diff --git a/lang/ghc/Makefile b/lang/ghc/Makefile
index 3258a75a12f8..5e6fda6901af 100644
--- a/lang/ghc/Makefile
+++ b/lang/ghc/Makefile
@@ -1,347 +1,345 @@
 PORTNAME=	ghc
 PORTVERSION=	${GHC_VERSION}
-PORTREVISION?=	0
+PORTREVISION?=	1
 CATEGORIES=	lang haskell
 MASTER_SITES=	https://www.haskell.org/ghc/dist/${PORTVERSION}/:source \
 		LOCAL/arrowd/:boot
 DISTFILES=	ghc-${PORTVERSION}-src${EXTRACT_SUFX}:source
 
 MAINTAINER=	haskell@FreeBSD.org
 COMMENT=	Compiler for the functional language Haskell
 WWW=		https://www.haskell.org/ghc/
 
 LICENSE=	BSD3CLAUSE
 LICENSE_FILE=	${WRKSRC}/LICENSE
 
 ONLY_FOR_ARCHS=		aarch64 amd64 i386
 
-BROKEN_FreeBSD_14_aarch64=	https://gitlab.haskell.org/ghc/ghc/-/issues/24580
-
 BUILD_DEPENDS=		${LOCALBASE}/lib/libgmp.so:math/gmp
 
 USES=			autoreconf compiler:c11 gmake \
 			ncurses perl5 python:build shebangfix \
 			tar:xz
 
 GNU_CONFIGURE=		yes
 CONFIGURE_ARGS+=	--docdir=${DOCSDIR}
 CONFIGURE_ENV=		ALEX=/usr/bin/true HAPPY=/usr/bin/true
 INSTALL_TARGET=		install-strip
 USE_LOCALE=		en_US.UTF-8
 USE_PERL5=		build
 NO_CCACHE=		yes
 OPTIONS_SUB=		yes
 SHEBANG_FILES=		${BOOT_SCRIPT}
 
 OPTIONS_DEFINE?=	DYNAMIC GMP PROFILE DOCS
 OPTIONS_DEFAULT=	DYNAMIC PROFILE GMP
 
 OPTIONS_GROUP=		BOOTSTRAP
 BOOTSTRAP_DESC=		Bootsrap using installed ghc
 OPTIONS_GROUP_BOOTSTRAP=BOOT
 
 BOOT_DESC=		Use installed GHC for bootstrapping
 DOCS_DESC=		Install HTML documentation
 DYNAMIC_DESC=		Add support for dynamic linking
 GMP_DESC=		Use GNU Multi-precision Library from Ports
 PROFILE_DESC=		Add support for performance profiling
 
 BOOT_CONFIGURE_ENV=	GHC=${LOCALBASE}/bin/ghc
 BOOT_CONFIGURE_ENV_OFF=	GHC=${BOOT_GHC} LLC=llc${BOOT_LLVM_VERSION} OPT=opt${BOOT_LLVM_VERSION}
 
 DOCS_BUILD_DEPENDS=	sphinx-build:textproc/py-sphinx \
 			xelatex:print/tex-xetex \
 			${LOCALBASE}/share/fonts/dejavu/DejaVuSans.ttf:x11-fonts/dejavu
 DOCS_VARS=		enable_docs=YES
 DOCS_VARS_OFF=		enable_docs=NO hadrian_docs_arg=--docs=none
 
 DYNAMIC_CONFIGURE_WITH=	system-libffi \
 			ffi-includes=${LOCALBASE}/include \
 			ffi-libraries=${LOCALBASE}/lib
 DYNAMIC_LIB_DEPENDS=	libffi.so:devel/libffi
 DYNAMIC_VARS=		enable_dynamic=YES hadrian_setting_dynamic=True
 DYNAMIC_VARS_OFF=	enable_dynamic=NO hadrian_setting_dynamic=False
 
 # do not replace this with GMP_CONFIGURE_WITH
 # it adds "--without-gmp-*" when the option is OFF, which results in "no" value
 # to be used as directory name
 GMP_CONFIGURE_ON=	--with-gmp-includes=${LOCALBASE}/include \
 			--with-gmp-libraries=${LOCALBASE}/lib
 GMP_LIB_DEPENDS=	libgmp.so:math/gmp
 GMP_VARS_OFF=		hadrian_gmp_arg=--bignum=native
 
 PROFILE_VARS=		enable_profile=YES hadrian_setting_profile=True
 PROFILE_VARS_OFF=	enable_profile=NO hadrian_setting_profile=False
 
 GHC_VERSION?=		9.6.4
 LLVM_VERSION?=		15
 BOOT_GHC_VERSION?=	9.2.8
 # LLVM version that bootstrap compiler uses
 BOOT_LLVM_VERSION?=	12
 
 BASE_PACKAGES?=		Cabal-3.10.1.0 array-0.5.6.0 base-4.18.2.0 binary-0.8.9.1 \
 			bytestring-0.11.5.3 containers-0.6.7 deepseq-1.4.8.1 \
 			directory-1.3.8.1 exceptions-0.10.7 filepath-1.4.200.1 \
 			ghc-${GHC_VERSION} ghc-bignum-1.3 ghc-compact-0.1.0.0 \
 			ghc-prim-0.10.0 haskeline-0.8.2.1 hpc-0.6.2.0 \
 			integer-gmp-1.1 mtl-2.3.1 parsec-3.1.16.1 pretty-1.1.3.6 \
 			process-1.6.17.0 stm-2.5.1.0 template-haskell-2.20.0.0 \
 			terminfo-0.4.1.6 text-2.0.2 time-1.12.2 \
 			transformers-0.6.1.0 unix-2.8.4.0 xhtml-3000.2.2.1
 
 .for pkg in ${BASE_PACKAGES}
 PLIST_SUB+=	${pkg:C/-([0-9.])+//:tu}_VERSION=${pkg:C/^([^\.]*-)+//}
 .endfor
 
 BUILD_MK=		DYNAMIC_GHC_PROGRAMS=${ENABLE_DYNAMIC} \
 			BUILD_PROF_LIBS=${ENABLE_PROFILE} \
 			HADDOCK_DOCS=YES \
 			BUILD_SPHINX_HTML=${ENABLE_DOCS} \
 			BUILD_SPHINX_PDF=NO \
 			SRC_HC_OPTS+="-I${NCURSESINC} -L${NCURSESLIB} -I${LOCALBASE}/include -L${LOCALBASE}/lib" \
 			SRC_CC_OPTS+="${CFLAGS}" \
 			EXTRA_HSC2HS_OPTS+="-I${LOCALBASE}/include --lflag=-L${LOCALBASE}/lib" \
 			EXTRA_LD_OPTS+="-L${LOCALBASE}/lib" \
 			libraries/terminfo_CONFIGURE_OPTS+="--configure-option=--with-curses-libraries=${NCURSESLIB}" \
 			${SLAVE_BUILD_MK} \
 			V=0
 
