[PATCH 0/3] remove driver-specific logic for AVX builds

[PATCH 0/3] remove driver-specific logic for AVX builds

[Bruce Richardson]

A number of drivers had special optimized AVX2 and AVX512 code paths
for performance reasons, and these tended to have copy-pasted logic
to build those files. Centralise that logic in the main
drivers/meson.build file to avoid duplication.

Bruce Richardson (3):
  build: add generalized AVX handling for drivers
  net/intel: use common AVX build code
  drivers/net: build use common AVX handling

 drivers/meson.build                | 30 ++++++++++++++++++++++++++++++
 drivers/net/bnxt/meson.build       | 10 +---------
 drivers/net/enic/meson.build       | 10 +---------
 drivers/net/intel/i40e/meson.build | 26 ++------------------------
 drivers/net/intel/iavf/meson.build | 25 ++-----------------------
 drivers/net/intel/ice/meson.build  | 25 ++-----------------------
 drivers/net/intel/idpf/meson.build | 25 ++-----------------------
 drivers/net/nfp/meson.build        | 10 +---------
 drivers/net/octeon_ep/meson.build  | 14 ++------------
 drivers/net/virtio/meson.build     |  9 +--------
 10 files changed, 44 insertions(+), 140 deletions(-)

--
2.43.0

[PATCH 1/3] build: add generalized AVX handling for drivers

[Bruce Richardson]

Add support to the top-level driver build file for AVX2 and AVX512
specific sources. This should simplify driver builds by avoiding the
need to constantly reimplement the same build logic

Signed-off-by: Bruce Richardson <bruce.richardson@intel.com>
---
 drivers/meson.build | 30 ++++++++++++++++++++++++++++++
 1 file changed, 30 insertions(+)

diff --git a/drivers/meson.build b/drivers/meson.build
index 05391a575d..c42c7764bf 100644
--- a/drivers/meson.build
+++ b/drivers/meson.build
@@ -126,6 +126,8 @@ foreach subpath:subdirs
         name = drv
         annotate_locks = true
         sources = []
+        sources_avx2 = []
+        sources_avx512 = []
         headers = []
         driver_sdk_headers = [] # public headers included by drivers
         objs = []
@@ -235,6 +237,34 @@ foreach subpath:subdirs
             dpdk_includes += include_directories(drv_path)
         endif
 
+        # handle avx2 and avx512 source files
+        if arch_subdir == 'x86'
+            if sources_avx2.length() > 0
+                avx2_lib = static_library(lib_name + '_avx2_lib',
+                        sources_avx2,
+                        dependencies: static_deps,
+                        include_directories: includes,
+                        c_args: [cflags, cc_avx2_flags])
+                objs += avx2_lib.extract_objects(sources_avx2)
+            endif
+            if sources_avx512.length() > 0
+                cflags += '-DCC_AVX512_SUPPORT'
+                avx512_args = [cflags, cc_avx512_flags]
+                if cc.has_argument('-march=skylake-avx512')
+                    avx512_args += '-march=skylake-avx512'
+                    if cc.has_argument('-Wno-overriding-option')
+                        avx512_args += '-Wno-overriding-option'
+                    endif
+                endif
+                avx512_lib = static_library(lib_name + '_avx512_lib',
+                        sources_avx512,
+                        dependencies: static_deps,
+                        include_directories: includes,
+                        c_args: avx512_args)
+                objs += avx512_lib.extract_objects(sources_avx512)
+            endif
+        endif
+
         # generate pmdinfo sources by building a temporary
         # lib and then running pmdinfogen on the contents of
         # that lib. The final lib reuses the object files and
-- 
2.43.0

[PATCH 2/3] net/intel: use common AVX build code

[Bruce Richardson]

Remove driver-specific build instructions for the AVX2 and AVX-512 code,
and rely instead on the generic driver build file.

Signed-off-by: Bruce Richardson <bruce.richardson@intel.com>
---
 drivers/net/intel/i40e/meson.build | 26 ++------------------------
 drivers/net/intel/iavf/meson.build | 25 ++-----------------------
 drivers/net/intel/ice/meson.build  | 25 ++-----------------------
 drivers/net/intel/idpf/meson.build | 25 ++-----------------------
 4 files changed, 8 insertions(+), 93 deletions(-)

diff --git a/drivers/net/intel/i40e/meson.build b/drivers/net/intel/i40e/meson.build
index 2973ed1a01..25a3d72714 100644
--- a/drivers/net/intel/i40e/meson.build
+++ b/drivers/net/intel/i40e/meson.build
@@ -40,35 +40,13 @@ includes += include_directories('base')
 
 if arch_subdir == 'x86'
     sources += files('i40e_rxtx_vec_sse.c')
+    sources_avx2 = files('i40e_rxtx_vec_avx2.c')
+    sources_avx512 = files('i40e_rxtx_vec_avx512.c')
 
     if is_windows and cc.get_id() != 'clang'
         cflags += ['-fno-asynchronous-unwind-tables']
     endif
 
