On Sat, Jan 16, 2021 at 3:39 PM Ruifeng Wang <ruifeng.w...@arm.com> wrote:
> > Subject: [PATCH 1/3] net/hinic: restore vectorised code
> >
> > Following make support removal, the vectorised code is not built anymore,
> > fix the build flag check.
> >
> > Fixes: 3cc6ecfdfe85 ("build: remove makefiles")
> > Cc: sta...@dpdk.org
> >
> > Signed-off-by: David Marchand <david.march...@redhat.com>
> > ---
> > drivers/net/hinic/hinic_pmd_rx.c | 6 +++---
> > drivers/net/hinic/hinic_pmd_tx.c | 10 +++++-----
> > 2 files changed, 8 insertions(+), 8 deletions(-)
> >
> > diff --git a/drivers/net/hinic/hinic_pmd_rx.c
> > b/drivers/net/hinic/hinic_pmd_rx.c
> > index a49769a863..842399cc4c 100644
> > --- a/drivers/net/hinic/hinic_pmd_rx.c
> > +++ b/drivers/net/hinic/hinic_pmd_rx.c
> > @@ -4,7 +4,7 @@
> >
> > #include <rte_ether.h>
> > #include <rte_mbuf.h>
> > -#ifdef __ARM64_NEON__
> > +#ifdef RTE_ARCH_ARM64
>
> We can test '__ARM_NEON' which will be defined by compilers.
> https://developer.arm.com/documentation/ihi0053/latest/
On the principle, I agree, but this was not what was tested before.
The activation was only gated by checking for the arm architecture.
See:
https://git.dpdk.org/dpdk/tree/drivers/net/hinic/Makefile?id=05b6eee7bdabf7c17ed69c44515e0cd7d6e1da23#n14
So I restored the test as it was before make removal.
What you propose can be done in a followup from the hinic developers,
but the minimal fix is still this current patch.
--
David Marchand
