[thirdparty/gcc.git] / gcc / config / arm / sync.md

;; Machine description for ARM processor synchronization primitives.
;; Copyright (C) 2010-2022 Free Software Foundation, Inc.
;; Written by Marcus Shawcroft (marcus.shawcroft@arm.com)
;; 64bit Atomics by Dave Gilbert (david.gilbert@linaro.org)
;;
;; This file is part of GCC.
;;
;; GCC is free software; you can redistribute it and/or modify it
;; under the terms of the GNU General Public License as published by
;; the Free Software Foundation; either version 3, or (at your option)
;; any later version.
;;
;; GCC is distributed in the hope that it will be useful, but
;; WITHOUT ANY WARRANTY; without even the implied warranty of
;; MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
;; General Public License for more details.
;;
;; You should have received a copy of the GNU General Public License
;; along with GCC; see the file COPYING3.  If not see
;; <http://www.gnu.org/licenses/>.  */

(define_mode_attr sync_predtab
  [(QI "TARGET_HAVE_LDREXBH && TARGET_HAVE_MEMORY_BARRIER")
   (HI "TARGET_HAVE_LDREXBH && TARGET_HAVE_MEMORY_BARRIER")
   (SI "TARGET_HAVE_LDREX && TARGET_HAVE_MEMORY_BARRIER")
   (DI "TARGET_HAVE_LDREXD && ARM_DOUBLEWORD_ALIGN
	&& TARGET_HAVE_MEMORY_BARRIER")])

(define_code_iterator syncop [plus minus ior xor and])

(define_code_attr sync_optab
  [(ior "or") (xor "xor") (and "and") (plus "add") (minus "sub")])

(define_mode_attr sync_sfx
  [(QI "b") (HI "h") (SI "") (DI "d")])

(define_expand "memory_barrier"
  [(set (match_dup 0)
	(unspec:BLK [(match_dup 0)] UNSPEC_MEMORY_BARRIER))]
  "TARGET_HAVE_MEMORY_BARRIER"
{
  operands[0] = gen_rtx_MEM (BLKmode, gen_rtx_SCRATCH (Pmode));
  MEM_VOLATILE_P (operands[0]) = 1;
})

(define_insn "*memory_barrier"
  [(set (match_operand:BLK 0 "" "")
	(unspec:BLK [(match_dup 0)] UNSPEC_MEMORY_BARRIER))]
  "TARGET_HAVE_MEMORY_BARRIER"
  {
    if (TARGET_HAVE_DMB)
      {
	return "dmb\\tish";
      }

    if (TARGET_HAVE_DMB_MCR)
      return "mcr\\tp15, 0, r0, c7, c10, 5";

    gcc_unreachable ();
  }
  [(set_attr "length" "4")
   (set_attr "conds" "unconditional")
   (set_attr "predicable" "no")])

(define_insn "atomic_load<mode>"
  [(set (match_operand:QHSI 0 "register_operand" "=r,r,l")
    (unspec_volatile:QHSI
      [(match_operand:QHSI 1 "arm_sync_memory_operand" "Q,Q,Q")
       (match_operand:SI 2 "const_int_operand" "n,Pf,n")]	;; model
      VUNSPEC_LDA))]
  "TARGET_HAVE_LDACQ"
  {
    if (aarch_mm_needs_acquire (operands[2]))
      {
	if (TARGET_THUMB1)
	  return "lda<sync_sfx>\t%0, %1";
	else
	  return "lda<sync_sfx>%?\t%0, %1";
      }
    else
      {
	if (TARGET_THUMB1)
	  return "ldr<sync_sfx>\t%0, %1";
	else
	  return "ldr<sync_sfx>%?\t%0, %1";
      }
  }
  [(set_attr "arch" "32,v8mb,any")
   (set_attr "predicable" "yes")])

