[Bug target/89290] [8/9 Regression] ICE in change_address_1, at emit-rtl.c:2286

jakub at gcc dot gnu.org gcc-bugzilla@gcc.gnu.org
Mon Feb 11 19:39:00 GMT 2019


https://gcc.gnu.org/bugzilla/show_bug.cgi?id=89290

Jakub Jelinek <jakub at gcc dot gnu.org> changed:

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

--- Comment #1 from Jakub Jelinek <jakub at gcc dot gnu.org> ---
The problem is that adjust_address with
(mem/c:TI (const:DI (unspec:DI [
                (symbol_ref:DI ("s") [flags 0x2a] <var_decl 0x7ffff7ffbb40 s>)
            ] UNSPEC_NTPOFF)) [1 s+0 S16 A64 AS1])
as first argument, E_DImode and 8 as last two doesn't validate.
We have originally:
(insn 7 6 10 2 (set (mem/c:TI (plus:DI (unspec:DI [
                        (const_int 0 [0])
                    ] UNSPEC_TP)
                (const:DI (unspec:DI [
                            (symbol_ref:DI ("s") [flags 0x2a]  <var_decl
0x7ff97a130b40 s>)
                        ] UNSPEC_NTPOFF))) [1 s+0 S16 A64])
        (reg:TI 83)) "pr89290.c":7:5 65 {*movti_internal}
     (nil))
then the ix86_rewrite_tls_address splitter rewrites this into:
(insn 12 6 10 2 (set (mem/c:TI (const:DI (unspec:DI [
                        (symbol_ref:DI ("s") [flags 0x2a]  <var_decl
0x7ff97a130b40 s>)
                    ] UNSPEC_NTPOFF)) [1 s+0 S16 A64 AS1])
        (reg:TI 83)) "pr89290.c":7:5 -1
     (nil))
and finally we ICE in the split_double_move splitter.


More information about the Gcc-bugs mailing list