i386: Only enable small loop unrolling in backend [PR 107602]
Commit Message
Hi,
Followed by the discussion in pr107602, -munroll-only-small-loops
Does not turns on/off -funroll-loops, and current check in
pass_rtl_unroll_loops::gate would cause -funroll-loops do not take
effect. Revert the change about targetm.loop_unroll_adjust and apply
the backend option change to strictly follow the rule that
-funroll-loops takes full control of loop unrolling, and
munroll-only-small-loops just change its behavior to unroll small size
loops.
Bootstrapped and regtested on x86-64-pc-linux-gnu.
Ok for trunk?
gcc/ChangeLog:
PR target/107602
* common/config/i386/i386-common.cc (ix86_optimization_table):
Enable loop unroll O2, disable -fweb and -frename-registers
by default.
* config/i386/i386-options.cc
(ix86_override_options_after_change):
Disable small loop unroll when funroll-loops enabled, reset
cunroll_grow_size when it is not explicitly enabled.
(ix86_option_override_internal): Call
ix86_override_options_after_change instead of calling
ix86_recompute_optlev_based_flags and ix86_default_align
separately.
* config/i386/i386.cc (ix86_loop_unroll_adjust): Adjust unroll
factor if -munroll-only-small-loops enabled.
* loop-init.cc (pass_rtl_unroll_loops::gate): Do not enable
loop unrolling for -O2-speed.
(pass_rtl_unroll_loops::execute): Rmove
targetm.loop_unroll_adjust check.
gcc/testsuite/ChangeLog:
PR target/107602
* gcc.target/i386/pr86270.c: Add -fno-unroll-loops.
* gcc.target/i386/pr93002.c: Likewise.
---
gcc/common/config/i386/i386-common.cc | 8 ++++++
gcc/config/i386/i386-options.cc | 34 ++++++++++++++++++++++---
gcc/config/i386/i386.cc | 18 ++++---------
gcc/loop-init.cc | 11 +++-----
gcc/testsuite/gcc.target/i386/pr86270.c | 2 +-
gcc/testsuite/gcc.target/i386/pr93002.c | 2 +-
6 files changed, 49 insertions(+), 26 deletions(-)
Comments
> -----Original Message-----
> From: Wang, Hongyu <hongyu.wang@intel.com>
> Sent: Saturday, November 19, 2022 2:26 PM
> To: gcc-patches@gcc.gnu.org
> Cc: richard.guenther@gmail.com; ubizjak@gmail.com; Liu, Hongtao
> <hongtao.liu@intel.com>
> Subject: [PATCH] i386: Only enable small loop unrolling in backend [PR 107602]
>
> Hi,
>
> Followed by the discussion in pr107602, -munroll-only-small-loops Does not
PR107692?
> turns on/off -funroll-loops, and current check in pass_rtl_unroll_loops::gate
> would cause -funroll-loops do not take effect. Revert the change about
> targetm.loop_unroll_adjust and apply the backend option change to strictly
> follow the rule that -funroll-loops takes full control of loop unrolling, and
> munroll-only-small-loops just change its behavior to unroll small size loops.
>
> Bootstrapped and regtested on x86-64-pc-linux-gnu.
>
> Ok for trunk?
>
> gcc/ChangeLog:
>
> PR target/107602
> * common/config/i386/i386-common.cc (ix86_optimization_table):
> Enable loop unroll O2, disable -fweb and -frename-registers
> by default.
> * config/i386/i386-options.cc
> (ix86_override_options_after_change):
> Disable small loop unroll when funroll-loops enabled, reset
> cunroll_grow_size when it is not explicitly enabled.
> (ix86_option_override_internal): Call
> ix86_override_options_after_change instead of calling
> ix86_recompute_optlev_based_flags and ix86_default_align
> separately.
> * config/i386/i386.cc (ix86_loop_unroll_adjust): Adjust unroll
> factor if -munroll-only-small-loops enabled.
> * loop-init.cc (pass_rtl_unroll_loops::gate): Do not enable
> loop unrolling for -O2-speed.
> (pass_rtl_unroll_loops::execute): Rmove
> targetm.loop_unroll_adjust check.
>
> gcc/testsuite/ChangeLog:
>
> PR target/107602
> * gcc.target/i386/pr86270.c: Add -fno-unroll-loops.
> * gcc.target/i386/pr93002.c: Likewise.
