24/03/2021 23:59, Wang, Yipeng1:
> From: kumar amber
> >
> > add _SSE42_ flag to enable compilation of
> > sse42 specific instructions only on supported architecture
> >
> > Signed-off-by: kumar amber <kumar.am...@intel.com>
> > ---
> > lib/librte_hash/rte_hash_crc.h | 16 ++++++++--------
> > 1 file changed, 8 insertions(+), 8 deletions(-)
> >
> > diff --git a/lib/librte_hash/rte_hash_crc.h b/lib/librte_hash/rte_hash_crc.h
> > index 3e131aa6bb..e9f063780c 100644
> > --- a/lib/librte_hash/rte_hash_crc.h
> > +++ b/lib/librte_hash/rte_hash_crc.h
> > @@ -358,7 +358,7 @@ crc32c_2words(uint64_t data, uint32_t init_val)
> > return crc;
> > }
> >
> > -#if defined(RTE_ARCH_X86)
> > +#if defined(RTE_ARCH_X86) && defined(__SSE42__)
> > static inline uint32_t
> > crc32c_sse42_u8(uint8_t data, uint32_t init_val) { @@ -404,7 +404,7 @@
> > crc32c_sse42_u64_mimic(uint64_t data, uint64_t init_val) } #endif
>
> ...
>
> > -#if defined RTE_ARCH_X86
> > +#if defined(RTE_ARCH_X86) && defined(__SSE42__)
> > if (likely(crc32_alg & CRC32_SSE42))
> > return crc32c_sse42_u64_mimic(data, init_val); #endif
> > --
> > 2.25.1
>
> [Wang, Yipeng]
> Hi, Kumar, thanks for the patch.
> I think the minimum required machine for x86 is sse4.2 compatible already. So
> I wonder if we really need this.
Yes, that's why I don't understand this patch.
> Also, I think the right way to check machine flag in DPDK should be:
> #If defined (RTE_MACHINE_CPUFLAG_SSE4_2)
These macros have been removed in DPDK 20.11.
> Instead of using compiler dependent macro.
Compiler macros are well standardized, it is OK.
