From: "Rong, Leyi" <leyi.rong@intel.com>
To: David Marchand <david.marchand@redhat.com>
Cc: "ferruh.yigit@xilinx.com" <ferruh.yigit@xilinx.com>,
"suanmingm@nvidia.com" <suanmingm@nvidia.com>,
"Wang, Yipeng1" <yipeng1.wang@intel.com>,
"zaoxingliu@gmail.com" <zaoxingliu@gmail.com>,
"Gobriel, Sameh" <sameh.gobriel@intel.com>,
"dev@dpdk.org" <dev@dpdk.org>
Subject: RE: [PATCH v3 1/2] member: implement NitroSketch mode
Date: Thu, 15 Sep 2022 13:02:14 +0000 [thread overview]
Message-ID: <MW4PR11MB577421F3624F28725635FCC9EC499@MW4PR11MB5774.namprd11.prod.outlook.com> (raw)
In-Reply-To: <CAJFAV8x+6ukkBWSB=BMJo1b2U70-e=cUmm9gitReCR=M17g_Dg@mail.gmail.com>
Hi David,
> -----Original Message-----
> From: David Marchand <david.marchand@redhat.com>
> Sent: Thursday, September 15, 2022 3:47 PM
> To: Rong, Leyi <leyi.rong@intel.com>
> Cc: ferruh.yigit@xilinx.com; suanmingm@nvidia.com; Wang, Yipeng1
> <yipeng1.wang@intel.com>; zaoxingliu@gmail.com; Gobriel, Sameh
> <sameh.gobriel@intel.com>; dev@dpdk.org
> Subject: Re: [PATCH v3 1/2] member: implement NitroSketch mode
>
> On Thu, Sep 15, 2022 at 4:15 AM Leyi Rong <leyi.rong@intel.com> wrote:
> >
> > Sketching algorithm provide high-fidelity approximate measurements and
> > appears as a promising alternative to traditional approaches such as
> > packet sampling.
> >
> > NitroSketch [1] is a software sketching framework that optimizes
> > performance, provides accuracy guarantees, and supports a variety of
> > sketches.
> >
> > This commit adds a new data structure called sketch into membership
> > library. This new data structure is an efficient way to profile the
> > traffic for heavy hitters. Also use min-heap structure to maintain the
> > top-k flow keys.
> >
> > [1] Zaoxing Liu, Ran Ben-Basat, Gil Einziger, Yaron Kassner, Vladimir
> > Braverman, Roy Friedman, Vyas Sekar, "NitroSketch: Robust and General
> > Sketch-based Monitoring in Software Switches", in ACM SIGCOMM 2019.
> > https://dl.acm.org/doi/pdf/10.1145/3341302.3342076
> >
> > Signed-off-by: Alan Liu <zaoxingliu@gmail.com>
> > Signed-off-by: Yipeng Wang <yipeng1.wang@intel.com>
> > Signed-off-by: Leyi Rong <leyi.rong@intel.com>
> > ---
> > lib/member/meson.build | 42 +-
> > lib/member/rte_member.c | 75 ++++
> > lib/member/rte_member.h | 154 ++++++-
> > lib/member/rte_member_heap.h | 424 ++++++++++++++++++
> > lib/member/rte_member_sketch.c | 594 ++++++++++++++++++++++++++
> > lib/member/rte_member_sketch.h | 97 +++++
> > lib/member/rte_member_sketch_avx512.c | 69 +++
> > lib/member/rte_member_sketch_avx512.h | 36 ++
> > lib/member/rte_xxh64_avx512.h | 117 +++++
> > lib/member/version.map | 9 +
> > 10 files changed, 1613 insertions(+), 4 deletions(-) create mode
> > 100644 lib/member/rte_member_heap.h create mode 100644
> > lib/member/rte_member_sketch.c create mode 100644
> > lib/member/rte_member_sketch.h create mode 100644
> > lib/member/rte_member_sketch_avx512.c
> > create mode 100644 lib/member/rte_member_sketch_avx512.h
> > create mode 100644 lib/member/rte_xxh64_avx512.h
> >
> > diff --git a/lib/member/meson.build b/lib/member/meson.build index
> > e06fddc240..8de0c09a6a 100644
> > --- a/lib/member/meson.build
> > +++ b/lib/member/meson.build
> > @@ -7,6 +7,46 @@ if is_windows
> > subdir_done()
> > endif
> >
> > -sources = files('rte_member.c', 'rte_member_ht.c',
> > 'rte_member_vbf.c')
> > +sources = files('rte_member.c', 'rte_member_ht.c',
> > +'rte_member_vbf.c', 'rte_member_sketch.c')
> > headers = files('rte_member.h')
> > deps += ['hash']
> > +includes += include_directories('../hash', '../ring')
> > +
> > +# compile AVX512 version if:
> > +# we are building 64-bit binary AND binutils can generate proper code
> > +if dpdk_conf.has('RTE_ARCH_X86_64') and binutils_ok
> > + # compile AVX512 version if either:
> > + # a. we have AVX512 supported in minimum instruction set
> > + # baseline
> > + # b. it's not minimum instruction set, but supported by
> > + # compiler
> > + #
> > + # in former case, just add avx512 C file to files list
> > + # in latter case, compile c file to static lib, using correct
> > + # compiler flags, and then have the .o file from static lib
> > + # linked into main lib.
