From: Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
To: Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>,
gcc Patches <gcc-patches@gcc.gnu.org>,
richard.sandiford@arm.com
Subject: Re: [aarch64] Code-gen for vector initialization involving constants
Date: Fri, 19 May 2023 16:26:09 +0530 [thread overview]
Message-ID: <CAAgBjMkzx9rB_WQD78jv18NiV+R+5bg9Jr3UFPonVmxTSSh+OA@mail.gmail.com> (raw)
In-Reply-To: <mptjzx5y4wj.fsf@arm.com>
[-- Attachment #1: Type: text/plain, Size: 7107 bytes --]
On Thu, 18 May 2023 at 22:04, Richard Sandiford
<richard.sandiford@arm.com> wrote:
>
> Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org> writes:
> > On Thu, 18 May 2023 at 13:37, Richard Sandiford
> > <richard.sandiford@arm.com> wrote:
> >>
> >> Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org> writes:
> >> > On Tue, 16 May 2023 at 00:29, Richard Sandiford
> >> > <richard.sandiford@arm.com> wrote:
> >> >>
> >> >> Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org> writes:
> >> >> > Hi Richard,
> >> >> > After committing the interleave+zip1 patch for vector initialization,
> >> >> > it seems to regress the s32 case for this patch:
> >> >> >
> >> >> > int32x4_t f_s32(int32_t x)
> >> >> > {
> >> >> > return (int32x4_t) { x, x, x, 1 };
> >> >> > }
> >> >> >
> >> >> > code-gen:
> >> >> > f_s32:
> >> >> > movi v30.2s, 0x1
> >> >> > fmov s31, w0
> >> >> > dup v0.2s, v31.s[0]
> >> >> > ins v30.s[0], v31.s[0]
> >> >> > zip1 v0.4s, v0.4s, v30.4s
> >> >> > ret
> >> >> >
> >> >> > instead of expected code-gen:
> >> >> > f_s32:
> >> >> > movi v31.2s, 0x1
> >> >> > dup v0.4s, w0
> >> >> > ins v0.s[3], v31.s[0]
> >> >> > ret
> >> >> >
> >> >> > Cost for fallback sequence: 16
> >> >> > Cost for interleave and zip sequence: 12
> >> >> >
> >> >> > For the above case, the cost for interleave+zip1 sequence is computed as:
> >> >> > halves[0]:
> >> >> > (set (reg:V2SI 96)
> >> >> > (vec_duplicate:V2SI (reg/v:SI 93 [ x ])))
> >> >> > cost = 8
> >> >> >
> >> >> > halves[1]:
> >> >> > (set (reg:V2SI 97)
> >> >> > (const_vector:V2SI [
> >> >> > (const_int 1 [0x1]) repeated x2
> >> >> > ]))
> >> >> > (set (reg:V2SI 97)
> >> >> > (vec_merge:V2SI (vec_duplicate:V2SI (reg/v:SI 93 [ x ]))
> >> >> > (reg:V2SI 97)
> >> >> > (const_int 1 [0x1])))
> >> >> > cost = 8
> >> >> >
> >> >> > followed by:
> >> >> > (set (reg:V4SI 95)
> >> >> > (unspec:V4SI [
> >> >> > (subreg:V4SI (reg:V2SI 96) 0)
> >> >> > (subreg:V4SI (reg:V2SI 97) 0)
> >> >> > ] UNSPEC_ZIP1))
> >> >> > cost = 4
> >> >> >
> >> >> > So the total cost becomes
> >> >> > max(costs[0], costs[1]) + zip1_insn_cost
> >> >> > = max(8, 8) + 4
> >> >> > = 12
> >> >> >
> >> >> > While the fallback rtl sequence is:
> >> >> > (set (reg:V4SI 95)
> >> >> > (vec_duplicate:V4SI (reg/v:SI 93 [ x ])))
> >> >> > cost = 8
> >> >> > (set (reg:SI 98)
> >> >> > (const_int 1 [0x1]))
> >> >> > cost = 4
> >> >> > (set (reg:V4SI 95)
> >> >> > (vec_merge:V4SI (vec_duplicate:V4SI (reg:SI 98))
> >> >> > (reg:V4SI 95)
> >> >> > (const_int 8 [0x8])))
> >> >> > cost = 4
> >> >> >
> >> >> > So total cost = 8 + 4 + 4 = 16, and we choose the interleave+zip1 sequence.
> >> >> >
> >> >> > I think the issue is probably that for the interleave+zip1 sequence we take
> >> >> > max(costs[0], costs[1]) to reflect that both halves are interleaved,
> >> >> > but for the fallback seq we use seq_cost, which assumes serial execution
> >> >> > of insns in the sequence.
> >> >> > For above fallback sequence,
> >> >> > set (reg:V4SI 95)
> >> >> > (vec_duplicate:V4SI (reg/v:SI 93 [ x ])))
> >> >> > and
> >> >> > (set (reg:SI 98)
> >> >> > (const_int 1 [0x1]))
> >> >> > could be executed in parallel, which would make it's cost max(8, 4) + 4 = 12.