> ---
> gcc/common/config/i386/i386-common.cc | 8 ++++++
> gcc/config/i386/i386-options.cc | 34 ++++++++++++++++++++++---
> gcc/config/i386/i386.cc | 18 ++++---------
> gcc/loop-init.cc | 11 +++-----
> gcc/testsuite/gcc.target/i386/pr86270.c | 2 +-
> gcc/testsuite/gcc.target/i386/pr93002.c | 2 +-
> 6 files changed, 49 insertions(+), 26 deletions(-)
>
> diff --git a/gcc/common/config/i386/i386-common.cc
> b/gcc/common/config/i386/i386-common.cc
> index 6ce2a588adc..660a977b68b 100644
> --- a/gcc/common/config/i386/i386-common.cc
> +++ b/gcc/common/config/i386/i386-common.cc
> @@ -1808,7 +1808,15 @@ static const struct default_options
> ix86_option_optimization_table[] =
> /* The STC algorithm produces the smallest code at -Os, for x86. */
> { OPT_LEVELS_2_PLUS, OPT_freorder_blocks_algorithm_, NULL,
> REORDER_BLOCKS_ALGORITHM_STC },
> +
> + /* Turn on -funroll-loops with -munroll-only-small-loops to enable small
> + loop unrolling at -O2. */
> + { OPT_LEVELS_2_PLUS_SPEED_ONLY, OPT_funroll_loops, NULL, 1 },
> { OPT_LEVELS_2_PLUS_SPEED_ONLY, OPT_munroll_only_small_loops, NULL,
> 1 },
> + /* Turns off -frename-registers and -fweb which are enabled by
> + funroll-loops. */
> + { OPT_LEVELS_ALL, OPT_frename_registers, NULL, 0 },
> + { OPT_LEVELS_ALL, OPT_fweb, NULL, 0 },
> /* Turn off -fschedule-insns by default. It tends to make the
> problem with not enough registers even worse. */
> { OPT_LEVELS_ALL, OPT_fschedule_insns, NULL, 0 }, diff --git
> a/gcc/config/i386/i386-options.cc b/gcc/config/i386/i386-options.cc index
> e5c77f3a84d..bc1d36e36a8 100644
> --- a/gcc/config/i386/i386-options.cc
> +++ b/gcc/config/i386/i386-options.cc
> @@ -1838,8 +1838,37 @@ ix86_recompute_optlev_based_flags (struct
> gcc_options *opts, void ix86_override_options_after_change (void) {
> + /* Default align_* from the processor table. */
> ix86_default_align (&global_options);
> +
> ix86_recompute_optlev_based_flags (&global_options, &global_options_set);
> +
> + /* Disable unrolling small loops when there's explicit
> + -f{,no}unroll-loop. */
> + if ((OPTION_SET_P (flag_unroll_loops))
> + || (OPTION_SET_P (flag_unroll_all_loops)
> + && flag_unroll_all_loops))
> + {
> + if (!OPTION_SET_P (ix86_unroll_only_small_loops))
> + ix86_unroll_only_small_loops = 0;
> + /* Re-enable -frename-registers and -fweb if funroll-loops
> + enabled. */
> + if (!OPTION_SET_P (flag_web))
> + flag_web = flag_unroll_loops;
> + if (!OPTION_SET_P (flag_rename_registers))
> + flag_rename_registers = flag_unroll_loops;
> + /* -fcunroll-grow-size default follws -f[no]-unroll-loops. */
> + if (!OPTION_SET_P (flag_cunroll_grow_size))
> + flag_cunroll_grow_size = flag_unroll_loops
> + || flag_peel_loops
> + || optimize >= 3;
> + }
> + else
> + {
> + if (!OPTION_SET_P (flag_cunroll_grow_size))
> + flag_cunroll_grow_size = flag_peel_loops || optimize >= 3;
> + }
> +
> }
>
> /* Clear stack slot assignments remembered from previous functions.
> @@ -2351,7 +2380,7 @@ ix86_option_override_internal (bool main_args_p,
>
> set_ix86_tune_features (opts, ix86_tune, opts->x_ix86_dump_tunes);
>
> - ix86_recompute_optlev_based_flags (opts, opts_set);
> + ix86_override_options_after_change ();
>
> ix86_tune_cost = processor_cost_table[ix86_tune];
> /* TODO: ix86_cost should be chosen at instruction or function granuality
> @@ -2382,9 +2411,6 @@ ix86_option_override_internal (bool main_args_p,
> || TARGET_64BIT_P (opts->x_ix86_isa_flags))
> opts->x_ix86_regparm = REGPARM_MAX;
>
> - /* Default align_* from the processor table. */
> - ix86_default_align (opts);
> -
> /* Provide default for -mbranch-cost= value. */
> SET_OPTION_IF_UNSET (opts, opts_set, ix86_branch_cost,
> ix86_tune_cost->branch_cost);
> diff --git a/gcc/config/i386/i386.cc b/gcc/config/i386/i386.cc index
> 292b32c5e99..10f8b4e91a5 100644
> --- a/gcc/config/i386/i386.cc
> +++ b/gcc/config/i386/i386.cc
> @@ -23829,20 +23829,12 @@ ix86_loop_unroll_adjust (unsigned nunroll,
> class loop *loop)
>
> /* Unroll small size loop when unroll factor is not explicitly
> specified. */
> - if (!(flag_unroll_loops
> - || flag_unroll_all_loops
> - || loop->unroll))
> + if (ix86_unroll_only_small_loops && !loop->unroll)
> {
> - nunroll = 1;
> -
> - /* Any explicit -f{no-}unroll-{all-}loops turns off
> - -munroll-only-small-loops. */
> - if (ix86_unroll_only_small_loops
> - && !OPTION_SET_P (flag_unroll_loops)
> - && loop->ninsns <= ix86_cost->small_unroll_ninsns)
> - nunroll = ix86_cost->small_unroll_factor;
> -
> - return nunroll;
> + if (loop->ninsns <= ix86_cost->small_unroll_ninsns)
> + return MIN (nunroll, ix86_cost->small_unroll_factor);
> + else
> + return 1;
> }
>
> if (!TARGET_ADJUST_UNROLL)
> diff --git a/gcc/loop-init.cc b/gcc/loop-init.cc index 9789efa1e11..f03c077cae1
> 100644
> --- a/gcc/loop-init.cc
> +++ b/gcc/loop-init.cc
> @@ -565,12 +565,10 @@ public:
> {}
>
> /* opt_pass methods: */
> - bool gate (function *fun) final override
> + bool gate (function *) final override
> {
> - return (flag_unroll_loops || flag_unroll_all_loops || cfun->has_unroll
> - || (targetm.loop_unroll_adjust
> - && optimize >= 2
> - && optimize_function_for_speed_p (fun)));
> + return (flag_unroll_loops || flag_unroll_all_loops
> + || cfun->has_unroll);
> }
>
> unsigned int execute (function *) final override; @@ -586,8 +584,7 @@
> pass_rtl_unroll_loops::execute (function *fun)
> if (dump_file)
> df_dump (dump_file);
>
> - if (flag_unroll_loops
> - || targetm.loop_unroll_adjust)
> + if (flag_unroll_loops)
> flags |= UAP_UNROLL;
> if (flag_unroll_all_loops)
> flags |= UAP_UNROLL_ALL;
> diff --git a/gcc/testsuite/gcc.target/i386/pr86270.c
> b/gcc/testsuite/gcc.target/i386/pr86270.c
> index cbc9fbb0450..98b012caf23 100644
> --- a/gcc/testsuite/gcc.target/i386/pr86270.c
> +++ b/gcc/testsuite/gcc.target/i386/pr86270.c
> @@ -1,5 +1,5 @@
> /* { dg-do compile } */
> -/* { dg-options "-O2 -mno-unroll-only-small-loops" } */
> +/* { dg-options "-O2 -fno-unroll-loops" } */
>
> int *a;
> long len;
> diff --git a/gcc/testsuite/gcc.target/i386/pr93002.c
> b/gcc/testsuite/gcc.target/i386/pr93002.c
> index f75a847f75d..7e2d869e17b 100644
> --- a/gcc/testsuite/gcc.target/i386/pr93002.c
> +++ b/gcc/testsuite/gcc.target/i386/pr93002.c
> @@ -1,6 +1,6 @@
> /* PR target/93002 */
> /* { dg-do compile } */
> -/* { dg-options "-O2 -mno-unroll-only-small-loops" } */
> +/* { dg-options "-O2 -fno-unroll-loops" } */
> /* { dg-final { scan-assembler-not "cmp\[^\n\r]*-1" } } */
>
> volatile int sink;
> --
> 2.18.1
On Mon, Nov 21, 2022 at 9:01 AM Liu, Hongtao via Gcc-patches
<gcc-patches@gcc.gnu.org> wrote:
>
>
>
> > -----Original Message-----
> > From: Wang, Hongyu <hongyu.wang@intel.com>
> > Sent: Saturday, November 19, 2022 2:26 PM
> > To: gcc-patches@gcc.gnu.org
> > Cc: richard.guenther@gmail.com; ubizjak@gmail.com; Liu, Hongtao
> > <hongtao.liu@intel.com>
> > Subject: [PATCH] i386: Only enable small loop unrolling in backend [PR 107602]
> >
> > Hi,
> >
> > Followed by the discussion in pr107602, -munroll-only-small-loops Does not
> PR107692?