-    i40e_avx2_lib = static_library('i40e_avx2_lib',
-            'i40e_rxtx_vec_avx2.c',
-            dependencies: [static_rte_ethdev, static_rte_kvargs, static_rte_hash],
-            include_directories: includes,
-            c_args: [cflags, cc_avx2_flags])
-    objs += i40e_avx2_lib.extract_objects('i40e_rxtx_vec_avx2.c')
-
-    if cc_has_avx512
-        cflags += ['-DCC_AVX512_SUPPORT']
-        avx512_args = cflags + cc_avx512_flags
-        if cc.has_argument('-march=skylake-avx512')
-            avx512_args += '-march=skylake-avx512'
-            if cc.has_argument('-Wno-overriding-option')
-                avx512_args += '-Wno-overriding-option'
-            endif
-        endif
-        i40e_avx512_lib = static_library('i40e_avx512_lib',
-                'i40e_rxtx_vec_avx512.c',
-                dependencies: [static_rte_ethdev,
-                    static_rte_kvargs, static_rte_hash],
-                include_directories: includes,
-                c_args: avx512_args)
-        objs += i40e_avx512_lib.extract_objects('i40e_rxtx_vec_avx512.c')
-    endif
 elif arch_subdir == 'ppc'
        sources += files('i40e_rxtx_vec_altivec.c')
 elif arch_subdir == 'arm'
diff --git a/drivers/net/intel/iavf/meson.build b/drivers/net/intel/iavf/meson.build
index f7eac7c57a..fc3e4d2f28 100644
--- a/drivers/net/intel/iavf/meson.build
+++ b/drivers/net/intel/iavf/meson.build
@@ -28,34 +28,13 @@ includes += include_directories('base')
 
 if arch_subdir == 'x86'
     sources += files('iavf_rxtx_vec_sse.c')
+    sources_avx2 = files('iavf_rxtx_vec_avx2.c')
+    sources_avx512 = files('iavf_rxtx_vec_avx512.c')
 
     if is_windows and cc.get_id() != 'clang'
         cflags += ['-fno-asynchronous-unwind-tables']
     endif
 
-    iavf_avx2_lib = static_library('iavf_avx2_lib',
-            'iavf_rxtx_vec_avx2.c',
-            dependencies: [static_rte_ethdev],
-            include_directories: includes,
-            c_args: [cflags, cc_avx2_flags])
-    objs += iavf_avx2_lib.extract_objects('iavf_rxtx_vec_avx2.c')
-
-    if cc_has_avx512
-        cflags += ['-DCC_AVX512_SUPPORT']
-        avx512_args = cflags + cc_avx512_flags
-        if cc.has_argument('-march=skylake-avx512')
-            avx512_args += '-march=skylake-avx512'
-            if cc.has_argument('-Wno-overriding-option')
-                avx512_args += '-Wno-overriding-option'
-            endif
-        endif
-        iavf_avx512_lib = static_library('iavf_avx512_lib',
-                'iavf_rxtx_vec_avx512.c',
-                dependencies: [static_rte_ethdev],
-                include_directories: includes,
-                c_args: avx512_args)
-        objs += iavf_avx512_lib.extract_objects('iavf_rxtx_vec_avx512.c')
-    endif
 elif arch_subdir == 'arm'
     sources += files('iavf_rxtx_vec_neon.c')
 endif
diff --git a/drivers/net/intel/ice/meson.build b/drivers/net/intel/ice/meson.build
index cbdf38c1c4..a33ec59272 100644
--- a/drivers/net/intel/ice/meson.build
+++ b/drivers/net/intel/ice/meson.build
@@ -34,34 +34,13 @@ endif
 
 if arch_subdir == 'x86'
     sources += files('ice_rxtx_vec_sse.c')
+    sources_avx2 = files('ice_rxtx_vec_avx2.c')
+    sources_avx512 = files('ice_rxtx_vec_avx512.c')
 
     if is_windows and cc.get_id() != 'clang'
         cflags += ['-fno-asynchronous-unwind-tables']
     endif
 
-    ice_avx2_lib = static_library('ice_avx2_lib',
-            'ice_rxtx_vec_avx2.c',
-            dependencies: [static_rte_ethdev, static_rte_hash],
-            include_directories: includes,
-            c_args: [cflags, cc_avx2_flags])
-    objs += ice_avx2_lib.extract_objects('ice_rxtx_vec_avx2.c')
-
-    if cc_has_avx512
-        cflags += ['-DCC_AVX512_SUPPORT']
-        avx512_args = cflags + cc_avx512_flags
-        if cc.has_argument('-march=skylake-avx512')
-            avx512_args += '-march=skylake-avx512'
-            if cc.has_argument('-Wno-overriding-option')
-                avx512_args += '-Wno-overriding-option'
-            endif
-        endif
-        ice_avx512_lib = static_library('ice_avx512_lib',
-                'ice_rxtx_vec_avx512.c',
-                dependencies: [static_rte_ethdev, static_rte_hash],
-                include_directories: includes,
-                c_args: avx512_args)
-        objs += ice_avx512_lib.extract_objects('ice_rxtx_vec_avx512.c')
-    endif
 endif
 
 sources += files(
diff --git a/drivers/net/intel/idpf/meson.build b/drivers/net/intel/idpf/meson.build
index 4b272d02b1..10465f0f36 100644
--- a/drivers/net/intel/idpf/meson.build
+++ b/drivers/net/intel/idpf/meson.build
@@ -19,29 +19,8 @@ sources = files(
 )
 
 if arch_subdir == 'x86' and dpdk_conf.get('RTE_IOVA_IN_MBUF') == 1
-    idpf_avx2_lib = static_library('idpf_avx2_lib',
-        'idpf_common_rxtx_avx2.c',
-        dependencies: [static_rte_ethdev, static_rte_hash],
-        include_directories: includes,
-        c_args: [cflags, cc_avx2_flags])
-    objs += idpf_avx2_lib.extract_objects('idpf_common_rxtx_avx2.c')
-
-    if cc_has_avx512
-        cflags += ['-DCC_AVX512_SUPPORT']
-        avx512_args = cflags + cc_avx512_flags
-        if cc.has_argument('-march=skylake-avx512')
-            avx512_args += '-march=skylake-avx512'
-            if cc.has_argument('-Wno-overriding-option')
-                avx512_args += '-Wno-overriding-option'
-            endif
-        endif
-        idpf_common_avx512_lib = static_library('idpf_common_avx512_lib',
-                'idpf_common_rxtx_avx512.c',
-                dependencies: static_rte_mbuf,
-                include_directories: includes,
-                c_args: avx512_args)
-        objs += idpf_common_avx512_lib.extract_objects('idpf_common_rxtx_avx512.c')
-    endif
+    sources_avx2 = files('idpf_common_rxtx_avx2.c')
+    sources_avx512 = files('idpf_common_rxtx_avx512.c')
 endif
 
 subdir('base')
-- 
2.43.0

[PATCH 3/3] drivers/net: build use common AVX handling

[Bruce Richardson]

Remove from remaining net drivers the special-case code to handle AVX2
or AVX512 specific files. These can be built instead using
drivers/meson.build.