> >> >>
> >> >> Agreed.
> >> >>
> >> >> A good-enough substitute for this might be to ignore scalar moves
> >> >> (for both alternatives) when costing for speed.
> >> > Thanks for the suggestions. Just wondering for aarch64, if there's an easy
> >> > way we can check if insn is a scalar move, similar to riscv's scalar_move_insn_p
> >> > that checks if get_attr_type(insn) is TYPE_VIMOVXV or TYPE_VFMOVFV ?
> >>
> >> It should be enough to check that the pattern is a SET:
> >>
> >> (a) whose SET_DEST has a scalar mode and
> >> (b) whose SET_SRC an aarch64_mov_operand
> > Hi Richard,
> > Thanks for the suggestions, the attached patch calls seq_cost to compute
> > cost for sequence and then subtracts cost of each scalar move insn from it.
> > Does that look OK ?
> > The patch is under bootstrap+test on aarch64-linux-gnu.
>
> Yeah, the patch looks reasonable (some comments below). The testing
> for this kind of patch is more than a formality though, so it would
> be good to wait to see if the tests pass.
>
> > [...]
> > diff --git a/gcc/config/aarch64/aarch64.cc b/gcc/config/aarch64/aarch64.cc
> > index 29dbacfa917..7efd896d364 100644
> > --- a/gcc/config/aarch64/aarch64.cc
> > +++ b/gcc/config/aarch64/aarch64.cc
> > @@ -22332,6 +22332,32 @@ aarch64_unzip_vector_init (machine_mode mode, rtx vals, bool even_p)
> > return gen_rtx_PARALLEL (new_mode, vec);
> > }
> >
> > +/* Return true if INSN is a scalar move. */
> > +
> > +static bool
> > +scalar_move_insn_p (rtx_insn *insn)
> > +{
> > + rtx set = single_set (insn);
> > + if (!set)
> > + return false;
> > + rtx src = SET_SRC (set);
> > + rtx dest = SET_DEST (set);
> > + return is_a<scalar_mode>(GET_MODE (dest)) && aarch64_mov_operand_p (src, GET_MODE (src));
>
> Long line.
>
> > +}
> > +
> > +/* Ignore cost for scalar moves from cost of sequence. This function is called
> > + for calculating sequence costs in aarch64_expand_vector_init. */
> > +
> > +static unsigned
> > +seq_cost_ignore_scalar_moves (rtx_insn *seq, bool speed)
>
> Maybe more readable as "ignoring" rather than "ignore".
>
> > +{
> > + unsigned cost = seq_cost (seq, speed);
> > + for (; seq; seq = NEXT_INSN (seq))
> > + if (scalar_move_insn_p (seq))
> > + cost -= insn_cost (seq, speed);
> > + return cost;
> > +}
> > +
>
> seq_cost uses set_rtx_cost rather than insn_cost for single sets.
>
> To avoid that kind of inconsistency, I think it'd better to duplicate and
> adjust seq_cost. Then scalar_move_insn_p can be passed the single set.
>
> > /* Expand a vector initialization sequence, such that TARGET is
> > initialized to contain VALS. */
> >
> > @@ -22367,7 +22393,7 @@ aarch64_expand_vector_init (rtx target, rtx vals)
> > halves[i] = gen_rtx_SUBREG (mode, tmp_reg, 0);
> > rtx_insn *rec_seq = get_insns ();
> > end_sequence ();
> > - costs[i] = seq_cost (rec_seq, !optimize_size);
> > + costs[i] = seq_cost_ignore_scalar_moves (rec_seq, !optimize_size);
> > emit_insn (rec_seq);
> > }
> >
> > @@ -22384,7 +22410,7 @@ aarch64_expand_vector_init (rtx target, rtx vals)
> > start_sequence ();
> > aarch64_expand_vector_init_fallback (target, vals);
> > rtx_insn *fallback_seq = get_insns ();
> > - unsigned fallback_seq_cost = seq_cost (fallback_seq, !optimize_size);
> > + unsigned fallback_seq_cost = seq_cost_ignore_scalar_moves (fallback_seq, !optimize_size);
>
> Long line.
Hi Richard,
Thanks for the suggestions. Does the attached patch look OK ?
Boostrap+test in progress on aarch64-linux-gnu.
Thanks,
Prathamesh
>
> Thanks,
> Richard
>
> > end_sequence ();
> >
> > emit_insn (seq_total_cost < fallback_seq_cost ? seq : fallback_seq);
[-- Attachment #2: ignore-scalar-move-2.txt --]
[-- Type: text/plain, Size: 2367 bytes --]
[aarch64] Ignore cost of scalar moves for seq in vector initialization.
gcc/ChangeLog:
* config/aarch64/aarch64.cc (scalar_move_insn_p): New function.
(seq_cost_ignoring_scalar_moves): Likewise.
