[Bug c/56024] New: ARM NEON polynomial types behave as if signed

[Bug c/56024] New: ARM NEON polynomial types behave as if signed

[Tim.Northover at arm dot com]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=56024

             Bug #: 56024
           Summary: ARM NEON polynomial types behave as if signed
    Classification: Unclassified
           Product: gcc
           Version: 4.8.0
            Status: UNCONFIRMED
          Severity: normal
          Priority: P3
         Component: c
        AssignedTo: unassigned@gcc.gnu.org
        ReportedBy: Tim.Northover@arm.com


(Apologies for any broken conventions, this is my first GCC bug).

The ARM NEON intrinsic header arm_neon.h defines two scalar types poly8_t and
poly16_t. The existence of these types is at least implicit in the vget_lane
and vset_lane intrinsics. However on GCC they seem to be treated as "signed" in
some sense with extremely dodgy mathematical consequences. For example:

poly8_t p = 0x80;
poly16_t q = (poly16_t)p;

The polynomial "q" has the value 0xff80. This means that the cast has converted
the polynomial "x^7" into "x^15 + x^14 + ... + x^7".

This behaviour is also incompatible with the ARM compiler's implementation of
these types (they are typedefs from the obvious unsigned types there).

[Bug c/56024] ARM NEON polynomial types behave as if signed

[rearnsha at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=56024

Richard Earnshaw <rearnsha at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Target|                            |arm-eabi, arm-linux-gnueabi
             Status|UNCONFIRMED                 |NEW
   Last reconfirmed|                            |2013-01-18
     Ever Confirmed|0                           |1

--- Comment #1 from Richard Earnshaw <rearnsha at gcc dot gnu.org> 2013-01-18 09:47:29 UTC ---
This certainly doesn't seem to be what a user would reasonably expect.

[Bug target/56024] ARM NEON polynomial types behave as if signed

[t.p.northover at gmail dot com]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=56024

Tim Northover <t.p.northover at gmail dot com> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |t.p.northover at gmail dot com

--- Comment #2 from Tim Northover <t.p.northover at gmail dot com> ---
Are there any plans to fix this soon? It's become relevant again on the Clang
mailing list. 

We don't want to switch to a sane type if GCC decides sanity is less important
than ABI-compatibility, but every so often it causes someone extra work to
handle the discrepancy.

[Bug target/56024] ARM NEON polynomial types behave as if signed

[StaffLeavers at arm dot com]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=56024

--- Comment #4 from StaffLeavers at arm dot com ---
tim.northover no longer works for ARM.

Your email will be forwarded to their line manager.


Please do not reply to this email.
If you need more information, please email real-postmaster@arm.com

Thank you.

[Bug target/56024] ARM NEON polynomial types behave as if signed

[StaffLeavers at arm dot com]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=56024

--- Comment #3 from StaffLeavers at arm dot com ---
tim.northover no longer works for ARM.

Your email will be forwarded to their line manager.


Please do not reply to this email.
If you need more information, please email real-postmaster@arm.com

Thank you.

[Bug target/56024] ARM NEON polynomial types behave as if signed

[StaffLeavers at arm dot com]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=56024

--- Comment #5 from StaffLeavers at arm dot com ---
tim.northover no longer works for ARM.

Your email will be forwarded to their line manager.


Please do not reply to this email.
If you need more information, please email real-postmaster@arm.com

Thank you.

[Bug target/56024] ARM NEON polynomial types behave as if signed

[StaffLeavers at arm dot com]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=56024

--- Comment #6 from StaffLeavers at arm dot com ---
tim.northover no longer works for ARM.

Your email will be forwarded to their line manager.


Please do not reply to this email.
If you need more information, please email real-postmaster@arm.com

Thank you.

[Bug target/56024] ARM NEON polynomial types behave as if signed

[StaffLeavers at arm dot com]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=56024

--- Comment #7 from StaffLeavers at arm dot com ---
tim.northover no longer works for ARM.

Your email will be forwarded to their line manager.


Please do not reply to this email.
If you need more information, please email real-postmaster@arm.com

Thank you.

[Bug target/56024] ARM NEON polynomial types behave as if signed

[StaffLeavers at arm dot com]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=56024

--- Comment #8 from StaffLeavers at arm dot com ---
tim.northover no longer works for ARM.

Your email will be forwarded to their line manager.


Please do not reply to this email.
If you need more information, please email real-postmaster@arm.com

Thank you.

[Bug target/56024] ARM NEON polynomial types behave as if signed

[t.p.northover at gmail dot com]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=56024

--- Comment #9 from Tim Northover <t.p.northover at gmail dot com> ---
Yay for infinite e-mail loops! Very sorry about that, it should be fixed now.