[Bug tree-optimization/102087] New: ICE on valid code at -O3 on x86_64-linux-gnu: in determine_exit_conditions, at tree-ssa-loop-manip.c:1049

[Bug tree-optimization/102087] New: ICE on valid code at -O3 on x86_64-linux-gnu: in determine_exit_conditions, at tree-ssa-loop-manip.c:1049

[zhendong.su at inf dot ethz.ch]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=102087

            Bug ID: 102087
           Summary: ICE on valid code at -O3 on x86_64-linux-gnu: in
                    determine_exit_conditions, at
                    tree-ssa-loop-manip.c:1049
           Product: gcc
           Version: unknown
            Status: UNCONFIRMED
          Severity: normal
          Priority: P3
         Component: tree-optimization
          Assignee: unassigned at gcc dot gnu.org
          Reporter: zhendong.su at inf dot ethz.ch
  Target Milestone: ---

[540] % gcctk -v
Using built-in specs.
COLLECT_GCC=gcctk
COLLECT_LTO_WRAPPER=/local/suz-local/software/local/gcc-trunk/libexec/gcc/x86_64-pc-linux-gnu/12.0.0/lto-wrapper
Target: x86_64-pc-linux-gnu
Configured with: ../gcc-trunk/configure --disable-bootstrap
--prefix=/local/suz-local/software/local/gcc-trunk --enable-languages=c,c++
--disable-werror --enable-multilib --with-system-zlib
Thread model: posix
Supported LTO compression algorithms: zlib
gcc version 12.0.0 20210826 (experimental) [master revision
352d5e85a70:96f76144c4d:5a6c626710ad2ac4baa2dba02fac0750177e3305] (GCC)
[541] %
[541] % gcctk -O2 small.c
[542] %
[542] % gcctk -O3 small.c
during GIMPLE pass: unrolljam
small.c: In function ‘main’:
small.c:4:5: internal compiler error: in determine_exit_conditions, at
tree-ssa-loop-manip.c:1049
    4 | int main() {
      |     ^~~~
0xfe4bea determine_exit_conditions
        ../../gcc-trunk/gcc/tree-ssa-loop-manip.c:1049
0xfe4bea tree_transform_and_unroll_loop(loop*, unsigned int, edge_def*,
tree_niter_desc*, void (*)(loop*, void*), void*)
        ../../gcc-trunk/gcc/tree-ssa-loop-manip.c:1253
0x19ed2eb tree_loop_unroll_and_jam
        ../../gcc-trunk/gcc/gimple-loop-jam.c:590
Please submit a full bug report,
with preprocessed source if appropriate.
Please include the complete backtrace with any bug report.
See <https://gcc.gnu.org/bugs/> for instructions.
[543] %
[543] % cat small.c
volatile int a[1];
unsigned b;
int c;
int main() {
  int d;
  for (; b > 1; b++)
    for (c = 0; c < 2; c++)
      for (d = 0; d < 2; d++)
        a[0];
  return 0;
}

[Bug tree-optimization/102087] [12 Regression] ICE on valid code at -O3 on x86_64-linux-gnu: in determine_exit_conditions, at tree-ssa-loop-manip.c:1049

[rguenth at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=102087

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

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Target|                            |x86_64-*-*
            Summary|ICE on valid code at -O3 on |[12 Regression] ICE on
                   |x86_64-linux-gnu: in        |valid code at -O3 on
                   |determine_exit_conditions,  |x86_64-linux-gnu: in
                   |at                          |determine_exit_conditions,
                   |tree-ssa-loop-manip.c:1049  |at
                   |                            |tree-ssa-loop-manip.c:1049
   Target Milestone|---                         |12.0
            Version|unknown                     |12.0

--- Comment #1 from Richard Biener <rguenth at gcc dot gnu.org> ---
This looks similar to PR102072