(aarch64_expand_vector_init): Call seq_cost_ignoring_scalar_moves.
diff --git a/gcc/config/aarch64/aarch64.cc b/gcc/config/aarch64/aarch64.cc
index 29dbacfa917..e611a7cca25 100644
--- a/gcc/config/aarch64/aarch64.cc
+++ b/gcc/config/aarch64/aarch64.cc
@@ -22332,6 +22332,43 @@ aarch64_unzip_vector_init (machine_mode mode, rtx vals, bool even_p)
return gen_rtx_PARALLEL (new_mode, vec);
}
+/* Return true if INSN is a scalar move. */
+
+static bool
+scalar_move_insn_p (const rtx_insn *insn)
+{
+ rtx set = single_set (insn);
+ if (!set)
+ return false;
+ rtx src = SET_SRC (set);
+ rtx dest = SET_DEST (set);
+ return is_a<scalar_mode>(GET_MODE (dest))
+ && aarch64_mov_operand_p (src, GET_MODE (src));
+}
+
+/* Similar to seq_cost, but ignore cost for scalar moves. This function
+ is called from aarch64_expand_vector_init. */
+
+static unsigned
+seq_cost_ignoring_scalar_moves (const rtx_insn *seq, bool speed)
+{
+ unsigned cost = 0;
+ rtx set;
+
+ for (; seq; seq = NEXT_INSN (seq))
+ if (NONDEBUG_INSN_P (seq)
+ && !scalar_move_insn_p (seq))
+ {
+ int this_cost = insn_cost (CONST_CAST_RTX_INSN (seq), speed);
+ if (this_cost > 0)
+ cost += this_cost;
+ else
+ cost++;
+ }
+
+ return cost;
+}
+
/* Expand a vector initialization sequence, such that TARGET is
initialized to contain VALS. */
@@ -22367,7 +22404,7 @@ aarch64_expand_vector_init (rtx target, rtx vals)
halves[i] = gen_rtx_SUBREG (mode, tmp_reg, 0);
rtx_insn *rec_seq = get_insns ();
end_sequence ();
- costs[i] = seq_cost (rec_seq, !optimize_size);
+ costs[i] = seq_cost_ignoring_scalar_moves (rec_seq, !optimize_size);
emit_insn (rec_seq);
}
@@ -22384,7 +22421,8 @@ aarch64_expand_vector_init (rtx target, rtx vals)
start_sequence ();
aarch64_expand_vector_init_fallback (target, vals);
rtx_insn *fallback_seq = get_insns ();
- unsigned fallback_seq_cost = seq_cost (fallback_seq, !optimize_size);
+ unsigned fallback_seq_cost
+ = seq_cost_ignoring_scalar_moves (fallback_seq, !optimize_size);
end_sequence ();
emit_insn (seq_total_cost < fallback_seq_cost ? seq : fallback_seq);
next prev parent reply other threads:[~2023-05-19 10:56 UTC|newest]
Thread overview: 31+ messages / expand[flat|nested] mbox.gz Atom feed top
2023-02-03 7:16 Prathamesh Kulkarni
2023-02-13 6:28 ` Prathamesh Kulkarni
2023-04-03 18:12 ` Prathamesh Kulkarni
2023-04-25 10:59 ` Richard Sandiford
2023-05-02 5:41 ` Prathamesh Kulkarni
2023-05-02 9:25 ` Richard Sandiford
2023-05-02 10:22 ` Prathamesh Kulkarni
2023-05-02 12:02 ` Richard Sandiford
2023-05-02 12:38 ` Prathamesh Kulkarni
2023-05-02 12:52 ` Richard Sandiford
2023-05-03 11:28 ` Prathamesh Kulkarni
2023-05-11 19:15 ` Richard Sandiford
2023-05-15 14:09 ` Prathamesh Kulkarni
2023-05-15 18:59 ` Richard Sandiford
2023-05-17 15:23 ` Prathamesh Kulkarni
2023-05-18 8:07 ` Richard Sandiford
2023-05-18 14:41 ` Prathamesh Kulkarni
2023-05-18 16:34 ` Richard Sandiford
2023-05-19 10:56 ` Prathamesh Kulkarni [this message]
2023-05-22 8:48 ` Richard Sandiford
2023-05-24 9:29 ` Prathamesh Kulkarni
2023-05-24 10:10 ` Richard Sandiford
2023-05-24 19:13 ` Prathamesh Kulkarni
2023-05-24 19:58 ` Richard Sandiford
2023-05-25 6:47 ` Prathamesh Kulkarni
2023-05-25 7:34 ` Richard Sandiford
2023-05-25 9:56 ` Prathamesh Kulkarni
2023-05-26 3:04 ` Prathamesh Kulkarni
2023-05-30 18:53 ` Richard Sandiford
2023-06-12 17:52 ` Prathamesh Kulkarni
2023-05-24 19:50 ` Prathamesh Kulkarni
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=CAAgBjMkzx9rB_WQD78jv18NiV+R+5bg9Jr3UFPonVmxTSSh+OA@mail.gmail.com \
--to=prathamesh.kulkarni@linaro.org \
--cc=gcc-patches@gcc.gnu.org \
--cc=richard.sandiford@arm.com \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for read-only IMAP folder(s) and NNTP newsgroup(s).