(define_insn "atomic_store<mode>"
  [(set (match_operand:QHSI 0 "memory_operand" "=Q,Q,Q")
    (unspec_volatile:QHSI
      [(match_operand:QHSI 1 "general_operand" "r,r,l")
       (match_operand:SI 2 "const_int_operand" "n,Pf,n")]	;; model
      VUNSPEC_STL))]
  "TARGET_HAVE_LDACQ"
  {
    if (aarch_mm_needs_release (operands[2]))
      {
	if (TARGET_THUMB1)
	  return "stl<sync_sfx>\t%1, %0";
	else
	  return "stl<sync_sfx>%?\t%1, %0";
      }
    else
      {
	if (TARGET_THUMB1)
	  return "str<sync_sfx>\t%1, %0";
	else
	  return "str<sync_sfx>%?\t%1, %0";
      }
  }
  [(set_attr "arch" "32,v8mb,any")
   (set_attr "predicable" "yes")])

;; An LDRD instruction usable by the atomic_loaddi expander on LPAE targets

(define_insn "arm_atomic_loaddi2_ldrd"
  [(set (match_operand:DI 0 "register_operand" "=r")
	(unspec_volatile:DI
	  [(match_operand:DI 1 "arm_sync_memory_operand" "Q")]
	    VUNSPEC_LDRD_ATOMIC))]
  "ARM_DOUBLEWORD_ALIGN && TARGET_HAVE_LPAE"
  "ldrd%?\t%0, %H0, %C1"
  [(set_attr "predicable" "yes")])

;; There are three ways to expand this depending on the architecture
;; features available.  As for the barriers, a load needs a barrier
;; after it on all non-relaxed memory models except when the load
;; has acquire semantics (for ARMv8-A).

(define_expand "atomic_loaddi"
  [(match_operand:DI 0 "s_register_operand")		;; val out
   (match_operand:DI 1 "mem_noofs_operand")		;; memory
   (match_operand:SI 2 "const_int_operand")]		;; model
  "(TARGET_HAVE_LDREXD || TARGET_HAVE_LPAE || TARGET_HAVE_LDACQEXD)
   && ARM_DOUBLEWORD_ALIGN"
{
  memmodel model = memmodel_from_int (INTVAL (operands[2]));

  /* For ARMv8-A we can use an LDAEXD to atomically load two 32-bit registers
     when acquire or stronger semantics are needed.  When the relaxed model is
     used this can be relaxed to a normal LDRD.  */
  if (TARGET_HAVE_LDACQEXD)
    {
      if (is_mm_relaxed (model))
	emit_insn (gen_arm_atomic_loaddi2_ldrd (operands[0], operands[1]));
      else
	emit_insn (gen_arm_load_acquire_exclusivedi (operands[0], operands[1]));

      DONE;
    }

  /* On LPAE targets LDRD and STRD accesses to 64-bit aligned
     locations are 64-bit single-copy atomic.  We still need barriers in the
     appropriate places to implement the ordering constraints.  */
  if (TARGET_HAVE_LPAE)
    emit_insn (gen_arm_atomic_loaddi2_ldrd (operands[0], operands[1]));
  else
    emit_insn (gen_arm_load_exclusivedi (operands[0], operands[1]));


  /* All non-relaxed models need a barrier after the load when load-acquire
     instructions are not available.  */
  if (!is_mm_relaxed (model))
    expand_mem_thread_fence (model);

  DONE;
})

(define_expand "atomic_compare_and_swap<mode>"
  [(match_operand:SI 0 "s_register_operand")		;; bool out
   (match_operand:QHSD 1 "s_register_operand")		;; val out
   (match_operand:QHSD 2 "mem_noofs_operand")		;; memory
   (match_operand:QHSD 3 "general_operand")		;; expected
   (match_operand:QHSD 4 "s_register_operand")		;; desired
   (match_operand:SI 5 "const_int_operand")		;; is_weak
   (match_operand:SI 6 "const_int_operand")		;; mod_s
   (match_operand:SI 7 "const_int_operand")]		;; mod_f
  "<sync_predtab>"
{
  arm_expand_compare_and_swap (operands);
  DONE;
})