Signed-off-by: Bruce Richardson <bruce.richardson@intel.com>
---
 drivers/net/bnxt/meson.build      | 10 +---------
 drivers/net/enic/meson.build      | 10 +---------
 drivers/net/nfp/meson.build       | 10 +---------
 drivers/net/octeon_ep/meson.build | 14 ++------------
 drivers/net/virtio/meson.build    |  9 +--------
 5 files changed, 6 insertions(+), 47 deletions(-)

diff --git a/drivers/net/bnxt/meson.build b/drivers/net/bnxt/meson.build
index fd82d0c409..dcca7df916 100644
--- a/drivers/net/bnxt/meson.build
+++ b/drivers/net/bnxt/meson.build
@@ -58,15 +58,7 @@ subdir('hcapi/cfa_v3')
 
 if arch_subdir == 'x86'
     sources += files('bnxt_rxtx_vec_sse.c')
-    # build AVX2 code with instruction set explicitly enabled for runtime selection
-    bnxt_avx2_lib = static_library('bnxt_avx2_lib',
-            'bnxt_rxtx_vec_avx2.c',
-            dependencies: [static_rte_ethdev,
-                static_rte_bus_pci,
-                static_rte_kvargs, static_rte_hash],
-            include_directories: includes,
-            c_args: [cflags, cc_avx2_flags])
-     objs += bnxt_avx2_lib.extract_objects('bnxt_rxtx_vec_avx2.c')
+    sources_avx2 = files('bnxt_rxtx_vec_avx2.c')
 elif arch_subdir == 'arm' and dpdk_conf.get('RTE_ARCH_64')
     sources += files('bnxt_rxtx_vec_neon.c')
 endif
diff --git a/drivers/net/enic/meson.build b/drivers/net/enic/meson.build
index cfe5ec170a..2b3052fae8 100644
--- a/drivers/net/enic/meson.build
+++ b/drivers/net/enic/meson.build
@@ -29,17 +29,9 @@ sources = files(
 deps += ['hash']
 includes += include_directories('base')
 
-# Build the avx2 handler for 64-bit X86 targets, even though 'machine'
-# may not. This is to support users who build for the min supported machine
-# and need to run the binary on newer CPUs too.
 if dpdk_conf.has('RTE_ARCH_X86_64')
     cflags += '-DENIC_RXTX_VEC'
-    enic_avx2_lib = static_library('enic_avx2_lib',
-            'enic_rxtx_vec_avx2.c',
-            dependencies: [static_rte_ethdev, static_rte_bus_pci],
-            include_directories: includes,
-            c_args: [cflags, cc_avx2_flags])
-    objs += enic_avx2_lib.extract_objects('enic_rxtx_vec_avx2.c')
+    sources_avx2 = files('enic_rxtx_vec_avx2.c')
 endif
 
 annotate_locks = false
diff --git a/drivers/net/nfp/meson.build b/drivers/net/nfp/meson.build
index 0a12b7dce7..a98b584042 100644
--- a/drivers/net/nfp/meson.build
+++ b/drivers/net/nfp/meson.build
@@ -52,19 +52,11 @@ cflags += no_wvla_cflag
 if arch_subdir == 'x86'
         includes += include_directories('../../common/nfp')
 
-        avx2_sources = files(
+        sources_avx2 = files(
                 'nfdk/nfp_nfdk_vec_avx2_dp.c',
                 'nfp_rxtx_vec_avx2.c',
         )
 
-        nfp_avx2_lib = static_library('nfp_avx2_lib',
-                avx2_sources,
-                dependencies: [static_rte_ethdev, static_rte_bus_pci],
-                include_directories: includes,
-                c_args: [cflags, cc_avx2_flags]
-        )
-
-        objs += nfp_avx2_lib.extract_all_objects(recursive: true)
 else
         sources += files(
                 'nfp_rxtx_vec_stub.c',
diff --git a/drivers/net/octeon_ep/meson.build b/drivers/net/octeon_ep/meson.build
index 1b34db3edc..9bf4627894 100644
--- a/drivers/net/octeon_ep/meson.build
+++ b/drivers/net/octeon_ep/meson.build
@@ -15,18 +15,8 @@ sources = files(
 
 if arch_subdir == 'x86'
     sources += files('cnxk_ep_rx_sse.c')
-    if cc.get_define('__AVX2__', args: machine_args) != ''
-        cflags += ['-DCC_AVX2_SUPPORT']
-        sources += files('cnxk_ep_rx_avx.c')
-    elif cc.has_multi_arguments(cc_avx2_flags)
-        cflags += ['-DCC_AVX2_SUPPORT']
-        otx_ep_avx2_lib = static_library('otx_ep_avx2_lib',
-                        'cnxk_ep_rx_avx.c',
-                        dependencies: [static_rte_ethdev, static_rte_pci, static_rte_bus_pci],
-                        include_directories: includes,
-                        c_args: [cflags, cc_avx2_flags])
-        objs += otx_ep_avx2_lib.extract_objects('cnxk_ep_rx_avx.c')
-    endif
+    cflags += ['-DCC_AVX2_SUPPORT']
+    sources_avx2 = files('cnxk_ep_rx_avx.c')
 endif
 
