> -----Original Message-----
> From: Richard Biener <richard.guent...@gmail.com>
> Sent: Tuesday, May 13, 2025 7:49 PM
> To: Uros Bizjak <ubiz...@gmail.com>
> Cc: Cui, Lili <lili....@intel.com>; gcc-patches@gcc.gnu.org; Liu, Hongtao
> <hongtao....@intel.com>
> Subject: Re: [PATCH] x86: Enable separate shrink wrapping
>
> On Tue, May 13, 2025 at 12:36 PM Uros Bizjak <ubiz...@gmail.com> wrote:
> >
> > On Tue, May 13, 2025 at 8:15 AM Cui, Lili <lili....@intel.com> wrote:
> > >
> > > From: Lili Cui <lili....@intel.com>
> > >
> > > Hi,
> > >
> > > This patch is to enale separate shrink wrapping for x86.
> > >
> > > Bootstrapped & regtested on x86-64-pc-linux-gnu.
> > >
> > > Ok for trunk?
> >
> > Unfortunately, the patched compiler fails to boot the latest linux kernel.
>
> Michael Matz also posted x86 separate shrink wrapping here:
> https://gcc.gnu.org/pipermail/gcc-patches/2024-July/657519.html
>
Thanks Richard. When I analyzed the 511.povray_r regressions, I found that some
hot functions in 511 often return early, and using separate shrink wrapping can
reduce the push and pop instructions before return. So, I created this patch.
But I didn't notice that Michael had already posted it. I will test and compare
the two patches, hope to find the optimal solution.
Lili.
> Richard.
>
> > Uros.
> >
> >
> >
> > Uros.
> > >
> > >
> > > This commit implements the target macros (TARGET_SHRINK_WRAP_*) that
> > > enable separate shrink wrapping for function prologues/epilogues in
> > > x86.
> > >
> > > When performing separate shrink wrapping, we choose to use mov
> > > instead of push/pop, because using push/pop is more complicated to
> > > handle rsp adjustment and may lose performance, so here we choose to
> > > use mov, which has a small impact on code size, but guarantees
> performance.
> > >
> > > Tested against SPEC CPU 2017, this change always has a net-positive
> > > effect on the dynamic instruction count. See the following table
> > > for the breakdown on how this reduces the number of dynamic
> > > instructions per workload on a like-for-like (with/without this commit):
> > >
> > > instruction count base with commit (commit-base)/commit
> > > 502.gcc_r 98666845943 96891561634 -1.80%
> > > 526.blender_r 6.21226E+11 6.12992E+11 -1.33%
> > > 520.omnetpp_r 1.1241E+11 1.11093E+11 -1.17%
> > > 500.perlbench_r 1271558717 1263268350 -0.65%
> > > 523.xalancbmk_r 2.20103E+11 2.18836E+11 -0.58%
> > > 531.deepsjeng_r 2.73591E+11 2.72114E+11 -0.54%
> > > 500.perlbench_r 64195557393 63881512409 -0.49%
> > > 541.leela_r 2.99097E+11 2.98245E+11 -0.29%
> > > 548.exchange2_r 1.27976E+11 1.27784E+11 -0.15%
> > > 527.cam4_r 88981458425 88887334679 -0.11%
> > > 554.roms_r 2.60072E+11 2.59809E+11 -0.10%
> > >
> > > gcc/ChangeLog:
> > >
> > > * config/i386/i386-protos.h (ix86_get_separate_components):
> > > New function.
> > > (ix86_components_for_bb): Likewise.
> > > (ix86_disqualify_components): Likewise.
> > > (ix86_emit_prologue_components): Likewise.
> > > (ix86_emit_epilogue_components): Likewise.
> > > (ix86_set_handled_components): Likewise.
> > > * config/i386/i386.cc (save_regs_using_push_pop):
> > > Encapsulate code.
> > > (ix86_compute_frame_layout):
> > > Handle save_regs_using_push_pop.
> > > (ix86_emit_save_regs_using_mov):
> > > Skip registers that are wrapped separately.
> > > (ix86_expand_prologue): Likewise.
> > > (ix86_emit_restore_regs_using_mov): Likewise.
> > > (ix86_expand_epilogue): Likewise.