 .include <bsd.port.pre.mk>
 
 .if ${GHC_VERSION:C/\..*//g} >= 9 && ${ARCH} != i386
 # hadrian build on i386 fails due to https://bugs.freebsd.org/bugzilla/show_bug.cgi?id=269570
 # relocation R_386_TLS_LE cannot be used with -shared
 USE_HADRIAN=		yes
 .endif
 
 .if ${GHC_VERSION:C/\..*//g} >= 9 && ${GHC_VERSION:C/[^\.]*\.//:C/\..*//g} >= 4
 BOOT_SCRIPT=	./boot.source
 IGNORE_i386=	requires Hadrian build which doesn't work on i386
 .else
 BOOT_SCRIPT=	./boot
 .endif
 
 .if ${SLAVE_PORT} != "yes"
 PORTDOCS=		*
 HADRIAN_PLAN=		${PATCHDIR}/plan-bootstrap-${BOOT_GHC_VERSION:C/\./_/g}.json
 .else
 HADRIAN_DOCS_ARG=	--docs=none
 HADRIAN_PLAN=		${MASTERDIR}/files/plan-bootstrap-${BOOT_GHC_VERSION:C/\./_/g}.json
 PLIST_SUB+=		GMP=
 .endif
 
 .ifdef USE_HADRIAN
 LIB_DEPENDS+=		libffi.so:devel/libffi
 .undef DYNAMIC_CONFIGURE_WITH
 CONFIGURE_ARGS+=	--with-system-libffi --with-ffi-includes=${LOCALBASE}/include --with-ffi-libraries=${LOCALBASE}/lib
 CONFIGURE_ENV+=		CC=${CC} CXX=${CXX}
 SHEBANG_FILES+=		hadrian/bootstrap/bootstrap.py
 HADRIAN_CMD=		${WRKSRC}/hadrian/bootstrap/_build/bin/hadrian ${HADRIAN_DOCS_ARG} ${HADRIAN_GMP_ARG}
 DO_MAKE_BUILD=		${SETENV} ${MAKE_ENV} ${HADRIAN_CMD} ${_MAKE_JOBS} --flavour=ports
 ALL_TARGET=		binary-dist-dir
 INSTALL_WRKSRC=		${WRKSRC}/_build/bindist/ghc-${GHC_VERSION}-${CONFIGURE_TARGET}
 INSTALL_TARGET=		install
 PLIST?=			${.CURDIR}/pkg-plist.hadrian
 
 .  if (empty(PORT_OPTIONS:MDYNAMIC) || empty(PORT_OPTIONS:MPROFILE))
 IGNORE=			Building with Hadrian requires both DYNAMIC and PROFILE for now
 .  endif
 .endif
 
 GHC_ARCH=		${ARCH:S/amd64/x86_64/:C/armv.*/arm/}
 CONFIGURE_TARGET=	${GHC_ARCH}-portbld-${OPSYS:tl}
 BOOT_DIR=		${WRKDIR}/ghc-${BOOT_GHC_VERSION}-boot
 BOOT_INSTALL_DIR=	${WRKDIR}/ghc-boot-install
 BOOT_GHC=		${BOOT_INSTALL_DIR}/bin/ghc-${BOOT_GHC_VERSION}
 
 PLIST_SUB+=		GHC_ARCH=${GHC_ARCH}
 
 .if ${BOOT_GHC_VERSION:C/\..*//g} >= 9
 BOOT_DIR=		${WRKDIR}/ghc-${BOOT_GHC_VERSION}-${CONFIGURE_TARGET}
 .endif
 
 # this is hacky but 92 is special as it is compiled with autotools on i386 and with Hadrian elsewhere
 .if ${SLAVE_PORT} == "yes" && ${PKGNAMESUFFIX} == "92"
 .  ifdef USE_HADRIAN
 PLIST=			${.CURDIR}/../ghc92/pkg-plist.hadrian
 .  else
 PLIST=			${.CURDIR}/../ghc92/pkg-plist
 .  endif
 .endif
 
 # This version of ncurses is needed by bootstrap compiler
 .if empty(PORT_OPTIONS:MBOOT)
 BUILD_DEPENDS+=	${LOCALBASE}/lib/compat/libncursesw.so.8:misc/compat12x
 .endif
 
 .if empty(PORT_OPTIONS:MBOOT)
 DISTFILES+=		ghc-${BOOT_GHC_VERSION}-boot-${ARCH}-freebsd${EXTRACT_SUFX}:boot
 .   if !defined(IGNORE_MISSING_HADRIAN) && defined(USE_HADRIAN)
 DISTFILES+=		hadrian-${GHC_VERSION}-boot.tar.gz:boot
 .   endif
 .endif # MBOOT
 
 .if ${ARCH} == aarch64 || ${ARCH:Marmv*}
 # ghc-8.10.x on arm requires devel/llvm10
 # CONFIGURE_TARGET must to be the same as the llvm triple
 CONFIGURE_TARGET=	${ARCH}-unknown-freebsd${"${ARCH:Maarch64}" != "":?:-gnueabihf}
 CONFIGURE_ARGS+=	--host=${CONFIGURE_TARGET}
 BUILD_DEPENDS+=		llc${LLVM_VERSION}:devel/llvm${LLVM_VERSION}
 RUN_DEPENDS+=		llc${LLVM_VERSION}:devel/llvm${LLVM_VERSION}
 
 # When GHC being compiled and GHC used for bootstrapping support different
 # LLVM versions, we have to pull in both. Luckily, this is relatively rare.
 .  if ${BOOT_LLVM_VERSION} != ${LLVM_VERSION}
 BUILD_DEPENDS+=		llc${BOOT_LLVM_VERSION}:devel/llvm${BOOT_LLVM_VERSION}
 .  endif
 .endif
 
 post-patch:
 #	Generate the build.mk file
 	${RM} -f ${WRKSRC}/mk/build.mk
 .for line in ${BUILD_MK}
 	${ECHO_CMD} ${line} >> ${WRKSRC}/mk/build.mk
 .endfor
 .ifdef USE_HADRIAN
 	${SED} -e 's|%%DYNAMIC%%|${HADRIAN_SETTING_DYNAMIC}|' \
 		-e 's|%%PROFILE%%|${HADRIAN_SETTING_PROFILE}|' \
 		${PATCHDIR}/UserSettings.hs > ${WRKSRC}/hadrian/src/UserSettings.hs
 .endif
 
 pre-configure:
 	# Call the bootstrap script
 	cd ${WRKSRC}/ && ${BOOT_SCRIPT}
 # If we are using bootstrap compiler, configure and install it into ${BOOT_DIR}
 .if empty(PORT_OPTIONS:MBOOT)
 	cd ${BOOT_DIR} && ${CONFIGURE_CMD} ${CONFIGURE_ENV} --prefix=${BOOT_INSTALL_DIR}
 	cd ${BOOT_DIR} && ${SETENVI} ${WRK_ENV} ${GMAKE} PACKAGES='' install
 .endif
 .ifdef USE_HADRIAN
 # Compile Hadrian
 	cd ${WRKSRC}/hadrian/bootstrap && \
 		./bootstrap.py -w ${BOOT_GHC} -s ${DISTDIR}/hadrian-${GHC_VERSION}-boot.tar.gz
 .endif
 