> > +
> > + #check if all required flags already enabled
> > + sketch_avx512_flags = ['__AVX512F__', '__AVX512DQ__',
> > + '__AVX512IFMA__']
> > +
> > + sketch_avx512_on = true
> > + foreach f:sketch_avx512_flags
> > + if cc.get_define(f, args: machine_args) == ''
> > + sketch_avx512_on = false
> > + endif
> > + endforeach
> > +
> > + if sketch_avx512_on == true
> > + cflags += ['-DCC_AVX512_SUPPORT']
> > + sources += files('rte_member_sketch_avx512.c')
> > + elif cc.has_multi_arguments('-mavx512f', '-mavx512dq', '-mavx512ifma')
> > + cflags += ['-DCC_AVX512_SUPPORT']
> > + cflags += ['-mavx512f', '-mavx512dq', '-mavx512ifma']
>
> No.
> Again, you can't push AVX512 flags to the variable cflags.
>
> On my laptop, running with Fedora 36:
>
> $ DPDK_TEST=member_autotest ./build-gcc/app/test/dpdk-test --no-huge -m
> 2048
> EAL: Detected CPU lcores: 12
> EAL: Detected NUMA nodes: 1
> EAL: Detected shared linkage of DPDK
> EAL: Multi-process socket /run/user/1000/dpdk/rte/mp_socket
> EAL: Selected IOVA mode 'VA'
> APP: HPET is not enabled, using TSC as default timer
> RTE>>member_autotest
> Expected error section begin...
> rte_member_create_vbf(): Membership vBF create with invalid parameters
> rte_member_create_vbf(): Membership vBF create with invalid parameters
> rte_member_create_vbf(): Membership vBF create with invalid parameters
> rte_member_create_ht(): Membership HT create with invalid parameters
> rte_member_create_ht(): Membership HT create with invalid parameters
> rte_member_create_ht(): Membership HT create with invalid parameters
> Expected error section end...
> rte_member_create_ht(): Hash table based filter created, the table has
> 65536 entries, 4096 buckets
> rte_member_create(): Creating a setsummary table with mode 0
> rte_member_create_ht(): Hash table based filter created, the table has
> 65536 entries, 4096 buckets
> rte_member_create(): Creating a setsummary table with mode 0
> rte_member_create_ht(): Hash table based filter created, the table has
> 65536 entries, 4096 buckets
> rte_member_create(): Creating a setsummary table with mode 0 Illegal
> instruction (core dumped)
>
> The reason is the same as I reported earlier, and this can be reproduced like this:
>
> $ touch lib/member/rte_member_sketch*
> $ ninja -C build-gcc -vv
> ninja: Entering directory `build-gcc'
> [1/9] ccache gcc -Ilib/member/libsketch_avx512_tmp.a.p -Ilib/member -
> I../lib/member -Ilib/hash -I../lib/hash -Ilib/ring -I../lib/ring -I.
