[Bug other/110744] New: [14 regression] cc.dg/tree-ssa/pr84512.c fails after r14-2267-gb8806f6ffbe72

[Bug other/110744] New: [14 regression] cc.dg/tree-ssa/pr84512.c fails after r14-2267-gb8806f6ffbe72

[seurer at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=110744

            Bug ID: 110744
           Summary: [14 regression] cc.dg/tree-ssa/pr84512.c fails after
                    r14-2267-gb8806f6ffbe72
           Product: gcc
           Version: 14.0
            Status: UNCONFIRMED
          Severity: normal
          Priority: P3
         Component: other
          Assignee: unassigned at gcc dot gnu.org
          Reporter: seurer at gcc dot gnu.org
  Target Milestone: ---

g:b8806f6ffbe72ed8fb7aba3b3a9196ec4d13e614, r14-2267-gb8806f6ffbe72e

make  -k check-gcc RUNTESTFLAGS="tree-ssa.exp=gcc.dg/tree-ssa/pr84512.c"
FAIL: gcc.dg/tree-ssa/pr84512.c scan-tree-dump optimized "return 285;"
# of expected passes            1
# of unexpected failures        1

commit b8806f6ffbe72ed8fb7aba3b3a9196ec4d13e614 (HEAD)
Author: Ju-Zhe Zhong <juzhe.zhong@rivai.ai>
Date:   Mon Jul 3 17:10:26 2023 +0800

    Middle-end: Change order of LEN_MASK_LOAD/LEN_MASK_STORE arguments


Hmmm.  There is PR84935 which looks to be the same sort of thing but this just
started occurring for me on powerpc64 with r14-2267-gb8806f6ffbe72.

[Bug other/110744] [14 regression] cc.dg/tree-ssa/pr84512.c fails after r14-2267-gb8806f6ffbe72

[juzhe.zhong at rivai dot ai]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=110744

--- Comment #1 from JuzheZhong <juzhe.zhong at rivai dot ai> ---
I am really sorry for causing the inconvience.
Working on it.

[Bug other/110744] [14 regression] cc.dg/tree-ssa/pr84512.c fails after r14-2267-gb8806f6ffbe72

[pan2.li at intel dot com]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=110744

--- Comment #2 from Li Pan <pan2.li at intel dot com> ---
Hi there,

Just try to reproduce this bug with powerPC cross compiler (sorry we don't have
a real powerPC) with the below options. Unfortunately, I failed to reproduce
this bug as above mentioned.

Could you please help to take a look if there is something missing or
incorrect?

1. Build cross compiler

../configure \
  --target=powerpc-unknown-elf \
  --prefix=${INSTALL_DIR} \
  --disable-shared \
  --enable-threads \
  --enable-tls \
  --enable-languages=c,c++ \
  --with-system-zlib \
  --with-newlib \
  --disable-libmudflap \
  --disable-libssp \
  --disable-libquadmath \
  --disable-libgomp \
  --enable-nls \
  --disable-tm-clone-registry \
  --enable-multilib \
  --src=`pwd`/../ \
  --enable-werror \

make -j $(nproc) all-gcc && make install-gcc

2. Compile the source file.

>> ./bin/powerpc-unknown-elf-gcc -mcpu=power10 -O3 -fdump-tree-optimized -c -S ../gcc/gcc/testsuite/gcc.dg/tree-ssa/pr84512.c -o -
        .file   "pr84512.c"
        .machine power10
        .section        ".text"
        .align 2
        .align 4
        .globl foo
        .type   foo, @function
foo:
.LFB0:
        li 3,285
        blr
.LFE0:
        .size   foo, .-foo
        .section        .eh_frame,"aw",@progbits
.Lframe1:
        .4byte  .LECIE1-.LSCIE1
.LSCIE1:
        .4byte  0
        .byte   0x3
        .string ""
        .byte   0x1
        .byte   0x7c
        .byte   0x41
        .byte   0xc
        .byte   0x1
        .byte   0
        .align 2
.LECIE1:
.LSFDE1:
        .4byte  .LEFDE1-.LASFDE1
.LASFDE1:
        .4byte  .LASFDE1-.Lframe1
        .4byte  .LFB0
        .4byte  .LFE0-.LFB0
        .align 2
.LEFDE1:
        .ident  "GCC: (GNU) 14.0.0 20230720 (experimental)"

>> cat pr84512.c.257t.optimized

;; Function foo (foo, funcdef_no=0, decl_uid=3445, cgraph_uid=1,
symbol_order=0)

int foo ()
{
  <bb 2> [local count: 97603129]:
  return 285;

}

[Bug other/110744] [14 regression] cc.dg/tree-ssa/pr84512.c fails after r14-2267-gb8806f6ffbe72

[juzhe.zhong at rivai dot ai]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=110744

--- Comment #3 from JuzheZhong <juzhe.zhong at rivai dot ai> ---
Kewen from IBM will soon send a patch to fix it.
Sorry for causing this issue.

Thanks.

[Bug other/110744] [14 regression] cc.dg/tree-ssa/pr84512.c fails after r14-2267-gb8806f6ffbe72

[linkw at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=110744

Kewen Lin <linkw at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
     Ever confirmed|0                           |1
             Status|UNCONFIRMED                 |ASSIGNED
   Last reconfirmed|                            |2023-07-20

--- Comment #4 from Kewen Lin <linkw at gcc dot gnu.org> ---
Assigning to myself as discussing with Juzhe.

