[dpdk-stable] [PATCH v2 2/5] net/hns3: fix build with sve enabled

[Ruifeng Wang]

> -----Original Message-----
> From: oulijun <oulijun at huawei.com>
> Sent: Saturday, January 9, 2021 10:12 AM
> To: Honnappa Nagarahalli <Honnappa.Nagarahalli at arm.com>; Ruifeng Wang
> <Ruifeng.Wang at arm.com>; Wei Hu (Xavier) <xavier.huwei at huawei.com>;
> Min Hu (Connor) <humin29 at huawei.com>; Yisen Zhuang
> <yisen.zhuang at huawei.com>; Huisong Li <lihuisong at huawei.com>;
> Chengchang Tang <tangchengchang at huawei.com>; Chengwen Feng
> <fengchengwen at huawei.com>
> Cc: dev at dpdk.org; vladimir.medvedkin at intel.com; jerinj at marvell.com;
> hemant.agrawal at nxp.com; nd <nd at arm.com>; stable at dpdk.org
> Subject: Re: [PATCH v2 2/5] net/hns3: fix build with sve enabled
> 
> 
> 在 2021/1/9 8:06, Honnappa Nagarahalli 写道:
> > <snip>
> >
> >>
> >> Building with SVE extension enabled stopped with error:
> >>
> >>   error: ACLE function ‘svwhilelt_b64_s32’ requires ISA extension ‘sve’
> >>     18 | #define PG64_256BIT  svwhilelt_b64(0, 4)
> >>
> >> This is caused by unintentional cflags reset.
> >> Fixed the issue by appending required flag to cflags instead of overriding it.
> >>
> >> Fixes: 952ebacce4f2 ("net/hns3: support SVE Rx")
> >> Cc: xavier.huwei at huawei.com
> >> Cc: stable at dpdk.org
> >>
> >> Signed-off-by: Ruifeng Wang <ruifeng.wang at arm.com>
> >> ---
> >>   drivers/net/hns3/meson.build | 2 +-
> >>   1 file changed, 1 insertion(+), 1 deletion(-)
> >>
> >> diff --git a/drivers/net/hns3/meson.build
> >> b/drivers/net/hns3/meson.build index 45cee34d9..798086357 100644
> >> --- a/drivers/net/hns3/meson.build
> >> +++ b/drivers/net/hns3/meson.build
> >> @@ -32,7 +32,7 @@ deps += ['hash']
> >>   if arch_subdir == 'arm' and dpdk_conf.get('RTE_ARCH_64')
> >>   	sources += files('hns3_rxtx_vec.c')
> >>   	if cc.get_define('__ARM_FEATURE_SVE', args: machine_args) != ''
> >> -		cflags = ['-DCC_SVE_SUPPORT']
> >> +		cflags += ['-DCC_SVE_SUPPORT']
> > This comment is unrelated to this patch. We need to be consistent with the
> macro definitions. Is '__ARM_FEATURE_SVE' not enough? If we need to
> define an additional flag, I would name it something like
> 'RTE_ARM_FEATURE_SVE'.
> >
> I think the __ARM_FEATURE_SVE is ok. if use the gcc version included SVE
> flag, it will be identified as __ARM_FEATURE_SVE. it is defined in the ARM
> SVE document.

Yes, we can rely on flags defined by compiler and no extra flag is needed.
I can update in next version to remove this section from meson file and replace CC_SVE_SUPPORT in code.
> >>   		sources += files('hns3_rxtx_vec_sve.c')
> >>   	endif
> >>   endif
> >> --
> >> 2.25.1
> >