;; Constraints of this pattern must be at least as strict as those of the
;; cbranchsi operations in thumb1.md and aim to be as permissive.
(define_insn_and_split "@atomic_compare_and_swap<CCSI:arch><NARROW:mode>_1"
  [(set (match_operand:CCSI 0 "cc_register_operand" "=&c,&l,&l")	;; bool out
	(unspec_volatile:CCSI [(const_int 0)] VUNSPEC_ATOMIC_CAS))
   (set (match_operand:SI 1 "s_register_operand" "=&r,&l,&l*h")	;; val out
	(zero_extend:SI
	  (match_operand:NARROW 2 "mem_noofs_operand" "+Ua,Ua,Ua")))	;; memory
   (set (match_dup 2)
	(unspec_volatile:NARROW
	  [(match_operand:SI 3 "arm_add_operand" "rIL,lILJ*h,*r")	;; expected
	   (match_operand:NARROW 4 "s_register_operand" "r,r,r")	;; desired
	   (match_operand:SI 5 "const_int_operand")		;; is_weak
	   (match_operand:SI 6 "const_int_operand")		;; mod_s
	   (match_operand:SI 7 "const_int_operand")]		;; mod_f
	  VUNSPEC_ATOMIC_CAS))
   (clobber (match_scratch:SI 8 "=&r,X,X"))]
  "<NARROW:sync_predtab>"
  "#"
  "&& reload_completed"
  [(const_int 0)]
  {
    arm_split_compare_and_swap (operands);
    DONE;
  }
  [(set_attr "arch" "32,v8mb,v8mb")])

(define_mode_attr cas_cmp_operand
  [(SI "arm_add_operand") (DI "cmpdi_operand")])
(define_mode_attr cas_cmp_str
  [(SI "rIL") (DI "rDi")])

;; Constraints of this pattern must be at least as strict as those of the
;; cbranchsi operations in thumb1.md and aim to be as permissive.
(define_insn_and_split "@atomic_compare_and_swap<CCSI:arch><SIDI:mode>_1"
  [(set (match_operand:CCSI 0 "cc_register_operand" "=&c,&l,&l")	;; bool out
	(unspec_volatile:CCSI [(const_int 0)] VUNSPEC_ATOMIC_CAS))
   (set (match_operand:SIDI 1 "s_register_operand" "=&r,&l,&l*h")	;; val out
	(match_operand:SIDI 2 "mem_noofs_operand" "+Ua,Ua,Ua"))	;; memory
   (set (match_dup 2)
	(unspec_volatile:SIDI
	  [(match_operand:SIDI 3 "<SIDI:cas_cmp_operand>" "<SIDI:cas_cmp_str>,lILJ*h,*r") ;; expect
	   (match_operand:SIDI 4 "s_register_operand" "r,r,r")	;; desired
	   (match_operand:SI 5 "const_int_operand")		;; is_weak
	   (match_operand:SI 6 "const_int_operand")		;; mod_s
	   (match_operand:SI 7 "const_int_operand")]		;; mod_f
	  VUNSPEC_ATOMIC_CAS))
   (clobber (match_scratch:SI 8 "=&r,X,X"))]
  "<SIDI:sync_predtab>"
  "#"
  "&& reload_completed"
  [(const_int 0)]
  {
    arm_split_compare_and_swap (operands);
    DONE;
  }
  [(set_attr "arch" "32,v8mb,v8mb")])

