[Bug middle-end/112976] New: expand_gimple_stmt_1 vs gimple_assign_nontemporal_move_p vs SSA_NAME on lhs

[Bug middle-end/112976] New: expand_gimple_stmt_1 vs gimple_assign_nontemporal_move_p vs SSA_NAME on lhs

[pinskia at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=112976

            Bug ID: 112976
           Summary: expand_gimple_stmt_1 vs
                    gimple_assign_nontemporal_move_p vs SSA_NAME on lhs
           Product: gcc
           Version: 14.0
            Status: UNCONFIRMED
          Keywords: internal-improvement
          Severity: normal
          Priority: P3
         Component: middle-end
          Assignee: unassigned at gcc dot gnu.org
          Reporter: pinskia at gcc dot gnu.org
  Target Milestone: ---

As far as I can tell if TREE_CODE (lhs) == SSA_NAME then
gimple_assign_nontemporal_move_p should always be false as it is never a store.

So all of the code for nontemporal in the else case of `TREE_CODE (lhs) !=
SSA_NAME` if in cfgexpand.cc is dead.

This has been there since r0-95521-g28ed065ef9f345 which added expand from
tuples directly.

We most likely should also have gimple_assign_set_nontemporal_move assert that
the gimple assign's lhs is not a SSA_NAME.

[Bug middle-end/112976] expand_gimple_stmt_1 vs gimple_assign_nontemporal_move_p vs SSA_NAME on lhs

[pinskia at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=112976

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

           What    |Removed                     |Added
----------------------------------------------------------------------------
           Assignee|unassigned at gcc dot gnu.org      |pinskia at gcc dot gnu.org
   Last reconfirmed|                            |2023-12-12
             Status|UNCONFIRMED                 |ASSIGNED
     Ever confirmed|0                           |1

--- Comment #1 from Andrew Pinski <pinskia at gcc dot gnu.org> ---
Mine for cleanup for GCC 15.

[Bug middle-end/112976] expand_gimple_stmt_1 vs gimple_assign_nontemporal_move_p vs SSA_NAME on lhs

[pinskia at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=112976

--- Comment #2 from Andrew Pinski <pinskia at gcc dot gnu.org> ---
Note gimple_assign_nontemporal_move_p is just for non temporal stores. There is
no code handling non-temporal loads (which do exist on some targets, aarch64
for an example). 
I will also add a comment to that effect above gimple_assign_nontemporal_move_p
too.

Plus I will also add to the verifier that gimple_assign_nontemporal_move_p is
only set for the case where LHS != SSA_NAME ....

[Bug middle-end/112976] expand_gimple_stmt_1 vs gimple_assign_nontemporal_move_p vs SSA_NAME on lhs

[pinskia at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=112976

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

           What    |Removed                     |Added
----------------------------------------------------------------------------
           See Also|                            |https://gcc.gnu.org/bugzill
                   |                            |a/show_bug.cgi?id=101852

--- Comment #3 from Andrew Pinski <pinskia at gcc dot gnu.org> ---
Note storent optab name is not documented.

[Bug middle-end/112976] expand_gimple_stmt_1 vs gimple_assign_nontemporal_move_p vs SSA_NAME on lhs

[pinskia at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=112976

--- Comment #4 from Andrew Pinski <pinskia at gcc dot gnu.org> ---
Created attachment 57978
  --> https://gcc.gnu.org/bugzilla/attachment.cgi?id=57978&action=edit
Patch 1/2

[Bug middle-end/112976] expand_gimple_stmt_1 vs gimple_assign_nontemporal_move_p vs SSA_NAME on lhs

[pinskia at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=112976

--- Comment #5 from Andrew Pinski <pinskia at gcc dot gnu.org> ---
Created attachment 57979
  --> https://gcc.gnu.org/bugzilla/attachment.cgi?id=57979&action=edit
Patch 2/2

[Bug middle-end/112976] expand_gimple_stmt_1 vs gimple_assign_nontemporal_move_p vs SSA_NAME on lhs

[pinskia at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=112976

--- Comment #6 from Andrew Pinski <pinskia at gcc dot gnu.org> ---
Will submit both patches once GCC 15 opens up.