[Bug tree-optimization/102087] [12 Regression] ICE on valid code at -O3 on x86_64-linux-gnu: in determine_exit_conditions, at tree-ssa-loop-manip.c:1049 since r12-3136-g3673dcf6d6baeb67

[marxin at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=102087

Martin Liška <marxin at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|UNCONFIRMED                 |NEW
            Summary|[12 Regression] ICE on      |[12 Regression] ICE on
                   |valid code at -O3 on        |valid code at -O3 on
                   |x86_64-linux-gnu: in        |x86_64-linux-gnu: in
                   |determine_exit_conditions,  |determine_exit_conditions,
                   |at                          |at
                   |tree-ssa-loop-manip.c:1049  |tree-ssa-loop-manip.c:1049
                   |                            |since
                   |                            |r12-3136-g3673dcf6d6baeb67
   Last reconfirmed|                            |2021-08-27
                 CC|                            |marxin at gcc dot gnu.org
     Ever confirmed|0                           |1

[Bug tree-optimization/102087] [12 Regression] ICE on valid code at -O3 on x86_64-linux-gnu: in determine_exit_conditions, at tree-ssa-loop-manip.c:1049 since r12-3136-g3673dcf6d6baeb67

[marxin at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=102087

--- Comment #2 from Martin Liška <marxin at gcc dot gnu.org> ---
Started with r12-3136-g3673dcf6d6baeb67.

[Bug tree-optimization/102087] [12 Regression] ICE on valid code at -O3 on x86_64-linux-gnu: in determine_exit_conditions, at tree-ssa-loop-manip.c:1049 since r12-3136-g3673dcf6d6baeb67

[guojiufu at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=102087

Jiu Fu Guo <guojiufu at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|NEW                         |RESOLVED
         Resolution|---                         |DUPLICATE

--- Comment #3 from Jiu Fu Guo <guojiufu at gcc dot gnu.org> ---
Yes, it is a duplicate of PR102072.

*** This bug has been marked as a duplicate of bug 102072 ***

[Bug tree-optimization/102087] [12 Regression] ICE on valid code at -O3 on x86_64-linux-gnu: in determine_exit_conditions, at tree-ssa-loop-manip.c:1049 since r12-3136-g3673dcf6d6baeb67

[hjl.tools at gmail dot com]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=102087

H.J. Lu <hjl.tools at gmail dot com> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
         Resolution|DUPLICATE                   |---
             Status|RESOLVED                    |REOPENED

--- Comment #4 from H.J. Lu <hjl.tools at gmail dot com> ---
Reopen.  This test isn't specific to armeb.

[Bug tree-optimization/102087] [12 Regression] ICE on valid code at -O3 on x86_64-linux-gnu: in determine_exit_conditions, at tree-ssa-loop-manip.c:1049 since r12-3136-g3673dcf6d6baeb67

[hjl.tools at gmail dot com]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=102087

H.J. Lu <hjl.tools at gmail dot com> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |clyon at gcc dot gnu.org

--- Comment #5 from H.J. Lu <hjl.tools at gmail dot com> ---
*** Bug 102072 has been marked as a duplicate of this bug. ***

[Bug tree-optimization/102087] [12 Regression] ICE on valid code at -O3 on x86_64-linux-gnu: in determine_exit_conditions, at tree-ssa-loop-manip.c:1049 since r12-3136-g3673dcf6d6baeb67