 .ifdef USE_HADRIAN
 pre-install:
 	cd ${INSTALL_WRKSRC} && ${CONFIGURE_CMD} ${CONFIGURE_ENV} --prefix=${PREFIX}
 .endif
 
 post-install:
 .ifdef USE_HADRIAN
 # Hadrian doesn't have --docdir
 	${MV} ${STAGEDIR}${DOCSDIR}-${GHC_VERSION} ${STAGEDIR}${DOCSDIR}
 # These includes are duplicated in lib/ghc-X.Y.Z/lib/<triple>/rts-X.Y.Z/include
 	${RM} -r ${STAGEDIR}${PREFIX}/include/*
 .endif
 	${FIND} ${STAGEDIR}${DOCSDIR} -name .buildinfo -delete
 # For some reason, INSTALL_TARGET=install-strip doesn't cause libraries to be stripped
 # Run strip on them manually
 	${FIND} ${STAGEDIR}${PREFIX}/lib/ghc-${GHC_VERSION} -type f -perm +111 -exec ${STRIP_CMD} {} +
 	${FIND} ${STAGEDIR}${PREFIX}/lib/ghc-${GHC_VERSION} -name '*.so' -exec ${STRIP_CMD} {} +
 	${RM} ${STAGEDIR}${PREFIX}/bin/haddock
 .if ${SLAVE_PORT} == "yes"
 .  ifndef USE_HADRIAN
 	${MV} ${STAGEDIR}${PREFIX}/bin/hsc2hs ${STAGEDIR}${PREFIX}/bin/hsc2hs-ghc-${GHC_VERSION}
 .  endif
 	${FIND} ${STAGEDIR}${PREFIX}/bin -not -type d -not -regex '.*-${GHC_VERSION}' -delete
 	${RM} -r ${STAGEDIR}${PREFIX}/lib/ghc-${GHC_VERSION}/html
 	${RM} -r ${STAGEDIR}${PREFIX}/lib/ghc-${GHC_VERSION}/latex
 .endif
 
 post-install-DOCS-off:
 # Docs for Haskell libraries are generated by Haddock, not sphinx, so we have
 # to pass WITH_HADDOCK=no to the build.mk to skip their generation.
 # However, we don't want to do that because this would cause Haddock itself not
 # to be built.
 	${RM} -r ${STAGEDIR}${DOCSDIR}/html
 
 .if exists(${BOOT_INSTALL_DIR}/bin/runhaskell)
 RUNHASKELL?=	${BOOT_INSTALL_DIR}/bin/runhaskell
 .else
 RUNHASKELL?=	${LOCALBASE}/bin/runhaskell
 .endif
 
 .PHONY: fixup-plist
 fixup-plist:
 	${RUNHASKELL} ${PATCHDIR}/fixup-plist.hs ${.CURDIR}/pkg-plist
 
 # Create a bootstrap compiler tar ball: run this in an interactive poudriere jail
 # Set all OPTIONS to OFF when generating bootstraps
 .PHONY: create-bootstrap
 create-bootstrap:
 .ifndef USE_HADRIAN
 	cd ${WRKSRC} \
 		&& ${ECHO_CMD} "BIN_DIST_NAME=ghc-${GHC_VERSION}-boot" >> mk/build.mk \
 		&& ${ECHO_CMD} "BIN_DIST_TAR=ghc-${GHC_VERSION}-boot.tar" >> mk/build.mk \
 		&& ${ECHO_CMD} "HADDOCK_DOCS=NO" >> mk/build.mk \
 		&& ${SETENVI} ${WRK_ENV} ${GMAKE} binary-dist TAR_COMP=xz \
 		&& ${MV} ${WRKSRC}/ghc-${GHC_VERSION}-boot-${GHC_ARCH}-portbld-freebsd.tar.xz /tmp/ghc-${GHC_VERSION}-boot-${ARCH}-freebsd.tar.xz
 .else
 	cd ${WRKSRC} \
 		&& ${HADRIAN_CMD} binary-dist-xz \
 		&& ${MV} ${WRKSRC}/_build/bindist/ghc-${GHC_VERSION}-${CONFIGURE_TARGET}.tar.xz /tmp/ghc-${GHC_VERSION}-boot-${ARCH}-freebsd.tar.xz
 .endif
 
 	@cd /tmp \
 		&& sha256 ghc-${GHC_VERSION}-boot-${ARCH}-freebsd.tar.xz \
 		&& ${ECHO_CMD} -n "SIZE (ghc-${GHC_VERSION}-boot-${ARCH}-freebsd.tar.xz) = " \
 		&& ${STAT} -f %z ghc-${GHC_VERSION}-boot-${ARCH}-freebsd.tar.xz
 
 # Much like create-bootstrap, just different naming and output format
 # Set DYNAMIC, GMP and PROFILE to ON, and DOCS to OFF when generating Stack bindist
 .PHONY: create-stack-bindist
 create-stack-bindist:
 .ifndef USE_HADRIAN
 	cd ${WRKSRC} \
 		&& ${SETENVI} ${WRK_ENV} ${GMAKE} binary-dist TAR_COMP=xz \
 		&& ${MV} ${WRKSRC}/ghc-${GHC_VERSION}-${GHC_ARCH}-portbld-freebsd.tar.xz /tmp/
 .else
 	cd ${WRKSRC} \
 		&& ${HADRIAN_CMD} binary-dist-xz \
 		&& ${MV} ${WRKSRC}/_build/bindist/ghc-${GHC_VERSION}-${GHC_ARCH}-portbld-freebsd.tar.xz /tmp/
 .endif
 
 	cd /tmp \
 		&& ${ECHO_CMD} "${GHC_VERSION}:" \
 		&& ${ECHO_CMD} "url: \"http://distcache.FreeBSD.org/local-distfiles/arrowd/stack-bindists/ghc-${GHC_VERSION}-${GHC_ARCH}-portbld-freebsd.tar.xz\"" \
 		&& ${ECHO_CMD} -n "content-length: " \
 		&& ${STAT} -f %z ghc-${GHC_VERSION}-${GHC_ARCH}-portbld-freebsd.tar.xz \
 		&& ${ECHO_CMD} -n "sha1: " \
 		&& sha1 -q ghc-${GHC_VERSION}-${GHC_ARCH}-portbld-freebsd.tar.xz \
 		&& ${ECHO_CMD} -n "sha256: " \
 		&& sha256 -q ghc-${GHC_VERSION}-${GHC_ARCH}-portbld-freebsd.tar.xz
 
