Thanks Jeff. > Or just ignore the case of an insn that is a toplevel USE/CLOBBER? I > bet we even have a function for that somewhere given how commonly we > need to do that.
I found there are something like "GET_CODE(x) == USE || GET_CODE(x) == CLOBBER" everywhere but lack of a function or macro. Thus, I add new macro for this as below v2. https://gcc.gnu.org/pipermail/gcc-patches/2023-August/626590.html Pan -----Original Message----- From: Jeff Law <jeffreya...@gmail.com> Sent: Tuesday, August 8, 2023 12:31 AM To: Richard Biener <richard.guent...@gmail.com>; Li, Pan2 <pan2...@intel.com> Cc: gcc-patches@gcc.gnu.org; juzhe.zh...@rivai.ai; kito.ch...@sifive.com; Wang, Yanzhang <yanzhang.w...@intel.com> Subject: Re: [PATCH v1] Mode-Switching: Fix SET_SRC ICE when only one operand On 8/7/23 07:13, Richard Biener wrote: > On Mon, Aug 7, 2023 at 2:30 PM Pan Li via Gcc-patches > <gcc-patches@gcc.gnu.org> wrote: >> >> From: Pan Li <pan2...@intel.com> >> >> In same cases, like gcc/testsuite/gcc.dg/pr78148.c in RISC-V, there will >> be only 1 operand when SET_SRC in create_pre_exit. For example as below. >> >> (insn 13 9 14 2 (clobber (reg/i:TI 10 a0)) >> "gcc/testsuite/gcc.dg/pr78148.c":24:1 -1 >> (expr_list:REG_UNUSED (reg/i:TI 10 a0) >> (nil))) >> >> Unfortunately, SET_SRC requires at least 2 operands and then Segment >> Fault here. This patch would like to fix this ICE by adding operand >> length check before SET_SRC. >> >> Signed-off-by: Pan Li <pan2...@intel.com> >> >> gcc/ChangeLog: >> >> * mode-switching.cc (create_pre_exit): Add operand length check. >> >> gcc/testsuite/ChangeLog: >> >> * gcc.target/riscv/mode-switch-ice-1.c: New test. >> --- >> gcc/mode-switching.cc | 2 ++ >> .../gcc.target/riscv/mode-switch-ice-1.c | 22 +++++++++++++++++++ >> 2 files changed, 24 insertions(+) >> create mode 100644 gcc/testsuite/gcc.target/riscv/mode-switch-ice-1.c >> >> diff --git a/gcc/mode-switching.cc b/gcc/mode-switching.cc >> index 64ae2bc29c3..cbf03c02159 100644 >> --- a/gcc/mode-switching.cc >> +++ b/gcc/mode-switching.cc >> @@ -411,6 +411,8 @@ create_pre_exit (int n_entities, int *entity_map, const >> int *num_modes) >> conflict with address reloads. */ >> if (copy_start >= ret_start >> && copy_start + copy_num <= ret_end >> + && GET_RTX_LENGTH (GET_CODE (return_copy_pat)) >> >= 2 >> + /* SET_SRC requires at least 2 operands. */ > > maybe check that return_copy_pat is a SET instead? Or just ignore the case of an insn that is a toplevel USE/CLOBBER? I bet we even have a function for that somewhere given how commonly we need to do that. jeff