[guojiufu at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=102087

--- Comment #6 from Jiu Fu Guo <guojiufu at gcc dot gnu.org> ---
Drafting patch to calculate three items: control, bound and cmp.
----
diff --git a/gcc/tree-ssa-loop-niter.c b/gcc/tree-ssa-loop-niter.c
index 7af92d1c893..c6e4b24fd83 100644
--- a/gcc/tree-ssa-loop-niter.c
+++ b/gcc/tree-ssa-loop-niter.c
@@ -1482,7 +1482,7 @@ number_of_iterations_until_wrap (class loop *, tree type,
affine_iv *iv0,
                                 affine_iv *iv1, class tree_niter_desc *niter)
 {
   tree niter_type = unsigned_type_for (type);
-  tree step, num, assumptions, may_be_zero;
+  tree step, num, assumptions, may_be_zero, span;
   wide_int high, low, max, min;

   may_be_zero = fold_build2 (LE_EXPR, boolean_type_node, iv1->base,
iv0->base);
@@ -1513,6 +1513,8 @@ number_of_iterations_until_wrap (class loop *, tree type,
affine_iv *iv0,
        low = wi::to_wide (iv0->base);
       else
        low = min;
+
+      niter->control = *iv1;
     }
   /* {base, -C} < n.  */
   else if (tree_int_cst_sign_bit (iv0->step) && integer_zerop (iv1->step))
@@ -1533,6 +1535,8 @@ number_of_iterations_until_wrap (class loop *, tree type,
affine_iv *iv0,
        high = wi::to_wide (iv1->base);
       else
        high = max;
+
+      niter->control = *iv0;
     }
   else
     return false;
@@ -1556,6 +1560,14 @@ number_of_iterations_until_wrap (class loop *, tree
type, affine_iv *iv0,
                                      niter->assumptions, assumptions);

   niter->control.no_overflow = false;
+  tree niter_m1 = fold_build2 (MINUS_EXPR, niter_type, niter->niter,
+                             build_int_cst (niter_type, 1));
+  span = fold_build2 (MULT_EXPR, niter_type, niter_m1,
+                     fold_convert (niter_type, niter->control.step));
+  niter->bound = fold_build2 (PLUS_EXPR, niter_type, span,
+                             fold_convert (niter_type, niter->control.base));
+  niter->bound = fold_convert (type, niter->bound);
+  niter->cmp = NE_EXPR;

   return true;
 }
------------------------

While this code may generate complicated niter->bound if the step is not +-1.

[Bug tree-optimization/102087] [12 Regression] ICE on valid code at -O3 on x86_64-linux-gnu: in determine_exit_conditions, at tree-ssa-loop-manip.c:1049 since r12-3136-g3673dcf6d6baeb67

[guojiufu at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=102087

--- Comment #7 from Jiu Fu Guo <guojiufu at gcc dot gnu.org> ---
If step is +-1, or if the 'iv base' is constant, the 'bound' would be
calculated as const.  

Otherwise, the 'bound' maybe something like: "(max - base) / step * step +
base". For this case, then runtime cost may need to calculate 'bound'. 
I'm wondering if it is beneficial to support this kind of case.

[Bug tree-optimization/102087] [12 Regression] ICE on valid code at -O3 on x86_64-linux-gnu: in determine_exit_conditions, at tree-ssa-loop-manip.c:1049 since r12-3136-g3673dcf6d6baeb67

[dcb314 at hotmail dot com]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=102087

David Binderman <dcb314 at hotmail dot com> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |dcb314 at hotmail dot com

--- Comment #8 from David Binderman <dcb314 at hotmail dot com> ---
I see this problem also, with this reduced C code:

char **Gif_ClipImage_gfi_0;
int Gif_ClipImage_y, Gif_ClipImage_shift;
void Gif_ClipImage(void) {
  for (; Gif_ClipImage_y >= Gif_ClipImage_shift; Gif_ClipImage_y++)
    Gif_ClipImage_gfi_0[Gif_ClipImage_shift] =
        Gif_ClipImage_gfi_0[Gif_ClipImage_y];
}

Flag -O3 -march=bdver2 required.

