[PATCH] Make hint ranges consistent (Was: For the 16-bit unallocated NOP hints, make disassembly representation consistent with their 32-bit counterparts)

[Artyom Skrobov]

Yes. Committed revision 193238.

Celebrating my first LLVM commit! :-)


-----Original Message-----
From: Richard Barton 
Sent: 22 October 2013 17:02
To: Artyom Skrobov; llvm-commits at cs.uiuc.edu
Subject: RE: [PATCH] Make hint ranges consistent (Was: For the 16-bit
unallocated NOP hints, make disassembly representation consistent with their
32-bit counterparts)

Hi Artyom

The additional tests look good, and your explanations are convincing. LGTM.

Do you have commit access?

Ta,
Rich

> -----Original Message-----
> From: Artyom Skrobov
> Sent: 21 October 2013 12:23
> To: Richard Barton; llvm-commits at cs.uiuc.edu
> Subject: RE: [PATCH] Make hint ranges consistent (Was: For the 16-bit
> unallocated NOP hints, make disassembly representation consistent with
> their 32-bit counterparts)
> 
> > A1 Hint:
> > * Should the hint syntax be conditional as well, to match with the
> > defined hint syntaxes? It looks like that whole space is unallocated,
> > so we may as well make it available.
> 
> The hints did support the conditional syntax from the beginning, although
it
> was never tested -- not even in my new tests.
> 
> These further changes on top of my patch should be sufficient:
> 
> [skipped]
> 
> > * Why not allow the hint syntax above immediate values of 0xf0? Should
> > the DBG instructions not now be re-implemented as hint aliases?
> 
> It could be done, but it's hard for me to see any benefit from doing so;
and
> it would come at the cost of complicating ARMInstPrinter even further.
> 
> > T2 Hint
> > * Why do we still need the InstAlias for the hint here? Can it not be
> > removed like the T1 version?
> 
> No, the InstAlias proves necessary for picking up the wide hint when the
> value is above #15.