(define_insn_and_split "atomic_exchange<mode>"
  [(set (match_operand:QHSD 0 "s_register_operand" "=&r,&r")	;; output
	(match_operand:QHSD 1 "mem_noofs_operand" "+Ua,Ua"))	;; memory
   (set (match_dup 1)
	(unspec_volatile:QHSD
	  [(match_operand:QHSD 2 "s_register_operand" "r,r")	;; input
	   (match_operand:SI 3 "const_int_operand" "")]		;; model
	  VUNSPEC_ATOMIC_XCHG))
   (clobber (reg:CC CC_REGNUM))
   (clobber (match_scratch:SI 4 "=&r,&l"))]
  "<sync_predtab>"
  "#"
  "&& reload_completed"
  [(const_int 0)]
  {
    arm_split_atomic_op (SET, operands[0], NULL, operands[1],
			 operands[2], operands[3], operands[4]);
    DONE;
  }
  [(set_attr "arch" "32,v8mb")])

;; The following mode and code attribute are defined here because they are
;; specific to atomics and are not needed anywhere else.

(define_mode_attr atomic_op_operand
  [(QI "reg_or_int_operand")
   (HI "reg_or_int_operand")
   (SI "reg_or_int_operand")
   (DI "s_register_operand")])

(define_mode_attr atomic_op_str
  [(QI "rn") (HI "rn") (SI "rn") (DI "r")])

(define_code_attr thumb1_atomic_op_str
  [(ior "l,l") (xor "l,l") (and "l,l") (plus "lIJL,r") (minus "lPd,lPd")])

(define_code_attr thumb1_atomic_newop_str
  [(ior "&l,&l") (xor "&l,&l") (and "&l,&l") (plus "&l,&r") (minus "&l,&l")])

;; Constraints of this pattern must be at least as strict as those of the non
;; atomic operations in thumb1.md and aim to be as permissive.
(define_insn_and_split "atomic_<sync_optab><mode>"
  [(set (match_operand:QHSD 0 "mem_noofs_operand" "+Ua,Ua,Ua")
	(unspec_volatile:QHSD
	  [(syncop:QHSD (match_dup 0)
	     (match_operand:QHSD 1 "<atomic_op_operand>" "<atomic_op_str>,<thumb1_atomic_op_str>"))
	   (match_operand:SI 2 "const_int_operand")]		;; model
	  VUNSPEC_ATOMIC_OP))
   (clobber (reg:CC CC_REGNUM))
   (clobber (match_scratch:QHSD 3 "=&r,<thumb1_atomic_newop_str>"))
   (clobber (match_scratch:SI 4 "=&r,&l,&l"))]
  "<sync_predtab>"
  "#"
  "&& reload_completed"
  [(const_int 0)]
  {
    arm_split_atomic_op (<CODE>, NULL, operands[3], operands[0],
			 operands[1], operands[2], operands[4]);
    DONE;
  }
  [(set_attr "arch" "32,v8mb,v8mb")])

;; Constraints of this pattern must be at least as strict as those of the non
;; atomic NANDs in thumb1.md and aim to be as permissive.
(define_insn_and_split "atomic_nand<mode>"
  [(set (match_operand:QHSD 0 "mem_noofs_operand" "+Ua,Ua")
	(unspec_volatile:QHSD
	  [(not:QHSD
	     (and:QHSD (match_dup 0)
	       (match_operand:QHSD 1 "<atomic_op_operand>" "<atomic_op_str>,l")))
	   (match_operand:SI 2 "const_int_operand")]		;; model
	  VUNSPEC_ATOMIC_OP))
   (clobber (reg:CC CC_REGNUM))
   (clobber (match_scratch:QHSD 3 "=&r,&l"))
   (clobber (match_scratch:SI 4 "=&r,&l"))]
  "<sync_predtab>"
  "#"
  "&& reload_completed"
  [(const_int 0)]
  {
    arm_split_atomic_op (NOT, NULL, operands[3], operands[0],
			 operands[1], operands[2], operands[4]);
    DONE;
  }
  [(set_attr "arch" "32,v8mb")])

;; 3 alternatives are needed to represent constraints after split from
;; thumb1_addsi3: (i) case where operand1 and destination can be in different
;; registers, (ii) case where they are in the same low register and (iii) case
;; when they are in the same register without restriction on the register.  We
;; disparage slightly alternatives that require copying the old value into the
;; register for the new value (see bind_old_new in arm_split_atomic_op).
(define_code_attr thumb1_atomic_fetch_op_str
  [(ior "l,l,l") (xor "l,l,l") (and "l,l,l") (plus "lL,?IJ,?r") (minus "lPd,lPd,lPd")])