 .PHONY: create-hadrian-bootstrap
 create-hadrian-bootstrap:
 .if !empty(PORT_OPTIONS:MBOOT)
 	@${ECHO_CMD} "Generating Hadrian bootstrap without bootstrap GHC isn't supported"
 	${FALSE}
 .endif
 	${MAKE} -C ${.CURDIR} patch build-depends USE_PACKAGE_DEPENDS_ONLY=yes IGNORE_MISSING_HADRIAN=yes
 # We always would need to create our own plan.json file and put it into PATCHDIR:
 # Predefined plans use integer-gmp, while we build bootstraps with integer-simple
 # Predefined plans aren't pretty-printed, so we can't easily patch them
 	cd ${WRKSRC}/hadrian/bootstrap && \
 		./bootstrap.py -w ${BOOT_GHC} --deps ${HADRIAN_PLAN} fetch -o /tmp/hadrian-${GHC_VERSION}-boot
 
 	@cd /tmp \
 		&& sha256 hadrian-${GHC_VERSION}-boot.tar.gz \
 		&& ${ECHO_CMD} -n "SIZE (hadrian-${GHC_VERSION}-boot.tar.gz) = " \
 		&& ${STAT} -f %z hadrian-${GHC_VERSION}-boot.tar.gz
 	@${ECHO_CMD}
 	@${ECHO_CMD} "Remember to check that hadrian bootstrap builds fine by running \"make check-hadrian-bootstrap\""
 
 .PHONY: check-hadrian-bootstrap
 check-hadrian-bootstrap:
 .if !empty(PORT_OPTIONS:MBOOT)
 	@${ECHO_CMD} "Generating Hadrian bootstrap without bootstrap GHC isn't supported"
 	${FALSE}
 .endif
 # Install bootstrap GHC
 	${MAKE_CMD} -C ${.CURDIR} pre-configure
 # Compile Hadrian
 	cd ${WRKSRC}/hadrian/bootstrap && \
 		./bootstrap.py -w ${BOOT_GHC} -s /tmp/hadrian-${GHC_VERSION}-boot.tar.gz
 
