Re: [PATCH v3 1/2] member: implement NitroSketch mode

[David Marchand <david.marchand@redhat.com>]

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



> +       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