From: "Medvedkin, Vladimir" <vladimir.medvedkin@intel.com>
To: "Ananyev, Konstantin" <konstantin.ananyev@intel.com>,
"dev@dpdk.org" <dev@dpdk.org>
Cc: "Chilikin, Andrey" <andrey.chilikin@intel.com>,
"Wang, Yipeng1" <yipeng1.wang@intel.com>,
"Gobriel, Sameh" <sameh.gobriel@intel.com>,
"Richardson, Bruce" <bruce.richardson@intel.com>
Subject: Re: [dpdk-dev] [PATCH 2/5] hash: enable gfni thash implementation
Date: Fri, 15 Oct 2021 11:13:55 +0200 [thread overview]
Message-ID: <cb0ef937-a6b5-335d-c7c2-fb1d9823f701@intel.com> (raw)
In-Reply-To: <DM6PR11MB4491A8F5710EF59EFC5711E59AB29@DM6PR11MB4491.namprd11.prod.outlook.com>
Hi Konstantin,
On 08/10/2021 13:31, Ananyev, Konstantin wrote:
>
>> This patch enables new GFNI Toeplitz hash in
>> predictable RSS library.
>>
>> Signed-off-by: Vladimir Medvedkin <vladimir.medvedkin@intel.com>
>> ---
>> lib/hash/rte_thash.c | 43 +++++++++++++++++++++++++++++++++++++++----
>> lib/hash/rte_thash.h | 19 +++++++++++++++++++
>> lib/hash/version.map | 1 +
>> 3 files changed, 59 insertions(+), 4 deletions(-)
>>
>> diff --git a/lib/hash/rte_thash.c b/lib/hash/rte_thash.c
>> index 07447f7..86a0e96 100644
>> --- a/lib/hash/rte_thash.c
>> +++ b/lib/hash/rte_thash.c
>> @@ -86,6 +86,8 @@ struct rte_thash_ctx {
>> uint32_t reta_sz_log; /** < size of the RSS ReTa in bits */
>> uint32_t subtuples_nb; /** < number of subtuples */
>> uint32_t flags;
>> + uint64_t *matrices;
>
> Comment, what is that, etc.
>
I'll rephrase the comment below.
>> + /**< rte_thash_complete_matrix generated matrices */
>> uint8_t hash_key[0];
>> };
>>
>> @@ -253,12 +255,25 @@ rte_thash_init_ctx(const char *name, uint32_t key_len, uint32_t reta_sz,
>> ctx->hash_key[i] = rte_rand();
>> }
>>
>> + if (rte_thash_gfni_supported) {
>
> I think it should be:
> if (rte_thash_gfni_supported && rte_vect_get_max_simd_bitwidth() >= RTE_VECT_SIMD_512)
>
>
Agree
>> + ctx->matrices = rte_zmalloc(NULL, key_len * sizeof(uint64_t),
>> + RTE_CACHE_LINE_SIZE);
>
> You can do it probably before allocation ctx, at the same place where te is allocated.
> Might be a bit nicer.
>
I'd prefer to keep allocation and initialization of matrices in one
place, below there is rte_thash_complete_matrix() which uses previously
generated ctx->hash_key.
>> + if (ctx->matrices == NULL)
>
> RTE_LOG(ERR, ...);
> rte_ernno = ENOMEM;
>
Agree
>> + goto free_ctx;
>> +
>> + rte_thash_complete_matrix(ctx->matrices, ctx->hash_key,
>> + key_len);
>> + }
>> +
>> te->data = (void *)ctx;
>> TAILQ_INSERT_TAIL(thash_list, te, next);
>>
>> rte_mcfg_tailq_write_unlock();
>>
>> return ctx;
>> +
>> +free_ctx:
>> + rte_free(ctx);
>> free_te:
>> rte_free(te);
>> exit:
>> @@ -372,6 +387,10 @@ generate_subkey(struct rte_thash_ctx *ctx, struct thash_lfsr *lfsr,
>> set_bit(ctx->hash_key, get_rev_bit_lfsr(lfsr), i);
>> }
>>
>> + if (rte_thash_gfni_supported)
>
> Here and in data-path functions, I think it would be better:
> if (ctx->matrices != NULL)
Agree
>> + rte_thash_complete_matrix(ctx->matrices, ctx->hash_key,
>> + ctx->key_len);
>> +
>> return 0;
>> }
>>
>> @@ -628,6 +647,16 @@ rte_thash_get_key(struct rte_thash_ctx *ctx)
>> return ctx->hash_key;
>> }
>>
>> +const uint64_t *
>> +rte_thash_get_gfni_matrices(struct rte_thash_ctx *ctx)
>> +{
>> + if (rte_thash_gfni_supported)
>> + return ctx->matrices;
>
> Why not just always:
> return ctx->matices;
> ?