[Bug tree-optimization/102087] [12 Regression] ICE on valid code at -O3 on x86_64-linux-gnu: in determine_exit_conditions, at tree-ssa-loop-manip.c:1049 since r12-3136-g3673dcf6d6baeb67

[dcb314 at hotmail dot com]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=102087

--- Comment #9 from David Binderman <dcb314 at hotmail dot com> ---
This second source code generates a similar but different error message:

void dnslabel_vector_dnslabel_to_dnsname_namestack(int bottom) {
  char **name;
  int top;
  while (bottom <= top) {
    char label = *name++;
    memcpy(0, label, label);
    bottom--;
  }
}

internal compiler error: in determine_exit_conditions, at
tree-ssa-loop-manip.c:1045

Some four lines different place in the compiler source code to the original.

[Bug tree-optimization/102087] [12 Regression] ICE on valid code at -O3 on x86_64-linux-gnu: in determine_exit_conditions, at tree-ssa-loop-manip.c:1049 since r12-3136-g3673dcf6d6baeb67

[guojiufu at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=102087

--- Comment #10 from Jiu Fu Guo <guojiufu at gcc dot gnu.org> ---
Drafted a patch:

diff --git a/gcc/tree-ssa-loop-niter.c b/gcc/tree-ssa-loop-niter.c
index 7af92d1c893..5c77c8b7d51 100644
--- a/gcc/tree-ssa-loop-niter.c
+++ b/gcc/tree-ssa-loop-niter.c
@@ -1482,7 +1482,7 @@ number_of_iterations_until_wrap (class loop *, tree type,
affine_iv *iv0,
                                 affine_iv *iv1, class tree_niter_desc *niter)
 {
   tree niter_type = unsigned_type_for (type);
-  tree step, num, assumptions, may_be_zero;
+  tree step, num, assumptions, may_be_zero, span;
   wide_int high, low, max, min;

   may_be_zero = fold_build2 (LE_EXPR, boolean_type_node, iv1->base,
iv0->base);
@@ -1513,6 +1513,8 @@ number_of_iterations_until_wrap (class loop *, tree type,
affine_iv *iv0,
        low = wi::to_wide (iv0->base);
       else
        low = min;
+
+      niter->control = *iv1;
     }
   /* {base, -C} < n.  */
   else if (tree_int_cst_sign_bit (iv0->step) && integer_zerop (iv1->step))
@@ -1533,6 +1535,8 @@ number_of_iterations_until_wrap (class loop *, tree type,
affine_iv *iv0,
        high = wi::to_wide (iv1->base);
       else
        high = max;
+
+      niter->control = *iv0;
     }
   else
     return false;
@@ -1556,6 +1560,14 @@ number_of_iterations_until_wrap (class loop *, tree
type, affine_iv *iv0,
                                      niter->assumptions, assumptions);

   niter->control.no_overflow = false;
+  niter->control.base = fold_build2 (MINUS_EXPR, niter_type,
+                                    niter->control.base, niter->control.step);
+  span = fold_build2 (MULT_EXPR, niter_type, niter->niter,
+                     fold_convert (niter_type, niter->control.step));
+  niter->bound = fold_build2 (PLUS_EXPR, niter_type, span,
+                             fold_convert (niter_type, niter->control.base));
+  niter->bound = fold_convert (type, niter->bound);
+  niter->cmp = NE_EXPR;

   return true;
 }


This patch supports the case even the 'bound' is not a const.

[Bug tree-optimization/102087] [12 Regression] ICE on valid code at -O3 on x86_64-linux-gnu: in determine_exit_conditions, at tree-ssa-loop-manip.c:1049 since r12-3136-g3673dcf6d6baeb67

[pinskia at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=102087

--- Comment #11 from Andrew Pinski <pinskia at gcc dot gnu.org> ---
*** Bug 102172 has been marked as a duplicate of this bug. ***

[Bug tree-optimization/102087] [12 Regression] ICE on valid code at -O3 on x86_64-linux-gnu: in determine_exit_conditions, at tree-ssa-loop-manip.c:1049 since r12-3136-g3673dcf6d6baeb67

[cvs-commit at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=102087

--- Comment #12 from CVS Commits <cvs-commit at gcc dot gnu.org> ---
The master branch has been updated by Jiu Fu Guo <guojiufu@gcc.gnu.org>:

https://gcc.gnu.org/g:3087d1b0a2cbe8880ccbbb67744726ff4e5d390e

commit r12-3791-g3087d1b0a2cbe8880ccbbb67744726ff4e5d390e
Author: Jiufu Guo <guojiufu@linux.ibm.com>
Date:   Wed Sep 22 13:20:29 2021 +0800

    Set bound/cmp/control for until wrap loop.

