* Re: [PATCH] build/x86: fix support for older compilers
2025-07-02 15:51 [PATCH] build/x86: fix support for older compilers Bruce Richardson
@ 2025-07-02 15:53 ` Bruce Richardson
2025-07-02 16:00 ` [PATCH v2] " Bruce Richardson
2025-07-03 10:21 ` [PATCH v3] " Bruce Richardson
2 siblings, 0 replies; 9+ messages in thread
From: Bruce Richardson @ 2025-07-02 15:53 UTC (permalink / raw)
To: dev; +Cc: Konstantin Ananyev, Vipin Varghese
On Wed, Jul 02, 2025 at 04:51:29PM +0100, Bruce Richardson wrote:
> Some older compilers e.g. gcc 8.5, do not support overriding
> -march=native with another architecture, leading to build warnings such
> as reported in Bugzilla (link below). Add a check for that case, and
> explicitly add the avx512 flags if necessary.
>
> Note: it appears that it is only the "native" flag that isn't
> overridden, which makes the issue hard to reproduce e.g. using
> godbolt.org, or on a modern machine. For example, testing with gcc 8.5
> on a haswell machine, using 'native' vs explicit 'haswell':
>
> gcc -march=native -march=skylake-avx512 -dM -E - < /dev/null | grep AVX
>
> gcc -march=haswell -march=skylake-avx512 -dM -E - < /dev/null | grep AVX
> ...
>
... and of course the output lines get stripped since they start with '#'
for '#define' :-(
> Bugzilla ID: 1736
> Fixes: e361ae3f59d3 ("build: reduce use of AVX compiler flags")
>
> Signed-off-by: Bruce Richardson <bruce.richardson@intel.com>
> ---
> config/x86/meson.build | 7 ++++++-
> 1 file changed, 6 insertions(+), 1 deletion(-)
>
> diff --git a/config/x86/meson.build b/config/x86/meson.build
> index e2ccfb6d12..0dcc5ddee4 100644
> --- a/config/x86/meson.build
> +++ b/config/x86/meson.build
> @@ -71,8 +71,13 @@ avx512_march_flag = '-march=x86-64-v4'
> if not cc.has_argument(avx512_march_flag)
> avx512_march_flag = '-march=skylake-avx512'
> endif
> +# workaround for older compilers, e.g. gcc 8.5 on RHEL 8.
> +# if march flag overriding doesn't work, explicitly add flags for AVX512.
> +if cc.get_define('__AVX512F__', args: [machine_args, avx512_march_flag]) == ''
> + avx512_march_flag = ['-mavx512f', '-mavx512bw', '-mavx512cd', '-mavx512dq', '-mavx512vl']
> +endif
> cc_avx512_flags = []
> -if (binutils_ok and cc.has_argument(avx512_march_flag)
> +if (binutils_ok and cc.has_multi_arguments(avx512_march_flag)
> and '-mno-avx512f' not in get_option('c_args'))
> # check if compiler is working with _mm512_extracti64x4_epi64
> # Ref: https://gcc.gnu.org/bugzilla/show_bug.cgi?id=82887
> --
> 2.48.1
>
^ permalink raw reply [flat|nested] 9+ messages in thread
* [PATCH v2] build/x86: fix support for older compilers
2025-07-02 15:51 [PATCH] build/x86: fix support for older compilers Bruce Richardson
2025-07-02 15:53 ` Bruce Richardson
@ 2025-07-02 16:00 ` Bruce Richardson
2025-07-02 18:22 ` Stephen Hemminger
2025-07-03 10:21 ` [PATCH v3] " Bruce Richardson
2 siblings, 1 reply; 9+ messages in thread
From: Bruce Richardson @ 2025-07-02 16:00 UTC (permalink / raw)
To: dev; +Cc: Bruce Richardson, Konstantin Ananyev, Vipin Varghese
Some older compilers e.g. gcc 8.5, do not support overriding
-march=native with another architecture, leading to build warnings such
as reported in Bugzilla (link below). Add a check for that case, and
explicitly add the avx512 flags if necessary.
