Re: [PATCH] Alpha: Emulate unaligned LDx_L/STx_C for data consistency

[Maciej W. Rozycki]

On Thu, 20 Feb 2025, Matt Turner wrote:

> > On Wed, Feb 19, 2025 at 7:46 AM Maciej W. Rozycki <macro@xxxxxxxxxxx> wrote:
> > >  #define OP_INT_MASK    ( 1L << 0x28 | 1L << 0x2c   /* ldl stl */       \
> > > +                       | 1L << 0x2a | 1L << 0x2e   /* ldl_l stl_c */   \
> > >                         | 1L << 0x29 | 1L << 0x2d   /* ldq stq */       \
> > > +                       | 1L << 0x2b | 1L << 0x2f   /* ldq_l stq_c */   \
> > >                         | 1L << 0x0c | 1L << 0x0d   /* ldwu stw */      \
> > >                         | 1L << 0x0a | 1L << 0x0e ) /* ldbu stb */
> > >
> > >  #define OP_WRITE_MASK  ( 1L << 0x26 | 1L << 0x27   /* sts stt */       \
> > >                         | 1L << 0x2c | 1L << 0x2d   /* stl stq */       \
> > > +                       | 1L << 0x2e | 1L << 0x2d   /* stl_c stq_c */   \
> >
> > stq_c should be 0x2f, not 0x2d. Looks like a copy-n-paste mistake.
> 
> The good news is that OP_WRITE_MASK appears to be unused going all the
> way back to the import into git, so this doesn't indicate a problem
> with any of the testing that's been done.

 Good catch, thank you, and I guess the lack of use is why things haven't 
broken.  I'll make a preparatory change in v2 and remove this macro then.

 FWIW it came with 2.1.36, already unused, so presumably a leftover from a 
WIP version.

 I'll fix the typo in the description as well, thank you.

  Maciej