> > turns on/off -funroll-loops, and current check in pass_rtl_unroll_loops::gate
> > would cause -funroll-loops do not take effect. Revert the change about
It's not necessarily right. unroll_factor will be set as 1 when
-fno-unroll-loops, which is exactly -fno-unroll-loops means.
> > targetm.loop_unroll_adjust and apply the backend option change to strictly
> > follow the rule that -funroll-loops takes full control of loop unrolling, and
> > munroll-only-small-loops just change its behavior to unroll small size loops.
> >
> > Bootstrapped and regtested on x86-64-pc-linux-gnu.
> >
> > Ok for trunk?
> >
> > gcc/ChangeLog:
> >
> > PR target/107602
> > * common/config/i386/i386-common.cc (ix86_optimization_table):
> > Enable loop unroll O2, disable -fweb and -frename-registers
> > by default.
> > * config/i386/i386-options.cc
> > (ix86_override_options_after_change):
> > Disable small loop unroll when funroll-loops enabled, reset
> > cunroll_grow_size when it is not explicitly enabled.
> > (ix86_option_override_internal): Call
> > ix86_override_options_after_change instead of calling
> > ix86_recompute_optlev_based_flags and ix86_default_align
> > separately.
> > * config/i386/i386.cc (ix86_loop_unroll_adjust): Adjust unroll
> > factor if -munroll-only-small-loops enabled.
> > * loop-init.cc (pass_rtl_unroll_loops::gate): Do not enable
> > loop unrolling for -O2-speed.
> > (pass_rtl_unroll_loops::execute): Rmove
> > targetm.loop_unroll_adjust check.
> >
> > gcc/testsuite/ChangeLog:
> >
> > PR target/107602
> > * gcc.target/i386/pr86270.c: Add -fno-unroll-loops.
> > * gcc.target/i386/pr93002.c: Likewise.
> > ---
> > gcc/common/config/i386/i386-common.cc | 8 ++++++
> > gcc/config/i386/i386-options.cc | 34 ++++++++++++++++++++++---
> > gcc/config/i386/i386.cc | 18 ++++---------
> > gcc/loop-init.cc | 11 +++-----
> > gcc/testsuite/gcc.target/i386/pr86270.c | 2 +-
> > gcc/testsuite/gcc.target/i386/pr93002.c | 2 +-
> > 6 files changed, 49 insertions(+), 26 deletions(-)
> >
> > diff --git a/gcc/common/config/i386/i386-common.cc
> > b/gcc/common/config/i386/i386-common.cc
> > index 6ce2a588adc..660a977b68b 100644
> > --- a/gcc/common/config/i386/i386-common.cc
> > +++ b/gcc/common/config/i386/i386-common.cc
> > @@ -1808,7 +1808,15 @@ static const struct default_options
> > ix86_option_optimization_table[] =
> > /* The STC algorithm produces the smallest code at -Os, for x86. */
> > { OPT_LEVELS_2_PLUS, OPT_freorder_blocks_algorithm_, NULL,
> > REORDER_BLOCKS_ALGORITHM_STC },
> > +
> > + /* Turn on -funroll-loops with -munroll-only-small-loops to enable small
> > + loop unrolling at -O2. */
> > + { OPT_LEVELS_2_PLUS_SPEED_ONLY, OPT_funroll_loops, NULL, 1 },
> > { OPT_LEVELS_2_PLUS_SPEED_ONLY, OPT_munroll_only_small_loops, NULL,
> > 1 },
> > + /* Turns off -frename-registers and -fweb which are enabled by
> > + funroll-loops. */
> > + { OPT_LEVELS_ALL, OPT_frename_registers, NULL, 0 },
> > + { OPT_LEVELS_ALL, OPT_fweb, NULL, 0 },
> > /* Turn off -fschedule-insns by default. It tends to make the
> > problem with not enough registers even worse. */
> > { OPT_LEVELS_ALL, OPT_fschedule_insns, NULL, 0 }, diff --git
> > a/gcc/config/i386/i386-options.cc b/gcc/config/i386/i386-options.cc index
> > e5c77f3a84d..bc1d36e36a8 100644
> > --- a/gcc/config/i386/i386-options.cc
> > +++ b/gcc/config/i386/i386-options.cc
> > @@ -1838,8 +1838,37 @@ ix86_recompute_optlev_based_flags (struct
> > gcc_options *opts, void ix86_override_options_after_change (void) {
> > + /* Default align_* from the processor table. */
> > ix86_default_align (&global_options);
> > +
> > ix86_recompute_optlev_based_flags (&global_options, &global_options_set);
> > +
> > + /* Disable unrolling small loops when there's explicit
> > + -f{,no}unroll-loop. */
> > + if ((OPTION_SET_P (flag_unroll_loops))
> > + || (OPTION_SET_P (flag_unroll_all_loops)
> > + && flag_unroll_all_loops))
> > + {
> > + if (!OPTION_SET_P (ix86_unroll_only_small_loops))
> > + ix86_unroll_only_small_loops = 0;
> > + /* Re-enable -frename-registers and -fweb if funroll-loops
> > + enabled. */
> > + if (!OPTION_SET_P (flag_web))
> > + flag_web = flag_unroll_loops;
> > + if (!OPTION_SET_P (flag_rename_registers))
> > + flag_rename_registers = flag_unroll_loops;
> > + /* -fcunroll-grow-size default follws -f[no]-unroll-loops. */
> > + if (!OPTION_SET_P (flag_cunroll_grow_size))
> > + flag_cunroll_grow_size = flag_unroll_loops
> > + || flag_peel_loops
> > + || optimize >= 3;
> > + }
> > + else
> > + {
> > + if (!OPTION_SET_P (flag_cunroll_grow_size))
> > + flag_cunroll_grow_size = flag_peel_loops || optimize >= 3;
> > + }
> > +
> > }
> >
> > /* Clear stack slot assignments remembered from previous functions.