Note: it appears that it is only the "native" flag that isn't
overridden, which makes the issue hard to reproduce e.g. using
godbolt.org, or on a modern machine. For example, testing with gcc 8.5
on a haswell machine, using 'native' vs explicit 'haswell':
gcc -march=native -march=skylake-avx512 -dM -E - < /dev/null | grep AVX512 | wc -l
0
gcc -march=haswell -march=skylake-avx512 -dM -E - < /dev/null | grep AVX512 | wc -l
5
Bugzilla ID: 1736
Fixes: e361ae3f59d3 ("build: reduce use of AVX compiler flags")
Signed-off-by: Bruce Richardson <bruce.richardson@intel.com>
---
config/x86/meson.build | 7 ++++++-
1 file changed, 6 insertions(+), 1 deletion(-)
diff --git a/config/x86/meson.build b/config/x86/meson.build
index e2ccfb6d12..0dcc5ddee4 100644
--- a/config/x86/meson.build
+++ b/config/x86/meson.build
@@ -71,8 +71,13 @@ avx512_march_flag = '-march=x86-64-v4'
if not cc.has_argument(avx512_march_flag)
avx512_march_flag = '-march=skylake-avx512'
endif
+# workaround for older compilers, e.g. gcc 8.5 on RHEL 8.
+# if march flag overriding doesn't work, explicitly add flags for AVX512.
+if cc.get_define('__AVX512F__', args: [machine_args, avx512_march_flag]) == ''
+ avx512_march_flag = ['-mavx512f', '-mavx512bw', '-mavx512cd', '-mavx512dq', '-mavx512vl']
+endif
cc_avx512_flags = []
-if (binutils_ok and cc.has_argument(avx512_march_flag)
+if (binutils_ok and cc.has_multi_arguments(avx512_march_flag)
and '-mno-avx512f' not in get_option('c_args'))
# check if compiler is working with _mm512_extracti64x4_epi64
# Ref: https://gcc.gnu.org/bugzilla/show_bug.cgi?id=82887
--
2.48.1
^ permalink raw reply [flat|nested] 9+ messages in thread
* Re: [PATCH v2] build/x86: fix support for older compilers
2025-07-02 16:00 ` [PATCH v2] " Bruce Richardson
@ 2025-07-02 18:22 ` Stephen Hemminger
2025-07-03 8:00 ` Bruce Richardson
0 siblings, 1 reply; 9+ messages in thread
From: Stephen Hemminger @ 2025-07-02 18:22 UTC (permalink / raw)
To: Bruce Richardson; +Cc: dev, Konstantin Ananyev, Vipin Varghese
On Wed, 2 Jul 2025 17:00:45 +0100
Bruce Richardson <bruce.richardson@intel.com> wrote:
> Some older compilers e.g. gcc 8.5, do not support overriding
> -march=native with another architecture, leading to build warnings such
> as reported in Bugzilla (link below). Add a check for that case, and
> explicitly add the avx512 flags if necessary.
>
> Note: it appears that it is only the "native" flag that isn't
> overridden, which makes the issue hard to reproduce e.g. using
> godbolt.org, or on a modern machine. For example, testing with gcc 8.5
> on a haswell machine, using 'native' vs explicit 'haswell':
>
> gcc -march=native -march=skylake-avx512 -dM -E - < /dev/null | grep AVX512 | wc -l
> 0
>
> gcc -march=haswell -march=skylake-avx512 -dM -E - < /dev/null | grep AVX512 | wc -l
> 5
>
> Bugzilla ID: 1736
> Fixes: e361ae3f59d3 ("build: reduce use of AVX compiler flags")
>
> Signed-off-by: Bruce Richardson <bruce.richardson@intel.com>
> ---
> config/x86/meson.build | 7 ++++++-
> 1 file changed, 6 insertions(+), 1 deletion(-)
>
> diff --git a/config/x86/meson.build b/config/x86/meson.build
> index e2ccfb6d12..0dcc5ddee4 100644
> --- a/config/x86/meson.build
> +++ b/config/x86/meson.build
> @@ -71,8 +71,13 @@ avx512_march_flag = '-march=x86-64-v4'
> if not cc.has_argument(avx512_march_flag)
> avx512_march_flag = '-march=skylake-avx512'
> endif
> +# workaround for older compilers, e.g. gcc 8.5 on RHEL 8.
