[Bug target/106531] New: -march=rv32iabmc should also enable zba, zbb, zbc, zbs

[Bug target/106531] New: -march=rv32iabmc should also enable zba, zbb, zbc, zbs

[pinskia at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=106531

            Bug ID: 106531
           Summary: -march=rv32iabmc should also enable zba, zbb, zbc, zbs
           Product: gcc
           Version: 12.0
            Status: UNCONFIRMED
          Severity: enhancement
          Priority: P3
         Component: target
          Assignee: pinskia at gcc dot gnu.org
          Reporter: pinskia at gcc dot gnu.org
  Target Milestone: ---
            Target: riscv

While doing some development, I Noticed that -march=rv32iabmc does not work
(binutils as fails but that is a different issue) and the zba, zbb, zbc, and
zbs extensions are not enabled like I think they should be.
I am filing this bug to remind me to submit the patch for this.

[Bug target/106531] -march=rv32iabmc should also enable zba, zbb, zbc, zbs

[andrew at sifive dot com]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=106531

Andrew Waterman <andrew at sifive dot com> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |andrew at sifive dot com

--- Comment #1 from Andrew Waterman <andrew at sifive dot com> ---
It turns out there is no such thing as the B extension; it was a working name
for a draft proposal a few years ago but wasn't (and presumably won't be)
ratified.  The correct behavior is to reject B but accept zba, zbb, zbc, and
zbs.

[Bug target/106531] -march=rv32iabmc should also enable zba, zbb, zbc, zbs

[pinskia at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=106531

--- Comment #2 from Andrew Pinski <pinskia at gcc dot gnu.org> ---
(In reply to Andrew Waterman from comment #1)
> It turns out there is no such thing as the B extension; it was a working
> name for a draft proposal a few years ago but wasn't (and presumably won't
> be) ratified.  The correct behavior is to reject B but accept zba, zbb, zbc,
> and zbs.

-march=rv32imac_zba_zbb_zbc_zbs works but that is just so bad.
Really b should just mean all 4 really; it should be a meta ISA extension :).
There is so much no documentation on what ISA names; that I had to just guess.

[Bug target/106531] -march=rv32iabmc should also enable zba, zbb, zbc, zbs

[andrew at sifive dot com]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=106531

--- Comment #3 from Andrew Waterman <andrew at sifive dot com> ---
The lengthy-ISA-name problem will be addressed to a great extent by the
forthcoming introduction of ISA profiles.  Although I agree the status quo is
ugly and overly verbose, I recommend against introducing GCC-specific syntax
for shortening ISA strings, since it'll quickly become legacy.

[Bug target/106531] -march=rv32iabmc should also enable zba, zbb, zbc, zbs

[pinskia at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=106531

Andrew Pinski <pinskia at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
           Assignee|pinskia at gcc dot gnu.org         |unassigned at gcc dot gnu.org
             Status|UNCONFIRMED                 |NEW
   Last reconfirmed|                            |2024-02-05
     Ever confirmed|0                           |1

--- Comment #4 from Andrew Pinski <pinskia at gcc dot gnu.org> ---
Looks like binutils is going to add b:
https://sourceware.org/pipermail/binutils/2024-February/132302.html

GCC most likely will be updated a little afterwards.

No longer working on this right now.

[Bug target/106531] -march=rv32iabmc should also enable zba, zbb, zbc, zbs

[andrew at sifive dot com]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=106531

--- Comment #5 from Andrew Waterman <andrew at sifive dot com> ---
Yeah, RISC-V International decided to define B = {Zba, Zbb, Zbs}: note, not
Zbc.

[Bug target/106531] -march=rv32iabmc should also enable zba, zbb, zbs

[pinskia at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=106531

Andrew Pinski <pinskia at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
            Summary|-march=rv32iabmc should     |-march=rv32iabmc should
                   |also enable zba, zbb, zbc,  |also enable zba, zbb, zbs
                   |zbs                         |

--- Comment #6 from Andrew Pinski <pinskia at gcc dot gnu.org> ---
(In reply to Andrew Waterman from comment #5)
> Yeah, RISC-V International decided to define B = {Zba, Zbb, Zbs}: note, not
> Zbc.

From my point of view that is a decent.