> > @@ -2351,7 +2380,7 @@ ix86_option_override_internal (bool main_args_p,
> >
> > set_ix86_tune_features (opts, ix86_tune, opts->x_ix86_dump_tunes);
> >
> > - ix86_recompute_optlev_based_flags (opts, opts_set);
> > + ix86_override_options_after_change ();
> >
> > ix86_tune_cost = processor_cost_table[ix86_tune];
> > /* TODO: ix86_cost should be chosen at instruction or function granuality
> > @@ -2382,9 +2411,6 @@ ix86_option_override_internal (bool main_args_p,
> > || TARGET_64BIT_P (opts->x_ix86_isa_flags))
> > opts->x_ix86_regparm = REGPARM_MAX;
> >
> > - /* Default align_* from the processor table. */
> > - ix86_default_align (opts);
> > -
> > /* Provide default for -mbranch-cost= value. */
> > SET_OPTION_IF_UNSET (opts, opts_set, ix86_branch_cost,
> > ix86_tune_cost->branch_cost);
> > diff --git a/gcc/config/i386/i386.cc b/gcc/config/i386/i386.cc index
> > 292b32c5e99..10f8b4e91a5 100644
> > --- a/gcc/config/i386/i386.cc
> > +++ b/gcc/config/i386/i386.cc
> > @@ -23829,20 +23829,12 @@ ix86_loop_unroll_adjust (unsigned nunroll,
> > class loop *loop)
> >
> > /* Unroll small size loop when unroll factor is not explicitly
> > specified. */
> > - if (!(flag_unroll_loops
> > - || flag_unroll_all_loops
> > - || loop->unroll))
> > + if (ix86_unroll_only_small_loops && !loop->unroll)
> > {
> > - nunroll = 1;
> > -
> > - /* Any explicit -f{no-}unroll-{all-}loops turns off
> > - -munroll-only-small-loops. */
> > - if (ix86_unroll_only_small_loops
> > - && !OPTION_SET_P (flag_unroll_loops)
> > - && loop->ninsns <= ix86_cost->small_unroll_ninsns)
> > - nunroll = ix86_cost->small_unroll_factor;
> > -
> > - return nunroll;
> > + if (loop->ninsns <= ix86_cost->small_unroll_ninsns)
> > + return MIN (nunroll, ix86_cost->small_unroll_factor);
> > + else
> > + return 1;
> > }
> >
> > if (!TARGET_ADJUST_UNROLL)
> > diff --git a/gcc/loop-init.cc b/gcc/loop-init.cc index 9789efa1e11..f03c077cae1
> > 100644
> > --- a/gcc/loop-init.cc
> > +++ b/gcc/loop-init.cc
> > @@ -565,12 +565,10 @@ public:
> > {}
> >
> > /* opt_pass methods: */
> > - bool gate (function *fun) final override
> > + bool gate (function *) final override
> > {
> > - return (flag_unroll_loops || flag_unroll_all_loops || cfun->has_unroll
> > - || (targetm.loop_unroll_adjust
> > - && optimize >= 2
> > - && optimize_function_for_speed_p (fun)));
> > + return (flag_unroll_loops || flag_unroll_all_loops
> > + || cfun->has_unroll);
> > }
> >
> > unsigned int execute (function *) final override; @@ -586,8 +584,7 @@
> > pass_rtl_unroll_loops::execute (function *fun)
> > if (dump_file)
> > df_dump (dump_file);
> >
> > - if (flag_unroll_loops
> > - || targetm.loop_unroll_adjust)
> > + if (flag_unroll_loops)
> > flags |= UAP_UNROLL;
> > if (flag_unroll_all_loops)
> > flags |= UAP_UNROLL_ALL;
> > diff --git a/gcc/testsuite/gcc.target/i386/pr86270.c
> > b/gcc/testsuite/gcc.target/i386/pr86270.c
> > index cbc9fbb0450..98b012caf23 100644
> > --- a/gcc/testsuite/gcc.target/i386/pr86270.c
> > +++ b/gcc/testsuite/gcc.target/i386/pr86270.c
> > @@ -1,5 +1,5 @@
> > /* { dg-do compile } */
> > -/* { dg-options "-O2 -mno-unroll-only-small-loops" } */
> > +/* { dg-options "-O2 -fno-unroll-loops" } */
> >
> > int *a;
> > long len;
> > diff --git a/gcc/testsuite/gcc.target/i386/pr93002.c
> > b/gcc/testsuite/gcc.target/i386/pr93002.c
> > index f75a847f75d..7e2d869e17b 100644
> > --- a/gcc/testsuite/gcc.target/i386/pr93002.c
> > +++ b/gcc/testsuite/gcc.target/i386/pr93002.c
> > @@ -1,6 +1,6 @@
> > /* PR target/93002 */
> > /* { dg-do compile } */
> > -/* { dg-options "-O2 -mno-unroll-only-small-loops" } */
> > +/* { dg-options "-O2 -fno-unroll-loops" } */
> > /* { dg-final { scan-assembler-not "cmp\[^\n\r]*-1" } } */
> >
> > volatile int sink;
> > --
> > 2.18.1
>
> It's not necessarily right. unroll_factor will be set as 1 when
> -fno-unroll-loops, which is exactly -fno-unroll-loops means.