> > > (ix86_get_separate_components): New function.
> > > (ix86_components_for_bb): Likewise.
> > > (ix86_disqualify_components): Likewise.
> > > (ix86_emit_prologue_components): Likewise.
> > > (ix86_emit_epilogue_components): Likewise.
> > > (ix86_set_handled_components): Likewise.
> > > (TARGET_SHRINK_WRAP_GET_SEPARATE_COMPONENTS): Define.
> > > (TARGET_SHRINK_WRAP_COMPONENTS_FOR_BB): Likewise.
> > > (TARGET_SHRINK_WRAP_DISQUALIFY_COMPONENTS): Likewise.
> > > (TARGET_SHRINK_WRAP_EMIT_PROLOGUE_COMPONENTS): Likewise.
> > > (TARGET_SHRINK_WRAP_EMIT_EPILOGUE_COMPONENTS): Likewise.
> > > (TARGET_SHRINK_WRAP_SET_HANDLED_COMPONENTS): Likewise.
> > > * config/i386/i386.h (struct machine_function):Add
> > > reg_is_wrapped_separately array for register wrapping
> > > information.
> > >
> > > gcc/testsuite/ChangeLog:
> > >
> > > * gcc.target/x86_64/abi/callabi/leaf-2.c: Adjust the test.
> > > * gcc.target/i386/interrupt-16.c: Likewise.
> > > * g++.target/i386/shrink_wrap_separate.c: New test.
> > > ---
> > > gcc/config/i386/i386-protos.h | 7 +
> > > gcc/config/i386/i386.cc | 261 +++++++++++++++---
> > > gcc/config/i386/i386.h | 1 +
> > > .../g++.target/i386/shrink_wrap_separate.c | 24 ++
> > > gcc/testsuite/gcc.target/i386/interrupt-16.c | 4 +-
> > > .../gcc.target/x86_64/abi/callabi/leaf-2.c | 2 +-
> > > 6 files changed, 257 insertions(+), 42 deletions(-) create mode
> > > 100644 gcc/testsuite/g++.target/i386/shrink_wrap_separate.c
> > >
> > > diff --git a/gcc/config/i386/i386-protos.h
> > > b/gcc/config/i386/i386-protos.h index e85b925704b..11d26e93973
> > > 100644
> > > --- a/gcc/config/i386/i386-protos.h
> > > +++ b/gcc/config/i386/i386-protos.h
> > > @@ -436,6 +436,13 @@ extern rtl_opt_pass
> > > *make_pass_align_tight_loops (gcc::context *); extern bool
> > > ix86_has_no_direct_extern_access; extern bool ix86_rpad_gate ();
> > >
> > > +extern sbitmap ix86_get_separate_components (void); extern sbitmap
> > > +ix86_components_for_bb (basic_block); extern void
> > > +ix86_disqualify_components (sbitmap, edge, sbitmap, bool); extern
> > > +void ix86_emit_prologue_components (sbitmap); extern void
> > > +ix86_emit_epilogue_components (sbitmap); extern void
> > > +ix86_set_handled_components (sbitmap);
> > > +
> > > /* In i386-expand.cc. */
> > > bool ix86_check_builtin_isa_match (unsigned int, HOST_WIDE_INT*,
> > > HOST_WIDE_INT*); diff --git
> > > a/gcc/config/i386/i386.cc b/gcc/config/i386/i386.cc index
> > > 3d629b06094..2e8485ddb8b 100644
> > > --- a/gcc/config/i386/i386.cc
> > > +++ b/gcc/config/i386/i386.cc
> > > @@ -6909,6 +6909,26 @@ ix86_pro_and_epilogue_can_use_push2pop2
> (int nregs)
> > > && (nregs + aligned) >= 3;
> > > }
> > >
> > > +/* Check if push pop should be used to save registers. */ static
> > > +bool save_regs_using_push_pop (HOST_WIDE_INT to_allocate) {
> > > + return ((!to_allocate && cfun->machine->frame.nregs <= 1)
> > > + || (TARGET_64BIT && to_allocate >= HOST_WIDE_INT_C
> (0x80000000))
> > > + /* If static stack checking is enabled and done with probes,
> > > + the registers need to be saved before allocating the frame.