(define_code_attr thumb1_atomic_fetch_newop_str
  [(ior "&l,&l,&l") (xor "&l,&l,&l") (and "&l,&l,&l") (plus "&l,&l,&r") (minus "&l,&l,&l")])

(define_code_attr thumb1_atomic_fetch_oldop_str
  [(ior "&r,&r,&r") (xor "&r,&r,&r") (and "&r,&r,&r") (plus "&l,&r,&r") (minus "&l,&l,&l")])

;; Constraints of this pattern must be at least as strict as those of the non
;; atomic operations in thumb1.md and aim to be as permissive.
(define_insn_and_split "atomic_fetch_<sync_optab><mode>"
  [(set (match_operand:QHSD 0 "s_register_operand" "=&r,<thumb1_atomic_fetch_oldop_str>")
	(match_operand:QHSD 1 "mem_noofs_operand" "+Ua,Ua,Ua,Ua"))
   (set (match_dup 1)
	(unspec_volatile:QHSD
	  [(syncop:QHSD (match_dup 1)
	     (match_operand:QHSD 2 "<atomic_op_operand>" "<atomic_op_str>,<thumb1_atomic_fetch_op_str>"))
	   (match_operand:SI 3 "const_int_operand")]		;; model
	  VUNSPEC_ATOMIC_OP))
   (clobber (reg:CC CC_REGNUM))
   (clobber (match_scratch:QHSD 4 "=&r,<thumb1_atomic_fetch_newop_str>"))
   (clobber (match_scratch:SI 5 "=&r,&l,&l,&l"))]
  "<sync_predtab>"
  "#"
  "&& reload_completed"
  [(const_int 0)]
  {
    arm_split_atomic_op (<CODE>, operands[0], operands[4], operands[1],
			 operands[2], operands[3], operands[5]);
    DONE;
  }
  [(set_attr "arch" "32,v8mb,v8mb,v8mb")])

;; Constraints of this pattern must be at least as strict as those of the non
;; atomic NANDs in thumb1.md and aim to be as permissive.
(define_insn_and_split "atomic_fetch_nand<mode>"
  [(set (match_operand:QHSD 0 "s_register_operand" "=&r,&r")
	(match_operand:QHSD 1 "mem_noofs_operand" "+Ua,Ua"))
   (set (match_dup 1)
	(unspec_volatile:QHSD
	  [(not:QHSD
	     (and:QHSD (match_dup 1)
	       (match_operand:QHSD 2 "<atomic_op_operand>" "<atomic_op_str>,l")))
	   (match_operand:SI 3 "const_int_operand")]		;; model
	  VUNSPEC_ATOMIC_OP))
   (clobber (reg:CC CC_REGNUM))
   (clobber (match_scratch:QHSD 4 "=&r,&l"))
   (clobber (match_scratch:SI 5 "=&r,&l"))]
  "<sync_predtab>"
  "#"
  "&& reload_completed"
  [(const_int 0)]
  {
    arm_split_atomic_op (NOT, operands[0], operands[4], operands[1],
			 operands[2], operands[3], operands[5]);
    DONE;
  }
  [(set_attr "arch" "32,v8mb")])

