DPDK patches and discussions
 help / color / mirror / Atom feed
From: Santosh Shukla <sshukla@mvista.com>
To: Yuanhan Liu <yuanhan.liu@linux.intel.com>
Cc: dpdk <dev@dpdk.org>
Subject: Re: [dpdk-dev] [PATCH v1] virtio: Use cpuflag for vector api
Date: Tue, 1 Mar 2016 11:40:41 +0530	[thread overview]
Message-ID: <CAAyOgsZJ0XDVNMMxXX06mXd2Zh09E8vJ=8-UwkmUgiuSFqURPA@mail.gmail.com> (raw)
In-Reply-To: <20160301055508.GL14300@yliu-dev.sh.intel.com>

On Tue, Mar 1, 2016 at 11:25 AM, Yuanhan Liu
<yuanhan.liu@linux.intel.com> wrote:
> On Mon, Feb 29, 2016 at 06:01:38PM +0530, Santosh Shukla wrote:
>> On Mon, Feb 29, 2016 at 9:57 AM, Yuanhan Liu
>> <yuanhan.liu@linux.intel.com> wrote:
>> > On Fri, Feb 26, 2016 at 02:21:02PM +0530, Santosh Shukla wrote:
>> >> Check cpuflag macro before using vectored api.
>> >> -virtio_recv_pkts_vec() uses _sse3__ simd instruction for now so added cpuflag.
>> >> - Also wrap other vectored freind api ie..
>> >> 1) virtqueue_enqueue_recv_refill_simple
>> >> 2) virtio_rxq_vec_setup
>> >>
>> > ...
>> >> diff --git a/drivers/net/virtio/virtio_rxtx_simple.c b/drivers/net/virtio/virtio_rxtx_simple.c
>> >> index 3a1de9d..be51d7c 100644
>> >> --- a/drivers/net/virtio/virtio_rxtx_simple.c
>> >> +++ b/drivers/net/virtio/virtio_rxtx_simple.c
>> >
>> > Hmm, why not wrapping the whole file, instead of just few functions?
>> >
>>
>> Better to refactor code and make arch specific. Current implementation
>> is temporary.
>
> I'm okay to refactor, if it's a clean one. But moving virtio __driver__
> stuff to EAL, sorry, it still doesn't make too much sense to me.
>

You misunderstood my comment, my arch specific meaning
driver/net/virtio/virtio_vec_<arch>.h

> For this case, let's make it simple so far, and consider it when we
> have another vec implementation, or better refactor comes out.
>
>> > Or maybe better, do a compile time check at the Makefile, something
>> > like:
>> >
>> >     if has_CPUFLAG_xxx
>> >         SRCS-$(CONFIG_RTE_LIBRTE_VIRTIO_PMD) += virtio_rxtx_simple.c
>> >     endif
>> >
>> Tried this approach but end up with link error,  If I try to fix below
>> link error then I will be ending up writing similar code,
>
> Sure you need the first part of your patch. Yes, it's similar code,
> but it has fewer "#ifdef " conditions, and more importantly, it
> leaves virtio_rxtx_simple.c alone.
>

Ok.,
>         --yliu
>
>> linker error snap:
>>
>> /work/santosh/thunder/nfs/dpdk/arm64-thunderx-linuxapp-gcc/lib/librte_pmd_virtio.a(virtio_rxtx.o):
>> In function `virtio_dev_rxtx_start':
>> virtio_rxtx.c:(.text+0x168c): undefined reference to
>> `virtqueue_enqueue_recv_refill_simple'
>> /work/santosh/thunder/nfs/dpdk/arm64-thunderx-linuxapp-gcc/lib/librte_pmd_virtio.a(virtio_rxtx.o):
>> In function `virtio_dev_rx_queue_setup':
>> virtio_rxtx.c:(.text+0x2364): undefined reference to `virtio_rxq_vec_setup'
>> /work/santosh/thunder/nfs/dpdk/arm64-thunderx-linuxapp-gcc/lib/librte_pmd_virtio.a(virtio_rxtx.o):
>> In function `virtio_dev_tx_queue_setup':
>> virtio_rxtx.c:(.text+0x2460): undefined reference to `virtio_xmit_pkts_simple'
>> virtio_rxtx.c:(.text+0x2464): undefined reference to `virtio_recv_pkts_vec'
>> virtio_rxtx.c:(.text+0x2468): undefined reference to `virtio_xmit_pkts_simple'
>> virtio_rxtx.c:(.text+0x246c): undefined reference to `virtio_recv_pkts_vec'

  reply	other threads:[~2016-03-01  6:10 UTC|newest]

Thread overview: 25+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2016-02-26  8:51 Santosh Shukla
2016-02-29  4:27 ` Yuanhan Liu
2016-02-29  5:33   ` Xie, Huawei
2016-02-29 12:31   ` Santosh Shukla
2016-03-01  5:55     ` Yuanhan Liu
2016-03-01  6:10       ` Santosh Shukla [this message]
2016-03-01  6:22         ` Yuanhan Liu
2016-02-29 12:58 ` [dpdk-dev] [PATCH v2] " Santosh Shukla
2016-03-01  5:59   ` Yuanhan Liu
2016-03-01  6:08     ` Santosh Shukla
2016-03-01  6:32       ` Yuanhan Liu
2016-03-01 10:07         ` Santosh Shukla
2016-03-01 10:02   ` [dpdk-dev] [PATCH v1 0/3] virtio vector and misc Santosh Shukla
2016-03-01 10:02     ` [dpdk-dev] [PATCH v1 1/3] virtio: use vector rx/tx for ssse cpuflag only Santosh Shukla
2016-03-01 10:02     ` [dpdk-dev] [PATCH v1 2/3] config: enable virtio for armv7/v8 Santosh Shukla
2016-03-01 10:02     ` [dpdk-dev] [PATCH v1 3/3] guide/release: add virtio for arm feature info Santosh Shukla
2016-03-02  8:32     ` [dpdk-dev] [PATCH v1 0/3] virtio vector and misc Yuanhan Liu
2016-03-02  8:41       ` Santosh Shukla
2016-03-03 13:26         ` Thomas Monjalon
2016-03-03 13:50           ` Santosh Shukla
2016-03-01  9:11 ` [dpdk-dev] [PATCH v1] virtio: Use cpuflag for vector api Qiu, Michael
2016-03-01  9:45   ` Santosh Shukla
2016-03-02  2:10     ` Qiu, Michael
2016-03-02  2:49       ` Yuanhan Liu
2016-03-04  6:36         ` Qiu, Michael

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='CAAyOgsZJ0XDVNMMxXX06mXd2Zh09E8vJ=8-UwkmUgiuSFqURPA@mail.gmail.com' \
    --to=sshukla@mvista.com \
    --cc=dev@dpdk.org \
    --cc=yuanhan.liu@linux.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).