 .include <bsd.port.post.mk>
diff --git a/lang/ghc/files/patch-rts_ARMOutlineAtomicsSymbols.h b/lang/ghc/files/patch-rts_ARMOutlineAtomicsSymbols.h
new file mode 100644
index 000000000000..2e956831a7f6
--- /dev/null
+++ b/lang/ghc/files/patch-rts_ARMOutlineAtomicsSymbols.h
@@ -0,0 +1,822 @@
+https://gitlab.haskell.org/ghc/ghc/-/merge_requests/12227
+
+https://gitlab.haskell.org/ghc/ghc/-/issues/24580
+
+--- rts/ARMOutlineAtomicsSymbols.h.orig	2024-01-08 13:24:59 UTC
++++ rts/ARMOutlineAtomicsSymbols.h
+@@ -10,583 +10,583 @@
+ #include <stdint.h>
+ #include <stdatomic.h>
+ 
+-uint8_t ghc___aarch64_cas1_relax(uint8_t old, uint8_t new, uint8_t* p);
+-uint8_t ghc___aarch64_cas1_relax(uint8_t old, uint8_t new, uint8_t* p) {
++uint8_t ghc___aarch64_cas1_relax(uint8_t old, uint8_t new, _Atomic uint8_t* p);
++uint8_t ghc___aarch64_cas1_relax(uint8_t old, uint8_t new, _Atomic uint8_t* p) {
+   atomic_compare_exchange_strong_explicit(p, &old, new, memory_order_relaxed, memory_order_relaxed); return old;
+ }
+ 
+-uint8_t ghc___aarch64_cas1_acq(uint8_t old, uint8_t new, uint8_t* p);
+-uint8_t ghc___aarch64_cas1_acq(uint8_t old, uint8_t new, uint8_t* p) {
++uint8_t ghc___aarch64_cas1_acq(uint8_t old, uint8_t new, _Atomic uint8_t* p);
++uint8_t ghc___aarch64_cas1_acq(uint8_t old, uint8_t new, _Atomic uint8_t* p) {
+   atomic_compare_exchange_strong_explicit(p, &old, new, memory_order_acquire, memory_order_acquire); return old;
+ }
+ 
+-uint8_t ghc___aarch64_cas1_acq_rel(uint8_t old, uint8_t new, uint8_t* p);
+-uint8_t ghc___aarch64_cas1_acq_rel(uint8_t old, uint8_t new, uint8_t* p) {
++uint8_t ghc___aarch64_cas1_acq_rel(uint8_t old, uint8_t new, _Atomic uint8_t* p);
++uint8_t ghc___aarch64_cas1_acq_rel(uint8_t old, uint8_t new, _Atomic uint8_t* p) {
+   atomic_compare_exchange_strong_explicit(p, &old, new, memory_order_acq_rel, memory_order_acquire); return old;
+ }
+ 
+-uint8_t ghc___aarch64_cas1_sync(uint8_t old, uint8_t new, uint8_t* p);
+-uint8_t ghc___aarch64_cas1_sync(uint8_t old, uint8_t new, uint8_t* p) {
++uint8_t ghc___aarch64_cas1_sync(uint8_t old, uint8_t new, _Atomic uint8_t* p);
++uint8_t ghc___aarch64_cas1_sync(uint8_t old, uint8_t new, _Atomic uint8_t* p) {
+   atomic_compare_exchange_strong_explicit(p, &old, new, memory_order_seq_cst, memory_order_seq_cst); return old;
+ }
+ 
+-uint16_t ghc___aarch64_cas2_relax(uint16_t old, uint16_t new, uint16_t* p);
+-uint16_t ghc___aarch64_cas2_relax(uint16_t old, uint16_t new, uint16_t* p) {
++uint16_t ghc___aarch64_cas2_relax(uint16_t old, uint16_t new, _Atomic uint16_t* p);
++uint16_t ghc___aarch64_cas2_relax(uint16_t old, uint16_t new, _Atomic uint16_t* p) {
+   atomic_compare_exchange_strong_explicit(p, &old, new, memory_order_relaxed, memory_order_relaxed); return old;
+ }
+ 
+-uint16_t ghc___aarch64_cas2_acq(uint16_t old, uint16_t new, uint16_t* p);
+-uint16_t ghc___aarch64_cas2_acq(uint16_t old, uint16_t new, uint16_t* p) {
++uint16_t ghc___aarch64_cas2_acq(uint16_t old, uint16_t new, _Atomic uint16_t* p);
++uint16_t ghc___aarch64_cas2_acq(uint16_t old, uint16_t new, _Atomic uint16_t* p) {
+   atomic_compare_exchange_strong_explicit(p, &old, new, memory_order_acquire, memory_order_acquire); return old;
+ }
+ 
+-uint16_t ghc___aarch64_cas2_acq_rel(uint16_t old, uint16_t new, uint16_t* p);
+-uint16_t ghc___aarch64_cas2_acq_rel(uint16_t old, uint16_t new, uint16_t* p) {
++uint16_t ghc___aarch64_cas2_acq_rel(uint16_t old, uint16_t new, _Atomic uint16_t* p);
++uint16_t ghc___aarch64_cas2_acq_rel(uint16_t old, uint16_t new, _Atomic uint16_t* p) {
+   atomic_compare_exchange_strong_explicit(p, &old, new, memory_order_acq_rel, memory_order_acquire); return old;
+ }
+ 
+-uint16_t ghc___aarch64_cas2_sync(uint16_t old, uint16_t new, uint16_t* p);
+-uint16_t ghc___aarch64_cas2_sync(uint16_t old, uint16_t new, uint16_t* p) {
++uint16_t ghc___aarch64_cas2_sync(uint16_t old, uint16_t new, _Atomic uint16_t* p);
++uint16_t ghc___aarch64_cas2_sync(uint16_t old, uint16_t new, _Atomic uint16_t* p) {
+   atomic_compare_exchange_strong_explicit(p, &old, new, memory_order_seq_cst, memory_order_seq_cst); return old;
+ }
+ 
+-uint32_t ghc___aarch64_cas4_relax(uint32_t old, uint32_t new, uint32_t* p);
+-uint32_t ghc___aarch64_cas4_relax(uint32_t old, uint32_t new, uint32_t* p) {
++uint32_t ghc___aarch64_cas4_relax(uint32_t old, uint32_t new, _Atomic uint32_t* p);
++uint32_t ghc___aarch64_cas4_relax(uint32_t old, uint32_t new, _Atomic uint32_t* p) {
+   atomic_compare_exchange_strong_explicit(p, &old, new, memory_order_relaxed, memory_order_relaxed); return old;
+ }
+ 
+-uint32_t ghc___aarch64_cas4_acq(uint32_t old, uint32_t new, uint32_t* p);
+-uint32_t ghc___aarch64_cas4_acq(uint32_t old, uint32_t new, uint32_t* p) {
++uint32_t ghc___aarch64_cas4_acq(uint32_t old, uint32_t new, _Atomic uint32_t* p);
++uint32_t ghc___aarch64_cas4_acq(uint32_t old, uint32_t new, _Atomic uint32_t* p) {
+   atomic_compare_exchange_strong_explicit(p, &old, new, memory_order_acquire, memory_order_acquire); return old;
+ }
+ 
+-uint32_t ghc___aarch64_cas4_acq_rel(uint32_t old, uint32_t new, uint32_t* p);