 if arch_subdir == 'arm'
diff --git a/drivers/net/virtio/meson.build b/drivers/net/virtio/meson.build
index c1c4a85bea..01bfb3c47d 100644
--- a/drivers/net/virtio/meson.build
+++ b/drivers/net/virtio/meson.build
@@ -27,15 +27,8 @@ cflags += no_wvla_cflag
 
 if arch_subdir == 'x86'
     if cc_has_avx512
-        cflags += ['-DCC_AVX512_SUPPORT']
         cflags += ['-DVIRTIO_RXTX_PACKED_VEC']
-        virtio_avx512_lib = static_library('virtio_avx512_lib',
-                'virtio_rxtx_packed.c',
-                dependencies: [static_rte_ethdev,
-                    static_rte_kvargs, static_rte_bus_pci],
-                include_directories: includes,
-                c_args: cflags + cc_avx512_flags)
-        objs += virtio_avx512_lib.extract_objects('virtio_rxtx_packed.c')
+        sources_avx512 = files('virtio_rxtx_packed.c')
         if (toolchain == 'gcc' and cc.version().version_compare('>=8.3.0'))
             cflags += '-DVIRTIO_GCC_UNROLL_PRAGMA'
         elif (toolchain == 'clang' and cc.version().version_compare('>=3.7.0'))
-- 
2.43.0

Re: [PATCH 3/3] drivers/net: build use common AVX handling

[David Marchand]

On Fri, Mar 14, 2025 at 6:24 PM Bruce Richardson
<bruce.richardson@intel.com> wrote:
> diff --git a/drivers/net/octeon_ep/meson.build b/drivers/net/octeon_ep/meson.build
> index 1b34db3edc..9bf4627894 100644
> --- a/drivers/net/octeon_ep/meson.build
> +++ b/drivers/net/octeon_ep/meson.build
> @@ -15,18 +15,8 @@ sources = files(
>
>  if arch_subdir == 'x86'
>      sources += files('cnxk_ep_rx_sse.c')
> -    if cc.get_define('__AVX2__', args: machine_args) != ''
> -        cflags += ['-DCC_AVX2_SUPPORT']
> -        sources += files('cnxk_ep_rx_avx.c')
> -    elif cc.has_multi_arguments(cc_avx2_flags)
> -        cflags += ['-DCC_AVX2_SUPPORT']
> -        otx_ep_avx2_lib = static_library('otx_ep_avx2_lib',
> -                        'cnxk_ep_rx_avx.c',
> -                        dependencies: [static_rte_ethdev, static_rte_pci, static_rte_bus_pci],
> -                        include_directories: includes,
> -                        c_args: [cflags, cc_avx2_flags])
> -        objs += otx_ep_avx2_lib.extract_objects('cnxk_ep_rx_avx.c')
> -    endif
> +    cflags += ['-DCC_AVX2_SUPPORT']

Could it be set by drivers/meson.build?
Similarly to AVX512.


> +    sources_avx2 = files('cnxk_ep_rx_avx.c')
>  endif
>
>  if arch_subdir == 'arm'

-- 
David Marchand

Re: [PATCH 3/3] drivers/net: build use common AVX handling

[Bruce Richardson]

On Fri, Mar 14, 2025 at 06:31:08PM +0100, David Marchand wrote:
> On Fri, Mar 14, 2025 at 6:24 PM Bruce Richardson
> <bruce.richardson@intel.com> wrote:
> > diff --git a/drivers/net/octeon_ep/meson.build b/drivers/net/octeon_ep/meson.build
> > index 1b34db3edc..9bf4627894 100644
> > --- a/drivers/net/octeon_ep/meson.build
> > +++ b/drivers/net/octeon_ep/meson.build
> > @@ -15,18 +15,8 @@ sources = files(
> >
> >  if arch_subdir == 'x86'
> >      sources += files('cnxk_ep_rx_sse.c')
> > -    if cc.get_define('__AVX2__', args: machine_args) != ''
> > -        cflags += ['-DCC_AVX2_SUPPORT']
> > -        sources += files('cnxk_ep_rx_avx.c')
> > -    elif cc.has_multi_arguments(cc_avx2_flags)
> > -        cflags += ['-DCC_AVX2_SUPPORT']
> > -        otx_ep_avx2_lib = static_library('otx_ep_avx2_lib',
> > -                        'cnxk_ep_rx_avx.c',
> > -                        dependencies: [static_rte_ethdev, static_rte_pci, static_rte_bus_pci],
> > -                        include_directories: includes,
> > -                        c_args: [cflags, cc_avx2_flags])
> > -        objs += otx_ep_avx2_lib.extract_objects('cnxk_ep_rx_avx.c')
> > -    endif
> > +    cflags += ['-DCC_AVX2_SUPPORT']
> 
> Could it be set by drivers/meson.build?
> Similarly to AVX512.
> 
Yes, but really I'd rather see it removed. It's now the same as the defines
for x86, since we always build AVX2 support when building for x86, because
all compilers we support can generate AVX2 code.