> > > */
> > > + || flag_stack_check == STATIC_BUILTIN_STACK_CHECK
> > > + /* If stack clash probing needs a loop, then it needs a
> > > + scratch register. But the returned register is only
> > > guaranteed
> > > + to be safe to use after register saves are complete. So if
> > > + stack clash protections are enabled and the allocated frame
> > > is
> > > + larger than the probe interval, then use pushes to save
> > > + callee saved registers. */
> > > + || (flag_stack_clash_protection
> > > + && !ix86_target_stack_probe ()
> > > + && to_allocate > get_probe_interval ())); }
> > > +
> > > /* Fill structure ix86_frame about frame of currently computed
> > > function. */
> > >
> > > static void
> > > @@ -7193,20 +7213,7 @@ ix86_compute_frame_layout (void)
> > > /* Size prologue needs to allocate. */
> > > to_allocate = offset - frame->sse_reg_save_offset;
> > >
> > > - if ((!to_allocate && frame->nregs <= 1)
> > > - || (TARGET_64BIT && to_allocate >= HOST_WIDE_INT_C (0x80000000))
> > > - /* If static stack checking is enabled and done with probes,
> > > - the registers need to be saved before allocating the frame. */
> > > - || flag_stack_check == STATIC_BUILTIN_STACK_CHECK
> > > - /* If stack clash probing needs a loop, then it needs a
> > > - scratch register. But the returned register is only guaranteed
> > > - to be safe to use after register saves are complete. So if
> > > - stack clash protections are enabled and the allocated frame is
> > > - larger than the probe interval, then use pushes to save
> > > - callee saved registers. */
> > > - || (flag_stack_clash_protection
> > > - && !ix86_target_stack_probe ()
> > > - && to_allocate > get_probe_interval ()))
> > > + if (save_regs_using_push_pop (to_allocate))
> > > frame->save_regs_using_mov = false;
> > >
> > > if (ix86_using_red_zone ()
> > > @@ -7664,7 +7671,9 @@ ix86_emit_save_regs_using_mov
> (HOST_WIDE_INT cfa_offset)
> > > for (regno = 0; regno < FIRST_PSEUDO_REGISTER; regno++)
> > > if (GENERAL_REGNO_P (regno) && ix86_save_reg (regno, true, true))
> > > {
> > > - ix86_emit_save_reg_using_mov (word_mode, regno, cfa_offset);
> > > + /* Skip shrink warp separate already processed registers. */
> > > + if (!cfun->machine->reg_is_wrapped_separately[regno])
> > > + ix86_emit_save_reg_using_mov (word_mode, regno,
> > > + cfa_offset);
> > > cfa_offset -= UNITS_PER_WORD;
> > > }
> > > }
> > > @@ -9227,6 +9236,18 @@ ix86_expand_prologue (void)
> > > && (! TARGET_STACK_PROBE
> > > || frame.stack_pointer_offset < CHECK_STACK_LIMIT))
> > > {
> > > + HOST_WIDE_INT allocate_offset;
> > > + /* If shrink wrap separate works, we will adjust the total
> > > offset of
> > > + rsp at the beginning. */
> > > + if (crtl->shrink_wrapped_separate)
> > > + {
> > > + allocate_offset = m->fs.sp_offset -
> > > frame.stack_pointer_offset;
> > > + pro_epilogue_adjust_stack (stack_pointer_rtx,
> > > stack_pointer_rtx,
> > > + GEN_INT (allocate_offset), -1,
> > > + m->fs.cfa_reg ==
> > > stack_pointer_rtx);
> > > + m->fs.sp_offset = cfun->machine->frame.stack_pointer_offset;
> > > + }
> > > +
> > > ix86_emit_save_regs_using_mov (frame.reg_save_offset);
> > > cfun->machine->red_zone_used = true;
> > > int_registers_saved = true; @@ -9806,30 +9827,36 @@
> > > ix86_emit_restore_regs_using_mov (HOST_WIDE_INT cfa_offset,
> > > for (regno = 0; regno < FIRST_PSEUDO_REGISTER; regno++)
> > > if (GENERAL_REGNO_P (regno) && ix86_save_reg (regno,