+-uint32_t ghc___aarch64_cas4_acq_rel(uint32_t old, uint32_t new, uint32_t* p) {
++uint32_t ghc___aarch64_cas4_acq_rel(uint32_t old, uint32_t new, _Atomic uint32_t* p);
++uint32_t ghc___aarch64_cas4_acq_rel(uint32_t old, uint32_t new, _Atomic uint32_t* p) {
+   atomic_compare_exchange_strong_explicit(p, &old, new, memory_order_acq_rel, memory_order_acquire); return old;
+ }
+ 
+-uint32_t ghc___aarch64_cas4_sync(uint32_t old, uint32_t new, uint32_t* p);
+-uint32_t ghc___aarch64_cas4_sync(uint32_t old, uint32_t new, uint32_t* p) {
++uint32_t ghc___aarch64_cas4_sync(uint32_t old, uint32_t new, _Atomic uint32_t* p);
++uint32_t ghc___aarch64_cas4_sync(uint32_t old, uint32_t new, _Atomic uint32_t* p) {
+   atomic_compare_exchange_strong_explicit(p, &old, new, memory_order_seq_cst, memory_order_seq_cst); return old;
+ }
+ 
+-uint64_t ghc___aarch64_cas8_relax(uint64_t old, uint64_t new, uint64_t* p);
+-uint64_t ghc___aarch64_cas8_relax(uint64_t old, uint64_t new, uint64_t* p) {
++uint64_t ghc___aarch64_cas8_relax(uint64_t old, uint64_t new, _Atomic uint64_t* p);
++uint64_t ghc___aarch64_cas8_relax(uint64_t old, uint64_t new, _Atomic uint64_t* p) {
+   atomic_compare_exchange_strong_explicit(p, &old, new, memory_order_relaxed, memory_order_relaxed); return old;
+ }
+ 
+-uint64_t ghc___aarch64_cas8_acq(uint64_t old, uint64_t new, uint64_t* p);
+-uint64_t ghc___aarch64_cas8_acq(uint64_t old, uint64_t new, uint64_t* p) {
++uint64_t ghc___aarch64_cas8_acq(uint64_t old, uint64_t new, _Atomic uint64_t* p);
++uint64_t ghc___aarch64_cas8_acq(uint64_t old, uint64_t new, _Atomic uint64_t* p) {
+   atomic_compare_exchange_strong_explicit(p, &old, new, memory_order_acquire, memory_order_acquire); return old;
+ }
+ 
+-uint64_t ghc___aarch64_cas8_acq_rel(uint64_t old, uint64_t new, uint64_t* p);
+-uint64_t ghc___aarch64_cas8_acq_rel(uint64_t old, uint64_t new, uint64_t* p) {
++uint64_t ghc___aarch64_cas8_acq_rel(uint64_t old, uint64_t new, _Atomic uint64_t* p);
++uint64_t ghc___aarch64_cas8_acq_rel(uint64_t old, uint64_t new, _Atomic uint64_t* p) {
+   atomic_compare_exchange_strong_explicit(p, &old, new, memory_order_acq_rel, memory_order_acquire); return old;
+ }
+ 
+-uint64_t ghc___aarch64_cas8_sync(uint64_t old, uint64_t new, uint64_t* p);
+-uint64_t ghc___aarch64_cas8_sync(uint64_t old, uint64_t new, uint64_t* p) {
++uint64_t ghc___aarch64_cas8_sync(uint64_t old, uint64_t new, _Atomic uint64_t* p);
++uint64_t ghc___aarch64_cas8_sync(uint64_t old, uint64_t new, _Atomic uint64_t* p) {
+   atomic_compare_exchange_strong_explicit(p, &old, new, memory_order_seq_cst, memory_order_seq_cst); return old;
+ }
+ 
+-uint8_t ghc___aarch64_swp1_relax(uint8_t v, uint8_t* p);
+-uint8_t ghc___aarch64_swp1_relax(uint8_t v, uint8_t* p) {
++uint8_t ghc___aarch64_swp1_relax(uint8_t v, _Atomic uint8_t* p);
++uint8_t ghc___aarch64_swp1_relax(uint8_t v, _Atomic uint8_t* p) {
+   return atomic_exchange_explicit(p, v, memory_order_relaxed);
+ }
+ 
+-uint8_t ghc___aarch64_swp1_acq(uint8_t v, uint8_t* p);
+-uint8_t ghc___aarch64_swp1_acq(uint8_t v, uint8_t* p) {
++uint8_t ghc___aarch64_swp1_acq(uint8_t v, _Atomic uint8_t* p);
++uint8_t ghc___aarch64_swp1_acq(uint8_t v, _Atomic uint8_t* p) {
+   return atomic_exchange_explicit(p, v, memory_order_acquire);
+ }
+ 
+-uint8_t ghc___aarch64_swp1_rel(uint8_t v, uint8_t* p);
+-uint8_t ghc___aarch64_swp1_rel(uint8_t v, uint8_t* p) {
++uint8_t ghc___aarch64_swp1_rel(uint8_t v, _Atomic uint8_t* p);
++uint8_t ghc___aarch64_swp1_rel(uint8_t v, _Atomic uint8_t* p) {
+   return atomic_exchange_explicit(p, v, memory_order_release);
+ }
+ 
+-uint8_t ghc___aarch64_swp1_acq_rel(uint8_t v, uint8_t* p);
+-uint8_t ghc___aarch64_swp1_acq_rel(uint8_t v, uint8_t* p) {
++uint8_t ghc___aarch64_swp1_acq_rel(uint8_t v, _Atomic uint8_t* p);
++uint8_t ghc___aarch64_swp1_acq_rel(uint8_t v, _Atomic uint8_t* p) {
+   return atomic_exchange_explicit(p, v, memory_order_acq_rel);
+ }
+ 
+-uint8_t ghc___aarch64_swp1_sync(uint8_t v, uint8_t* p);
+-uint8_t ghc___aarch64_swp1_sync(uint8_t v, uint8_t* p) {
++uint8_t ghc___aarch64_swp1_sync(uint8_t v, _Atomic uint8_t* p);
++uint8_t ghc___aarch64_swp1_sync(uint8_t v, _Atomic uint8_t* p) {
+   return atomic_exchange_explicit(p, v, memory_order_seq_cst);
+ }
+ 
+-uint16_t ghc___aarch64_swp2_relax(uint16_t v, uint16_t* p);
+-uint16_t ghc___aarch64_swp2_relax(uint16_t v, uint16_t* p) {
++uint16_t ghc___aarch64_swp2_relax(uint16_t v, _Atomic uint16_t* p);
++uint16_t ghc___aarch64_swp2_relax(uint16_t v, _Atomic uint16_t* p) {
+   return atomic_exchange_explicit(p, v, memory_order_relaxed);
+ }
+ 
+-uint16_t ghc___aarch64_swp2_acq(uint16_t v, uint16_t* p);
+-uint16_t ghc___aarch64_swp2_acq(uint16_t v, uint16_t* p) {
++uint16_t ghc___aarch64_swp2_acq(uint16_t v, _Atomic uint16_t* p);
++uint16_t ghc___aarch64_swp2_acq(uint16_t v, _Atomic uint16_t* p) {
+   return atomic_exchange_explicit(p, v, memory_order_acquire);
+ }
+ 
+-uint16_t ghc___aarch64_swp2_rel(uint16_t v, uint16_t* p);
+-uint16_t ghc___aarch64_swp2_rel(uint16_t v, uint16_t* p) {
++uint16_t ghc___aarch64_swp2_rel(uint16_t v, _Atomic uint16_t* p);
++uint16_t ghc___aarch64_swp2_rel(uint16_t v, _Atomic uint16_t* p) {
