From: Richard Sandiford <richard.sandiford@arm.com>
To: Robin Dapp <rdapp@linux.ibm.com>
Cc: gcc-patches@gcc.gnu.org
Subject: Re: [PATCH 1/7] ifcvt: Check if cmovs are needed.
Date: Thu, 14 Oct 2021 09:45:00 +0100 [thread overview]
Message-ID: <mpt4k9kgflf.fsf@arm.com> (raw)
In-Reply-To: <b974902c-9cfe-b861-e3c8-cdbc950004ee@linux.ibm.com> (Robin Dapp's message of "Wed, 15 Sep 2021 10:39:54 +0200")
Hi Robin,
Thanks for the update and sorry for the late response.
Robin Dapp <rdapp@linux.ibm.com> writes:
> Hi Richard,
>
>> Don't we still need this code (without the REG_DEAD handling) for the
>> case in which…
>>
>>> + /* As we are transforming
>>> + if (x > y)
>>> + {
>>> + a = b;
>>> + c = d;
>>> + }
>>> + into
>>> + a = (x > y) ...
>>> + c = (x > y) ...
>>> +
>>> + we potentially check x > y before every set.
>>> + Even though the check might be removed by subsequent passes, this means
>>> + that we cannot transform
>>> + if (x > y)
>>> + {
>>> + x = y;
>>> + ...
>>> + }
>>> + into
>>> + x = (x > y) ...
>>> + ...
>>> + since this would invalidate x and the following to-be-removed checks.
>>> + Therefore we introduce a temporary every time we are about to
>>> + overwrite a variable used in the check. Costing of a sequence with
>>> + these is going to be inaccurate so only use temporaries when
>>> + needed. */
>>> + if (reg_overlap_mentioned_p (target, cond))
>>> + temp = gen_reg_rtx (GET_MODE (target));
>>
>> …this code triggers? I don't see otherwise how later uses of x would
>> pick up “temp” instead of the original target. E.g. suppose we had:
>>
>> if (x > y)
>> {
>> x = …;
>> z = x; // x does not die here
>> }
>>
>> Without the loop, it looks like z would pick up the old value of x
>> (used in the comparison) instead of the new one.
>
> getting back to this now. I re-added handling of the situation you
> mentioned (even though I didn't manage to trigger it myself).
Yeah, the only reliable way to test for this might be an rtx test
(see testsuite/gcc.dg/rtl for some examples). A test isn't necessary
though.
My only remaining concern is that bb_ok_for_noce_convert_multiple_sets
allows subreg sources:
if (!(REG_P (src)
|| (GET_CODE (src) == SUBREG && REG_P (SUBREG_REG (src))
&& subreg_lowpart_p (src))))
return false;
These subregs could also require rewrites. It looks like the original
code checks for overlaps correctly, but doesn't necessarily deal with
a hit correctly. So this is at least partly pre-existing.
I think the way of handling that is as follows:
> diff --git a/gcc/ifcvt.c b/gcc/ifcvt.c
> index 017944f4f79..f1448667732 100644
> --- a/gcc/ifcvt.c
> +++ b/gcc/ifcvt.c
> @@ -98,6 +98,8 @@ static int dead_or_predicable (basic_block, basic_block, basic_block,
> edge, int);
> static void noce_emit_move_insn (rtx, rtx);
> static rtx_insn *block_has_only_trap (basic_block);
> +static void need_cmov_or_rewire (basic_block, hash_set<rtx_insn *> *,
> + hash_map<rtx_insn *, int> *);
> \f
> /* Count the number of non-jump active insns in BB. */
>
> @@ -3203,6 +3205,11 @@ noce_convert_multiple_sets (struct noce_if_info *if_info)
> auto_vec<rtx_insn *> unmodified_insns;
> int count = 0;
>
> + hash_set<rtx_insn *> need_no_cmov;
> + hash_map<rtx_insn *, int> rewired_src;
> +
> + need_cmov_or_rewire (then_bb, &need_no_cmov, &rewired_src);
> +
> FOR_BB_INSNS (then_bb, insn)
> {
> /* Skip over non-insns. */
> @@ -3213,26 +3220,47 @@ noce_convert_multiple_sets (struct noce_if_info *if_info)
> gcc_checking_assert (set);
>
> rtx target = SET_DEST (set);
> - rtx temp = gen_reg_rtx (GET_MODE (target));
> - rtx new_val = SET_SRC (set);
> + rtx temp;
> +
> + int *ii = rewired_src.get (insn);
> + rtx new_val = ii == NULL ? SET_SRC (set) : temporaries[*ii];
(1) here use something like:
rtx new_val = SET_SRC (set);
if (int *ii = rewired_src.get (insn))
new_val = simplify_replace_rtx (new_val, targets[*ii],
temporaries[*ii]);
> rtx old_val = target;
>
> - /* If we were supposed to read from an earlier write in this block,
> - we've changed the register allocation. Rewire the read. While
> - we are looking, also try to catch a swap idiom. */
> - for (int i = count - 1; i >= 0; --i)
> - if (reg_overlap_mentioned_p (new_val, targets[i]))
> - {
> - /* Catch a "swap" style idiom. */
> - if (find_reg_note (insn, REG_DEAD, new_val) != NULL_RTX)
> - /* The write to targets[i] is only live until the read
> - here. As the condition codes match, we can propagate
> - the set to here. */
> - new_val = SET_SRC (single_set (unmodified_insns[i]));
> - else
> - new_val = temporaries[i];
> - break;
> - }
> + /* As we are transforming
> + if (x > y)
> + {
> + a = b;
> + c = d;
> + }
> + into
> + a = (x > y) ...