    In patch r12-3136, niter->control, niter->bound and niter->cmp are
    derived from number_of_iterations_lt.  While for 'until wrap condition',
    the calculation in number_of_iterations_lt is not align the requirements
    on the define of them and requirements in determine_exit_conditions.

    This patch calculate niter->control, niter->bound and niter->cmp in
    number_of_iterations_until_wrap.

    gcc/ChangeLog:

    2021-09-22  Jiufu Guo  <guojiufu@linux.ibm.com>

            PR tree-optimization/102087
            * tree-ssa-loop-niter.c (number_of_iterations_until_wrap):
            Update bound/cmp/control for niter.

    gcc/testsuite/ChangeLog:

    2021-09-22  Jiufu Guo  <guojiufu@linux.ibm.com>

            * gcc.dg/pr102087.c: New test.
            PR tree-optimization/102087

[Bug tree-optimization/102087] [12 Regression] ICE on valid code at -O3 on x86_64-linux-gnu: in determine_exit_conditions, at tree-ssa-loop-manip.c:1049 since r12-3136-g3673dcf6d6baeb67

[dcb314 at hotmail dot com]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=102087

--- Comment #13 from David Binderman <dcb314 at hotmail dot com> ---
The code in comment 8 still seems to fail:

$ /home/dcb/gcc/results/bin/gcc -c -O3 -w -march=bdver2 bug760.c 
bug760.c: In function ‘Gif_ClipImage’:
bug760.c:3:1: error: type mismatch in binary expression
    3 | Gif_ClipImage() {
      | ^~~~~~~~~~~~~
unsigned int

int

int

_18 = Gif_ClipImage_gfi_1.0_1 + -1;
bug760.c:3:1: error: type mismatch in binary expression
unsigned int

int

int

_12 = Gif_ClipImage_gfi_1.0_1 + -1;
during GIMPLE pass: aprefetch
bug760.c:3:1: internal compiler error: verify_gimple failed
0xdac31a verify_gimple_in_cfg(function*, bool)
        ../../trunk.git/gcc/tree-cfg.c:5576

$ /home/dcb/gcc/results/bin/gcc -v 
Using built-in specs.
COLLECT_GCC=/home/dcb/gcc/results/bin/gcc
COLLECT_LTO_WRAPPER=/home/dcb/gcc/results.20210924/libexec/gcc/x86_64-pc-linux-g
nu/12.0.0/lto-wrapper
Target: x86_64-pc-linux-gnu
Configured with: ../trunk.git/configure --prefix=/home/dcb/gcc/results.20210924 
--disable-bootstrap --disable-multilib --disable-werror
--with-pkgversion=29c928
57039d0a10 --enable-checking=df,extra,fold,rtl,yes
--enable-languages=c,c++,fort
ran
Thread model: posix
Supported LTO compression algorithms: zlib zstd
gcc version 12.0.0 20210924 (experimental) (29c92857039d0a10)

[Bug tree-optimization/102087] [12 Regression] ICE on valid code at -O3 on x86_64-linux-gnu: in determine_exit_conditions, at tree-ssa-loop-manip.c:1049 since r12-3136-g3673dcf6d6baeb67

[rguenth at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=102087

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

           What    |Removed                     |Added
----------------------------------------------------------------------------
           Priority|P3                          |P1
             Status|REOPENED                    |NEW

--- Comment #14 from Richard Biener <rguenth at gcc dot gnu.org> ---
Confirmed.