Not that exactly, -fno-unroll-loops previously will prevent the pass
from running, and on the current trunk the pass still runs.
Actually I think the implementation on trunk is a bit tricky since we
use a target hook without return value to execute the pass. Also the
logic !OPTION_SET_P looks quite like a workaround. IMHO this is also
not good for maintenance.
Hongtao Liu via Gcc-patches <gcc-patches@gcc.gnu.org> 于2022年11月21日周一 09:33写道：
>
> On Mon, Nov 21, 2022 at 9:01 AM Liu, Hongtao via Gcc-patches
> <gcc-patches@gcc.gnu.org> wrote:
> >
> >
> >
> > > -----Original Message-----
> > > From: Wang, Hongyu <hongyu.wang@intel.com>
> > > Sent: Saturday, November 19, 2022 2:26 PM
> > > To: gcc-patches@gcc.gnu.org
> > > Cc: richard.guenther@gmail.com; ubizjak@gmail.com; Liu, Hongtao
> > > <hongtao.liu@intel.com>
> > > Subject: [PATCH] i386: Only enable small loop unrolling in backend [PR 107602]
> > >
> > > Hi,
> > >
> > > Followed by the discussion in pr107602, -munroll-only-small-loops Does not
> > PR107692?
> > > turns on/off -funroll-loops, and current check in pass_rtl_unroll_loops::gate
> > > would cause -funroll-loops do not take effect. Revert the change about
> It's not necessarily right. unroll_factor will be set as 1 when
> -fno-unroll-loops, which is exactly -fno-unroll-loops means.
> > > targetm.loop_unroll_adjust and apply the backend option change to strictly
> > > follow the rule that -funroll-loops takes full control of loop unrolling, and
> > > munroll-only-small-loops just change its behavior to unroll small size loops.
> > >
> > > Bootstrapped and regtested on x86-64-pc-linux-gnu.
> > >
> > > Ok for trunk?
> > >
> > > gcc/ChangeLog:
> > >
> > > PR target/107602
> > > * common/config/i386/i386-common.cc (ix86_optimization_table):
> > > Enable loop unroll O2, disable -fweb and -frename-registers
> > > by default.
> > > * config/i386/i386-options.cc
> > > (ix86_override_options_after_change):
> > > Disable small loop unroll when funroll-loops enabled, reset
> > > cunroll_grow_size when it is not explicitly enabled.
> > > (ix86_option_override_internal): Call
> > > ix86_override_options_after_change instead of calling
> > > ix86_recompute_optlev_based_flags and ix86_default_align
> > > separately.
> > > * config/i386/i386.cc (ix86_loop_unroll_adjust): Adjust unroll
> > > factor if -munroll-only-small-loops enabled.
> > > * loop-init.cc (pass_rtl_unroll_loops::gate): Do not enable
> > > loop unrolling for -O2-speed.
> > > (pass_rtl_unroll_loops::execute): Rmove
> > > targetm.loop_unroll_adjust check.
> > >
> > > gcc/testsuite/ChangeLog:
> > >
> > > PR target/107602
> > > * gcc.target/i386/pr86270.c: Add -fno-unroll-loops.
> > > * gcc.target/i386/pr93002.c: Likewise.
> > > ---
> > > gcc/common/config/i386/i386-common.cc | 8 ++++++
> > > gcc/config/i386/i386-options.cc | 34 ++++++++++++++++++++++---
> > > gcc/config/i386/i386.cc | 18 ++++---------
> > > gcc/loop-init.cc | 11 +++-----
> > > gcc/testsuite/gcc.target/i386/pr86270.c | 2 +-
> > > gcc/testsuite/gcc.target/i386/pr93002.c | 2 +-
> > > 6 files changed, 49 insertions(+), 26 deletions(-)
> > >
> > > diff --git a/gcc/common/config/i386/i386-common.cc
> > > b/gcc/common/config/i386/i386-common.cc
> > > index 6ce2a588adc..660a977b68b 100644
> > > --- a/gcc/common/config/i386/i386-common.cc
> > > +++ b/gcc/common/config/i386/i386-common.cc
> > > @@ -1808,7 +1808,15 @@ static const struct default_options
> > > ix86_option_optimization_table[] =
> > > /* The STC algorithm produces the smallest code at -Os, for x86. */
> > > { OPT_LEVELS_2_PLUS, OPT_freorder_blocks_algorithm_, NULL,
> > > REORDER_BLOCKS_ALGORITHM_STC },
> > > +
> > > + /* Turn on -funroll-loops with -munroll-only-small-loops to enable small
> > > + loop unrolling at -O2. */
> > > + { OPT_LEVELS_2_PLUS_SPEED_ONLY, OPT_funroll_loops, NULL, 1 },
> > > { OPT_LEVELS_2_PLUS_SPEED_ONLY, OPT_munroll_only_small_loops, NULL,
> > > 1 },
> > > + /* Turns off -frename-registers and -fweb which are enabled by
> > > + funroll-loops. */
> > > + { OPT_LEVELS_ALL, OPT_frename_registers, NULL, 0 },
> > > + { OPT_LEVELS_ALL, OPT_fweb, NULL, 0 },
> > > /* Turn off -fschedule-insns by default. It tends to make the
> > > problem with not enough registers even worse. */
> > > { OPT_LEVELS_ALL, OPT_fschedule_insns, NULL, 0 }, diff --git
> > > a/gcc/config/i386/i386-options.cc b/gcc/config/i386/i386-options.cc index
> > > e5c77f3a84d..bc1d36e36a8 100644
> > > --- a/gcc/config/i386/i386-options.cc
> > > +++ b/gcc/config/i386/i386-options.cc
> > > @@ -1838,8 +1838,37 @@ ix86_recompute_optlev_based_flags (struct
> > > gcc_options *opts, void ix86_override_options_after_change (void) {
