[Bug target/110268] New: arm MVE intrinsics support broken with LTO

[Bug target/110268] New: arm MVE intrinsics support broken with LTO

[clyon at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=110268

            Bug ID: 110268
           Summary: arm MVE intrinsics support broken with LTO
           Product: gcc
           Version: 14.0
            Status: UNCONFIRMED
          Severity: normal
          Priority: P3
         Component: target
          Assignee: unassigned at gcc dot gnu.org
          Reporter: clyon at gcc dot gnu.org
  Target Milestone: ---

Since I rewrote (a lot of) MVE intrinsics, use of LTO is broken:

$ cat ~/t.c
#include <arm_mve.h>
int main(void)
{
  return vaddvq(vdupq_n_s8 (1));
}

# no LTO, OK:
$ arm-none-eabi-gcc ~/t.c -march=armv8.1-m.main+mve -mfloat-abi=hard
-specs=rdimon.specs

# with LTO, broken:
$ arm-none-eabi-gcc ~/t.c -march=armv8.1-m.main+mve -mfloat-abi=hard
-specs=rdimon.specs -flto
lto1: fatal error: target specific builtin not available
compilation terminated.
lto-wrapper: fatal error: arm-none-eabi-gcc returned 1 exit status
compilation terminated.
/arm-none-eabi/bin/../lib/gcc/arm-none-eabi/14.0.0/../../../../arm-none-eabi/bin/ld:
error: lto-wrapper failed
collect2: error: ld returned 1 exit status

[Bug target/110268] [14 Regression] arm MVE intrinsics support broken with LTO

[acoplan at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=110268

Alex Coplan <acoplan at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
     Ever confirmed|0                           |1
   Last reconfirmed|                            |2023-06-15
             Status|UNCONFIRMED                 |NEW

--- Comment #1 from Alex Coplan <acoplan at gcc dot gnu.org> ---
Confirmed.

[Bug target/110268] [14 Regression] arm MVE intrinsics support broken with LTO

[clyon at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=110268

--- Comment #2 from Christophe Lyon <clyon at gcc dot gnu.org> ---
This regression appeared after the patch that re-implements vdupq, but the
issue is likely more generic.
velo r

I tried to update arm_init_mve_builtins() with:

+ if (in_lto_p)
+   {
+     arm_mve::handle_arm_mve_types_h ();
+     arm_mve::handle_arm_mve_h (false);
+   }

but that's not sufficient, we still get the same error although the MVE
intrinsics are registered.

I noticed that the error happens in unpack_ts_function_decl_value_fields() with
fcode==885, not sure where it comes from?

Besides, we'll also have to check how to compute the right value for
handle_arm_mve_h's parameter.

[Bug target/110268] [14 Regression] arm MVE intrinsics support broken with LTO

[rguenth at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=110268

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

           What    |Removed                     |Added
----------------------------------------------------------------------------
   Target Milestone|---                         |14.0

--- Comment #3 from Richard Biener <rguenth at gcc dot gnu.org> ---
At stream-out time we stream the MD specific DECL_FUNCTION_CODE and at
stream-in time we expect the backend has a FUNCTION_DECL corresponding to that
code for us.  The error means that targetm.builtin_decl doesn't know about the
builtin
(or did not implement "materializing" it).

[Bug target/110268] [14 Regression] arm MVE intrinsics support broken with LTO

[rguenth at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=110268

--- Comment #4 from Richard Biener <rguenth at gcc dot gnu.org> ---
tree
arm_builtin_decl (unsigned code, bool initialize_p ATTRIBUTE_UNUSED)
{ 
  unsigned subcode = code >> ARM_BUILTIN_SHIFT;
  switch (code & ARM_BUILTIN_CLASS)
    {
    case ARM_BUILTIN_GENERAL:
      return arm_general_builtin_decl (subcode);
    case ARM_BUILTIN_MVE:
      return error_mark_node;

^^^

I guess that's the problem.  MVE builtins are not handled at all?

[Bug target/110268] [14 Regression] arm MVE intrinsics support broken with LTO

[clyon at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=110268

Christophe Lyon <clyon at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|NEW                         |ASSIGNED

--- Comment #5 from Christophe Lyon <clyon at gcc dot gnu.org> ---
Indeed, a small update to arm_builtin_decl() does the trick.

Investigating how to decide whether to call
arm_mve::handle_arm_mve_h (false);
or
arm_mve::handle_arm_mve_h (true);

which is normally controlled at preprocessing time depending on
__ARM_MVE_PRESERVE_USER_NAMESPACE

[Bug target/110268] [14 Regression] arm MVE intrinsics support broken with LTO

[cvs-commit at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=110268

--- Comment #6 from CVS Commits <cvs-commit at gcc dot gnu.org> ---
The master branch has been updated by Christophe Lyon <clyon@gcc.gnu.org>:

https://gcc.gnu.org/g:eca10aaa3954af3dab56eccc208c90273c2b1732

commit r14-2418-geca10aaa3954af3dab56eccc208c90273c2b1732
Author: Christophe Lyon <christophe.lyon@linaro.org>
Date:   Mon Jun 26 14:39:47 2023 +0000

    arm: Fix MVE intrinsics support with LTO (PR target/110268)

    After the recent MVE intrinsics re-implementation, LTO stopped working
    because the intrinsics would no longer be defined.

    The main part of the patch is simple and similar to what we do for
    AArch64:
    - call handle_arm_mve_h() from arm_init_mve_builtins to declare the
      intrinsics when the compiler is in LTO mode
    - actually implement arm_builtin_decl for MVE.

    It was just a bit tricky to handle __ARM_MVE_PRESERVE_USER_NAMESPACE:
    its value in the user code cannot be guessed at LTO time, so we always
    have to assume that it was not defined.  The led to a few fixes in the
    way we register MVE builtins as placeholders or not.  Without this
    patch, we would just omit some versions of the inttrinsics when
    __ARM_MVE_PRESERVE_USER_NAMESPACE is true. In fact, like for the C/C++
    placeholders, we need to always keep entries for all of them to ensure
    that we have a consistent numbering scheme.

    2023-06-26  Christophe Lyon   <christophe.lyon@linaro.org>

            PR target/110268
            gcc/
            * config/arm/arm-builtins.cc (arm_init_mve_builtins): Handle LTO.
            (arm_builtin_decl): Hahndle MVE builtins.
            * config/arm/arm-mve-builtins.cc (builtin_decl): New function.
            (add_unique_function): Fix handling of
            __ARM_MVE_PRESERVE_USER_NAMESPACE.
            (add_overloaded_function): Likewise.
            * config/arm/arm-protos.h (builtin_decl): New declaration.

            gcc/testsuite/
            * gcc.target/arm/pr110268-1.c: New test.
            * gcc.target/arm/pr110268-2.c: New test.

[Bug target/110268] [14 Regression] arm MVE intrinsics support broken with LTO

[clyon at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=110268

Christophe Lyon <clyon at gcc dot gnu.org> changed:

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

--- Comment #7 from Christophe Lyon <clyon at gcc dot gnu.org> ---
Fixed.