[dpdk-dev] [PATCH v2] mk: gcc -march support for intel processors code names

Pattan, Reshma reshma.pattan at intel.com
Mon Oct 10 17:21:57 CEST 2016


Hi Thomas

> -----Original Message-----
> From: Thomas Monjalon [mailto:thomas.monjalon at 6wind.com]
> Sent: Monday, October 10, 2016 3:26 PM
> To: Liu, Yong <yong.liu at intel.com>; Pattan, Reshma
> <reshma.pattan at intel.com>
> Cc: dev at dpdk.org
> Subject: Re: [dpdk-dev] [PATCH v2] mk: gcc -march support for intel processors
> code names
> 
> 2016-09-28 10:54, Liu, Yong:
> > Tested-by: Yong Liu <yong.liu at intel.com>
> >             FangFang Wei <fangfangx.wei at intel.com>
> [...]
> >    Description: Build test on different distributions
> >    Command / instruction:
> >      Verify build pass on listed distributions.
> >
> >      OS                   GCC            Kernel
> >      Ubuntu 16.04         5.4.0          4.4.0-36-generic
> >      Fedora23             5.3.1          4.2.3-300
> >      Fedora24             6.1.1          4.6.4-301
> >      Ubuntu 12.04         4.6.3          3.8.0-29
> >      Ubuntu 12.04 i686    4.6.3          3.8.0-29
> >      Ubuntu 14.04         4.8.4          3.16.0-30
> >      Ubuntu 14.04 i686    4.8.4          3.16.0-30
> >      Fedora18             4.7.2          3.6.10-4
> >      Fedora18 i686        4.7.2          3.6.10-4
> >      Fedora20             4.8.2          3.15.6-200
> >      Fedora20 i686        4.8.3          3.11.0
> >      Suse11SP2            4.5.1          3.0.13-0.2
> >      Suse12SP3            4.7.2          3.7.10-1.1
> >      RHEL7.0              4.8.2          3.10.0-123
> >      RHEL7.2              4.8.5          3.10.0-327
> >      CentOS7.0            4.8.5          3.10.0-327
> >      FreeBSD10.0          4.8.4          10.0-RELEASE
> >      FreeBSD10.3          4.8.5          10.3-RELEASE
> 
> I do not understand this test.
> This patch is dropping a lot of optimizations with compilers older than 4.9 !
> 
> Why not recommend GCC 4.9 and keep the graceful degradation for older
> versions, at least for one more year, even if it is not optimal for newer
> architectures?

I am ok with the idea, so should I send latest patch by keeping the code in  mk/toolchain/gcc/rte.toolchain-compat.mk untouched/intact?

Thanks,
Reshma


More information about the dev mailing list