>
Agree
>> +
>> + rte_errno = ENOTSUP;
>> + return NULL;
>> +}
>> +
>> static inline uint8_t
>> read_unaligned_byte(uint8_t *ptr, unsigned int len, unsigned int offset)
>> {
>> @@ -739,11 +768,17 @@ rte_thash_adjust_tuple(struct rte_thash_ctx *ctx,
>> attempts = RTE_MIN(attempts, 1U << (h->tuple_len - ctx->reta_sz_log));
>>
>> for (i = 0; i < attempts; i++) {
>> - for (j = 0; j < (tuple_len / 4); j++)
>> - tmp_tuple[j] =
>> - rte_be_to_cpu_32(*(uint32_t *)&tuple[j * 4]);
>> + if (rte_thash_gfni_supported)
> if (ctx->matrices)
>
>> + hash = rte_thash_gfni(ctx->matrices, tuple, tuple_len);
>> + else {
>> + for (j = 0; j < (tuple_len / 4); j++)
>> + tmp_tuple[j] =
>> + rte_be_to_cpu_32(
>> + *(uint32_t *)&tuple[j * 4]);
>> +
>> + hash = rte_softrss(tmp_tuple, tuple_len / 4, hash_key);
>> + }
>>
>> - hash = rte_softrss(tmp_tuple, tuple_len / 4, hash_key);
>> adj_bits = rte_thash_get_complement(h, hash, desired_value);
>>
>> /*
>> diff --git a/lib/hash/rte_thash.h b/lib/hash/rte_thash.h
>> index e3f1fc6..6e6861c 100644
>> --- a/lib/hash/rte_thash.h
>> +++ b/lib/hash/rte_thash.h
>> @@ -410,6 +410,25 @@ const uint8_t *
>> rte_thash_get_key(struct rte_thash_ctx *ctx);
>>
>> /**
>> + * Get a pointer to the toeplitz hash matrices contained in the context.
>> + * These matrices could be used with fast toeplitz hash implementation if
>> + * CPU supports GFNI.
>> + * Matrices changes after each addition of a helper.
>> + *
>> + * @warning
>> + * @b EXPERIMENTAL: this API may change without prior notice.
>> + *
>> + * @param ctx
>> + * Thash context
>> + * @return
>> + * A pointer to the toeplitz hash key matrices on success
>> + * NULL if GFNI is not supported.
>> + */
>> +__rte_experimental
>> +const uint64_t *
>> +rte_thash_get_gfni_matrices(struct rte_thash_ctx *ctx);
>> +
>> +/**
>> * Function prototype for the rte_thash_adjust_tuple
>> * to check if adjusted tuple could be used.
