public inbox for gcc-patches@gcc.gnu.org
 help / color / mirror / Atom feed
From: Hans-Peter Nilsson <hp@axis.com>
To: <gcc-patches@gcc.gnu.org>
Subject: [cris-decc0 4/14] cris.md: Post-reload, split/generate clobberless memory destination moves
Date: Wed, 12 Feb 2020 06:54:00 -0000	[thread overview]
Message-ID: <202002120654.01C6sYjq026507@ignucius.se.axis.com> (raw)

* config/cris/cris.md ("movsi"): For memory destination
post-reload, generate clobberless variant.
("*mov_tomem<mode>_split"): New split.
("*mov_tomem<mode>"): New insn.
("enabled", mov_tomem_enabled): Define and use to exclude "x" ->
"Q>m" for less-than-SImode.

In preparation for compare-elimination (for it to be obviously
useful), we have to have some common insn in-between that
doesn't clobber condition-codes.  A move to memory is an obvious
choice.  Note the FIXME: we can do this for a zero source too;
later.
---
 gcc/config/cris/cris.md | 47 +++++++++++++++++++++++++++++++++++++++++++++++
 1 file changed, 47 insertions(+)

diff --git a/gcc/config/cris/cris.md b/gcc/config/cris/cris.md
index 0b42197a9..3de2f5a13 100644
--- a/gcc/config/cris/cris.md
+++ b/gcc/config/cris/cris.md
@@ -147,6 +147,7 @@ (define_delay (eq_attr "slottable" "has_return_slot")
 	(not (match_test "dead_or_set_regno_p (insn, CRIS_SRP_REGNUM)")))
    (nil) (nil)])
 
+(define_attr "enabled" "no,yes" (const_string "yes"))
 \f
 ;; Iterator definitions.
 
@@ -445,6 +446,15 @@ (define_expand "movsi"
       && operands[1] != const0_rtx
       && can_create_pseudo_p ())
     operands[1] = force_reg (SImode, operands[1]);
+
+   /* At post-reload time, we'll get here for e.g. split multi-mode insns
+      with a memory destination.  Go directly to the clobber-less variant.
+      FIXME: Also applies to zero source.  */
+   if (MEM_P (operands[0]) && reload_completed)
+     {
+        emit_insn (gen_rtx_SET (operands[0], operands[1]));
+        DONE;
+     }
 })
 
 (define_insn "*movsi_internal"
@@ -650,6 +660,43 @@ (define_insn "movsf"
    move %1,%0
    move %1,%0"
   [(set_attr "slottable" "yes,yes,yes,yes,yes,no,no,no,yes,yes,yes,no,yes,no")])
+
+;; Post-reload, for memory destinations, split the clobber-variant and
+;; get rid of the clobber.
+
+(define_split ;; "*mov_tomem<mode>_split"
+  [(set (match_operand:BWD 0 "memory_operand")
+	(match_operand:BWD 1 "nonmemory_operand"))
+   (clobber (reg:CC CRIS_CC0_REGNUM))]
+  "reload_completed"
+  [(set (match_dup 0) (match_dup 1))]
+  "")
+
+;; Exclude moving special-registers to memory from matching for
+;; less-than-SImode, as they are SImode only (or actually, the size of
+;; the register, but the ones free for "x" are naturally SImode; see
+;; special measures taken for reload).
+;; This might be a belt-and-suspenders thing, as a move from special
+;; register to memory in less-than-SImode should not have made it here.
+
+(define_mode_attr mov_tomem_enabled
+  [(SI "yes,yes,yes,yes,yes,yes")
+   (HI "yes,yes,no,yes,yes,no")
+   (QI "yes,yes,no,yes,yes,no")])
+
+(define_insn "*mov_tomem<mode>"
+  [(set (match_operand:BWD 0 "memory_operand"   "=Q>,Q>,Q>,m,m,m")
+	(match_operand:BWD 1 "nonmemory_operand" "M, r, x, M,r,x"))]
+  "reload_completed"
+  "@
+   clear<m> %0
+   move<m> %1,%0
+   move %1,%0
+   clear<m> %0
+   move<m> %1,%0
+   move %1,%0"
+  [(set_attr "slottable" "yes,yes,yes,no,no,no")
+   (set_attr "enabled" "<mov_tomem_enabled>")])
 \f
 ;; Movem patterns.  Primarily for use in function prologue and epilogue.
 ;; Unfortunately, movem stores R0 in the highest memory location, thus
-- 
2.11.0

brgds, H-P

                 reply	other threads:[~2020-02-12  6:54 UTC|newest]

Thread overview: [no followups] expand[flat|nested]  mbox.gz  Atom feed

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=202002120654.01C6sYjq026507@ignucius.se.axis.com \
    --to=hp@axis.com \
    --cc=gcc-patches@gcc.gnu.org \
    /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).