> + c = (x > y) ...
> +
> + we potentially check x > y before every set.
> + Even though the check might be removed by subsequent passes, this means
> + that we cannot transform
> + if (x > y)
> + {
> + x = y;
> + ...
> + }
> + into
> + x = (x > y) ...
> + ...
> + since this would invalidate x and the following to-be-removed checks.
> + Therefore we introduce a temporary every time we are about to
> + overwrite a variable used in the check. Costing of a sequence with
> + these is going to be inaccurate so only use temporaries when
> + needed. */
> + if (reg_overlap_mentioned_p (target, cond))
> + temp = gen_reg_rtx (GET_MODE (target));
> + else
> + temp = target;
> +
> + /* We have identified swap-style idioms in check_need_cmovs. A normal
> + set will need to be a cmov while the first instruction of a swap-style
> + idiom can be a regular move. This helps with costing. */
> + bool need_cmov = !need_no_cmov.contains (insn);
>
> /* If we had a non-canonical conditional jump (i.e. one where
> the fallthrough is to the "else" case) we need to reverse
> @@ -3275,16 +3303,29 @@ noce_convert_multiple_sets (struct noce_if_info *if_info)
> old_val = lowpart_subreg (dst_mode, old_val, src_mode);
> }
>
> - /* Actually emit the conditional move. */
> - rtx temp_dest = noce_emit_cmove (if_info, temp, cond_code,
> - x, y, new_val, old_val);
> + rtx temp_dest = NULL_RTX;
>
> - /* If we failed to expand the conditional move, drop out and don't
> - try to continue. */
> - if (temp_dest == NULL_RTX)
> + if (need_cmov)
> {
> - end_sequence ();
> - return FALSE;
> + /* Actually emit the conditional move. */
> + temp_dest = noce_emit_cmove (if_info, temp, cond_code,
> + x, y, new_val, old_val);
> +
> + /* If we failed to expand the conditional move, drop out and don't
> + try to continue. */
> + if (temp_dest == NULL_RTX)
> + {
> + end_sequence ();
> + return FALSE;
> + }
> + }
> + else
> + {
> + if (if_info->then_else_reversed)
> + noce_emit_move_insn (temp, old_val);
> + else
> + noce_emit_move_insn (temp, new_val);
> + temp_dest = temp;
> }
>
> /* Bookkeeping. */
> @@ -3808,6 +3849,87 @@ check_cond_move_block (basic_block bb,
> return TRUE;
> }
>
> +/* Find local swap-style idioms in BB and mark the first insn (1)
> + that is only a temporary as not needing a conditional move as
> + it is going to be dead afterwards anyway.
> +
> + (1) int tmp = a;
> + a = b;
> + b = tmp;
> +
> + ifcvt
> + -->
> +
> + tmp = a;
> + a = cond ? b : a_old;
> + b = cond ? tmp : b_old;
> +
> + Additionally, store the index of insns like (2) when a subsequent
> + SET reads from their destination.
> +
> + (2) int c = a;
> + int d = c;
> +
> + ifcvt
> + -->
> +
> + c = cond ? a : c_old;
> + d = cond ? d : c; // Need to use c rather than c_old here.
> +*/
> +
> +static void
> +need_cmov_or_rewire (basic_block bb,
> + hash_set<rtx_insn *> *need_no_cmov,
> + hash_map<rtx_insn *, int> *rewired_src)
> +{
> + rtx_insn *insn;
> + int count = 0;
> + auto_vec<rtx_insn *> insns;
> + auto_vec<rtx> dests;
> +
> + /* Iterate over all SETs, storing the destinations
> + in DEST.