> > > + /* Default align_* from the processor table. */
> > > ix86_default_align (&global_options);
> > > +
> > > ix86_recompute_optlev_based_flags (&global_options, &global_options_set);
> > > +
> > > + /* Disable unrolling small loops when there's explicit
> > > + -f{,no}unroll-loop. */
> > > + if ((OPTION_SET_P (flag_unroll_loops))
> > > + || (OPTION_SET_P (flag_unroll_all_loops)
> > > + && flag_unroll_all_loops))
> > > + {
> > > + if (!OPTION_SET_P (ix86_unroll_only_small_loops))
> > > + ix86_unroll_only_small_loops = 0;
> > > + /* Re-enable -frename-registers and -fweb if funroll-loops
> > > + enabled. */
> > > + if (!OPTION_SET_P (flag_web))
> > > + flag_web = flag_unroll_loops;
> > > + if (!OPTION_SET_P (flag_rename_registers))
> > > + flag_rename_registers = flag_unroll_loops;
> > > + /* -fcunroll-grow-size default follws -f[no]-unroll-loops. */
> > > + if (!OPTION_SET_P (flag_cunroll_grow_size))
> > > + flag_cunroll_grow_size = flag_unroll_loops
> > > + || flag_peel_loops
> > > + || optimize >= 3;
> > > + }
> > > + else
> > > + {
> > > + if (!OPTION_SET_P (flag_cunroll_grow_size))
> > > + flag_cunroll_grow_size = flag_peel_loops || optimize >= 3;
> > > + }
> > > +
> > > }
> > >
> > > /* Clear stack slot assignments remembered from previous functions.
> > > @@ -2351,7 +2380,7 @@ ix86_option_override_internal (bool main_args_p,
> > >
> > > set_ix86_tune_features (opts, ix86_tune, opts->x_ix86_dump_tunes);
> > >
> > > - ix86_recompute_optlev_based_flags (opts, opts_set);
> > > + ix86_override_options_after_change ();
> > >
> > > ix86_tune_cost = processor_cost_table[ix86_tune];
> > > /* TODO: ix86_cost should be chosen at instruction or function granuality
> > > @@ -2382,9 +2411,6 @@ ix86_option_override_internal (bool main_args_p,
> > > || TARGET_64BIT_P (opts->x_ix86_isa_flags))
> > > opts->x_ix86_regparm = REGPARM_MAX;
> > >
> > > - /* Default align_* from the processor table. */
> > > - ix86_default_align (opts);
> > > -
> > > /* Provide default for -mbranch-cost= value. */
> > > SET_OPTION_IF_UNSET (opts, opts_set, ix86_branch_cost,
> > > ix86_tune_cost->branch_cost);
> > > diff --git a/gcc/config/i386/i386.cc b/gcc/config/i386/i386.cc index
> > > 292b32c5e99..10f8b4e91a5 100644
> > > --- a/gcc/config/i386/i386.cc
> > > +++ b/gcc/config/i386/i386.cc
> > > @@ -23829,20 +23829,12 @@ ix86_loop_unroll_adjust (unsigned nunroll,
> > > class loop *loop)
> > >
> > > /* Unroll small size loop when unroll factor is not explicitly
> > > specified. */
> > > - if (!(flag_unroll_loops
> > > - || flag_unroll_all_loops
> > > - || loop->unroll))
> > > + if (ix86_unroll_only_small_loops && !loop->unroll)
> > > {
> > > - nunroll = 1;
> > > -
> > > - /* Any explicit -f{no-}unroll-{all-}loops turns off
> > > - -munroll-only-small-loops. */
> > > - if (ix86_unroll_only_small_loops
> > > - && !OPTION_SET_P (flag_unroll_loops)
> > > - && loop->ninsns <= ix86_cost->small_unroll_ninsns)
> > > - nunroll = ix86_cost->small_unroll_factor;
> > > -
> > > - return nunroll;
> > > + if (loop->ninsns <= ix86_cost->small_unroll_ninsns)
> > > + return MIN (nunroll, ix86_cost->small_unroll_factor);
> > > + else
> > > + return 1;
> > > }
> > >
> > > if (!TARGET_ADJUST_UNROLL)
> > > diff --git a/gcc/loop-init.cc b/gcc/loop-init.cc index 9789efa1e11..f03c077cae1
> > > 100644
> > > --- a/gcc/loop-init.cc
> > > +++ b/gcc/loop-init.cc
> > > @@ -565,12 +565,10 @@ public:
> > > {}
> > >
> > > /* opt_pass methods: */
> > > - bool gate (function *fun) final override
> > > + bool gate (function *) final override
> > > {
> > > - return (flag_unroll_loops || flag_unroll_all_loops || cfun->has_unroll
> > > - || (targetm.loop_unroll_adjust
> > > - && optimize >= 2
> > > - && optimize_function_for_speed_p (fun)));
> > > + return (flag_unroll_loops || flag_unroll_all_loops
> > > + || cfun->has_unroll);
> > > }
> > >
> > > unsigned int execute (function *) final override; @@ -586,8 +584,7 @@
> > > pass_rtl_unroll_loops::execute (function *fun)
> > > if (dump_file)
> > > df_dump (dump_file);
> > >
> > > - if (flag_unroll_loops
> > > - || targetm.loop_unroll_adjust)
> > > + if (flag_unroll_loops)
> > > flags |= UAP_UNROLL;
> > > if (flag_unroll_all_loops)
> > > flags |= UAP_UNROLL_ALL;
> > > diff --git a/gcc/testsuite/gcc.target/i386/pr86270.c
> > > b/gcc/testsuite/gcc.target/i386/pr86270.c
> > > index cbc9fbb0450..98b012caf23 100644
> > > --- a/gcc/testsuite/gcc.target/i386/pr86270.c
> > > +++ b/gcc/testsuite/gcc.target/i386/pr86270.c
> > > @@ -1,5 +1,5 @@
> > > /* { dg-do compile } */
> > > -/* { dg-options "-O2 -mno-unroll-only-small-loops" } */
> > > +/* { dg-options "-O2 -fno-unroll-loops" } */
> > >
> > > int *a;
