     - Enabling HAVE_MOVE_PMD on arm64 : ~ 8x speed up
> > > > > > >     - Enabling HAVE_MOVE_PUD on arm64 : ~19x speed up
> > > > > > >
> > > > > > >           Altogether, HAVE_MOVE_PMD and HAVE_MOVE_PUD
> > > > > > >           give a total of ~150x speed up on arm64.
> > > > > >
> > > > > > Is there a *real* workload that benefit from HAVE_MOVE_PUD?
> > > > > >
> > > > > We have a Java garbage collector under development which requires
> > > > > moving physical pages of multi-gigabyte heap using mremap. During this
> > > > > move, the application threads have to be paused for correctness. It is
> > > > > critical to keep this pause as short as possible to avoid jitters
> > > > > during user interaction. This is where HAVE_MOVE_PUD will greatly
> > > > > help.
> > > >
> > > > Any chance to quantify the effect of mremap() with and without
> > > > HAVE_MOVE_PUD?
> > > >
> > > > I doubt it's a major contributor to the GC pause. I expect you need to
> > > > move tens of gigs to get sizable effect. And if your GC routinely moves
> > > > tens of gigs, maybe problem somewhere else?
> > > >
> > > > I'm asking for numbers, because increase in complexity comes with cost.
> > > > If it doesn't provide an substantial benefit to a real workload
> > > > maintaining the code forever doesn't make sense.
> > >
> > mremap is indeed the biggest contributor to the GC pause. It has to
> > take place in what is typically known as a 'stop-the-world' pause,
> > wherein all application threads are paused. During this pause the GC
> > thread flips the GC roots (threads' stacks, globals etc.), and then
> > resumes threads along with concurrent compaction of the heap.This
> > GC-root flip differs depending on which compaction algorithm is being
> > used.
> >
> > In our case it involves updating object references in threads' stacks
> > and remapping java heap to a different location. The threads' stacks
> > can be handled in parallel with the mremap. Therefore, the dominant
> > factor is indeed the cost of mremap. From patches 2 and 4, it is clear
> > that remapping 1GB without this optimization will take ~9ms on arm64.
> >
> > Although this mremap has to happen only once every GC cycle, and the
> > typical size is also not going to be more than a GB or 2, pausing
> > application threads for ~9ms is guaranteed to cause jitters. OTOH,
> > with this optimization, mremap is reduced to ~60us, which is a totally
> > acceptable pause time.
> >
> > Unfortunately, implementation of the new GC algorithm hasn't yet
> > reached the point where I can quantify the effect of this
> > optimization. But I can confirm that without this optimization the new
> > GC will not be approved.
>
> IIUC, the 9ms -> 90us improvement attributed to combination HAVE_MOVE_PMD
> and HAVE_MOVE_PUD, right? I expect HAVE_MOVE_PMD to be reasonable for some
> workloads, but marginal benefit of HAVE_MOVE_PUD is in doubt. Do you see
> it's useful for your workload?
>
Yes, 9ms -> 90us is when both are combined. The past experience has
been that even ~1ms long stop-the-world pause is prone to cause
jitters. HAVE_MOVE_PMD takes us only this far. So HAVE_MOVE_PUD is
required to bring the mremap cost to acceptable level.

Ideally, I was hoping that the functionality of HAVE_MOVE_PMD can be
extended to all levels of the hierarchical page table, and in the
process simplify the implementation. But unfortunately, that doesn't
seem to be possible from patch 3.

> --
>  Kirill A. Shutemov

Powered by blists - more mailing lists