> +# if march flag overriding doesn't work, explicitly add flags for AVX512.
> +if cc.get_define('__AVX512F__', args: [machine_args, avx512_march_flag]) == ''
> + avx512_march_flag = ['-mavx512f', '-mavx512bw', '-mavx512cd', '-mavx512dq', '-mavx512vl']
> +endif
RHEL 8.5 reached end of extended support cycle (EOL).
Why are we still supporting it?
^ permalink raw reply [flat|nested] 9+ messages in thread
* Re: [PATCH v2] build/x86: fix support for older compilers
2025-07-02 18:22 ` Stephen Hemminger
@ 2025-07-03 8:00 ` Bruce Richardson
2025-07-03 8:12 ` David Marchand
0 siblings, 1 reply; 9+ messages in thread
From: Bruce Richardson @ 2025-07-03 8:00 UTC (permalink / raw)
To: Stephen Hemminger; +Cc: dev, Konstantin Ananyev, Vipin Varghese
On Wed, Jul 02, 2025 at 11:22:42AM -0700, Stephen Hemminger wrote:
> On Wed, 2 Jul 2025 17:00:45 +0100
> Bruce Richardson <bruce.richardson@intel.com> wrote:
>
> > Some older compilers e.g. gcc 8.5, do not support overriding
> > -march=native with another architecture, leading to build warnings such
> > as reported in Bugzilla (link below). Add a check for that case, and
> > explicitly add the avx512 flags if necessary.
> >
> > Note: it appears that it is only the "native" flag that isn't
> > overridden, which makes the issue hard to reproduce e.g. using
> > godbolt.org, or on a modern machine. For example, testing with gcc 8.5
> > on a haswell machine, using 'native' vs explicit 'haswell':
> >
> > gcc -march=native -march=skylake-avx512 -dM -E - < /dev/null | grep AVX512 | wc -l
> > 0
> >
> > gcc -march=haswell -march=skylake-avx512 -dM -E - < /dev/null | grep AVX512 | wc -l
> > 5
> >
> > Bugzilla ID: 1736
> > Fixes: e361ae3f59d3 ("build: reduce use of AVX compiler flags")
> >
> > Signed-off-by: Bruce Richardson <bruce.richardson@intel.com>
> > ---
> > config/x86/meson.build | 7 ++++++-
> > 1 file changed, 6 insertions(+), 1 deletion(-)
> >
> > diff --git a/config/x86/meson.build b/config/x86/meson.build
> > index e2ccfb6d12..0dcc5ddee4 100644
> > --- a/config/x86/meson.build
> > +++ b/config/x86/meson.build
> > @@ -71,8 +71,13 @@ avx512_march_flag = '-march=x86-64-v4'
> > if not cc.has_argument(avx512_march_flag)
> > avx512_march_flag = '-march=skylake-avx512'
> > endif
> > +# workaround for older compilers, e.g. gcc 8.5 on RHEL 8.
> > +# if march flag overriding doesn't work, explicitly add flags for AVX512.
> > +if cc.get_define('__AVX512F__', args: [machine_args, avx512_march_flag]) == ''
> > + avx512_march_flag = ['-mavx512f', '-mavx512bw', '-mavx512cd', '-mavx512dq', '-mavx512vl']
> > +endif
>
> RHEL 8.5 reached end of extended support cycle (EOL).
> Why are we still supporting it?
Folks working for Redhat can answer better than me, but according to [1]
it's not EOL yet. Also, we support it in DPDK because users are still using
it - we had two separate people report seeing the issue on RHEL 8 on RC2,
so likely a reasonable number will be using it with the actual release.