> > > long len;
> > > diff --git a/gcc/testsuite/gcc.target/i386/pr93002.c
> > > b/gcc/testsuite/gcc.target/i386/pr93002.c
> > > index f75a847f75d..7e2d869e17b 100644
> > > --- a/gcc/testsuite/gcc.target/i386/pr93002.c
> > > +++ b/gcc/testsuite/gcc.target/i386/pr93002.c
> > > @@ -1,6 +1,6 @@
> > > /* PR target/93002 */
> > > /* { dg-do compile } */
> > > -/* { dg-options "-O2 -mno-unroll-only-small-loops" } */
> > > +/* { dg-options "-O2 -fno-unroll-loops" } */
> > > /* { dg-final { scan-assembler-not "cmp\[^\n\r]*-1" } } */
> > >
> > > volatile int sink;
> > > --
> > > 2.18.1
> >
>
>
> --
> BR,
> Hongtao
On 11/18/22 23:25, Hongyu Wang via Gcc-patches wrote:
> Hi,
>
> Followed by the discussion in pr107602, -munroll-only-small-loops
> Does not turns on/off -funroll-loops, and current check in
> pass_rtl_unroll_loops::gate would cause -funroll-loops do not take
> effect. Revert the change about targetm.loop_unroll_adjust and apply
> the backend option change to strictly follow the rule that
> -funroll-loops takes full control of loop unrolling, and
> munroll-only-small-loops just change its behavior to unroll small size
> loops.
>
> Bootstrapped and regtested on x86-64-pc-linux-gnu.
>
> Ok for trunk?
>
> gcc/ChangeLog:
>
> PR target/107602
> * common/config/i386/i386-common.cc (ix86_optimization_table):
> Enable loop unroll O2, disable -fweb and -frename-registers
> by default.
> * config/i386/i386-options.cc
> (ix86_override_options_after_change):
> Disable small loop unroll when funroll-loops enabled, reset
> cunroll_grow_size when it is not explicitly enabled.
> (ix86_option_override_internal): Call
> ix86_override_options_after_change instead of calling
> ix86_recompute_optlev_based_flags and ix86_default_align
> separately.
> * config/i386/i386.cc (ix86_loop_unroll_adjust): Adjust unroll
> factor if -munroll-only-small-loops enabled.
> * loop-init.cc (pass_rtl_unroll_loops::gate): Do not enable
> loop unrolling for -O2-speed.
> (pass_rtl_unroll_loops::execute): Rmove
> targetm.loop_unroll_adjust check.
The reversion of the loop-init.cc changes is fine. The x86 maintainers
will need to chime in on the rest. Consider installing the loop-init.cc
reversion immediately as the current state has regressed s390 and
potentially other targets.
jeff
On Tue, Nov 22, 2022 at 1:41 AM Jeff Law via Gcc-patches
<gcc-patches@gcc.gnu.org> wrote:
>
>
> On 11/18/22 23:25, Hongyu Wang via Gcc-patches wrote:
> > Hi,
> >
> > Followed by the discussion in pr107602, -munroll-only-small-loops
> > Does not turns on/off -funroll-loops, and current check in
> > pass_rtl_unroll_loops::gate would cause -funroll-loops do not take
> > effect. Revert the change about targetm.loop_unroll_adjust and apply
> > the backend option change to strictly follow the rule that
> > -funroll-loops takes full control of loop unrolling, and
> > munroll-only-small-loops just change its behavior to unroll small size
> > loops.
> >
> > Bootstrapped and regtested on x86-64-pc-linux-gnu.
> >
> > Ok for trunk?
> >
> > gcc/ChangeLog:
> >
> > PR target/107602
> > * common/config/i386/i386-common.cc (ix86_optimization_table):
> > Enable loop unroll O2, disable -fweb and -frename-registers
> > by default.
> > * config/i386/i386-options.cc
> > (ix86_override_options_after_change):
> > Disable small loop unroll when funroll-loops enabled, reset
> > cunroll_grow_size when it is not explicitly enabled.
> > (ix86_option_override_internal): Call
> > ix86_override_options_after_change instead of calling
> > ix86_recompute_optlev_based_flags and ix86_default_align
> > separately.
> > * config/i386/i386.cc (ix86_loop_unroll_adjust): Adjust unroll
> > factor if -munroll-only-small-loops enabled.
> > * loop-init.cc (pass_rtl_unroll_loops::gate): Do not enable
> > loop unrolling for -O2-speed.
> > (pass_rtl_unroll_loops::execute): Rmove
> > targetm.loop_unroll_adjust check.
> The reversion of the loop-init.cc changes is fine. The x86 maintainers
> will need to chime in on the rest. Consider installing the loop-init.cc
> reversion immediately as the current state has regressed s390 and
> potentially other targets.
x86 part is ok.
>
>
> jeff
>
Hi Jeff,
> The reversion of the loop-init.cc changes is fine. The x86 maintainers
> will need to chime in on the rest. Consider installing the loop-init.cc
> reversion immediately as the current state has regressed s390 and
> potentially other targets.
I've posted a patch in
https://gcc.gnu.org/pipermail/gcc-patches/2022-November/606478.html
to change the rs6000 target and there is a discussion ongoing. If we
draw a conclusion that we only want to make changes in the backend, I
will install this one.