> maybe_eh_return, true))
> > > {
> > > - rtx reg = gen_rtx_REG (word_mode, regno);
> > > - rtx mem;
> > > - rtx_insn *insn;
> > > -
> > > - mem = choose_baseaddr (cfa_offset, NULL);
> > > - mem = gen_frame_mem (word_mode, mem);
> > > - insn = emit_move_insn (reg, mem);
> > >
> > > - if (m->fs.cfa_reg == crtl->drap_reg && regno == REGNO (crtl-
> >drap_reg))
> > > + /* Skip shrink warp separate already processed registers. */
> > > + if (!cfun->machine->reg_is_wrapped_separately[regno])
> > > {
> > > - /* Previously we'd represented the CFA as an expression
> > > - like *(%ebp - 8). We've just popped that value from
> > > - the stack, which means we need to reset the CFA to
> > > - the drap register. This will remain until we restore
> > > - the stack pointer. */
> > > - add_reg_note (insn, REG_CFA_DEF_CFA, reg);
> > > - RTX_FRAME_RELATED_P (insn) = 1;
> > > + rtx reg = gen_rtx_REG (word_mode, regno);
> > > + rtx mem;
> > > + rtx_insn *insn;
> > >
> > > - /* This means that the DRAP register is valid for addressing.
> > > */
> > > - m->fs.drap_valid = true;
> > > - }
> > > - else
> > > - ix86_add_cfa_restore_note (NULL, reg, cfa_offset);
> > > + mem = choose_baseaddr (cfa_offset, NULL);
> > > + mem = gen_frame_mem (word_mode, mem);
> > > + insn = emit_move_insn (reg, mem);
> > >
> > > + if (m->fs.cfa_reg == crtl->drap_reg
> > > + && regno == REGNO (crtl->drap_reg))
> > > + {
> > > + /* Previously we'd represented the CFA as an expression
> > > + like *(%ebp - 8). We've just popped that value from
> > > + the stack, which means we need to reset the CFA to
> > > + the drap register. This will remain until we restore
> > > + the stack pointer. */
> > > + add_reg_note (insn, REG_CFA_DEF_CFA, reg);
> > > + RTX_FRAME_RELATED_P (insn) = 1;
> > > +
> > > + /* This means that the DRAP register is valid for
> > > addressing.
> > > + */
> > > + m->fs.drap_valid = true;
> > > + }
> > > + else
> > > + ix86_add_cfa_restore_note (NULL, reg, cfa_offset);
> > > + }
> > > cfa_offset -= UNITS_PER_WORD;
> > > }
> > > }
> > > @@ -10108,10 +10135,11 @@ ix86_expand_epilogue (int style)
> > > less work than reloading sp and popping the register. */
> > > else if (!sp_valid_at (frame.hfp_save_offset) && frame.nregs <= 1)
> > > restore_regs_via_mov = true;
> > > - else if (TARGET_EPILOGUE_USING_MOVE
> > > - && cfun->machine->use_fast_prologue_epilogue
> > > - && (frame.nregs > 1
> > > - || m->fs.sp_offset != reg_save_offset))
> > > + else if (crtl->shrink_wrapped_separate
> > > + || (TARGET_EPILOGUE_USING_MOVE
> > > + && cfun->machine->use_fast_prologue_epilogue
> > > + && (frame.nregs > 1
> > > + || m->fs.sp_offset != reg_save_offset)))
> > > restore_regs_via_mov = true;
> > > else if (frame_pointer_needed
> > > && !frame.nregs
> > > @@ -28065,6 +28093,161 @@ ix86_cannot_copy_insn_p (rtx_insn *insn)
> > > #undef TARGET_DOCUMENTATION_NAME #define
> TARGET_DOCUMENTATION_NAME
> > > "x86"
> > >
> > > +/* Implement TARGET_SHRINK_WRAP_GET_SEPARATE_COMPONENTS. */
> > > +sbitmap ix86_get_separate_components (void) {
> > > + HOST_WIDE_INT offset, to_allocate;
> > > + sbitmap components = sbitmap_alloc (FIRST_PSEUDO_REGISTER);
> > > + bitmap_clear (components);
> > > +
> > > + offset = cfun->machine->frame.stack_pointer_offset;
> > > + to_allocate = offset - cfun->machine->frame.sse_reg_save_offset;
> > > + /* When PPX is enabled, disable shrink wrap separate.