/Bruce
[1] https://en.wikipedia.org/wiki/Red_Hat_Enterprise_Linux#Version_history_and_timeline
^ permalink raw reply [flat|nested] 9+ messages in thread
* Re: [PATCH v2] build/x86: fix support for older compilers
2025-07-03 8:00 ` Bruce Richardson
@ 2025-07-03 8:12 ` David Marchand
2025-07-03 16:10 ` Stephen Hemminger
0 siblings, 1 reply; 9+ messages in thread
From: David Marchand @ 2025-07-03 8:12 UTC (permalink / raw)
To: Bruce Richardson
Cc: Stephen Hemminger, dev, Konstantin Ananyev, Vipin Varghese
On Thu, Jul 3, 2025 at 10:01 AM Bruce Richardson
<bruce.richardson@intel.com> wrote:
>
> On Wed, Jul 02, 2025 at 11:22:42AM -0700, Stephen Hemminger wrote:
> > On Wed, 2 Jul 2025 17:00:45 +0100
> > Bruce Richardson <bruce.richardson@intel.com> wrote:
> >
> > > Some older compilers e.g. gcc 8.5, do not support overriding
> > > -march=native with another architecture, leading to build warnings such
> > > as reported in Bugzilla (link below). Add a check for that case, and
> > > explicitly add the avx512 flags if necessary.
> > >
> > > Note: it appears that it is only the "native" flag that isn't
> > > overridden, which makes the issue hard to reproduce e.g. using
> > > godbolt.org, or on a modern machine. For example, testing with gcc 8.5
> > > on a haswell machine, using 'native' vs explicit 'haswell':
> > >
> > > gcc -march=native -march=skylake-avx512 -dM -E - < /dev/null | grep AVX512 | wc -l
> > > 0
> > >
> > > gcc -march=haswell -march=skylake-avx512 -dM -E - < /dev/null | grep AVX512 | wc -l
> > > 5
> > >
> > > Bugzilla ID: 1736
> > > Fixes: e361ae3f59d3 ("build: reduce use of AVX compiler flags")
> > >
> > > Signed-off-by: Bruce Richardson <bruce.richardson@intel.com>
> > > ---
> > > config/x86/meson.build | 7 ++++++-
> > > 1 file changed, 6 insertions(+), 1 deletion(-)
> > >
> > > diff --git a/config/x86/meson.build b/config/x86/meson.build
> > > index e2ccfb6d12..0dcc5ddee4 100644
> > > --- a/config/x86/meson.build
> > > +++ b/config/x86/meson.build
> > > @@ -71,8 +71,13 @@ avx512_march_flag = '-march=x86-64-v4'
> > > if not cc.has_argument(avx512_march_flag)
> > > avx512_march_flag = '-march=skylake-avx512'
> > > endif
> > > +# workaround for older compilers, e.g. gcc 8.5 on RHEL 8.
> > > +# if march flag overriding doesn't work, explicitly add flags for AVX512.
> > > +if cc.get_define('__AVX512F__', args: [machine_args, avx512_march_flag]) == ''
> > > + avx512_march_flag = ['-mavx512f', '-mavx512bw', '-mavx512cd', '-mavx512dq', '-mavx512vl']
> > > +endif
> >
> > RHEL 8.5 reached end of extended support cycle (EOL).
> > Why are we still supporting it?
>
> Folks working for Redhat can answer better than me, but according to [1]
> it's not EOL yet. Also, we support it in DPDK because users are still using
RHEL 8 is not EOL.
--
David Marchand
^ permalink raw reply [flat|nested] 9+ messages in thread
* Re: [PATCH v2] build/x86: fix support for older compilers
2025-07-03 8:12 ` David Marchand
@ 2025-07-03 16:10 ` Stephen Hemminger
0 siblings, 0 replies; 9+ messages in thread
From: Stephen Hemminger @ 2025-07-03 16:10 UTC (permalink / raw)
To: David Marchand; +Cc: Bruce Richardson, dev, Konstantin Ananyev, Vipin Varghese
On Thu, 3 Jul 2025 10:12:10 +0200
David Marchand <david.marchand@redhat.com> wrote:
> On Thu, Jul 3, 2025 at 10:01 AM Bruce Richardson
> <bruce.richardson@intel.com> wrote:
> >
> > On Wed, Jul 02, 2025 at 11:22:42AM -0700, Stephen Hemminger wrote:
> > > On Wed, 2 Jul 2025 17:00:45 +0100
> > > Bruce Richardson <bruce.richardson@intel.com> wrote:
> > >
> > > > Some older compilers e.g. gcc 8.5, do not support overriding
> > > > -march=native with another architecture, leading to build warnings such
> > > > as reported in Bugzilla (link below). Add a check for that case, and
> > > > explicitly add the avx512 flags if necessary.