/Bruce

[PATCH v2 0/4] remove driver-specific logic for AVX builds

[Bruce Richardson]

A number of drivers had special optimized AVX2 and AVX512 code paths
for performance reasons, and these tended to have copy-pasted logic
to build those files. Centralise that logic in the main
drivers/meson.build file to avoid duplication.

v2: add patch 4 to remove use of unnecessary CC_AVX2_SUPPORT flag

Bruce Richardson (4):
  build: add generalized AVX handling for drivers
  net/intel: use common AVX build code
  drivers/net: build use common AVX handling
  drivers/net: remove AVX2 build-time define

 drivers/meson.build                   | 30 +++++++++++++++++++++++++++
 drivers/net/bnxt/bnxt_ethdev.c        |  2 --
 drivers/net/bnxt/meson.build          | 10 +--------
 drivers/net/enic/meson.build          | 10 +--------
 drivers/net/intel/i40e/meson.build    | 26 ++---------------------
 drivers/net/intel/iavf/meson.build    | 25 ++--------------------
 drivers/net/intel/ice/meson.build     | 25 ++--------------------
 drivers/net/intel/idpf/meson.build    | 25 ++--------------------
 drivers/net/nfp/meson.build           | 10 +--------
 drivers/net/octeon_ep/meson.build     | 13 +-----------
 drivers/net/octeon_ep/otx_ep_ethdev.c |  4 ----
 drivers/net/virtio/meson.build        |  9 +-------
 12 files changed, 43 insertions(+), 146 deletions(-)

--
2.43.0

[PATCH v2 1/4] build: add generalized AVX handling for drivers

[Bruce Richardson]

Add support to the top-level driver build file for AVX2 and AVX512
specific sources. This should simplify driver builds by avoiding the
need to constantly reimplement the same build logic

Signed-off-by: Bruce Richardson <bruce.richardson@intel.com>
---
 drivers/meson.build | 30 ++++++++++++++++++++++++++++++
 1 file changed, 30 insertions(+)

diff --git a/drivers/meson.build b/drivers/meson.build
index 05391a575d..c42c7764bf 100644
--- a/drivers/meson.build
+++ b/drivers/meson.build
@@ -126,6 +126,8 @@ foreach subpath:subdirs
         name = drv
         annotate_locks = true
         sources = []
+        sources_avx2 = []
+        sources_avx512 = []
         headers = []
         driver_sdk_headers = [] # public headers included by drivers
         objs = []
@@ -235,6 +237,34 @@ foreach subpath:subdirs
             dpdk_includes += include_directories(drv_path)
         endif
 
+        # handle avx2 and avx512 source files
+        if arch_subdir == 'x86'
+            if sources_avx2.length() > 0
+                avx2_lib = static_library(lib_name + '_avx2_lib',
+                        sources_avx2,
+                        dependencies: static_deps,
+                        include_directories: includes,
+                        c_args: [cflags, cc_avx2_flags])
+                objs += avx2_lib.extract_objects(sources_avx2)
+            endif
+            if sources_avx512.length() > 0
+                cflags += '-DCC_AVX512_SUPPORT'
+                avx512_args = [cflags, cc_avx512_flags]
+                if cc.has_argument('-march=skylake-avx512')
+                    avx512_args += '-march=skylake-avx512'
+                    if cc.has_argument('-Wno-overriding-option')
+                        avx512_args += '-Wno-overriding-option'
+                    endif
+                endif
+                avx512_lib = static_library(lib_name + '_avx512_lib',
+                        sources_avx512,
+                        dependencies: static_deps,
+                        include_directories: includes,
+                        c_args: avx512_args)
+                objs += avx512_lib.extract_objects(sources_avx512)
+            endif
+        endif
+
         # generate pmdinfo sources by building a temporary
         # lib and then running pmdinfogen on the contents of
         # that lib. The final lib reuses the object files and
-- 
2.43.0

[PATCH v2 2/4] net/intel: use common AVX build code

[Bruce Richardson]

Remove driver-specific build instructions for the AVX2 and AVX-512 code,
and rely instead on the generic driver build file.

Signed-off-by: Bruce Richardson <bruce.richardson@intel.com>
---
 drivers/net/intel/i40e/meson.build | 26 ++------------------------
 drivers/net/intel/iavf/meson.build | 25 ++-----------------------
 drivers/net/intel/ice/meson.build  | 25 ++-----------------------
 drivers/net/intel/idpf/meson.build | 25 ++-----------------------
 4 files changed, 8 insertions(+), 93 deletions(-)

diff --git a/drivers/net/intel/i40e/meson.build b/drivers/net/intel/i40e/meson.build
index 2973ed1a01..25a3d72714 100644
--- a/drivers/net/intel/i40e/meson.build
+++ b/drivers/net/intel/i40e/meson.build
@@ -40,35 +40,13 @@ includes += include_directories('base')
 
 if arch_subdir == 'x86'
     sources += files('i40e_rxtx_vec_sse.c')
+    sources_avx2 = files('i40e_rxtx_vec_avx2.c')
+    sources_avx512 = files('i40e_rxtx_vec_avx512.c')
 
     if is_windows and cc.get_id() != 'clang'
         cflags += ['-fno-asynchronous-unwind-tables']
     endif
 
-    i40e_avx2_lib = static_library('i40e_avx2_lib',
-            'i40e_rxtx_vec_avx2.c',
-            dependencies: [static_rte_ethdev, static_rte_kvargs, static_rte_hash],
-            include_directories: includes,
-            c_args: [cflags, cc_avx2_flags])
-    objs += i40e_avx2_lib.extract_objects('i40e_rxtx_vec_avx2.c')
-
-    if cc_has_avx512
-        cflags += ['-DCC_AVX512_SUPPORT']
-        avx512_args = cflags + cc_avx512_flags
-        if cc.has_argument('-march=skylake-avx512')
-            avx512_args += '-march=skylake-avx512'
-            if cc.has_argument('-Wno-overriding-option')
-                avx512_args += '-Wno-overriding-option'
-            endif
-        endif
-        i40e_avx512_lib = static_library('i40e_avx512_lib',
-                'i40e_rxtx_vec_avx512.c',
-                dependencies: [static_rte_ethdev,
-                    static_rte_kvargs, static_rte_hash],
-                include_directories: includes,
-                c_args: avx512_args)
-        objs += i40e_avx512_lib.extract_objects('i40e_rxtx_vec_avx512.c')
-    endif
 elif arch_subdir == 'ppc'
        sources += files('i40e_rxtx_vec_altivec.c')
 elif arch_subdir == 'arm'
diff --git a/drivers/net/intel/iavf/meson.build b/drivers/net/intel/iavf/meson.build
index f7eac7c57a..fc3e4d2f28 100644
--- a/drivers/net/intel/iavf/meson.build
+++ b/drivers/net/intel/iavf/meson.build
@@ -28,34 +28,13 @@ includes += include_directories('base')
 
