[Bug tree-optimization/126862] New: [17 Regression] ICE: verify_ssa failed for PHI node during GIMPLE pass: slp with -O3 -ftree-slp-vectorize -fvect-cost-model=unlimited

[bouncy12578 at gmail dot com via Gcc-bugs]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=126862

            Bug ID: 126862
           Summary: [17 Regression] ICE: verify_ssa failed for PHI node
                    during GIMPLE pass: slp with -O3 -ftree-slp-vectorize
                    -fvect-cost-model=unlimited
           Product: gcc
           Version: 17.0
            Status: UNCONFIRMED
          Severity: normal
          Priority: P3
         Component: tree-optimization
          Assignee: unassigned at gcc dot gnu.org
          Reporter: bouncy12578 at gmail dot com
  Target Milestone: ---

Created attachment 65332
  --> https://gcc.gnu.org/bugzilla/attachment.cgi?id=65332&action=edit
preprocessed crash.c

Compiler Explorer: https://godbolt.org/z/h6rxThGY3

When compiling a CSmith generated file with GCC 17.0.0 (x86_64-pc-linux-gnu), I
hit an Internal Compiler Error (verify_ssa failed) during the SLP vectorization
pass under unlimited model.


Compiler version: 

Using built-in specs.
COLLECT_GCC=/data/lenovo/compiler-build/gcc/gcc-install/bin/gcc
COLLECT_LTO_WRAPPER=/data/lenovo/compiler-build/gcc/gcc-install/libexec/gcc/x86_64-pc-linux-gnu/17.0.0/lto-wrapper
Target: x86_64-pc-linux-gnu
Configured with: ../gcc/configure
--prefix=/data/lenovo/compiler-build/gcc/gcc-install --disable-bootstrap
--enable-languages=c,c++ --disable-multilib --with-isl --enable-checking=all :
(reconfigured) ../gcc/configure
--prefix=/data/lenovo/compiler-build/gcc/gcc-install --disable-bootstrap
--enable-languages=c,c++ --disable-multilib --with-isl --enable-checking=all
Thread model: posix
Supported LTO compression algorithms: zlib
gcc version 17.0.0 20260812 (experimental) (GCC)

Stack trace: 

gcc crash.c -O3 -ftree-slp-vectorize -fvect-cost-model=unlimited -freport-bug

crash.c: In function ‘f’:
crash.c:14:10: error: definition in block 25 does not dominate use in block 24
   14 | uint32_t f() {
      |          ^
for SSA_NAME: _38 in statement:
d_lsm.5_52 = PHI <0(21), _38(24)>
PHI argument
_38
for PHI node
d_lsm.5_52 = PHI <0(21), _38(24)>
during GIMPLE pass: slp
crash.c:14:10: internal compiler error: verify_ssa failed
0x2f1e2eb internal_error(char const*, ...)
        ../../gcc/gcc/diagnostic-global-context.cc:787
0x196ee8f verify_ssa(bool, bool)
        ../../gcc/gcc/tree-ssa.cc:1226
0x1565cc4 execute_function_todo
        ../../gcc/gcc/passes.cc:2107
0x156641e execute_todo
        ../../gcc/gcc/passes.cc:2152
/data/lenovo/compiler-build/gcc/gcc-install/libexec/gcc/x86_64-pc-linux-gnu/17.0.0/cc1
-quiet -imultiarch x86_64-linux-gnu crash.c -quiet -dumpdir a- -dumpbase
crash.c -dumpbase-ext .c -mtune=generic -march=x86-64 -O3 -ftree-slp-vectorize
-fvect-cost-model=unlimited -freport-bug -o /tmp/cc7nMfe3.s
Please submit a full bug report, with preprocessed source.
Please include the complete backtrace with any bug report.
See <https://gcc.gnu.org/bugs/> for instructions.
Preprocessed source stored into /tmp/cc7qnOks.out file, please attach this to
your bugreport.

crash.c

#include <stdint.h>
static uint16_t
(safe_mod_func_uint16_t_u_u)(uint16_t ui1, uint16_t ui2 ){
  return
    (ui2 == 0) ?
    ((ui1)) :
    (ui1 % ui2);
}
struct a {};
int32_t b[4][5];
int16_t c[5];
uint16_t d;
int32_t *e(struct a, uint16_t, uint32_t, int32_t *, int64_t);
uint32_t f() {
  int32_t g[4][5];
  for (b[3][3] = 0; b[3][3] <= 3; b[3][3]++) {
    struct a h;
    e(h, 0, g[2][4], &b[3][3], b[9][4]);
  }
}
int32_t *e(struct a, uint16_t, uint32_t, int32_t *k, int64_t) {
  int32_t i[4];
  uint16_t *j = &d;
  if (c[4] ^= safe_mod_func_uint16_t_u_u(++*j, 0), *i)
    *j = 0;
  return k;
}