[Bug tree-optimization/102087] [12 Regression] ICE on valid code at -O3 on x86_64-linux-gnu: in determine_exit_conditions, at tree-ssa-loop-manip.c:1049 since r12-3136-g3673dcf6d6baeb67

[dcb314 at hotmail dot com]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=102087

--- Comment #15 from David Binderman <dcb314 at hotmail dot com> ---
Since this bug depends on the setting of -march,
I tried out all the possible legal settings of that value.

alderlake
amdfam10
during GIMPLE pass: aprefetch
athlon-fx
during GIMPLE pass: aprefetch
athlon64
during GIMPLE pass: aprefetch
athlon64-sse3
during GIMPLE pass: aprefetch
atom
barcelona
during GIMPLE pass: aprefetch
bdver1
during GIMPLE pass: aprefetch
bdver2
during GIMPLE pass: aprefetch
bdver3
during GIMPLE pass: aprefetch
bdver4
during GIMPLE pass: aprefetch
bonnell
broadwell
btver1
during GIMPLE pass: aprefetch
btver2
during GIMPLE pass: aprefetch
cannonlake
cascadelake
cooperlake
core-avx-i
core-avx2
core2
corei7
corei7-avx
eden-x2
during GIMPLE pass: aprefetch
eden-x4
during GIMPLE pass: aprefetch
goldmont
goldmont-plus
haswell
celake-client
icelake-server
ivybridge
k8
during GIMPLE pass: aprefetch
k8-sse3
during GIMPLE pass: aprefetch
knl
knm
nano
during GIMPLE pass: aprefetch
nano-1000
during GIMPLE pass: aprefetch
nano-2000
during GIMPLE pass: aprefetch
nano-3000
during GIMPLE pass: aprefetch
nano-x2
during GIMPLE pass: aprefetch
nano-x4
during GIMPLE pass: aprefetch
native
during GIMPLE pass: aprefetch
nehalem
nocona
opteron
during GIMPLE pass: aprefetch
opteron-sse3
during GIMPLE pass: aprefetch
rocketlake
sandybridge
sapphirerapids
silvermont
skylake
skylake-avx512
slm
tigerlake
tremont
westmere
x86-64
x86-64-v2
x86-64-v3
x86-64-v4
znver1
znver2
znver3

I generated this list with this command:

$ for i in `cat ~/arch.list`; do echo $i; /home/dcb/gcc/results/bin/gcc -c -O3
-march=$i bug760.c 2>&1 | fgrep GIMPLE ; done

[Bug tree-optimization/102087] [12 Regression] ICE on valid code at -O3 on x86_64-linux-gnu: in determine_exit_conditions, at tree-ssa-loop-manip.c:1049 since r12-3136-g3673dcf6d6baeb67

[guojiufu at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=102087

--- Comment #16 from Jiu Fu Guo <guojiufu at gcc dot gnu.org> ---
Thanks David, Richard,

~/gcc/install/gcc-mainline-base-debug/bin/gcc -v
Using built-in specs.
COLLECT_GCC=/home/guojiufu/gcc/install/gcc-mainline-base-debug/bin/gcc
COLLECT_LTO_WRAPPER=/home/guojiufu/gcc/install/gcc-mainline-base-debug/libexec/gcc/x86_64-pc-linux-gnu/12.0.0/lto-wrapper
Target: x86_64-pc-linux-gnu
Configured with: /home/guojiufu/gcc/gcc-mainline-base/configure
--prefix=/home/guojiufu/gcc/install/gcc-mainline-base-debug --disable-bootstrap
--disable-multilib --disable-werror --with-pkgversion=29c92857039d0a10
--enable-checking=df,extra,fold,rtl,yes --enable-languages=c,c++,fortran :
(reconfigured) /home/guojiufu/gcc/gcc-mainline-base/configure
--prefix=/home/guojiufu/gcc/install/gcc-mainline-base-debug --disable-bootstrap
--disable-multilib --disable-werror --with-pkgversion=29c92857039d0a10
--enable-checking=df,extra,fold,rtl,yes --enable-languages=c,c++,fortran,lto
--no-create --no-recursion
Thread model: posix
Supported LTO compression algorithms: zlib
gcc version 12.0.0 20210922 (experimental) (29c92857039d0a10) 