;; Constraints of this pattern must be at least as strict as those of the non
;; atomic operations in thumb1.md and aim to be as permissive.
(define_insn_and_split "atomic_<sync_optab>_fetch<mode>"
  [(set (match_operand:QHSD 0 "s_register_operand" "=&r,<thumb1_atomic_newop_str>")
	(syncop:QHSD
	  (match_operand:QHSD 1 "mem_noofs_operand" "+Ua,Ua,Ua")
	  (match_operand:QHSD 2 "<atomic_op_operand>" "<atomic_op_str>,<thumb1_atomic_op_str>")))
   (set (match_dup 1)
	(unspec_volatile:QHSD
	  [(match_dup 1) (match_dup 2)
	   (match_operand:SI 3 "const_int_operand")]		;; model
	  VUNSPEC_ATOMIC_OP))
   (clobber (reg:CC CC_REGNUM))
   (clobber (match_scratch:SI 4 "=&r,&l,&l"))]
  "<sync_predtab>"
  "#"
  "&& reload_completed"
  [(const_int 0)]
  {
    arm_split_atomic_op (<CODE>, NULL, operands[0], operands[1],
			 operands[2], operands[3], operands[4]);
    DONE;
  }
  [(set_attr "arch" "32,v8mb,v8mb")])

;; Constraints of this pattern must be at least as strict as those of the non
;; atomic NANDs in thumb1.md and aim to be as permissive.
(define_insn_and_split "atomic_nand_fetch<mode>"
  [(set (match_operand:QHSD 0 "s_register_operand" "=&r,&l")
	(not:QHSD
	  (and:QHSD
	    (match_operand:QHSD 1 "mem_noofs_operand" "+Ua,Ua")
	    (match_operand:QHSD 2 "<atomic_op_operand>" "<atomic_op_str>,l"))))
   (set (match_dup 1)
	(unspec_volatile:QHSD
	  [(match_dup 1) (match_dup 2)
	   (match_operand:SI 3 "const_int_operand")]		;; model
	  VUNSPEC_ATOMIC_OP))
   (clobber (reg:CC CC_REGNUM))
   (clobber (match_scratch:SI 4 "=&r,&l"))]
  "<sync_predtab>"
  "#"
  "&& reload_completed"
  [(const_int 0)]
  {
    arm_split_atomic_op (NOT, NULL, operands[0], operands[1],
			 operands[2], operands[3], operands[4]);
    DONE;
  }
  [(set_attr "arch" "32,v8mb")])

(define_insn "arm_load_exclusive<mode>"
  [(set (match_operand:SI 0 "s_register_operand" "=r,r")
        (zero_extend:SI
	  (unspec_volatile:NARROW
	    [(match_operand:NARROW 1 "mem_noofs_operand" "Ua,Ua")]
	    VUNSPEC_LL)))]
  "TARGET_HAVE_LDREXBH"
  "@
   ldrex<sync_sfx>%?\t%0, %C1
   ldrex<sync_sfx>\t%0, %C1"
  [(set_attr "arch" "32,v8mb")
   (set_attr "predicable" "yes")])

(define_insn "arm_load_acquire_exclusive<mode>"
  [(set (match_operand:SI 0 "s_register_operand" "=r,r")
        (zero_extend:SI
	  (unspec_volatile:NARROW
	    [(match_operand:NARROW 1 "mem_noofs_operand" "Ua,Ua")]
	    VUNSPEC_LAX)))]
  "TARGET_HAVE_LDACQ"
  "@
   ldaex<sync_sfx>%?\\t%0, %C1
   ldaex<sync_sfx>\\t%0, %C1"
  [(set_attr "arch" "32,v8mb")
   (set_attr "predicable" "yes")])

(define_insn "arm_load_exclusivesi"
  [(set (match_operand:SI 0 "s_register_operand" "=r,r")
	(unspec_volatile:SI
	  [(match_operand:SI 1 "mem_noofs_operand" "Ua,Ua")]
	  VUNSPEC_LL))]
  "TARGET_HAVE_LDREX"
  "@
   ldrex%?\t%0, %C1
   ldrex\t%0, %C1"
  [(set_attr "arch" "32,v8mb")
   (set_attr "predicable" "yes")])