@@ -1808,7 +1808,15 @@ static const struct default_options ix86_option_optimization_table[] =
/* The STC algorithm produces the smallest code at -Os, for x86. */
{ OPT_LEVELS_2_PLUS, OPT_freorder_blocks_algorithm_, NULL,
REORDER_BLOCKS_ALGORITHM_STC },
+
+ /* Turn on -funroll-loops with -munroll-only-small-loops to enable small
+ loop unrolling at -O2. */
+ { OPT_LEVELS_2_PLUS_SPEED_ONLY, OPT_funroll_loops, NULL, 1 },
{ OPT_LEVELS_2_PLUS_SPEED_ONLY, OPT_munroll_only_small_loops, NULL, 1 },
+ /* Turns off -frename-registers and -fweb which are enabled by
+ funroll-loops. */
+ { OPT_LEVELS_ALL, OPT_frename_registers, NULL, 0 },
+ { OPT_LEVELS_ALL, OPT_fweb, NULL, 0 },
/* Turn off -fschedule-insns by default. It tends to make the
problem with not enough registers even worse. */
{ OPT_LEVELS_ALL, OPT_fschedule_insns, NULL, 0 },
@@ -1838,8 +1838,37 @@ ix86_recompute_optlev_based_flags (struct gcc_options *opts,
void
ix86_override_options_after_change (void)
{
+ /* Default align_* from the processor table. */
ix86_default_align (&global_options);
+
ix86_recompute_optlev_based_flags (&global_options, &global_options_set);
+
+ /* Disable unrolling small loops when there's explicit
+ -f{,no}unroll-loop. */
+ if ((OPTION_SET_P (flag_unroll_loops))
+ || (OPTION_SET_P (flag_unroll_all_loops)
+ && flag_unroll_all_loops))
+ {
+ if (!OPTION_SET_P (ix86_unroll_only_small_loops))
+ ix86_unroll_only_small_loops = 0;
+ /* Re-enable -frename-registers and -fweb if funroll-loops
+ enabled. */
+ if (!OPTION_SET_P (flag_web))
+ flag_web = flag_unroll_loops;
+ if (!OPTION_SET_P (flag_rename_registers))
+ flag_rename_registers = flag_unroll_loops;
+ /* -fcunroll-grow-size default follws -f[no]-unroll-loops. */
+ if (!OPTION_SET_P (flag_cunroll_grow_size))
+ flag_cunroll_grow_size = flag_unroll_loops
+ || flag_peel_loops
+ || optimize >= 3;
+ }
+ else
+ {
+ if (!OPTION_SET_P (flag_cunroll_grow_size))
+ flag_cunroll_grow_size = flag_peel_loops || optimize >= 3;
+ }
+
}
/* Clear stack slot assignments remembered from previous functions.
@@ -2351,7 +2380,7 @@ ix86_option_override_internal (bool main_args_p,
set_ix86_tune_features (opts, ix86_tune, opts->x_ix86_dump_tunes);
- ix86_recompute_optlev_based_flags (opts, opts_set);
+ ix86_override_options_after_change ();
ix86_tune_cost = processor_cost_table[ix86_tune];
/* TODO: ix86_cost should be chosen at instruction or function granuality
@@ -2382,9 +2411,6 @@ ix86_option_override_internal (bool main_args_p,
|| TARGET_64BIT_P (opts->x_ix86_isa_flags))
opts->x_ix86_regparm = REGPARM_MAX;
- /* Default align_* from the processor table. */
- ix86_default_align (opts);
-
/* Provide default for -mbranch-cost= value. */
SET_OPTION_IF_UNSET (opts, opts_set, ix86_branch_cost,
ix86_tune_cost->branch_cost);
@@ -23829,20 +23829,12 @@ ix86_loop_unroll_adjust (unsigned nunroll, class loop *loop)
/* Unroll small size loop when unroll factor is not explicitly
specified. */
- if (!(flag_unroll_loops
- || flag_unroll_all_loops
- || loop->unroll))
+ if (ix86_unroll_only_small_loops && !loop->unroll)
{
- nunroll = 1;
-
- /* Any explicit -f{no-}unroll-{all-}loops turns off
- -munroll-only-small-loops. */
- if (ix86_unroll_only_small_loops
- && !OPTION_SET_P (flag_unroll_loops)
- && loop->ninsns <= ix86_cost->small_unroll_ninsns)
- nunroll = ix86_cost->small_unroll_factor;
-
- return nunroll;
+ if (loop->ninsns <= ix86_cost->small_unroll_ninsns)
+ return MIN (nunroll, ix86_cost->small_unroll_factor);
+ else
+ return 1;
}
if (!TARGET_ADJUST_UNROLL)
@@ -565,12 +565,10 @@ public:
{}
/* opt_pass methods: */
- bool gate (function *fun) final override
+ bool gate (function *) final override
{
- return (flag_unroll_loops || flag_unroll_all_loops || cfun->has_unroll
- || (targetm.loop_unroll_adjust
- && optimize >= 2
- && optimize_function_for_speed_p (fun)));
+ return (flag_unroll_loops || flag_unroll_all_loops
+ || cfun->has_unroll);
}
unsigned int execute (function *) final override;
@@ -586,8 +584,7 @@ pass_rtl_unroll_loops::execute (function *fun)
if (dump_file)
df_dump (dump_file);
- if (flag_unroll_loops
- || targetm.loop_unroll_adjust)
+ if (flag_unroll_loops)
flags |= UAP_UNROLL;
if (flag_unroll_all_loops)
flags |= UAP_UNROLL_ALL;
@@ -1,5 +1,5 @@
/* { dg-do compile } */
-/* { dg-options "-O2 -mno-unroll-only-small-loops" } */
+/* { dg-options "-O2 -fno-unroll-loops" } */
int *a;
long len;
@@ -1,6 +1,6 @@
/* PR target/93002 */
/* { dg-do compile } */
-/* { dg-options "-O2 -mno-unroll-only-small-loops" } */
+/* { dg-options "-O2 -fno-unroll-loops" } */
/* { dg-final { scan-assembler-not "cmp\[^\n\r]*-1" } } */
volatile int sink;