> > > >
> > > > Note: it appears that it is only the "native" flag that isn't
> > > > overridden, which makes the issue hard to reproduce e.g. using
> > > > godbolt.org, or on a modern machine. For example, testing with gcc 8.5
> > > > on a haswell machine, using 'native' vs explicit 'haswell':
> > > >
> > > > gcc -march=native -march=skylake-avx512 -dM -E - < /dev/null | grep AVX512 | wc -l
> > > > 0
> > > >
> > > > gcc -march=haswell -march=skylake-avx512 -dM -E - < /dev/null | grep AVX512 | wc -l
> > > > 5
> > > >
> > > > Bugzilla ID: 1736
> > > > Fixes: e361ae3f59d3 ("build: reduce use of AVX compiler flags")
> > > >
> > > > Signed-off-by: Bruce Richardson <bruce.richardson@intel.com>
> > > > ---
> > > > config/x86/meson.build | 7 ++++++-
> > > > 1 file changed, 6 insertions(+), 1 deletion(-)
> > > >
> > > > diff --git a/config/x86/meson.build b/config/x86/meson.build
> > > > index e2ccfb6d12..0dcc5ddee4 100644
> > > > --- a/config/x86/meson.build
> > > > +++ b/config/x86/meson.build
> > > > @@ -71,8 +71,13 @@ avx512_march_flag = '-march=x86-64-v4'
> > > > if not cc.has_argument(avx512_march_flag)
> > > > avx512_march_flag = '-march=skylake-avx512'
> > > > endif
> > > > +# workaround for older compilers, e.g. gcc 8.5 on RHEL 8.
> > > > +# if march flag overriding doesn't work, explicitly add flags for AVX512.
> > > > +if cc.get_define('__AVX512F__', args: [machine_args, avx512_march_flag]) == ''
> > > > + avx512_march_flag = ['-mavx512f', '-mavx512bw', '-mavx512cd', '-mavx512dq', '-mavx512vl']
> > > > +endif
> > >
> > > RHEL 8.5 reached end of extended support cycle (EOL).
> > > Why are we still supporting it?
> >
> > Folks working for Redhat can answer better than me, but according to [1]
> > it's not EOL yet. Also, we support it in DPDK because users are still using
>
> RHEL 8 is not EOL.
>
>
Right but 8.5 is, did compiler get fixed in later 8.X versions?
^ permalink raw reply [flat|nested] 9+ messages in thread
* [PATCH v3] build/x86: fix support for older compilers
2025-07-02 15:51 [PATCH] build/x86: fix support for older compilers Bruce Richardson
2025-07-02 15:53 ` Bruce Richardson
2025-07-02 16:00 ` [PATCH v2] " Bruce Richardson
@ 2025-07-03 10:21 ` Bruce Richardson
2025-07-03 13:23 ` Thomas Monjalon
2 siblings, 1 reply; 9+ messages in thread
From: Bruce Richardson @ 2025-07-03 10:21 UTC (permalink / raw)
To: dev; +Cc: Bruce Richardson, Konstantin Ananyev, Vipin Varghese
Some older compilers e.g. gcc 8.5, do not support overriding
-march=native with another architecture, leading to build warnings such
as reported in Bugzilla (link below). Add a check for that case, and
explicitly add the avx512 flags if necessary.