[Bug middle-end/112976] expand_gimple_stmt_1 vs gimple_assign_nontemporal_move_p vs SSA_NAME on lhs

[pinskia at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=112976

--- Comment #7 from Andrew Pinski <pinskia at gcc dot gnu.org> ---
Patches posted:
https://gcc.gnu.org/pipermail/gcc-patches/2024-April/650085.html
https://gcc.gnu.org/pipermail/gcc-patches/2024-April/650084.html

[Bug middle-end/112976] expand_gimple_stmt_1 vs gimple_assign_nontemporal_move_p vs SSA_NAME on lhs

[cvs-commit at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=112976

--- Comment #8 from GCC Commits <cvs-commit at gcc dot gnu.org> ---
The trunk branch has been updated by Andrew Pinski <pinskia@gcc.gnu.org>:

https://gcc.gnu.org/g:e3a7f359c18bf347f6ac8fcda05e9839fac5bd62

commit r15-72-ge3a7f359c18bf347f6ac8fcda05e9839fac5bd62
Author: Andrew Pinski <quic_apinski@quicinc.com>
Date:   Wed Apr 17 14:12:17 2024 -0700

    Add verification of gimple_assign_nontemporal_move_p [PR112976]

    Currently the middle-end only knows how to support temporal stores
    (the undocumented storent optab) so let's verify that the only time
    we set nontemporal_move on an assign is if the the lhs is not a
    gimple reg.

    Bootstrapped and tested on x86_64-linux-gnu no regressions.

    gcc/ChangeLog:

            PR middle-end/112976
            * tree-cfg.cc (verify_gimple_assign): Verify that
            nontmporal moves are stores.
            * gimple.h (struct gimple): Note that only
            nontemporal stores are supported.

    Signed-off-by: Andrew Pinski <quic_apinski@quicinc.com>

[Bug middle-end/112976] expand_gimple_stmt_1 vs gimple_assign_nontemporal_move_p vs SSA_NAME on lhs

[cvs-commit at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=112976

--- Comment #9 from GCC Commits <cvs-commit at gcc dot gnu.org> ---
The trunk branch has been updated by Andrew Pinski <pinskia@gcc.gnu.org>:

https://gcc.gnu.org/g:8614d60233a64afd7e28ae7af2ab74c4a5b06010

commit r15-73-g8614d60233a64afd7e28ae7af2ab74c4a5b06010
Author: Andrew Pinski <quic_apinski@quicinc.com>
Date:   Wed Apr 17 14:30:06 2024 -0700

    Remove support for nontemporal stores with ssa_names on lhs [PR112976]

    When cfgexpand was changed to support expanding from tuple gimple
    (r0-95521-g28ed065ef9f345), the code was added to support
    doing nontemporal stores with LHS of a SSA_NAME but that will
    never be a nontemporal store.
    This patch removes that and asserts that expanding with a LHS
    of a SSA_NAME is not a nontemporal store.

    Bootstrapped and tested on x86_64-linux-gnu.

    gcc/ChangeLog:

            PR middle-end/112976
            * cfgexpand.cc (expand_gimple_stmt_1): Remove
            support for expanding nontemporal "moves" with
            ssa names on the LHS.

    Signed-off-by: Andrew Pinski <quic_apinski@quicinc.com>

[Bug middle-end/112976] expand_gimple_stmt_1 vs gimple_assign_nontemporal_move_p vs SSA_NAME on lhs

[pinskia at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=112976

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

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|ASSIGNED                    |RESOLVED
   Target Milestone|---                         |15.0
         Resolution|---                         |FIXED

--- Comment #10 from Andrew Pinski <pinskia at gcc dot gnu.org> ---
Fixed.