~/gcc/install/gcc-mainline-base-debug/bin/gcc -c -O3 -w  ~/temp/t.c
-march=opteron
~/gcc/install/gcc-mainline-base-debug/bin/gcc -c -O3 -w -march=bdver2
~/temp/t.c

cat ~/temp/t.c
char **Gif_ClipImage_gfi_0;
int Gif_ClipImage_y, Gif_ClipImage_shift;
void Gif_ClipImage(void) {
  for (; Gif_ClipImage_y >= Gif_ClipImage_shift; Gif_ClipImage_y++)
    Gif_ClipImage_gfi_0[Gif_ClipImage_shift] =
        Gif_ClipImage_gfi_0[Gif_ClipImage_y];
}

I build both 3087d1b0a2c and the latest trunk, the case could pass.

[Bug tree-optimization/102087] [12 Regression] ICE on valid code at -O3 on x86_64-linux-gnu: in determine_exit_conditions, at tree-ssa-loop-manip.c:1049 since r12-3136-g3673dcf6d6baeb67

[asolokha at gmx dot com]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=102087

--- Comment #17 from Arseny Solokha <asolokha at gmx dot com> ---
(In reply to David Binderman from comment #13)
> The code in comment 8 still seems to fail:

It might make sense to file a new PR for that, I guess.

[Bug tree-optimization/102087] [12 Regression] ICE on valid code at -O3 on x86_64-linux-gnu: in determine_exit_conditions, at tree-ssa-loop-manip.c:1049 since r12-3136-g3673dcf6d6baeb67

[dcb314 at hotmail dot com]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=102087

--- Comment #18 from David Binderman <dcb314 at hotmail dot com> ---
(In reply to Arseny Solokha from comment #17)
> (In reply to David Binderman from comment #13)
> > The code in comment 8 still seems to fail:
> 
> It might make sense to file a new PR for that, I guess.

I checked, and the code in comment 8 works fine, since sometime
before 20211105.

[Bug tree-optimization/102087] [12 Regression] ICE on valid code at -O3 on x86_64-linux-gnu: in determine_exit_conditions, at tree-ssa-loop-manip.c:1049 since r12-3136-g3673dcf6d6baeb67

[guojiufu at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=102087

Jiu Fu Guo <guojiufu at gcc dot gnu.org> changed:

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

--- Comment #19 from Jiu Fu Guo <guojiufu at gcc dot gnu.org> ---
So, mark this as resolved.

[Bug tree-optimization/102087] [12 Regression] ICE on valid code at -O3 on x86_64-linux-gnu: in determine_exit_conditions, at tree-ssa-loop-manip.c:1049 since r12-3136-g3673dcf6d6baeb67

[dcb314 at hotmail dot com]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=102087

--- Comment #20 from David Binderman <dcb314 at hotmail dot com> ---
This code still seems broken:

**Gif_ClipImage_gfi_0;
Gif_ClipImage_gfi_1, Gif_ClipImage_y, Gif_ClipImage_shift;
Gif_ClipImage() {
  Gif_ClipImage_y = Gif_ClipImage_gfi_1 - 1;
  for (; Gif_ClipImage_y >= Gif_ClipImage_shift; Gif_ClipImage_y++)
    Gif_ClipImage_gfi_0[Gif_ClipImage_shift] =
        Gif_ClipImage_gfi_0[Gif_ClipImage_y];
}

bug760.c: In function ‘Gif_ClipImage’:
bug760.c:3:1: error: type mismatch in binary expression
    3 | Gif_ClipImage() {
      | ^~~~~~~~~~~~~
unsigned int

int

int

_18 = Gif_ClipImage_gfi_1.0_1 + -1;
bug760.c:3:1: error: type mismatch in binary expression
unsigned int

int

int

_12 = Gif_ClipImage_gfi_1.0_1 + -1;
during GIMPLE pass: aprefetch
bug760.c:3:1: internal compiler error: verify_gimple failed
0xde2f5a verify_gimple_in_cfg(function*, bool)