 if arch_subdir == 'x86'
     sources += files('iavf_rxtx_vec_sse.c')
+    sources_avx2 = files('iavf_rxtx_vec_avx2.c')
+    sources_avx512 = files('iavf_rxtx_vec_avx512.c')
 
     if is_windows and cc.get_id() != 'clang'
         cflags += ['-fno-asynchronous-unwind-tables']
     endif
 
-    iavf_avx2_lib = static_library('iavf_avx2_lib',
-            'iavf_rxtx_vec_avx2.c',
-            dependencies: [static_rte_ethdev],
-            include_directories: includes,
-            c_args: [cflags, cc_avx2_flags])
-    objs += iavf_avx2_lib.extract_objects('iavf_rxtx_vec_avx2.c')
-
-    if cc_has_avx512
-        cflags += ['-DCC_AVX512_SUPPORT']
-        avx512_args = cflags + cc_avx512_flags
-        if cc.has_argument('-march=skylake-avx512')
-            avx512_args += '-march=skylake-avx512'
-            if cc.has_argument('-Wno-overriding-option')
-                avx512_args += '-Wno-overriding-option'
-            endif
-        endif
-        iavf_avx512_lib = static_library('iavf_avx512_lib',
-                'iavf_rxtx_vec_avx512.c',
-                dependencies: [static_rte_ethdev],
-                include_directories: includes,
-                c_args: avx512_args)
-        objs += iavf_avx512_lib.extract_objects('iavf_rxtx_vec_avx512.c')
-    endif
 elif arch_subdir == 'arm'
     sources += files('iavf_rxtx_vec_neon.c')
 endif
diff --git a/drivers/net/intel/ice/meson.build b/drivers/net/intel/ice/meson.build
index cbdf38c1c4..a33ec59272 100644
--- a/drivers/net/intel/ice/meson.build
+++ b/drivers/net/intel/ice/meson.build
@@ -34,34 +34,13 @@ endif
 
 if arch_subdir == 'x86'
     sources += files('ice_rxtx_vec_sse.c')
+    sources_avx2 = files('ice_rxtx_vec_avx2.c')
+    sources_avx512 = files('ice_rxtx_vec_avx512.c')
 
     if is_windows and cc.get_id() != 'clang'
         cflags += ['-fno-asynchronous-unwind-tables']
     endif
 
-    ice_avx2_lib = static_library('ice_avx2_lib',
-            'ice_rxtx_vec_avx2.c',
-            dependencies: [static_rte_ethdev, static_rte_hash],
-            include_directories: includes,
-            c_args: [cflags, cc_avx2_flags])
-    objs += ice_avx2_lib.extract_objects('ice_rxtx_vec_avx2.c')
-
-    if cc_has_avx512
-        cflags += ['-DCC_AVX512_SUPPORT']
-        avx512_args = cflags + cc_avx512_flags
-        if cc.has_argument('-march=skylake-avx512')
-            avx512_args += '-march=skylake-avx512'
-            if cc.has_argument('-Wno-overriding-option')
-                avx512_args += '-Wno-overriding-option'
-            endif
-        endif
-        ice_avx512_lib = static_library('ice_avx512_lib',
-                'ice_rxtx_vec_avx512.c',
-                dependencies: [static_rte_ethdev, static_rte_hash],
-                include_directories: includes,
-                c_args: avx512_args)
-        objs += ice_avx512_lib.extract_objects('ice_rxtx_vec_avx512.c')
-    endif
 endif
 
 sources += files(
diff --git a/drivers/net/intel/idpf/meson.build b/drivers/net/intel/idpf/meson.build
index 4b272d02b1..10465f0f36 100644
--- a/drivers/net/intel/idpf/meson.build
+++ b/drivers/net/intel/idpf/meson.build
@@ -19,29 +19,8 @@ sources = files(
 )
 
 if arch_subdir == 'x86' and dpdk_conf.get('RTE_IOVA_IN_MBUF') == 1
-    idpf_avx2_lib = static_library('idpf_avx2_lib',
-        'idpf_common_rxtx_avx2.c',
-        dependencies: [static_rte_ethdev, static_rte_hash],
-        include_directories: includes,
-        c_args: [cflags, cc_avx2_flags])
-    objs += idpf_avx2_lib.extract_objects('idpf_common_rxtx_avx2.c')
-
-    if cc_has_avx512
-        cflags += ['-DCC_AVX512_SUPPORT']
-        avx512_args = cflags + cc_avx512_flags
-        if cc.has_argument('-march=skylake-avx512')
-            avx512_args += '-march=skylake-avx512'
-            if cc.has_argument('-Wno-overriding-option')
-                avx512_args += '-Wno-overriding-option'
-            endif
-        endif
-        idpf_common_avx512_lib = static_library('idpf_common_avx512_lib',
-                'idpf_common_rxtx_avx512.c',
-                dependencies: static_rte_mbuf,
-                include_directories: includes,
-                c_args: avx512_args)
-        objs += idpf_common_avx512_lib.extract_objects('idpf_common_rxtx_avx512.c')
-    endif
+    sources_avx2 = files('idpf_common_rxtx_avx2.c')
+    sources_avx512 = files('idpf_common_rxtx_avx512.c')
 endif
 
 subdir('base')
-- 
2.43.0

[PATCH v2 3/4] drivers/net: build use common AVX handling

[Bruce Richardson]

Remove from remaining net drivers the special-case code to handle AVX2
or AVX512 specific files. These can be built instead using
drivers/meson.build.