+   return atomic_exchange_explicit(p, v, memory_order_release);
+ }
+ 
+-uint16_t ghc___aarch64_swp2_acq_rel(uint16_t v, uint16_t* p);
+-uint16_t ghc___aarch64_swp2_acq_rel(uint16_t v, uint16_t* p) {
++uint16_t ghc___aarch64_swp2_acq_rel(uint16_t v, _Atomic uint16_t* p);
++uint16_t ghc___aarch64_swp2_acq_rel(uint16_t v, _Atomic uint16_t* p) {
+   return atomic_exchange_explicit(p, v, memory_order_acq_rel);
+ }
+ 
+-uint16_t ghc___aarch64_swp2_sync(uint16_t v, uint16_t* p);
+-uint16_t ghc___aarch64_swp2_sync(uint16_t v, uint16_t* p) {
++uint16_t ghc___aarch64_swp2_sync(uint16_t v, _Atomic uint16_t* p);
++uint16_t ghc___aarch64_swp2_sync(uint16_t v, _Atomic uint16_t* p) {
+   return atomic_exchange_explicit(p, v, memory_order_seq_cst);
+ }
+ 
+-uint32_t ghc___aarch64_swp4_relax(uint32_t v, uint32_t* p);
+-uint32_t ghc___aarch64_swp4_relax(uint32_t v, uint32_t* p) {
++uint32_t ghc___aarch64_swp4_relax(uint32_t v, _Atomic uint32_t* p);
++uint32_t ghc___aarch64_swp4_relax(uint32_t v, _Atomic uint32_t* p) {
+   return atomic_exchange_explicit(p, v, memory_order_relaxed);
+ }
+ 
+-uint32_t ghc___aarch64_swp4_acq(uint32_t v, uint32_t* p);
+-uint32_t ghc___aarch64_swp4_acq(uint32_t v, uint32_t* p) {
++uint32_t ghc___aarch64_swp4_acq(uint32_t v, _Atomic uint32_t* p);
++uint32_t ghc___aarch64_swp4_acq(uint32_t v, _Atomic uint32_t* p) {
+   return atomic_exchange_explicit(p, v, memory_order_acquire);
+ }
+ 
+-uint32_t ghc___aarch64_swp4_rel(uint32_t v, uint32_t* p);
+-uint32_t ghc___aarch64_swp4_rel(uint32_t v, uint32_t* p) {
++uint32_t ghc___aarch64_swp4_rel(uint32_t v, _Atomic uint32_t* p);
++uint32_t ghc___aarch64_swp4_rel(uint32_t v, _Atomic uint32_t* p) {
+   return atomic_exchange_explicit(p, v, memory_order_release);
+ }
+ 
+-uint32_t ghc___aarch64_swp4_acq_rel(uint32_t v, uint32_t* p);
+-uint32_t ghc___aarch64_swp4_acq_rel(uint32_t v, uint32_t* p) {
++uint32_t ghc___aarch64_swp4_acq_rel(uint32_t v, _Atomic uint32_t* p);
++uint32_t ghc___aarch64_swp4_acq_rel(uint32_t v, _Atomic uint32_t* p) {
+   return atomic_exchange_explicit(p, v, memory_order_acq_rel);
+ }
+ 
+-uint32_t ghc___aarch64_swp4_sync(uint32_t v, uint32_t* p);
+-uint32_t ghc___aarch64_swp4_sync(uint32_t v, uint32_t* p) {
++uint32_t ghc___aarch64_swp4_sync(uint32_t v, _Atomic uint32_t* p);
++uint32_t ghc___aarch64_swp4_sync(uint32_t v, _Atomic uint32_t* p) {
+   return atomic_exchange_explicit(p, v, memory_order_seq_cst);
+ }
+ 
+-uint64_t ghc___aarch64_swp8_relax(uint64_t v, uint64_t* p);
+-uint64_t ghc___aarch64_swp8_relax(uint64_t v, uint64_t* p) {
++uint64_t ghc___aarch64_swp8_relax(uint64_t v, _Atomic uint64_t* p);
++uint64_t ghc___aarch64_swp8_relax(uint64_t v, _Atomic uint64_t* p) {
+   return atomic_exchange_explicit(p, v, memory_order_relaxed);
+ }
+ 
+-uint64_t ghc___aarch64_swp8_acq(uint64_t v, uint64_t* p);
+-uint64_t ghc___aarch64_swp8_acq(uint64_t v, uint64_t* p) {
++uint64_t ghc___aarch64_swp8_acq(uint64_t v, _Atomic uint64_t* p);
++uint64_t ghc___aarch64_swp8_acq(uint64_t v, _Atomic uint64_t* p) {
+   return atomic_exchange_explicit(p, v, memory_order_acquire);
+ }
+ 
+-uint64_t ghc___aarch64_swp8_rel(uint64_t v, uint64_t* p);
+-uint64_t ghc___aarch64_swp8_rel(uint64_t v, uint64_t* p) {
++uint64_t ghc___aarch64_swp8_rel(uint64_t v, _Atomic uint64_t* p);
++uint64_t ghc___aarch64_swp8_rel(uint64_t v, _Atomic uint64_t* p) {
+   return atomic_exchange_explicit(p, v, memory_order_release);
+ }
+ 
+-uint64_t ghc___aarch64_swp8_acq_rel(uint64_t v, uint64_t* p);
+-uint64_t ghc___aarch64_swp8_acq_rel(uint64_t v, uint64_t* p) {
++uint64_t ghc___aarch64_swp8_acq_rel(uint64_t v, _Atomic uint64_t* p);
++uint64_t ghc___aarch64_swp8_acq_rel(uint64_t v, _Atomic uint64_t* p) {
+   return atomic_exchange_explicit(p, v, memory_order_acq_rel);
+ }
+ 
+-uint64_t ghc___aarch64_swp8_sync(uint64_t v, uint64_t* p);
+-uint64_t ghc___aarch64_swp8_sync(uint64_t v, uint64_t* p) {
++uint64_t ghc___aarch64_swp8_sync(uint64_t v, _Atomic uint64_t* p);
++uint64_t ghc___aarch64_swp8_sync(uint64_t v, _Atomic uint64_t* p) {
+   return atomic_exchange_explicit(p, v, memory_order_seq_cst);
+ }
+ 
+-uint8_t ghc___aarch64_ldadd1_relax(uint8_t v, uint8_t* p);
+-uint8_t ghc___aarch64_ldadd1_relax(uint8_t v, uint8_t* p) {
++uint8_t ghc___aarch64_ldadd1_relax(uint8_t v, _Atomic uint8_t* p);
++uint8_t ghc___aarch64_ldadd1_relax(uint8_t v, _Atomic uint8_t* p) {
+   return atomic_fetch_add_explicit(p, v, memory_order_relaxed);
+ }
+ 
+-uint8_t ghc___aarch64_ldadd1_acq(uint8_t v, uint8_t* p);
+-uint8_t ghc___aarch64_ldadd1_acq(uint8_t v, uint8_t* p) {
++uint8_t ghc___aarch64_ldadd1_acq(uint8_t v, _Atomic uint8_t* p);
++uint8_t ghc___aarch64_ldadd1_acq(uint8_t v, _Atomic uint8_t* p) {
+   return atomic_fetch_add_explicit(p, v, memory_order_acquire);
+ }
+ 
+-uint8_t ghc___aarch64_ldadd1_rel(uint8_t v, uint8_t* p);
+-uint8_t ghc___aarch64_ldadd1_rel(uint8_t v, uint8_t* p) {
++uint8_t ghc___aarch64_ldadd1_rel(uint8_t v, _Atomic uint8_t* p);
++uint8_t ghc___aarch64_ldadd1_rel(uint8_t v, _Atomic uint8_t* p) {
+   return atomic_fetch_add_explicit(p, v, memory_order_release);
+ }
+ 