> > > + It is a trade-off. */
> > > + if (TARGET_APX_PPX && !crtl->calls_eh_return)
> > > + return components;
> > > +
> > > + /* Since shrink wrapping separate uses MOV instead of PUSH/POP
> > > + We need to disable shrink wrap separate when move is
> > > + prohibited. */ if (save_regs_using_push_pop (to_allocate))
> > > + return components;
> > > +
> > > + for (unsigned int regno = 0; regno < FIRST_PSEUDO_REGISTER; regno++)
> > > + if (GENERAL_REGNO_P (regno) && ix86_save_reg (regno, true, true))
> > > + {
> > > + /* We can only wrap registers that have small operand offsets.
> > > + For large offsets a pseudo register might be needed which
> > > + cannot be created during the shrink wrapping pass. */
> > > + if (IN_RANGE (offset, -0x8000, 0x7fff))
> > > + bitmap_set_bit (components, regno);
> > > + offset += UNITS_PER_WORD;
> > > + }
> > > +
> > > + /* Don't mess with the following frame pointer. */ if
> > > + (frame_pointer_needed)
> > > + bitmap_clear_bit (components, HARD_FRAME_POINTER_REGNUM);
> > > +
> > > + if (crtl->drap_reg)
> > > + bitmap_clear_bit (components, REGNO (crtl->drap_reg));
> > > +
> > > + if (pic_offset_table_rtx)
> > > + bitmap_clear_bit (components, REAL_PIC_OFFSET_TABLE_REGNUM);
> > > +
> > > + return components;
> > > +}
> > > +
> > > +/* Implement TARGET_SHRINK_WRAP_COMPONENTS_FOR_BB. */
> sbitmap
> > > +ix86_components_for_bb (basic_block bb) {
> > > + bitmap in = DF_LIVE_IN (bb);
> > > + bitmap gen = &DF_LIVE_BB_INFO (bb)->gen;
> > > + bitmap kill = &DF_LIVE_BB_INFO (bb)->kill;
> > > +
> > > + sbitmap components = sbitmap_alloc (FIRST_PSEUDO_REGISTER);
> > > + bitmap_clear (components);
> > > +
> > > + function_abi_aggregator callee_abis; rtx_insn *insn;
> > > + FOR_BB_INSNS (bb, insn)
> > > + if (CALL_P (insn))
> > > + callee_abis.note_callee_abi (insn_callee_abi (insn));
> > > + HARD_REG_SET extra_caller_saves = callee_abis.caller_save_regs
> > > + (*crtl->abi);
> > > +
> > > + /* GPRs are used in a bb if they are in the IN, GEN, or KILL
> > > + sets. */ for (unsigned int regno = 0; regno < FIRST_PSEUDO_REGISTER;
> regno++)
> > > + if (!fixed_regs[regno]
> > > + && (TEST_HARD_REG_BIT (extra_caller_saves, regno)
> > > + || bitmap_bit_p (in, regno)
> > > + || bitmap_bit_p (gen, regno)
> > > + || bitmap_bit_p (kill, regno)))
> > > + bitmap_set_bit (components, regno);
> > > +
> > > + return components;
> > > +}
> > > +
> > > +/* Implement TARGET_SHRINK_WRAP_DISQUALIFY_COMPONENTS. */
> void
> > > +ix86_disqualify_components (sbitmap, edge, sbitmap, bool) {
> > > + /* Nothing to do for i386. */
> > > +}
> > > +
> > > +/* Implement TARGET_SHRINK_WRAP_EMIT_PROLOGUE_COMPONENTS.