>> * Generally it is some kind of lookup function to check
>> diff --git a/lib/hash/version.map b/lib/hash/version.map
>> index cecf922..3eda695 100644
>> --- a/lib/hash/version.map
>> +++ b/lib/hash/version.map
>> @@ -43,6 +43,7 @@ EXPERIMENTAL {
>> rte_thash_find_existing;
>> rte_thash_free_ctx;
>> rte_thash_get_complement;
>> + rte_thash_get_gfni_matrices;
>> rte_thash_get_helper;
>> rte_thash_get_key;
>> rte_thash_gfni_supported;
>> --
>> 2.7.4
>
--
Regards,
Vladimir
next prev parent reply other threads:[~2021-10-15 9:14 UTC|newest]
Thread overview: 72+ messages / expand[flat|nested] mbox.gz Atom feed top
2021-09-06 16:03 [dpdk-dev] [PATCH 0/5] optimized Toeplitz hash implementation Vladimir Medvedkin
2021-09-06 16:03 ` [dpdk-dev] [PATCH 1/5] hash: add new toeplitz " Vladimir Medvedkin
2021-10-07 18:23 ` Ananyev, Konstantin
2021-10-08 11:19 ` Ananyev, Konstantin
2021-10-15 9:11 ` Medvedkin, Vladimir
2021-10-15 10:55 ` Ananyev, Konstantin
2021-10-15 13:09 ` Medvedkin, Vladimir
2021-09-06 16:03 ` [dpdk-dev] [PATCH 2/5] hash: enable gfni thash implementation Vladimir Medvedkin
2021-10-08 11:31 ` Ananyev, Konstantin
2021-10-15 9:13 ` Medvedkin, Vladimir [this message]
2021-09-06 16:03 ` [dpdk-dev] [PATCH 3/5] doc/hash: update documentation for the thash library Vladimir Medvedkin
2021-09-06 16:03 ` [dpdk-dev] [PATCH 4/5] test/thash: add tests for a new Toeplitz hash function Vladimir Medvedkin
2021-09-07 0:35 ` Stephen Hemminger
2021-09-08 13:59 ` Medvedkin, Vladimir
2021-09-06 16:03 ` [dpdk-dev] [PATCH 5/5] test/thash: add performance tests for the Toeplitz hash Vladimir Medvedkin
2021-10-15 9:30 ` [dpdk-dev] [PATCH v2 0/5] optimized Toeplitz hash implementation Vladimir Medvedkin
2021-10-20 18:20 ` [dpdk-dev] [PATCH v3 " Vladimir Medvedkin
2021-10-21 17:18 ` [dpdk-dev] [PATCH v4 " Vladimir Medvedkin
2021-10-21 18:54 ` [dpdk-dev] [PATCH v5 " Vladimir Medvedkin
2021-10-21 18:54 ` [dpdk-dev] [PATCH v5 1/5] hash: add new toeplitz " Vladimir Medvedkin
2021-10-25 17:05 ` Thomas Monjalon
2021-10-21 18:54 ` [dpdk-dev] [PATCH v5 2/5] hash: enable gfni thash implementation Vladimir Medvedkin
2021-10-21 18:54 ` [dpdk-dev] [PATCH v5 3/5] doc/hash: update documentation for the thash library Vladimir Medvedkin
2021-10-25 17:04 ` Thomas Monjalon
2021-10-26 20:30 ` Medvedkin, Vladimir
2021-10-21 18:54 ` [dpdk-dev] [PATCH v5 4/5] test/thash: add tests for a new Toeplitz hash function Vladimir Medvedkin
2021-10-21 18:54 ` [dpdk-dev] [PATCH v5 5/5] test/thash: add performance tests for the Toeplitz hash Vladimir Medvedkin
2021-10-25 17:02 ` Thomas Monjalon
2021-10-26 20:29 ` Medvedkin, Vladimir
2021-10-27 8:29 ` Thomas Monjalon
2021-10-27 15:48 ` Medvedkin, Vladimir
2021-10-25 17:27 ` Stephen Hemminger
2021-10-26 20:31 ` Medvedkin, Vladimir
2021-10-21 17:18 ` [dpdk-dev] [PATCH v4 1/5] hash: add new toeplitz hash implementation Vladimir Medvedkin
2021-10-21 17:18 ` [dpdk-dev] [PATCH v4 2/5] hash: enable gfni thash implementation Vladimir Medvedkin
2021-10-21 17:18 ` [dpdk-dev] [PATCH v4 3/5] doc/hash: update documentation for the thash library Vladimir Medvedkin
2021-10-21 17:18 ` [dpdk-dev] [PATCH v4 4/5] test/thash: add tests for a new Toeplitz hash function Vladimir Medvedkin
2021-10-21 17:18 ` [dpdk-dev] [PATCH v4 5/5] test/thash: add performance tests for the Toeplitz hash Vladimir Medvedkin