[Bug other/110744] [14 regression] gcc.dg/tree-ssa/pr84512.c fails after r14-2267-gb8806f6ffbe72

[linkw at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=110744

--- Comment #5 from Kewen Lin <linkw at gcc dot gnu.org> ---
(In reply to Li Pan from comment #2)
> Hi there,
> 
> Just try to reproduce this bug with powerPC cross compiler (sorry we don't
> have a real powerPC) with the below options. Unfortunately, I failed to
> reproduce this bug as above mentioned.

Cfarm (https://cfarm.tetaneutral.net/) recently gets one Power10 machine, you
can have a try there. :) But yeah, for this kind of compilation issue, cross
compiler is normally enough.

> 
> Could you please help to take a look if there is something missing or
> incorrect?
> 
> 1. Build cross compiler
> 
> ../configure \
>   --target=powerpc-unknown-elf \

I can reproduce it with --target=powerpc64le-unknown-linux-gnu.

powerpc is 32bit Power, unfortunately lxvl and stxvl (vector with length)
hardware instructions require 64bit GPR (for holding length in some of its high
bits), so partial vector is not enabled there.

>   --prefix=${INSTALL_DIR} \
>   --disable-shared \
>   --enable-threads \
>   --enable-tls \
>   --enable-languages=c,c++ \
>   --with-system-zlib \
>   --with-newlib \
>   --disable-libmudflap \
>   --disable-libssp \
>   --disable-libquadmath \
>   --disable-libgomp \
>   --enable-nls \
>   --disable-tm-clone-registry \
>   --enable-multilib \
>   --src=`pwd`/../ \
>   --enable-werror \
> 
> make -j $(nproc) all-gcc && make install-gcc
>

[Bug other/110744] [14 regression] gcc.dg/tree-ssa/pr84512.c fails after r14-2267-gb8806f6ffbe72

[linkw at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=110744

--- Comment #6 from Kewen Lin <linkw at gcc dot gnu.org> ---
The root cause is that the length and bias handling about LEN_STORE in sccvn
was missed to be updated, the below diff can fix the failure.

diff --git a/gcc/tree-ssa-sccvn.cc b/gcc/tree-ssa-sccvn.cc
index 11061a374a2..c0b3ec420c5 100644
--- a/gcc/tree-ssa-sccvn.cc
+++ b/gcc/tree-ssa-sccvn.cc
@@ -3299,11 +3299,14 @@ vn_reference_lookup_3 (ao_ref *ref, tree vuse, void
*data_,
            return (void *)-1;
          break;
        case IFN_LEN_STORE:
-         len = gimple_call_arg (call, 2);
-         bias = gimple_call_arg (call, 4);
-         if (!tree_fits_uhwi_p (len) || !tree_fits_shwi_p (bias))
-           return (void *)-1;
-         break;
+         {
+           int len_index = internal_fn_len_index (fn);
+           len = gimple_call_arg (call, len_index);
+           bias = gimple_call_arg (call, len_index + 1);
+           if (!tree_fits_uhwi_p (len) || !tree_fits_shwi_p (bias))
+             return (void *) -1;
+           break;
+         }
        default:
          return (void *)-1;
        }

Fully testing it.

[Bug other/110744] [14 regression] gcc.dg/tree-ssa/pr84512.c fails after r14-2267-gb8806f6ffbe72

[pan2.li at intel dot com]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=110744

--- Comment #7 from Li Pan <pan2.li at intel dot com> ---
Thanks a lot for the explanation, Kewen. 

Looks you are taking care of this already, anything is required from my-side
please feel free to let me know.

[Bug other/110744] [14 regression] gcc.dg/tree-ssa/pr84512.c fails after r14-2267-gb8806f6ffbe72

[rguenth at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=110744

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

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

[Bug other/110744] [14 regression] gcc.dg/tree-ssa/pr84512.c fails after r14-2267-gb8806f6ffbe72

[cvs-commit at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=110744

--- Comment #8 from CVS Commits <cvs-commit at gcc dot gnu.org> ---
The master branch has been updated by Kewen Lin <linkw@gcc.gnu.org>:

https://gcc.gnu.org/g:a6654c08fde11890d621fa7831180d410054568a

commit r14-2694-ga6654c08fde11890d621fa7831180d410054568a
Author: Kewen Lin <linkw@linux.ibm.com>
Date:   Fri Jul 21 00:18:19 2023 -0500

    sccvn: Correct the index of bias for IFN_LEN_STORE [PR110744]

    Commit r14-2267-gb8806f6ffbe72e adjusts the arguments order
    of LEN_STORE from {len,vector,bias} to {len,bias,vector},
    in order to make them consistent with LEN_MASK_STORE and
    MASK_STORE.  But it missed to update the related handlings
    in tree-ssa-sccvn.cc, it caused the failure shown in PR
    110744.  This patch is to fix the related handlings with
    the correct index.

            PR tree-optimization/110744

    gcc/ChangeLog:

            * tree-ssa-sccvn.cc (vn_reference_lookup_3): Correct the index of
bias
            operand for ifn IFN_LEN_STORE.

[Bug tree-optimization/110744] [14 regression] gcc.dg/tree-ssa/pr84512.c fails after r14-2267-gb8806f6ffbe72

[linkw at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=110744

Kewen Lin <linkw at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
          Component|other                       |tree-optimization
         Resolution|---                         |FIXED
             Status|ASSIGNED                    |RESOLVED

--- Comment #9 from Kewen Lin <linkw at gcc dot gnu.org> ---
Should be fixed on trunk.