> */ void
> > > +ix86_emit_prologue_components (sbitmap components) {
> > > + HOST_WIDE_INT cfa_offset;
> > > + struct machine_function *m = cfun->machine;
> > > +
> > > + cfa_offset = m->frame.reg_save_offset + m->fs.sp_offset
> > > + - m->frame.stack_pointer_offset;
> > > + for (unsigned int regno = 0; regno < FIRST_PSEUDO_REGISTER; regno++)
> > > + if (GENERAL_REGNO_P (regno) && ix86_save_reg (regno, true, true))
> > > + {
> > > + if (bitmap_bit_p (components, regno))
> > > + ix86_emit_save_reg_using_mov (word_mode, regno, cfa_offset);
> > > + cfa_offset -= UNITS_PER_WORD;
> > > + }
> > > +}
> > > +
> > > +/* Implement TARGET_SHRINK_WRAP_EMIT_EPILOGUE_COMPONENTS.
> */ void
> > > +ix86_emit_epilogue_components (sbitmap components) {
> > > + HOST_WIDE_INT cfa_offset;
> > > + struct machine_function *m = cfun->machine;
> > > + cfa_offset = m->frame.reg_save_offset + m->fs.sp_offset
> > > + - m->frame.stack_pointer_offset;
> > > +
> > > + for (unsigned int regno = 0; regno < FIRST_PSEUDO_REGISTER; regno++)
> > > + if (GENERAL_REGNO_P (regno) && ix86_save_reg (regno, true, true))
> > > + {
> > > + if (bitmap_bit_p (components, regno))
> > > + {
> > > + rtx reg = gen_rtx_REG (word_mode, regno);
> > > + rtx mem;
> > > + rtx_insn *insn;
> > > +
> > > + mem = choose_baseaddr (cfa_offset, NULL);
> > > + mem = gen_frame_mem (word_mode, mem);
> > > + insn = emit_move_insn (reg, mem);
> > > +
> > > + RTX_FRAME_RELATED_P (insn) = 1;
> > > + add_reg_note (insn, REG_CFA_RESTORE, reg);
> > > + }
> > > + cfa_offset -= UNITS_PER_WORD;
> > > + }
> > > +}
> > > +
> > > +void
> > > +ix86_set_handled_components (sbitmap components) {
> > > + for (unsigned int regno = 0; regno < FIRST_PSEUDO_REGISTER; regno++)
> > > + if (bitmap_bit_p (components, regno))
> > > + {
> > > + cfun->machine->reg_is_wrapped_separately[regno] = true;
> > > + cfun->machine->use_fast_prologue_epilogue = true;
> > > + cfun->machine->frame.save_regs_using_mov = true;
> > > + }
> > > +}
> > > +
> > > +#undef TARGET_SHRINK_WRAP_GET_SEPARATE_COMPONENTS
> > > +#define TARGET_SHRINK_WRAP_GET_SEPARATE_COMPONENTS
> > > +ix86_get_separate_components #undef
> > > +TARGET_SHRINK_WRAP_COMPONENTS_FOR_BB
> > > +#define TARGET_SHRINK_WRAP_COMPONENTS_FOR_BB
> ix86_components_for_bb
> > > +#undef TARGET_SHRINK_WRAP_DISQUALIFY_COMPONENTS
> > > +#define TARGET_SHRINK_WRAP_DISQUALIFY_COMPONENTS
> > > +ix86_disqualify_components #undef
> > > +TARGET_SHRINK_WRAP_EMIT_PROLOGUE_COMPONENTS
> > > +#define TARGET_SHRINK_WRAP_EMIT_PROLOGUE_COMPONENTS \
> > > + ix86_emit_prologue_components
> > > +#undef TARGET_SHRINK_WRAP_EMIT_EPILOGUE_COMPONENTS
> > > +#define TARGET_SHRINK_WRAP_EMIT_EPILOGUE_COMPONENTS \
> > > + ix86_emit_epilogue_components
> > > +#undef TARGET_SHRINK_WRAP_SET_HANDLED_COMPONENTS
> > > +#define TARGET_SHRINK_WRAP_SET_HANDLED_COMPONENTS
> > > +ix86_set_handled_components
> > > +
> > > struct gcc_target targetm = TARGET_INITIALIZER;
> > >
> > > #include "gt-i386.h"
> > > diff --git a/gcc/config/i386/i386.h b/gcc/config/i386/i386.h index
> > > 18fa97a9eb0..e89e57f8278 100644
