[dpdk-dev] [PATCH v3 04/18] net/i40e: add checks for max SIMD bitwidth
Ananyev, Konstantin
konstantin.ananyev at intel.com
Thu Oct 8 18:14:25 CEST 2020
> Hi Konstantin,
>
>
> >-----Original Message-----
> >From: Ananyev, Konstantin <konstantin.ananyev at intel.com>
> >Sent: Thursday 8 October 2020 16:22
> >To: Power, Ciara <ciara.power at intel.com>; dev at dpdk.org
> >Cc: Power, Ciara <ciara.power at intel.com>; Xing, Beilei
> ><beilei.xing at intel.com>; Guo, Jia <jia.guo at intel.com>
> >Subject: RE: [dpdk-dev] [PATCH v3 04/18] net/i40e: add checks for max SIMD
> >bitwidth
> >
> >
> >>
> >> When choosing a vector path to take, an extra condition must be
> >> satisfied to ensure the max SIMD bitwidth allows for the CPU enabled
> >> path.
> >>
> >> Cc: Beilei Xing <beilei.xing at intel.com>
> >> Cc: Jeff Guo <jia.guo at intel.com>
> >>
> >> Signed-off-by: Ciara Power <ciara.power at intel.com>
> >> ---
> >> drivers/net/i40e/i40e_rxtx.c | 19 +++++++++++++------
> >> 1 file changed, 13 insertions(+), 6 deletions(-)
> >>
> >> diff --git a/drivers/net/i40e/i40e_rxtx.c
> >> b/drivers/net/i40e/i40e_rxtx.c index 60b33d20a1..9b535b52fa 100644
> >> --- a/drivers/net/i40e/i40e_rxtx.c
> >> +++ b/drivers/net/i40e/i40e_rxtx.c
> >> @@ -3098,7 +3098,8 @@ static eth_rx_burst_t
> >> i40e_get_latest_rx_vec(bool scatter) { #if defined(RTE_ARCH_X86) &&
> >> defined(CC_AVX2_SUPPORT)
> >> - if (rte_cpu_get_flag_enabled(RTE_CPUFLAG_AVX2))
> >> + if (rte_cpu_get_flag_enabled(RTE_CPUFLAG_AVX2) &&
> >> + rte_get_max_simd_bitwidth() >=
> >RTE_MAX_256_SIMD)
> >> return scatter ? i40e_recv_scattered_pkts_vec_avx2 :
> >> i40e_recv_pkts_vec_avx2;
> >
> >Hmm, but that means - if user will set --simd-bitwidth=128 we'll select
> >scalar function, right?
> >Even though sse one is available.
> >Is that what we really want in that case?
> >
>
> If the max SIMD is 128, the second return in this function is used, which I believe is SSE:
>
> return scatter ? i40e_recv_scattered_pkts_vec :
> i40e_recv_pkts_vec;
>
> And that function is only called if the max SIMD is >=128, scalar is used otherwise.
>
> Am I missing something else here?
Nope, you are right, that was me not reading code properly 😊
Acked-by: Konstantin Ananyev <konstantin.ananyev at intel.com>
>
> Thanks,
> Ciara
>
> >> #endif
> >> @@ -3115,7 +3116,8 @@ i40e_get_recommend_rx_vec(bool scatter)
> >> * use of AVX2 version to later plaforms, not all those that could
> >> * theoretically run it.
> >> */
> >> - if (rte_cpu_get_flag_enabled(RTE_CPUFLAG_AVX512F))
> >> + if (rte_cpu_get_flag_enabled(RTE_CPUFLAG_AVX512F) &&
> >> + rte_get_max_simd_bitwidth() >=
> >RTE_MAX_256_SIMD)
> >> return scatter ? i40e_recv_scattered_pkts_vec_avx2 :
> >> i40e_recv_pkts_vec_avx2;
> >> #endif
> >> @@ -3154,7 +3156,8 @@ i40e_set_rx_function(struct rte_eth_dev *dev)
> >> }
> >> }
> >>
> >> - if (ad->rx_vec_allowed) {
> >> + if (ad->rx_vec_allowed && rte_get_max_simd_bitwidth()
> >> + >= RTE_MAX_128_SIMD) {
> >> /* Vec Rx path */
> >> PMD_INIT_LOG(DEBUG, "Vector Rx path will be used on
> >port=%d.",
> >> dev->data->port_id);
> >> @@ -3268,7 +3271,8 @@ static eth_tx_burst_t
> >> i40e_get_latest_tx_vec(void)
> >> {
> >> #if defined(RTE_ARCH_X86) && defined(CC_AVX2_SUPPORT)
> >> - if (rte_cpu_get_flag_enabled(RTE_CPUFLAG_AVX2))
> >> + if (rte_cpu_get_flag_enabled(RTE_CPUFLAG_AVX2) &&
> >> + rte_get_max_simd_bitwidth() >=
> >RTE_MAX_256_SIMD)
> >> return i40e_xmit_pkts_vec_avx2;
> >> #endif
> >> return i40e_xmit_pkts_vec;
> >> @@ -3283,7 +3287,8 @@ i40e_get_recommend_tx_vec(void)
> >> * use of AVX2 version to later plaforms, not all those that could
> >> * theoretically run it.
> >> */
> >> - if (rte_cpu_get_flag_enabled(RTE_CPUFLAG_AVX512F))
> >> + if (rte_cpu_get_flag_enabled(RTE_CPUFLAG_AVX512F) &&
> >> + rte_get_max_simd_bitwidth() >=
> >RTE_MAX_256_SIMD)
> >> return i40e_xmit_pkts_vec_avx2;
> >> #endif
> >> return i40e_xmit_pkts_vec;
> >> @@ -3311,7 +3316,9 @@ i40e_set_tx_function(struct rte_eth_dev *dev)
> >> }
> >>
> >> if (ad->tx_simple_allowed) {
> >> - if (ad->tx_vec_allowed) {
> >> + if (ad->tx_vec_allowed &&
> >> + rte_get_max_simd_bitwidth()
> >> + >= RTE_MAX_128_SIMD) {
> >> PMD_INIT_LOG(DEBUG, "Vector tx finally be used.");
> >> if (ad->use_latest_vec)
> >> dev->tx_pkt_burst =
> >> --
> >> 2.17.1
More information about the dev
mailing list