2021-10-20 18:20 ` [dpdk-dev] [PATCH v3 1/5] hash: add new toeplitz hash implementation Vladimir Medvedkin
2021-10-21 9:42 ` Ananyev, Konstantin
2021-10-21 17:17 ` Medvedkin, Vladimir
2021-10-20 18:20 ` [dpdk-dev] [PATCH v3 2/5] hash: enable gfni thash implementation Vladimir Medvedkin
2021-10-21 9:46 ` Ananyev, Konstantin
2021-10-20 18:20 ` [dpdk-dev] [PATCH v3 3/5] doc/hash: update documentation for the thash library Vladimir Medvedkin
2021-10-20 18:20 ` [dpdk-dev] [PATCH v3 4/5] test/thash: add tests for a new Toeplitz hash function Vladimir Medvedkin
2021-10-20 18:20 ` [dpdk-dev] [PATCH v3 5/5] test/thash: add performance tests for the Toeplitz hash Vladimir Medvedkin
2021-10-15 9:30 ` [dpdk-dev] [PATCH v2 1/5] hash: add new toeplitz hash implementation Vladimir Medvedkin
2021-10-15 16:58 ` Stephen Hemminger
2021-10-18 10:40 ` Ananyev, Konstantin
2021-10-19 1:15 ` Stephen Hemminger
2021-10-19 15:42 ` Medvedkin, Vladimir
2021-10-18 11:08 ` Medvedkin, Vladimir
2021-10-15 9:30 ` [dpdk-dev] [PATCH v2 2/5] hash: enable gfni thash implementation Vladimir Medvedkin
2021-10-15 9:30 ` [dpdk-dev] [PATCH v2 3/5] doc/hash: update documentation for the thash library Vladimir Medvedkin
2021-10-15 9:30 ` [dpdk-dev] [PATCH v2 4/5] test/thash: add tests for a new Toeplitz hash function Vladimir Medvedkin
2021-10-15 9:30 ` [dpdk-dev] [PATCH v2 5/5] test/thash: add performance tests for the Toeplitz hash Vladimir Medvedkin
2021-10-26 20:32 ` [dpdk-dev] [PATCH v6 0/4] optimized Toeplitz hash implementation Vladimir Medvedkin
2021-10-26 20:32 ` [dpdk-dev] [PATCH v6 1/4] hash: add new toeplitz " Vladimir Medvedkin
2021-10-26 20:32 ` [dpdk-dev] [PATCH v6 2/4] hash: add bulk " Vladimir Medvedkin
2021-10-26 20:32 ` [dpdk-dev] [PATCH v6 3/4] hash: enable gfni thash implementation Vladimir Medvedkin
2021-10-26 20:32 ` [dpdk-dev] [PATCH v6 4/4] test/thash: add performance tests for the Toeplitz hash Vladimir Medvedkin
2021-10-27 16:16 ` [dpdk-dev] [PATCH v7 0/4] optimized Toeplitz hash implementation Vladimir Medvedkin
2021-10-27 16:16 ` [dpdk-dev] [PATCH v7 1/4] hash: add new toeplitz " Vladimir Medvedkin
2021-10-27 16:16 ` [dpdk-dev] [PATCH v7 2/4] hash: add bulk " Vladimir Medvedkin
2021-10-27 16:16 ` [dpdk-dev] [PATCH v7 3/4] hash: enable gfni thash implementation Vladimir Medvedkin
2021-10-27 16:16 ` [dpdk-dev] [PATCH v7 4/4] test/thash: add performance tests for the Toeplitz hash Vladimir Medvedkin
2021-11-02 18:38 ` [dpdk-dev] [PATCH v8 0/4] optimized Toeplitz hash implementation Vladimir Medvedkin
2021-11-04 10:20 ` Thomas Monjalon
2021-11-02 18:38 ` [dpdk-dev] [PATCH v8 1/4] hash: add new toeplitz " Vladimir Medvedkin
2021-11-02 18:38 ` [dpdk-dev] [PATCH v8 2/4] hash: add bulk " Vladimir Medvedkin
2021-11-02 18:38 ` [dpdk-dev] [PATCH v8 3/4] hash: enable gfni thash implementation Vladimir Medvedkin
2021-11-02 18:38 ` [dpdk-dev] [PATCH v8 4/4] test/thash: add performance tests for the Toeplitz hash Vladimir Medvedkin
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=cb0ef937-a6b5-335d-c7c2-fb1d9823f701@intel.com \
--to=vladimir.medvedkin@intel.com \
--cc=andrey.chilikin@intel.com \
--cc=bruce.richardson@intel.com \
--cc=dev@dpdk.org \
--cc=konstantin.ananyev@intel.com \
--cc=sameh.gobriel@intel.com \
--cc=yipeng1.wang@intel.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).