On Thu, Jun 20, 2024 at 01:40:42PM +0200, Mattias Rönnblom wrote: > On 2024-06-20 11:28, Bruce Richardson wrote: > > On Thu, Jun 20, 2024 at 09:24:47AM +0200, Mattias Rönnblom wrote: > > > Add missing <emmintrin.h> include, to get the _mm_cvtsi128_si64 > > > prototype. > > > > > > Signed-off-by: Mattias Rönnblom <mattias.ronnb...@ericsson.com> > > > --- > > > drivers/net/fm10k/fm10k_rxtx_vec.c | 1 + > > > 1 file changed, 1 insertion(+) > > > > > > diff --git a/drivers/net/fm10k/fm10k_rxtx_vec.c > > > b/drivers/net/fm10k/fm10k_rxtx_vec.c > > > index 2b6914b1da..d417b31bbb 100644 > > > --- a/drivers/net/fm10k/fm10k_rxtx_vec.c > > > +++ b/drivers/net/fm10k/fm10k_rxtx_vec.c > > > @@ -10,6 +10,7 @@ > > > #include "base/fm10k_type.h" > > > #include <tmmintrin.h> > > > +#include <emmintrin.h> > > Beyond my ack of this patch, a small suggestion is to just include > > rte_vect.h rather than trying to include specific x86-intrinsics headers. > > > > My ack remains with or without taking on board this suggestion. > > > > /Bruce > > I will do that, and hope it will magically solve the > _mm_cvtsi128_si64-on-32-bit-x86 issue.
I was looking at that, and it does solve it in my testing. There are a lot of drivers that have just "tmmintrin.h" included. Changing all of those to rte_vect.h allows 32bit to build with your other changes applied. /Bruce