> -I.. -Iconfig -I../config -Ilib/eal/include -I../lib/eal/include -Ilib/eal/linux/include -
> I../lib/eal/linux/include -Ilib/eal/x86/include -I../lib/eal/x86/include -
> Ilib/eal/common -I../lib/eal/common -Ilib/eal -I../lib/eal -Ilib/kvargs -
> I../lib/kvargs -Ilib/metrics -I../lib/metrics -Ilib/telemetry -I../lib/telemetry -
> fdiagnostics-color=always -D_FILE_OFFSET_BITS=64 -Wall -Winvalid-pch -
> Wextra -Werror -O3 -g -include rte_config.h -Wcast-qual -Wdeprecated -
> Wformat -Wformat-nonliteral -Wformat-security -Wmissing-declarations -
> Wmissing-prototypes -Wnested-externs -Wold-style-definition -Wpointer-arith -
> Wsign-compare -Wstrict-prototypes -Wundef -Wwrite-strings -Wno-address-of-
> packed-member -Wno-packed-not-aligned -Wno-missing-field-initializers -Wno-
> zero-length-bounds -D_GNU_SOURCE -fPIC -march=native -
> DALLOW_EXPERIMENTAL_API -DALLOW_INTERNAL_API -Wno-format-
> truncation -DCC_AVX512_SUPPORT -mavx512f -mavx512dq -mavx512ifma -MD
> -MQ lib/member/libsketch_avx512_tmp.a.p/rte_member_sketch_avx512.c.o -
> MF lib/member/libsketch_avx512_tmp.a.p/rte_member_sketch_avx512.c.o.d -o
> lib/member/libsketch_avx512_tmp.a.p/rte_member_sketch_avx512.c.o -
> c ../lib/member/rte_member_sketch_avx512.c
> [2/9] ccache gcc -Ilib/librte_member.a.p -Ilib -I../lib -Ilib/hash -I../lib/hash -
> Ilib/ring -I../lib/ring -Ilib/member -I../lib/member -I. -I.. -Iconfig -I../config -
> Ilib/eal/include -I../lib/eal/include -Ilib/eal/linux/include -I../lib/eal/linux/include
> -Ilib/eal/x86/include -I../lib/eal/x86/include -Ilib/eal/common -
> I../lib/eal/common -Ilib/eal -I../lib/eal -Ilib/kvargs -I../lib/kvargs -Ilib/metrics -
> I../lib/metrics -Ilib/telemetry -I../lib/telemetry -Ilib/net -I../lib/net -Ilib/mbuf -
> I../lib/mbuf -Ilib/mempool -I../lib/mempool -Ilib/rcu -I../lib/rcu -fdiagnostics-
> color=always -D_FILE_OFFSET_BITS=64 -Wall -Winvalid-pch -Wextra -Werror -
> O3 -g -include rte_config.h -Wcast-qual -Wdeprecated -Wformat -Wformat-
> nonliteral -Wformat-security -Wmissing-declarations -Wmissing-prototypes -
> Wnested-externs -Wold-style-definition -Wpointer-arith -Wsign-compare -
> Wstrict-prototypes -Wundef -Wwrite-strings -Wno-address-of-packed-member
> -Wno-packed-not-aligned -Wno-missing-field-initializers -Wno-zero-length-
> bounds -D_GNU_SOURCE -fPIC -march=native -DALLOW_EXPERIMENTAL_API -
> DALLOW_INTERNAL_API -Wno-format-truncation -DCC_AVX512_SUPPORT -
> mavx512f -mavx512dq -mavx512ifma -
> DRTE_LOG_DEFAULT_LOGTYPE=lib.member -MD -MQ
> lib/librte_member.a.p/member_rte_member.c.o -MF
> lib/librte_member.a.p/member_rte_member.c.o.d -o
> lib/librte_member.a.p/member_rte_member.c.o -
> c ../lib/member/rte_member.c
>
> Here, the rte_member.c file is compiled with "-mavx512f -mavx512dq -
> mavx512ifma".
> The compiler might insert AVX512 instruction in this generic code.