Signed-off-by: Bruce Richardson <bruce.richardson@intel.com>
---
 drivers/net/bnxt/meson.build      | 10 +---------
 drivers/net/enic/meson.build      | 10 +---------
 drivers/net/nfp/meson.build       | 10 +---------
 drivers/net/octeon_ep/meson.build | 14 ++------------
 drivers/net/virtio/meson.build    |  9 +--------
 5 files changed, 6 insertions(+), 47 deletions(-)

diff --git a/drivers/net/bnxt/meson.build b/drivers/net/bnxt/meson.build
index fd82d0c409..dcca7df916 100644
--- a/drivers/net/bnxt/meson.build
+++ b/drivers/net/bnxt/meson.build
@@ -58,15 +58,7 @@ subdir('hcapi/cfa_v3')
 
 if arch_subdir == 'x86'
     sources += files('bnxt_rxtx_vec_sse.c')
-    # build AVX2 code with instruction set explicitly enabled for runtime selection
-    bnxt_avx2_lib = static_library('bnxt_avx2_lib',
-            'bnxt_rxtx_vec_avx2.c',
-            dependencies: [static_rte_ethdev,
-                static_rte_bus_pci,
-                static_rte_kvargs, static_rte_hash],
-            include_directories: includes,
-            c_args: [cflags, cc_avx2_flags])
-     objs += bnxt_avx2_lib.extract_objects('bnxt_rxtx_vec_avx2.c')
+    sources_avx2 = files('bnxt_rxtx_vec_avx2.c')
 elif arch_subdir == 'arm' and dpdk_conf.get('RTE_ARCH_64')
     sources += files('bnxt_rxtx_vec_neon.c')
 endif
diff --git a/drivers/net/enic/meson.build b/drivers/net/enic/meson.build
index cfe5ec170a..2b3052fae8 100644
--- a/drivers/net/enic/meson.build
+++ b/drivers/net/enic/meson.build
@@ -29,17 +29,9 @@ sources = files(
 deps += ['hash']
 includes += include_directories('base')
 
-# Build the avx2 handler for 64-bit X86 targets, even though 'machine'
-# may not. This is to support users who build for the min supported machine
-# and need to run the binary on newer CPUs too.
 if dpdk_conf.has('RTE_ARCH_X86_64')
     cflags += '-DENIC_RXTX_VEC'
-    enic_avx2_lib = static_library('enic_avx2_lib',
-            'enic_rxtx_vec_avx2.c',
-            dependencies: [static_rte_ethdev, static_rte_bus_pci],
-            include_directories: includes,
-            c_args: [cflags, cc_avx2_flags])
-    objs += enic_avx2_lib.extract_objects('enic_rxtx_vec_avx2.c')
+    sources_avx2 = files('enic_rxtx_vec_avx2.c')
 endif
 
 annotate_locks = false
diff --git a/drivers/net/nfp/meson.build b/drivers/net/nfp/meson.build
index 0a12b7dce7..a98b584042 100644
--- a/drivers/net/nfp/meson.build
+++ b/drivers/net/nfp/meson.build
@@ -52,19 +52,11 @@ cflags += no_wvla_cflag
 if arch_subdir == 'x86'
         includes += include_directories('../../common/nfp')
 
-        avx2_sources = files(
+        sources_avx2 = files(
                 'nfdk/nfp_nfdk_vec_avx2_dp.c',
                 'nfp_rxtx_vec_avx2.c',
         )
 
-        nfp_avx2_lib = static_library('nfp_avx2_lib',
-                avx2_sources,
-                dependencies: [static_rte_ethdev, static_rte_bus_pci],
-                include_directories: includes,
-                c_args: [cflags, cc_avx2_flags]
-        )
-
-        objs += nfp_avx2_lib.extract_all_objects(recursive: true)
 else
         sources += files(
                 'nfp_rxtx_vec_stub.c',
diff --git a/drivers/net/octeon_ep/meson.build b/drivers/net/octeon_ep/meson.build
index 1b34db3edc..9bf4627894 100644
--- a/drivers/net/octeon_ep/meson.build
+++ b/drivers/net/octeon_ep/meson.build
@@ -15,18 +15,8 @@ sources = files(
 
 if arch_subdir == 'x86'
     sources += files('cnxk_ep_rx_sse.c')
-    if cc.get_define('__AVX2__', args: machine_args) != ''
-        cflags += ['-DCC_AVX2_SUPPORT']
-        sources += files('cnxk_ep_rx_avx.c')
-    elif cc.has_multi_arguments(cc_avx2_flags)
-        cflags += ['-DCC_AVX2_SUPPORT']
-        otx_ep_avx2_lib = static_library('otx_ep_avx2_lib',
-                        'cnxk_ep_rx_avx.c',
-                        dependencies: [static_rte_ethdev, static_rte_pci, static_rte_bus_pci],
-                        include_directories: includes,
-                        c_args: [cflags, cc_avx2_flags])
-        objs += otx_ep_avx2_lib.extract_objects('cnxk_ep_rx_avx.c')
-    endif
+    cflags += ['-DCC_AVX2_SUPPORT']
+    sources_avx2 = files('cnxk_ep_rx_avx.c')
 endif
 