> + - If we hit a SET that reads from a destination
> + that we have seen before and the corresponding register
> + is dead afterwards, the register does not need to be
> + moved conditionally.
> + - If we encounter a previously changed register,
> + rewire the read to the original source. */
> + FOR_BB_INSNS (bb, insn)
> + {
> + rtx set, src, dest;
> +
> + if (!active_insn_p (insn))
> + continue;
> +
> + set = single_set (insn);
> + if (set == NULL_RTX)
> + continue;
> +
> + src = SET_SRC (set);
> + dest = SET_DEST (set);
> +
> + /* Check if the current SET's source is the same
> + as any previously seen destination.
> + This is quadratic but the number of insns in BB
> + is bounded by PARAM_MAX_RTL_IF_CONVERSION_INSNS. */
(2) Insert:
if (SUBREG_P (src))
src = SUBREG_REG (src);
here.
OK with those changes if that works. Let me know if they don't —
I'll try to be quicker with the next review.
Thanks,
Richard
> + if (REG_P (src))
> + for (int i = count - 1; i >= 0; --i)
> + if (reg_overlap_mentioned_p (src, dests[i]))
> + {
> + if (find_reg_note (insn, REG_DEAD, src) != NULL_RTX)
> + need_no_cmov->add (insns[i]);
> + else
> + rewired_src->put (insn, i);
> + }
> +
> + insns.safe_push (insn);
> + dests.safe_push (dest);
> +
> + count++;
> + }
> +}
> +
> /* Given a basic block BB suitable for conditional move conversion,
> a condition COND, and pointer maps THEN_VALS and ELSE_VALS containing
> the register values depending on COND, emit the insns in the block as
next prev parent reply other threads:[~2021-10-14 8:45 UTC|newest]
Thread overview: 31+ messages / expand[flat|nested] mbox.gz Atom feed top
2021-06-25 16:08 [PATCH 0/7] ifcvt: Convert multiple Robin Dapp
2021-06-25 16:08 ` [PATCH 1/7] ifcvt: Check if cmovs are needed Robin Dapp
2021-07-15 20:10 ` Richard Sandiford
2021-07-22 12:06 ` Robin Dapp
2021-07-26 19:08 ` Richard Sandiford
2021-09-15 8:39 ` Robin Dapp
2021-10-14 8:45 ` Richard Sandiford [this message]
2021-10-14 14:20 ` Robin Dapp
2021-10-14 14:32 ` Richard Sandiford
2021-10-18 11:40 ` Robin Dapp
2021-11-03 8:55 ` Robin Dapp
2021-11-05 15:33 ` Richard Sandiford
2021-11-12 13:00 ` Robin Dapp
2021-11-30 16:36 ` Richard Sandiford
2021-06-25 16:09 ` [PATCH 2/7] ifcvt: Allow constants for noce_convert_multiple Robin Dapp
2021-07-15 20:25 ` Richard Sandiford
2021-06-25 16:09 ` [PATCH 3/7] ifcvt: Improve costs handling " Robin Dapp
2021-07-15 20:42 ` Richard Sandiford
2021-07-22 12:07 ` Robin Dapp
2021-07-26 19:10 ` Richard Sandiford
2021-06-25 16:09 ` [PATCH 4/7] ifcvt/optabs: Allow using a CC comparison for emit_conditional_move Robin Dapp
2021-07-15 20:54 ` Richard Sandiford
2021-07-22 12:07 ` Robin Dapp
2021-07-26 19:31 ` Richard Sandiford
2021-07-27 20:49 ` Robin Dapp
2021-08-06 12:14 ` Richard Sandiford
2021-06-25 16:09 ` [PATCH 5/7] ifcvt: Try re-using CC for conditional moves Robin Dapp
2021-07-22 12:12 ` Robin Dapp
2021-06-25 16:09 ` [PATCH 6/7] testsuite/s390: Add tests for noce_convert_multiple Robin Dapp
2021-06-25 16:09 ` [PATCH 7/7] s390: Increase costs for load on condition and change movqicc expander Robin Dapp
2021-07-13 12:42 ` [PATCH 0/7] ifcvt: Convert multiple Robin Dapp
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=mpt4k9kgflf.fsf@arm.com \
--to=richard.sandiford@arm.com \
--cc=gcc-patches@gcc.gnu.org \
--cc=rdapp@linux.ibm.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).