| Message ID | mptle934ue9.fsf@arm.com |
|---|---|
| State | New |
| Headers | show |
| Series | Keep track of the FUNCTION_BEG note | expand |
On 1/5/24 09:28, Richard Sandiford wrote: > function.cc emits a NOTE_FUNCTION_BEG after all arguments have > been copied to pseudos. It then records this note in parm_birth_insn. > Various other pieces of code use this insn as a convenient place to > insert things at the start of the function. > > However, cfgexpand later changes parm_birth_insn as follows: > > /* If we emitted any instructions for setting up the variables, > emit them before the FUNCTION_START note. */ > if (var_seq) > { > emit_insn_before (var_seq, parm_birth_insn); > > /* In expand_function_end we'll insert the alloca save/restore > before parm_birth_insn. We've just insertted an alloca call. > Adjust the pointer to match. */ > parm_birth_insn = var_seq; > } > > But the FUNCTION_BEG note is still useful for things that aren't > sensitive to stack allocation, and it has the advantage that > (unlike the var_seq above) it is never deleted or combined. > This patch adds a separate variable to track it. > > Tested on aarch64-linux-gnu, where it's needed for fixing PR113196. > OK to install? > > Richard > > > gcc/ > * emit-rtl.h (rtl_data::x_function_beg_note): New member variable. > (function_beg_insn): New macro. > * function.cc (expand_function_start): Initialize function_beg_insn. OK jeff
diff --git a/gcc/emit-rtl.h b/gcc/emit-rtl.h index f749ca9f2a0..34f44cb2990 100644 --- a/gcc/emit-rtl.h +++ b/gcc/emit-rtl.h @@ -141,6 +141,9 @@ struct GTY(()) rtl_data { If stack grows up, this is the address for the next slot. */ poly_int64 x_frame_offset; + /* The function's FUNCTION_BEG note. */ + rtx_insn *x_function_beg_insn; + /* Insn after which register parms and SAVE_EXPRs are born, if nonopt. */ rtx_insn *x_parm_birth_insn; @@ -323,6 +326,7 @@ struct GTY(()) rtl_data { #define return_label (crtl->x_return_label) #define naked_return_label (crtl->x_naked_return_label) #define stack_slot_list (crtl->x_stack_slot_list) +#define function_beg_insn (crtl->x_function_beg_insn) #define parm_birth_insn (crtl->x_parm_birth_insn) #define frame_offset (crtl->x_frame_offset) #define stack_check_probe_note (crtl->x_stack_check_probe_note) diff --git a/gcc/function.cc b/gcc/function.cc index de356f7fba3..5ffd438475e 100644 --- a/gcc/function.cc +++ b/gcc/function.cc @@ -5202,7 +5202,7 @@ expand_function_start (tree subr) gcc_assert (NOTE_P (get_last_insn ())); - parm_birth_insn = get_last_insn (); + function_beg_insn = parm_birth_insn = get_last_insn (); /* If the function receives a non-local goto, then store the bits we need to restore the frame pointer. */
function.cc emits a NOTE_FUNCTION_BEG after all arguments have been copied to pseudos. It then records this note in parm_birth_insn. Various other pieces of code use this insn as a convenient place to insert things at the start of the function. However, cfgexpand later changes parm_birth_insn as follows: /* If we emitted any instructions for setting up the variables, emit them before the FUNCTION_START note. */ if (var_seq) { emit_insn_before (var_seq, parm_birth_insn); /* In expand_function_end we'll insert the alloca save/restore before parm_birth_insn. We've just insertted an alloca call. Adjust the pointer to match. */ parm_birth_insn = var_seq; } But the FUNCTION_BEG note is still useful for things that aren't sensitive to stack allocation, and it has the advantage that (unlike the var_seq above) it is never deleted or combined. This patch adds a separate variable to track it. Tested on aarch64-linux-gnu, where it's needed for fixing PR113196. OK to install? Richard gcc/ * emit-rtl.h (rtl_data::x_function_beg_note): New member variable. (function_beg_insn): New macro. * function.cc (expand_function_start): Initialize function_beg_insn. --- gcc/emit-rtl.h | 4 ++++ gcc/function.cc | 2 +- 2 files changed, 5 insertions(+), 1 deletion(-)