 if arch_subdir == 'arm'
diff --git a/drivers/net/virtio/meson.build b/drivers/net/virtio/meson.build
index c1c4a85bea..01bfb3c47d 100644
--- a/drivers/net/virtio/meson.build
+++ b/drivers/net/virtio/meson.build
@@ -27,15 +27,8 @@ cflags += no_wvla_cflag
 
 if arch_subdir == 'x86'
     if cc_has_avx512
-        cflags += ['-DCC_AVX512_SUPPORT']
         cflags += ['-DVIRTIO_RXTX_PACKED_VEC']
-        virtio_avx512_lib = static_library('virtio_avx512_lib',
-                'virtio_rxtx_packed.c',
-                dependencies: [static_rte_ethdev,
-                    static_rte_kvargs, static_rte_bus_pci],
-                include_directories: includes,
-                c_args: cflags + cc_avx512_flags)
-        objs += virtio_avx512_lib.extract_objects('virtio_rxtx_packed.c')
+        sources_avx512 = files('virtio_rxtx_packed.c')
         if (toolchain == 'gcc' and cc.version().version_compare('>=8.3.0'))
             cflags += '-DVIRTIO_GCC_UNROLL_PRAGMA'
         elif (toolchain == 'clang' and cc.version().version_compare('>=3.7.0'))
-- 
2.43.0

[PATCH v2 4/4] drivers/net: remove AVX2 build-time define

[Bruce Richardson]

Since all supported compilers can generate AVX2 code, we will always
enable the build of the AVX2 files on x86. This means that
CC_AVX2_SUPPORT is always true on x86, so it can be removed and a
regular "#ifdef RTE_ARCH_x86" used in its place.

Signed-off-by: Bruce Richardson <bruce.richardson@intel.com>
---
 drivers/net/bnxt/bnxt_ethdev.c        | 2 --
 drivers/net/octeon_ep/meson.build     | 1 -
 drivers/net/octeon_ep/otx_ep_ethdev.c | 4 ----
 3 files changed, 7 deletions(-)

diff --git a/drivers/net/bnxt/bnxt_ethdev.c b/drivers/net/bnxt/bnxt_ethdev.c
index 144d4377bd..a318604c20 100644
--- a/drivers/net/bnxt/bnxt_ethdev.c
+++ b/drivers/net/bnxt/bnxt_ethdev.c
@@ -3238,8 +3238,6 @@ static const struct {
 #if defined(RTE_ARCH_X86)
 	{bnxt_crx_pkts_vec,		"Vector SSE"},
 	{bnxt_recv_pkts_vec,		"Vector SSE"},
-#endif
-#if defined(RTE_ARCH_X86) && defined(CC_AVX2_SUPPORT)
 	{bnxt_crx_pkts_vec_avx2,	"Vector AVX2"},
 	{bnxt_recv_pkts_vec_avx2,	"Vector AVX2"},
 #endif
diff --git a/drivers/net/octeon_ep/meson.build b/drivers/net/octeon_ep/meson.build
index 9bf4627894..a4a7663d1d 100644
--- a/drivers/net/octeon_ep/meson.build
+++ b/drivers/net/octeon_ep/meson.build
@@ -15,7 +15,6 @@ sources = files(
 
 if arch_subdir == 'x86'
     sources += files('cnxk_ep_rx_sse.c')
-    cflags += ['-DCC_AVX2_SUPPORT']
     sources_avx2 = files('cnxk_ep_rx_avx.c')
 endif
 
diff --git a/drivers/net/octeon_ep/otx_ep_ethdev.c b/drivers/net/octeon_ep/otx_ep_ethdev.c
index 8b14734b0c..10f2f8a2e0 100644
--- a/drivers/net/octeon_ep/otx_ep_ethdev.c
+++ b/drivers/net/octeon_ep/otx_ep_ethdev.c
@@ -91,11 +91,9 @@ otx_ep_set_rx_func(struct rte_eth_dev *eth_dev)
 		eth_dev->rx_pkt_burst = &cnxk_ep_recv_pkts;
 #ifdef RTE_ARCH_X86
 		eth_dev->rx_pkt_burst = &cnxk_ep_recv_pkts_sse;
-#ifdef CC_AVX2_SUPPORT
 		if (rte_vect_get_max_simd_bitwidth() >= RTE_VECT_SIMD_256 &&
 		    rte_cpu_get_flag_enabled(RTE_CPUFLAG_AVX2) == 1)
 			eth_dev->rx_pkt_burst = &cnxk_ep_recv_pkts_avx;
-#endif
 #elif defined(RTE_ARCH_ARM64)
 		eth_dev->rx_pkt_burst = &cnxk_ep_recv_pkts_neon;
 #endif
@@ -105,11 +103,9 @@ otx_ep_set_rx_func(struct rte_eth_dev *eth_dev)
 		eth_dev->rx_pkt_burst = &cn9k_ep_recv_pkts;
 #ifdef RTE_ARCH_X86
 		eth_dev->rx_pkt_burst = &cn9k_ep_recv_pkts_sse;
-#ifdef CC_AVX2_SUPPORT
 		if (rte_vect_get_max_simd_bitwidth() >= RTE_VECT_SIMD_256 &&
 		    rte_cpu_get_flag_enabled(RTE_CPUFLAG_AVX2) == 1)
 			eth_dev->rx_pkt_burst = &cn9k_ep_recv_pkts_avx;
-#endif
 #elif defined(RTE_ARCH_ARM64)
 		eth_dev->rx_pkt_burst = &cn9k_ep_recv_pkts_neon;
 #endif
-- 
2.43.0