>
> [3/9] ccache gcc -Ilib/librte_member.a.p -Ilib -I../lib -Ilib/hash -I../lib/hash -
> Ilib/ring -I../lib/ring -Ilib/member -I../lib/member -I. -I.. -Iconfig -I../config -
> Ilib/eal/include -I../lib/eal/include -Ilib/eal/linux/include -I../lib/eal/linux/include
> -Ilib/eal/x86/include -I../lib/eal/x86/include -Ilib/eal/common -
> I../lib/eal/common -Ilib/eal -I../lib/eal -Ilib/kvargs -I../lib/kvargs -Ilib/metrics -
> I../lib/metrics -Ilib/telemetry -I../lib/telemetry -Ilib/net -I../lib/net -Ilib/mbuf -
> I../lib/mbuf -Ilib/mempool -I../lib/mempool -Ilib/rcu -I../lib/rcu -fdiagnostics-
> color=always -D_FILE_OFFSET_BITS=64 -Wall -Winvalid-pch -Wextra -Werror -
> O3 -g -include rte_config.h -Wcast-qual -Wdeprecated -Wformat -Wformat-
> nonliteral -Wformat-security -Wmissing-declarations -Wmissing-prototypes -
> Wnested-externs -Wold-style-definition -Wpointer-arith -Wsign-compare -
> Wstrict-prototypes -Wundef -Wwrite-strings -Wno-address-of-packed-member
> -Wno-packed-not-aligned -Wno-missing-field-initializers -Wno-zero-length-
> bounds -D_GNU_SOURCE -fPIC -march=native -DALLOW_EXPERIMENTAL_API -
> DALLOW_INTERNAL_API -Wno-format-truncation -DCC_AVX512_SUPPORT -
> mavx512f -mavx512dq -mavx512ifma -
> DRTE_LOG_DEFAULT_LOGTYPE=lib.member -MD -MQ
> lib/librte_member.a.p/member_rte_member_sketch.c.o -MF
> lib/librte_member.a.p/member_rte_member_sketch.c.o.d -o
> lib/librte_member.a.p/member_rte_member_sketch.c.o -
> c ../lib/member/rte_member_sketch.c
>
> Idem.
>
> [4/9] rm -f lib/member/libsketch_avx512_tmp.a && gcc-ar csrDT
> lib/member/libsketch_avx512_tmp.a
> lib/member/libsketch_avx512_tmp.a.p/rte_member_sketch_avx512.c.o
> [5/9] rm -f lib/librte_member.a && gcc-ar csrD lib/librte_member.a
> lib/librte_member.a.p/member_rte_member.c.o
> lib/librte_member.a.p/member_rte_member_ht.c.o
> lib/librte_member.a.p/member_rte_member_vbf.c.o
> lib/librte_member.a.p/member_rte_member_sketch.c.o
> lib/member/libsketch_avx512_tmp.a.p/rte_member_sketch_avx512.c.o
> [6/9] /usr/bin/meson --internal exe --capture lib/member.sym_chk --
> /home/dmarchan/git/pub/dpdk.org/buildtools/check-symbols.sh
> /home/dmarchan/git/pub/dpdk.org/lib/member/version.map
> lib/librte_member.a
> [7/9] gcc -o lib/librte_member.so.23.0
> lib/member/libsketch_avx512_tmp.a.p/rte_member_sketch_avx512.c.o
> lib/librte_member.a.p/member_rte_member.c.o
> lib/librte_member.a.p/member_rte_member_ht.c.o
> lib/librte_member.a.p/member_rte_member_vbf.c.o
> lib/librte_member.a.p/member_rte_member_sketch.c.o -Wl,--as-needed -Wl,--
> no-undefined -Wl,-O1 -shared -fPIC -Wl,--start-group
> -Wl,-soname,librte_member.so.23 -Wl,--no-as-needed -pthread -lm -ldl -lnuma -
> lfdt -larchive '-Wl,-rpath,$ORIGIN/'
> -Wl,-rpath-link,/home/dmarchan/git/pub/dpdk.org/build-gcc/lib
> lib/librte_eal.so.23.0 lib/librte_kvargs.so.23.0
> lib/librte_telemetry.so.23.0 lib/librte_hash.so.23.0
> lib/librte_net.so.23.0 lib/librte_mbuf.so.23.0
> lib/librte_mempool.so.23.0 lib/librte_ring.so.23.0
> lib/librte_rcu.so.23.0
> -Wl,--version-
> script=/home/dmarchan/git/pub/dpdk.org/lib/member/version.map
> /usr/lib64/libbsd.so -Wl,--end-group
> [8/9] /usr/bin/meson --internal symbolextractor
> /home/dmarchan/git/pub/dpdk.org/build-gcc lib/librte_member.so.23.0
> lib/librte_member.so.23.0
> lib/librte_member.so.23.0.p/librte_member.so.23.0.symbols
>
Catch it, will fix in v4.