(define_insn "arm_load_acquire_exclusivesi"
  [(set (match_operand:SI 0 "s_register_operand" "=r,r")
	(unspec_volatile:SI
	  [(match_operand:SI 1 "mem_noofs_operand" "Ua,Ua")]
	  VUNSPEC_LAX))]
  "TARGET_HAVE_LDACQ"
  "@
   ldaex%?\t%0, %C1
   ldaex\t%0, %C1"
  [(set_attr "arch" "32,v8mb")
   (set_attr "predicable" "yes")])

(define_insn "arm_load_exclusivedi"
  [(set (match_operand:DI 0 "s_register_operand" "=r")
	(unspec_volatile:DI
	  [(match_operand:DI 1 "mem_noofs_operand" "Ua")]
	  VUNSPEC_LL))]
  "TARGET_HAVE_LDREXD"
  "ldrexd%?\t%0, %H0, %C1"
  [(set_attr "predicable" "yes")])

(define_insn "arm_load_acquire_exclusivedi"
  [(set (match_operand:DI 0 "s_register_operand" "=r")
	(unspec_volatile:DI
	  [(match_operand:DI 1 "mem_noofs_operand" "Ua")]
	  VUNSPEC_LAX))]
  "TARGET_HAVE_LDACQEXD && ARM_DOUBLEWORD_ALIGN"
  "ldaexd%?\t%0, %H0, %C1"
  [(set_attr "predicable" "yes")])

(define_insn "arm_store_exclusive<mode>"
  [(set (match_operand:SI 0 "s_register_operand" "=&r")
	(unspec_volatile:SI [(const_int 0)] VUNSPEC_SC))
   (set (match_operand:QHSD 1 "mem_noofs_operand" "=Ua")
	(unspec_volatile:QHSD
	  [(match_operand:QHSD 2 "s_register_operand" "r")]
	  VUNSPEC_SC))]
  "<sync_predtab>"
  {
    if (<MODE>mode == DImode)
      {
	/* The restrictions on target registers in ARM mode are that the two
	   registers are consecutive and the first one is even; Thumb is
	   actually more flexible, but DI should give us this anyway.
	   Note that the 1st register always gets the
	   lowest word in memory.  */
	gcc_assert ((REGNO (operands[2]) & 1) == 0 || TARGET_THUMB2);
	return "strexd%?\t%0, %2, %H2, %C1";
      }
    if (TARGET_THUMB1)
      return "strex<sync_sfx>\t%0, %2, %C1";
    else
      return "strex<sync_sfx>%?\t%0, %2, %C1";
  }
  [(set_attr "predicable" "yes")])

(define_insn "arm_store_release_exclusivedi"
  [(set (match_operand:SI 0 "s_register_operand" "=&r")
	(unspec_volatile:SI [(const_int 0)] VUNSPEC_SLX))
   (set (match_operand:DI 1 "mem_noofs_operand" "=Ua")
	(unspec_volatile:DI
	  [(match_operand:DI 2 "s_register_operand" "r")]
	  VUNSPEC_SLX))]
  "TARGET_HAVE_LDACQEXD && ARM_DOUBLEWORD_ALIGN"
  {
    /* See comment in arm_store_exclusive<mode> above.  */
    gcc_assert ((REGNO (operands[2]) & 1) == 0 || TARGET_THUMB2);
    return "stlexd%?\t%0, %2, %H2, %C1";
  }
  [(set_attr "predicable" "yes")])

(define_insn "arm_store_release_exclusive<mode>"
  [(set (match_operand:SI 0 "s_register_operand" "=&r,&r")
	(unspec_volatile:SI [(const_int 0)] VUNSPEC_SLX))
   (set (match_operand:QHSI 1 "mem_noofs_operand" "=Ua,Ua")
	(unspec_volatile:QHSI
	  [(match_operand:QHSI 2 "s_register_operand" "r,r")]
	  VUNSPEC_SLX))]
  "TARGET_HAVE_LDACQ"
  "@
   stlex<sync_sfx>%?\t%0, %2, %C1
   stlex<sync_sfx>\t%0, %2, %C1"
  [(set_attr "arch" "32,v8mb")
   (set_attr "predicable" "yes")])