> From: Konstantin Ananyev [mailto:[email protected]]
> Sent: Thursday, 23 October 2025 10.51
>
> > -#define RTE_MBUF_DIRECT(mb) \
> > - (!((mb)->ol_flags & (RTE_MBUF_F_INDIRECT | RTE_MBUF_F_EXTERNAL)))
> > + *
> > + * Note: Macro optimized for code size.
> > + *
> > + * The plain macro would be:
> > + * #define RTE_MBUF_DIRECT(mb) \
> > + * (!((mb)->ol_flags & (RTE_MBUF_F_INDIRECT |
> RTE_MBUF_F_EXTERNAL)))
> > + *
> > + * The flags RTE_MBUF_F_INDIRECT and RTE_MBUF_F_EXTERNAL are both in
> the
> > MSB (most significant
> > + * byte) of the 64-bit ol_flags field, so we only compare this one
> byte instead of all
> > 64 bits.
> > + *
> > + * E.g., GCC version 16.0.0 20251019 (experimental) generates the
> following code
> > for x86-64.
> > + *
> > + * With the plain macro, 17 bytes of instructions:
> > + * movabs rax,0x6000000000000000 // 10 bytes
> > + * and rax,QWORD PTR [rdi+0x18] // 4 bytes
> > + * sete al // 3 bytes
> > + * With this optimized macro, only 7 bytes of instructions:
> > + * test BYTE PTR [rdi+0x1f],0x60 // 4 bytes
> > + * sete al // 3 bytes
> > + */
> > +#if RTE_BYTE_ORDER == RTE_LITTLE_ENDIAN
> > +/* On little endian architecture, the MSB of a 64-bit integer is at
> byte offset 7. */
> > +#define RTE_MBUF_DIRECT(mb) !(((const char *)(&(mb)-
> >ol_flags))[7] & 0x60)
> > +#elif RTE_BYTE_ORDER == RTE_BIG_ENDIAN
> > +/* On big endian architecture, the MSB of a 64-bit integer is at
> byte offset 0. */
> > +#define RTE_MBUF_DIRECT(mb) !(((const char *)(&(mb)-
> >ol_flags))[0] & 0x60)
>
> A stupid q: why then not simply do:
> (mb->ol_flags >> 56) & 0x60
> then?
> Should help to all these LE/BE casts, etc.
GCC is too stupid for that too.
Playing around with Godbolt shows that
return !((char)(p[3] >> 56) & 0x60);
becomes
movzx eax,BYTE PTR [rdi+0x1f] // 4 bytes
test al,0x60 // 2 bytes
Instead of simply
test BYTE PTR [rdi+0x1f],0x60 // 4 bytes
Good suggestion, though.
>
> > +#endif
> > +/* Verify the optimization above. */
> > +static_assert((RTE_MBUF_F_INDIRECT | RTE_MBUF_F_EXTERNAL) ==
> > UINT64_C(0x60) << (7 * CHAR_BIT),
> > + "(RTE_MBUF_F_INDIRECT | RTE_MBUF_F_EXTERNAL) is not 0x60 at
> MSB");
> >
> > /** Uninitialized or unspecified port. */
> > #define RTE_MBUF_PORT_INVALID UINT16_MAX
> > --
> > 2.43.0