>
>
> > + sketch_avx512_tmp = static_library('sketch_avx512_tmp',
> > + 'rte_member_sketch_avx512.c',
> > + include_directories: includes,
> > + dependencies: static_rte_eal,
> > + c_args: cflags)
> > + objs +=
> sketch_avx512_tmp.extract_objects('rte_member_sketch_avx512.c')
> > + endif
> > +endif
>
>
> --
> David Marchand
next prev parent reply other threads:[~2022-09-15 13:02 UTC|newest]
Thread overview: 34+ messages / expand[flat|nested] mbox.gz Atom feed top
2022-08-10 7:45 [PATCH 0/2] introduce NitroSketch Mode into membership library Leyi Rong
2022-08-10 7:45 ` [PATCH 1/2] member: implement NitroSketch mode Leyi Rong
2022-08-16 8:59 ` Ferruh Yigit
2022-08-26 2:32 ` Rong, Leyi
2022-08-10 7:45 ` [PATCH 2/2] test/member: add functional and perf tests for sketch Leyi Rong
2022-08-11 2:33 ` Suanming Mou
2022-08-26 2:26 ` Rong, Leyi
2022-08-31 6:46 ` [PATCH v2 0/2] introduce NitroSketch Mode into membership library Leyi Rong
2022-08-31 6:46 ` [PATCH v2 1/2] member: implement NitroSketch mode Leyi Rong
2022-09-13 14:56 ` David Marchand
2022-09-14 9:42 ` Rong, Leyi
2022-08-31 6:46 ` [PATCH v2 2/2] test/member: add functional and perf tests for sketch Leyi Rong
2022-09-15 2:14 ` [PATCH v3 0/2] introduce NitroSketch Mode into membership library Leyi Rong
2022-09-15 2:14 ` [PATCH v3 1/2] member: implement NitroSketch mode Leyi Rong
2022-09-15 7:47 ` David Marchand
2022-09-15 13:02 ` Rong, Leyi [this message]
2022-09-15 2:14 ` [PATCH v3 2/2] test/member: add functional and perf tests for sketch Leyi Rong
2022-09-15 7:30 ` Jiang, YuX
2022-09-15 7:41 ` David Marchand
2022-09-15 13:03 ` [PATCH v4 0/2] introduce NitroSketch Mode into membership library Leyi Rong
2022-09-15 13:03 ` [PATCH v4 1/2] member: implement NitroSketch mode Leyi Rong
2022-09-15 14:12 ` Bruce Richardson
2022-09-16 2:30 ` Rong, Leyi
2022-09-15 13:03 ` [PATCH v4 2/2] test/member: add functional and perf tests for sketch Leyi Rong
2022-09-16 3:03 ` [PATCH v5 0/2] introduce NitroSketch Mode into membership library Leyi Rong
2022-09-16 3:03 ` [PATCH v5 1/2] member: implement NitroSketch mode Leyi Rong
2022-09-22 3:00 ` Jiang, YuX
2022-10-03 12:37 ` Thomas Monjalon
2022-10-08 3:36 ` Rong, Leyi
2022-09-16 3:03 ` [PATCH v5 2/2] test/member: add functional and perf tests for sketch Leyi Rong
2022-10-09 23:13 ` [PATCH v5 0/2] introduce NitroSketch Mode into membership library Thomas Monjalon
2022-10-10 2:36 ` Rong, Leyi
2022-10-10 7:30 ` Thomas Monjalon
2022-10-10 8:28 ` Rong, Leyi
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=MW4PR11MB577421F3624F28725635FCC9EC499@MW4PR11MB5774.namprd11.prod.outlook.com \
--to=leyi.rong@intel.com \
--cc=david.marchand@redhat.com \
--cc=dev@dpdk.org \
--cc=ferruh.yigit@xilinx.com \
--cc=sameh.gobriel@intel.com \
--cc=suanmingm@nvidia.com \
--cc=yipeng1.wang@intel.com \
--cc=zaoxingliu@gmail.com \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).