Flag -O3 -march=bdver2 required.

[Bug tree-optimization/102087] [12 Regression] ICE on valid code at -O3 on x86_64-linux-gnu: in determine_exit_conditions, at tree-ssa-loop-manip.c:1049 since r12-3136-g3673dcf6d6baeb67

[marxin at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=102087

Martin Liška <marxin at gcc dot gnu.org> changed:

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

--- Comment #21 from Martin Liška <marxin at gcc dot gnu.org> ---
> _12 = Gif_ClipImage_gfi_1.0_1 + -1;
> during GIMPLE pass: aprefetch
> bug760.c:3:1: internal compiler error: verify_gimple failed
> 0xde2f5a verify_gimple_in_cfg(function*, bool)
> 
> Flag -O3 -march=bdver2 required.

Confirmed.

[Bug tree-optimization/102087] [12 Regression] ICE on valid code at -O3 on x86_64-linux-gnu: in determine_exit_conditions, at tree-ssa-loop-manip.c:1049 since r12-3136-g3673dcf6d6baeb67

[guojiufu at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=102087

--- Comment #22 from Jiu Fu Guo <guojiufu at gcc dot gnu.org> ---
(In reply to Martin Liška from comment #21)
> > _12 = Gif_ClipImage_gfi_1.0_1 + -1;
> > during GIMPLE pass: aprefetch
> > bug760.c:3:1: internal compiler error: verify_gimple failed
> > 0xde2f5a verify_gimple_in_cfg(function*, bool)
> > 
> > Flag -O3 -march=bdver2 required.
> 
> Confirmed.

Hi David and Martin, Thanks.  I can reproduce the failure. 
Just drafting a patch and will send it out for review after regtest.

[Bug tree-optimization/102087] [12 Regression] ICE on valid code at -O3 on x86_64-linux-gnu: in determine_exit_conditions, at tree-ssa-loop-manip.c:1049 since r12-3136-g3673dcf6d6baeb67

[cvs-commit at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=102087

--- Comment #23 from CVS Commits <cvs-commit at gcc dot gnu.org> ---
The master branch has been updated by Jiu Fu Guo <guojiufu@gcc.gnu.org>:

https://gcc.gnu.org/g:634de54f9c421b7069865d0d7365ad97412f34bd

commit r12-6834-g634de54f9c421b7069865d0d7365ad97412f34bd
Author: Jiufu Guo <guojiufu@linux.ibm.com>
Date:   Fri Jan 21 17:03:50 2022 +0800

    Update the type of control.base after changed

    This patch correct the type of niter->control.base, when it is updated
    as a PLUS expr.
    During build PLUS expr, the result type should align with the type of
    the operands.

            PR tree-optimization/102087

    gcc/ChangeLog:

            * tree-ssa-loop-niter.cc (number_of_iterations_until_wrap):
            Correct PLUS result type.

    gcc/testsuite/ChangeLog:

            * gcc.dg/pr102087_1.c: New test.

[Bug tree-optimization/102087] [12 Regression] ICE on valid code at -O3 on x86_64-linux-gnu: in determine_exit_conditions, at tree-ssa-loop-manip.c:1049 since r12-3136-g3673dcf6d6baeb67

[rguenth at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=102087

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

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

--- Comment #24 from Richard Biener <rguenth at gcc dot gnu.org> ---
Assuming fixed again.