[Bug target/105879] New: DI mode constants are incorrectly loaded into registers on big-endian xtensa

[Bug target/105879] New: DI mode constants are incorrectly loaded into registers on big-endian xtensa

[jcmvbkbc at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=105879

            Bug ID: 105879
           Summary: DI mode constants are incorrectly loaded into
                    registers on big-endian xtensa
           Product: gcc
           Version: 13.0
            Status: UNCONFIRMED
          Severity: normal
          Priority: P3
         Component: target
          Assignee: unassigned at gcc dot gnu.org
          Reporter: jcmvbkbc at gcc dot gnu.org
  Target Milestone: ---

A lot of testsuite tests are failing in big-endian xtensa configuration, e.g.
gcc.c-torture/execute/20001031-1.c
For the following part

long long t4 (void)
{ 
  long long i;
  t3 ((i = 4096) + 0x7fffffffULL);
  return i;
}

in the generated assembly I see the following:

        .literal .LC7, -2147479553
        .literal .LC8, t3@PLT
        .literal .LC9, 4096
        .align  4
        .global t4
        .type   t4, @function
t4:
        entry   sp, 32
        movi.n  a3, 0
        l32r    a10, .LC7
        mov.n   a11, a3
        l32r    a8, .LC8
        callx8  a8
        l32r    a2, .LC9
        retw.n

(i.e. 0 in a11), while it should've been the following (i.e. 0 in a10):

        .literal .LC7, -2147479553
        .literal .LC8, t3@PLT
        .literal .LC9, 4096
        .align  4
        .global t4
        .type   t4, @function
t4:
        entry   sp, 32
        movi.n  a2, 0
        mov.n   a10, a2
        l32r    a11, .LC7
        l32r    a8, .LC8
        callx8  a8
        l32r    a3, .LC9
        retw.n

This regression clearly traces back to the commit 64a54505ec82 ("gcc: xtensa:
rearrange DI mode constant loading") and is present in gcc-11.1+ and gcc-12.1

[Bug target/105879] DI mode constants are incorrectly loaded into registers on big-endian xtensa

[jcmvbkbc at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=105879

jcmvbkbc at gcc dot gnu.org changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|UNCONFIRMED                 |ASSIGNED
     Ever confirmed|0                           |1
           Assignee|unassigned at gcc dot gnu.org      |jcmvbkbc at gcc dot gnu.org
   Last reconfirmed|                            |2022-06-08

[Bug target/105879] DI mode constants are incorrectly loaded into registers on big-endian xtensa

[cvs-commit at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=105879

--- Comment #1 from CVS Commits <cvs-commit at gcc dot gnu.org> ---
The master branch has been updated by Max Filippov <jcmvbkbc@gcc.gnu.org>:

https://gcc.gnu.org/g:e94c6dbfb57a862dd8a8685eabc4886ad1aaea25

commit r13-1015-ge94c6dbfb57a862dd8a8685eabc4886ad1aaea25
Author: Max Filippov <jcmvbkbc@gmail.com>
Date:   Tue Jun 7 21:01:01 2022 -0700

    gcc: xtensa: fix PR target/105879

    split_double operates with the 'word that comes first in memory in the
    target' terminology, while gen_lowpart operates with the 'value
    representing some low-order bits of X' terminology. They are not
    equivalent and must be dealt with differently on little- and big-endian
    targets.

    gcc/
            PR target/105879
            * config/xtensa/xtensa.md (movdi): Rename 'first' and 'second'
            to 'lowpart' and 'highpart' so that they match 'gen_lowpart' and
            'gen_highpart' bitwise semantics and fix order of highpart and
            lowpart depending on target endianness.

[Bug target/105879] DI mode constants are incorrectly loaded into registers on big-endian xtensa

[cvs-commit at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=105879

--- Comment #2 from CVS Commits <cvs-commit at gcc dot gnu.org> ---
The releases/gcc-12 branch has been updated by Max Filippov
<jcmvbkbc@gcc.gnu.org>:

https://gcc.gnu.org/g:777a80af43eebb7506937cf04b54f17c4fce2b8e

commit r12-8466-g777a80af43eebb7506937cf04b54f17c4fce2b8e
Author: Max Filippov <jcmvbkbc@gmail.com>
Date:   Tue Jun 7 21:01:01 2022 -0700

    gcc: xtensa: fix PR target/105879

    split_double operates with the 'word that comes first in memory in the
    target' terminology, while gen_lowpart operates with the 'value
    representing some low-order bits of X' terminology. They are not
    equivalent and must be dealt with differently on little- and big-endian
    targets.

    gcc/
            PR target/105879
            * config/xtensa/xtensa.md (movdi): Rename 'first' and 'second'
            to 'lowpart' and 'highpart' so that they match 'gen_lowpart' and
            'gen_highpart' bitwise semantics and fix order of highpart and
            lowpart depending on target endianness.

    (cherry picked from commit e94c6dbfb57a862dd8a8685eabc4886ad1aaea25)

[Bug target/105879] DI mode constants are incorrectly loaded into registers on big-endian xtensa

[cvs-commit at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=105879

--- Comment #3 from CVS Commits <cvs-commit at gcc dot gnu.org> ---
The releases/gcc-11 branch has been updated by Max Filippov
<jcmvbkbc@gcc.gnu.org>:

https://gcc.gnu.org/g:ee9cd6d6e504cb67b27b1a5e39f965bafd62653b

commit r11-10054-gee9cd6d6e504cb67b27b1a5e39f965bafd62653b
Author: Max Filippov <jcmvbkbc@gmail.com>
Date:   Tue Jun 7 21:01:01 2022 -0700

    gcc: xtensa: fix PR target/105879

    split_double operates with the 'word that comes first in memory in the
    target' terminology, while gen_lowpart operates with the 'value
    representing some low-order bits of X' terminology. They are not
    equivalent and must be dealt with differently on little- and big-endian
    targets.

    gcc/
            PR target/105879
            * config/xtensa/xtensa.md (movdi): Rename 'first' and 'second'
            to 'lowpart' and 'highpart' so that they match 'gen_lowpart' and
            'gen_highpart' bitwise semantics and fix order of highpart and
            lowpart depending on target endianness.

    (cherry picked from commit e94c6dbfb57a862dd8a8685eabc4886ad1aaea25)

[Bug target/105879] DI mode constants are incorrectly loaded into registers on big-endian xtensa

[jcmvbkbc at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=105879

jcmvbkbc at gcc dot gnu.org changed:

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

--- Comment #4 from jcmvbkbc at gcc dot gnu.org ---
Fix committed to master and gcc-11 and gcc-12 branches.