diff options
| author |   Damjan Marion <damarion@cisco.com> | 2016-05-11 23:07:18 +0200 |
|---|---|---|
| committer | Damjan Marion <damarion@cisco.com> | 2016-05-19 18:14:38 +0200 |
| commit | 1c80e831b728ab378949714d5059a0b5b1822a0a (patch) | |
| tree | e4f3ecfee143f7dee0e9905570d41ca3ee345c83 /vpp/vnet/main.c | |
| parent | 82e29c455833b5b12e04c89d2dec1106b499e6b0 (diff) | |
Add support for multiple microarchitectures in single binary
* compiler -march= parameter is changed from native to corei7
so code is always genereted with instructions which are available
on the Nehalem microarchitecture (up to SSE4.2)
* compiler -mtune= parameter is added so code is optimized for
corei7-avx which equals to Sandy Bridge microarchitecture
* set of macros is added which allows run-time detection of available
cpu instructions (e.g. clib_cpu_supports_avx())
* set of macros is added which allows us to clone graph node funcitons
where cloned function is optmized for different microarchitecture
Those macros are using following attributes:
__attribute__((flatten))
__attribute__((target("arch=core-avx2)))
I.e. If applied to foo_node_fn() macro will generate cloned
functions foo_node_fn_avx2() and foo_node_fn_avx512() (future)
It will also generate function void * foo_node_fn_multiarch_select()
which detects available instruction set and returns pointer to the
best matching function clone.
Change-Id: I2dce0ac92a5ede95fcb56f47f3d1f3c4c040bac0
Signed-off-by: Damjan Marion <damarion@cisco.com>
Diffstat (limited to 'vpp/vnet/main.c')
| -rw-r--r-- | vpp/vnet/main.c | 19 |
1 files changed, 10 insertions, 9 deletions
diff --git a/vpp/vnet/main.c b/vpp/vnet/main.c index f707994ba54..2540083982b 100644 --- a/vpp/vnet/main.c +++ b/vpp/vnet/main.c @@ -12,6 +12,8 @@ * See the License for the specific language governing permissions and * limitations under the License. */ + +#include <vppinfra/cpu.h> #include <vlib/vlib.h> #include <vlib/unix/unix.h> #include <vnet/plugin/plugin.h> @@ -59,33 +61,32 @@ int main (int argc, char * argv[]) u32 size; void vlib_set_get_handoff_structure_cb (void *cb); -#if __x86_64__ && !defined(__clang__) - __builtin_cpu_init (); +#if __x86_64__ const char * msg = "ERROR: This binary requires CPU with %s extensions.\n"; #define _(a,b) \ - if (!__builtin_cpu_supports(a)) \ + if (!clib_cpu_supports_ ## a ()) \ { \ fprintf(stderr, msg, b); \ exit(1); \ } #if __AVX2__ - _("avx2", "AVX2") + _(avx2, "AVX2") #endif #if __AVX__ - _("avx", "AVX") + _(avx, "AVX") #endif #if __SSE4_2__ - _("sse4.2", "SSE4.2") + _(sse42, "SSE4.2") #endif #if __SSE4_1__ - _("sse4.1", "SSE4.1") + _(sse41, "SSE4.1") #endif #if __SSSE3__ - _("ssse3", "SSSE3") + _(ssse3, "SSSE3") #endif #if __SSE3__ - _("sse3", "SSE3") + _(sse3, "SSE3") #endif #undef _ #endif |