[Bug tree-optimization/114799] [14 Regression] ICE: tree check: expected class 'type', have 'exceptional' (error_mark) in useless_type_conversion_p, at gimple-expr.cc:85 with -O2 -fno-vect-cost-model since r14-9316

["jakub at gcc dot gnu.org" <gcc-bugzilla@gcc.gnu.org>]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=114799

--- Comment #5 from Jakub Jelinek <jakub at gcc dot gnu.org> ---
_10 is created during pattern recognition:
patt_10 = (unsigned short) y.0_1;

#0  vect_convert_input (vinfo=0x3a883e0, stmt_info=0x3be5410,
type=<integer_type 0x7fffea2db540>, unprom=0x7fffffffcc80, vectype=<vector_type
0x7fffea2db930>, subtype=optab_default)
    at ../../gcc/tree-vect-patterns.cc:926
#1  0x00000000026dcede in vect_convert_inputs (vinfo=0x3a883e0,
stmt_info=0x3be5410, n=2, result=0x7fffffffcc50, type=<integer_type
0x7fffea2db540>, unprom=0x7fffffffcc80, 
    vectype=<vector_type 0x7fffea2db930>, subtype=optab_default) at
../../gcc/tree-vect-patterns.cc:977
#2  0x00000000026e35f5 in vect_recog_over_widening_pattern (vinfo=0x3a883e0,
last_stmt_info=0x3be5410, type_out=0x7fffffffce48) at
../../gcc/tree-vect-patterns.cc:3155
#3  0x00000000026f250a in vect_pattern_recog_1 (vinfo=0x3a883e0,
recog_func=0x385ae30 <vect_vect_recog_func_ptrs+48>, stmt_info=0x3be5410) at
../../gcc/tree-vect-patterns.cc:7181
#4  0x00000000026f29fb in vect_pattern_recog (vinfo=0x3a883e0) at
../../gcc/tree-vect-patterns.cc:7338
#5  0x00000000011663ac in vect_slp_analyze_bb_1 (bb_vinfo=0x3a883e0, n_stmts=9,
fatal=@0x7fffffffd2cb: false, Python Exception <class 'gdb.error'>: There is no
member or method named m_vecpfx.
dataref_groups=0x7fffffffd760) at ../../gcc/tree-vect-slp.cc:7722
#6  0x00000000011668b6 in vect_slp_region (bbs=..., datarefs=..., Python
Exception <class 'gdb.error'>: There is no member or method named m_vecpfx.
dataref_groups=0x7fffffffd760, n_stmts=9, orig_loop=0x0) at
../../gcc/tree-vect-slp.cc:7831
#7  0x000000000116791f in vect_slp_bbs (bbs=..., orig_loop=0x0) at
../../gcc/tree-vect-slp.cc:8042

Unfortunately, I'm afraid I know next to nothing about SLP handling of
patterns, just loop vectorization :(