> > > --- a/gcc/config/i386/i386.h
> > > +++ b/gcc/config/i386/i386.h
> > > @@ -2817,6 +2817,7 @@ struct GTY(()) machine_function {
> > > /* Cached initial frame layout for the current function. */
> > > struct ix86_frame frame;
> > >
> > > + bool reg_is_wrapped_separately[FIRST_PSEUDO_REGISTER];
> > > /* For -fsplit-stack support: A stack local which holds a pointer to
> > > the stack arguments for a function with a variable number of
> > > arguments. This is set at the start of the function and is
> > > used diff --git
> > > a/gcc/testsuite/g++.target/i386/shrink_wrap_separate.c
> > > b/gcc/testsuite/g++.target/i386/shrink_wrap_separate.c
> > > new file mode 100644
> > > index 00000000000..8be04822ac3
> > > --- /dev/null
> > > +++ b/gcc/testsuite/g++.target/i386/shrink_wrap_separate.c
> > > @@ -0,0 +1,24 @@
> > > +/* { dg-do compile } */
> > > +/* { dg-options "-O2 -fdump-rtl-pro_and_epilogue" } */ typedef
> > > +struct a b; typedef double c; struct a {
> > > + b *d;
> > > + b *e;
> > > +};
> > > +struct f {
> > > + c g;
> > > +};
> > > +inline bool h(c i, b *m) {
> > > + b *j = m->e;
> > > + for (; m->e; j = j->d)
> > > + if (h(i, j))
> > > + return 0;
> > > + return m;
> > > +}
> > > +bool k() {
> > > + f *l;
> > > + b *n;
> > > + h(l->g, n);
> > > +}
> > > +/* { dg-final { scan-rtl-dump "The components we wrap separately
> > > +are \[sep 40 41 42 43\]" "pro_and_epilogue" } } */
> > > diff --git a/gcc/testsuite/gcc.target/i386/interrupt-16.c
> > > b/gcc/testsuite/gcc.target/i386/interrupt-16.c
> > > index cb45ba54e3d..ca4441b3aee 100644
> > > --- a/gcc/testsuite/gcc.target/i386/interrupt-16.c
> > > +++ b/gcc/testsuite/gcc.target/i386/interrupt-16.c
> > > @@ -18,5 +18,5 @@ foo (int i)
> > > /* { dg-final { scan-assembler-not "(push|pop)(l|q)\[\\t
> > > \]*%(r|e)bp" } } */
> > > /* { dg-final { scan-assembler-not "(push|pop)l\[\\t \]*%edi" {
> > > target ia32 } } } */
> > > /* { dg-final { scan-assembler-not "(push|pop)q\[\\t \]*%r\[0-9\]+"
> > > { target { ! ia32 } } } } */
> > > -/* { dg-final { scan-assembler-times "pushq\[\\t \]*%rdi" 1 {
> > > target { ! ia32 } } } } */
> > > -/* { dg-final { scan-assembler-times "popq\[\\t \]*%rdi" 1 { target
> > > { ! ia32 } } } } */
> > > +/* { dg-final { scan-assembler-times
> > > +"(pushq.*%rdi|subq.*\\\$8,.*%rsp)" 1 { target { ! ia32 } } } } */
> > > +/* { dg-final { scan-assembler-times
> > > +"(popq.*%rdi|addq.*\\\$8,.*%rsp)" 1 { target { ! ia32 } } } } */
> > > diff --git a/gcc/testsuite/gcc.target/x86_64/abi/callabi/leaf-2.c
> > > b/gcc/testsuite/gcc.target/x86_64/abi/callabi/leaf-2.c
> > > index 5f3d3e166af..46fc4648dbd 100644
> > > --- a/gcc/testsuite/gcc.target/x86_64/abi/callabi/leaf-2.c
> > > +++ b/gcc/testsuite/gcc.target/x86_64/abi/callabi/leaf-2.c
> > > @@ -1,5 +1,5 @@
> > > /* { dg-do compile } */
> > > -/* { dg-options "-O2 -fno-tree-vectorize -mabi=sysv" } */
> > > +/* { dg-options "-O2 -fno-tree-vectorize -mabi=sysv
> > > +-fno-shrink-wrap-separate" } */
> > >
> > > extern int glb1, gbl2, gbl3;
> > >
> > > --
> > > 2.34.1
> > >