Note: it appears that it is only the "native" flag that isn't
overridden, which makes the issue hard to reproduce e.g. using
godbolt.org, or on a modern machine. For example, testing with gcc 8.5
on a haswell machine, using 'native' vs explicit 'haswell':
gcc -march=native -march=skylake-avx512 -dM -E - < /dev/null | grep AVX512 | wc -l
0
gcc -march=haswell -march=skylake-avx512 -dM -E - < /dev/null | grep AVX512 | wc -l
5
Bugzilla ID: 1736
Fixes: e361ae3f59d3 ("build: reduce use of AVX compiler flags")
Signed-off-by: Bruce Richardson <bruce.richardson@intel.com>
---
V3: fix build when using meson 0.57 which doesn't flatten the sub-arrays
V2: fix commit log
---
config/x86/meson.build | 15 ++++++++++-----
1 file changed, 10 insertions(+), 5 deletions(-)
diff --git a/config/x86/meson.build b/config/x86/meson.build
index e2ccfb6d12..bd135eb02d 100644
--- a/config/x86/meson.build
+++ b/config/x86/meson.build
@@ -67,26 +67,31 @@ if is_linux or cc.get_id() == 'gcc'
endif
endif
-avx512_march_flag = '-march=x86-64-v4'
+avx512_march_flag = ['-march=x86-64-v4']
if not cc.has_argument(avx512_march_flag)
- avx512_march_flag = '-march=skylake-avx512'
+ avx512_march_flag = ['-march=skylake-avx512']
+endif
+# workaround for older compilers, e.g. gcc 8.5 on RHEL 8.
+# if march flag overriding doesn't work, explicitly add flags for AVX512.
+if cc.get_define('__AVX512F__', args: machine_args + avx512_march_flag) == ''
+ avx512_march_flag = ['-mavx512f', '-mavx512bw', '-mavx512cd', '-mavx512dq', '-mavx512vl']
endif
cc_avx512_flags = []
-if (binutils_ok and cc.has_argument(avx512_march_flag)
+if (binutils_ok and cc.has_multi_arguments(avx512_march_flag)
and '-mno-avx512f' not in get_option('c_args'))
# check if compiler is working with _mm512_extracti64x4_epi64
# Ref: https://gcc.gnu.org/bugzilla/show_bug.cgi?id=82887
code = '''#include <immintrin.h>
void test(__m512i zmm){
__m256i ymm = _mm512_extracti64x4_epi64(zmm, 0);}'''
- result = cc.compiles(code, args : [avx512_march_flag], name : 'AVX512 checking')
+ result = cc.compiles(code, args : avx512_march_flag, name : 'AVX512 checking')
if result == false
machine_args += '-mno-avx512f'
warning('Broken _mm512_extracti64x4_epi64, disabling AVX512 support')
else
cc_has_avx512 = true
if cc.get_define('__AVX512F__', args: machine_args) == ''
- cc_avx512_flags = [avx512_march_flag]
+ cc_avx512_flags = avx512_march_flag
if cc.has_argument('-Wno-overriding-option')
cc_avx512_flags += '-Wno-overriding-option'
endif
--
2.48.1
^ permalink raw reply [flat|nested] 9+ messages in thread
* Re: [PATCH v3] build/x86: fix support for older compilers
2025-07-03 10:21 ` [PATCH v3] " Bruce Richardson
@ 2025-07-03 13:23 ` Thomas Monjalon
0 siblings, 0 replies; 9+ messages in thread
From: Thomas Monjalon @ 2025-07-03 13:23 UTC (permalink / raw)
To: Bruce Richardson
Cc: dev, Konstantin Ananyev, Vipin Varghese, Alaa Duhaidy, Khadem Ullah
03/07/2025 12:21, Bruce Richardson:
> Some older compilers e.g. gcc 8.5, do not support overriding
> -march=native with another architecture, leading to build warnings such
> as reported in Bugzilla (link below). Add a check for that case, and
> explicitly add the avx512 flags if necessary.
>
> Note: it appears that it is only the "native" flag that isn't
> overridden, which makes the issue hard to reproduce e.g. using
> godbolt.org, or on a modern machine. For example, testing with gcc 8.5
> on a haswell machine, using 'native' vs explicit 'haswell':
>
> gcc -march=native -march=skylake-avx512 -dM -E - < /dev/null | grep AVX512 | wc -l
> 0
>
> gcc -march=haswell -march=skylake-avx512 -dM -E - < /dev/null | grep AVX512 | wc -l
> 5
>
> Bugzilla ID: 1736
> Fixes: e361ae3f59d3 ("build: reduce use of AVX compiler flags")
>
> Signed-off-by: Bruce Richardson <bruce.richardson@intel.com>
Applied, thanks.
^ permalink raw reply [flat|nested] 9+ messages in thread