[Bug target/91886] [10 regression] powerpc64 impossible constraint in asm

[Bug target/91886] [10 regression] powerpc64 impossible constraint in asm

[segher at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=91886

--- Comment #36 from Segher Boessenkool <segher at gcc dot gnu.org> ---
(In reply to Rich Felker from comment #34)
> Per the IBM docs, LE/elfv2 (which they confusingly equate)

Where do you see this, btw?  The introduction of the ABI doc says

> The OpenPOWER ELF V2 ABI is intended for use in little- and big-endian environments.

and I don't see it contradict that elsewhere.

[Bug target/91886] [10 regression] powerpc64 impossible constraint in asm

[segher at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=91886

--- Comment #37 from Segher Boessenkool <segher at gcc dot gnu.org> ---
Oh, hrm, I am looking at an older version.  Ugh.

[Bug target/91886] [10 regression] powerpc64 impossible constraint in asm

[rguenth at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=91886

Richard Biener <rguenth at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
           Priority|P3                          |P1

--- Comment #38 from Richard Biener <rguenth at gcc dot gnu.org> ---
This PR is a mess ... for one "ws" is said to be invalid but then there's
talk about still allowing it for backward compatibility.  But I see no action
here?

P1.  If it ends up the last release blocker I'll turn this bug into INVALID.

[Bug target/91886] [10 regression] powerpc64 impossible constraint in asm

[segher at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=91886

--- Comment #39 from Segher Boessenkool <segher at gcc dot gnu.org> ---
commit 07fe4af4d51d74b63a76ea632d4db01d1f69f037
Author: Segher Boessenkool <segher@kernel.crashing.org>
Date:   Wed Mar 18 21:58:45 2020 +0000

    rs6000: Add back some w* constraints (PR91886)

    In May and June last year I deleted many of our (vector) constraints.
    We can now just use "wa" for those, together with some other
    conditions, which can be per alternative using the "enabled" attribute
    (which in turn primarily uses the "isa" attribute).

    But, it turns out that Clang implements some of those constraints as
    well, and at least musl uses some of them.  It is easy for us to add
    those contraints back (as undocumented aliases to "wa", which always
    did mean the same thing for valid inline assembler code), so do that.

    gcc/
        * config/rs6000/constraints.md (wd, wf, wi, ws, ww): New undocumented
        aliases for "wa".


(I forgot to mark it appropriately, whoops).

[Bug target/91886] [10 regression] powerpc64 impossible constraint in asm

[jakub at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=91886

Jakub Jelinek <jakub at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |jakub at gcc dot gnu.org

--- Comment #40 from Jakub Jelinek <jakub at gcc dot gnu.org> ---
Thus fixed in r10-7269-g07fe4af4d51d74b63a76ea632d4db01d1f69f037 then?

[Bug target/91886] [10 regression] powerpc64 impossible constraint in asm

[segher at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=91886

--- Comment #41 from Segher Boessenkool <segher at gcc dot gnu.org> ---
Fixed.

[Bug target/91886] [10 regression] powerpc64 impossible constraint in asm

[segher at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=91886

Segher Boessenkool <segher at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
         Resolution|---                         |FIXED
             Status|NEW                         |RESOLVED

--- Comment #42 from Segher Boessenkool <segher at gcc dot gnu.org> ---
I said FIXED.  Grr, BZ.