+-uint8_t ghc___aarch64_ldadd1_acq_rel(uint8_t v, uint8_t* p);
+-uint8_t ghc___aarch64_ldadd1_acq_rel(uint8_t v, uint8_t* p) {
++uint8_t ghc___aarch64_ldadd1_acq_rel(uint8_t v, _Atomic uint8_t* p);
++uint8_t ghc___aarch64_ldadd1_acq_rel(uint8_t v, _Atomic uint8_t* p) {
+   return atomic_fetch_add_explicit(p, v, memory_order_acq_rel);
+ }
+ 
+-uint8_t ghc___aarch64_ldadd1_sync(uint8_t v, uint8_t* p);
+-uint8_t ghc___aarch64_ldadd1_sync(uint8_t v, uint8_t* p) {
++uint8_t ghc___aarch64_ldadd1_sync(uint8_t v, _Atomic uint8_t* p);
++uint8_t ghc___aarch64_ldadd1_sync(uint8_t v, _Atomic uint8_t* p) {
+   return atomic_fetch_add_explicit(p, v, memory_order_seq_cst);
+ }
+ 
+-uint16_t ghc___aarch64_ldadd2_relax(uint16_t v, uint16_t* p);
+-uint16_t ghc___aarch64_ldadd2_relax(uint16_t v, uint16_t* p) {
++uint16_t ghc___aarch64_ldadd2_relax(uint16_t v, _Atomic uint16_t* p);
++uint16_t ghc___aarch64_ldadd2_relax(uint16_t v, _Atomic uint16_t* p) {
+   return atomic_fetch_add_explicit(p, v, memory_order_relaxed);
+ }
+ 
+-uint16_t ghc___aarch64_ldadd2_acq(uint16_t v, uint16_t* p);
+-uint16_t ghc___aarch64_ldadd2_acq(uint16_t v, uint16_t* p) {
++uint16_t ghc___aarch64_ldadd2_acq(uint16_t v, _Atomic uint16_t* p);
++uint16_t ghc___aarch64_ldadd2_acq(uint16_t v, _Atomic uint16_t* p) {
+   return atomic_fetch_add_explicit(p, v, memory_order_acquire);
+ }
+ 
+-uint16_t ghc___aarch64_ldadd2_rel(uint16_t v, uint16_t* p);
+-uint16_t ghc___aarch64_ldadd2_rel(uint16_t v, uint16_t* p) {
++uint16_t ghc___aarch64_ldadd2_rel(uint16_t v, _Atomic uint16_t* p);
++uint16_t ghc___aarch64_ldadd2_rel(uint16_t v, _Atomic uint16_t* p) {
+   return atomic_fetch_add_explicit(p, v, memory_order_release);
+ }
+ 
+-uint16_t ghc___aarch64_ldadd2_acq_rel(uint16_t v, uint16_t* p);
+-uint16_t ghc___aarch64_ldadd2_acq_rel(uint16_t v, uint16_t* p) {
++uint16_t ghc___aarch64_ldadd2_acq_rel(uint16_t v, _Atomic uint16_t* p);
++uint16_t ghc___aarch64_ldadd2_acq_rel(uint16_t v, _Atomic uint16_t* p) {
+   return atomic_fetch_add_explicit(p, v, memory_order_acq_rel);
+ }
+ 
+-uint16_t ghc___aarch64_ldadd2_sync(uint16_t v, uint16_t* p);
+-uint16_t ghc___aarch64_ldadd2_sync(uint16_t v, uint16_t* p) {
++uint16_t ghc___aarch64_ldadd2_sync(uint16_t v, _Atomic uint16_t* p);
++uint16_t ghc___aarch64_ldadd2_sync(uint16_t v, _Atomic uint16_t* p) {
+   return atomic_fetch_add_explicit(p, v, memory_order_seq_cst);
+ }
+ 
+-uint32_t ghc___aarch64_ldadd4_relax(uint32_t v, uint32_t* p);
+-uint32_t ghc___aarch64_ldadd4_relax(uint32_t v, uint32_t* p) {
++uint32_t ghc___aarch64_ldadd4_relax(uint32_t v, _Atomic uint32_t* p);
++uint32_t ghc___aarch64_ldadd4_relax(uint32_t v, _Atomic uint32_t* p) {
+   return atomic_fetch_add_explicit(p, v, memory_order_relaxed);
+ }
+ 
+-uint32_t ghc___aarch64_ldadd4_acq(uint32_t v, uint32_t* p);
+-uint32_t ghc___aarch64_ldadd4_acq(uint32_t v, uint32_t* p) {
++uint32_t ghc___aarch64_ldadd4_acq(uint32_t v, _Atomic uint32_t* p);
++uint32_t ghc___aarch64_ldadd4_acq(uint32_t v, _Atomic uint32_t* p) {
+   return atomic_fetch_add_explicit(p, v, memory_order_acquire);
+ }
+ 
+-uint32_t ghc___aarch64_ldadd4_rel(uint32_t v, uint32_t* p);
+-uint32_t ghc___aarch64_ldadd4_rel(uint32_t v, uint32_t* p) {
++uint32_t ghc___aarch64_ldadd4_rel(uint32_t v, _Atomic uint32_t* p);
++uint32_t ghc___aarch64_ldadd4_rel(uint32_t v, _Atomic uint32_t* p) {
+   return atomic_fetch_add_explicit(p, v, memory_order_release);
+ }
+ 
+-uint32_t ghc___aarch64_ldadd4_acq_rel(uint32_t v, uint32_t* p);
+-uint32_t ghc___aarch64_ldadd4_acq_rel(uint32_t v, uint32_t* p) {
++uint32_t ghc___aarch64_ldadd4_acq_rel(uint32_t v, _Atomic uint32_t* p);
++uint32_t ghc___aarch64_ldadd4_acq_rel(uint32_t v, _Atomic uint32_t* p) {
+   return atomic_fetch_add_explicit(p, v, memory_order_acq_rel);
+ }
+ 
+-uint32_t ghc___aarch64_ldadd4_sync(uint32_t v, uint32_t* p);
+-uint32_t ghc___aarch64_ldadd4_sync(uint32_t v, uint32_t* p) {
++uint32_t ghc___aarch64_ldadd4_sync(uint32_t v, _Atomic uint32_t* p);
++uint32_t ghc___aarch64_ldadd4_sync(uint32_t v, _Atomic uint32_t* p) {
+   return atomic_fetch_add_explicit(p, v, memory_order_seq_cst);
+ }
+ 
+-uint64_t ghc___aarch64_ldadd8_relax(uint64_t v, uint64_t* p);
+-uint64_t ghc___aarch64_ldadd8_relax(uint64_t v, uint64_t* p) {
++uint64_t ghc___aarch64_ldadd8_relax(uint64_t v, _Atomic uint64_t* p);
++uint64_t ghc___aarch64_ldadd8_relax(uint64_t v, _Atomic uint64_t* p) {
+   return atomic_fetch_add_explicit(p, v, memory_order_relaxed);
+ }
+ 
+-uint64_t ghc___aarch64_ldadd8_acq(uint64_t v, uint64_t* p);
+-uint64_t ghc___aarch64_ldadd8_acq(uint64_t v, uint64_t* p) {
++uint64_t ghc___aarch64_ldadd8_acq(uint64_t v, _Atomic uint64_t* p);
++uint64_t ghc___aarch64_ldadd8_acq(uint64_t v, _Atomic uint64_t* p) {
+   return atomic_fetch_add_explicit(p, v, memory_order_acquire);
+ }
+ 
+-uint64_t ghc___aarch64_ldadd8_rel(uint64_t v, uint64_t* p);
+-uint64_t ghc___aarch64_ldadd8_rel(uint64_t v, uint64_t* p) {
++uint64_t ghc___aarch64_ldadd8_rel(uint64_t v, _Atomic uint64_t* p);
++uint64_t ghc___aarch64_ldadd8_rel(uint64_t v, _Atomic uint64_t* p) {
+   return atomic_fetch_add_explicit(p, v, memory_order_release);
+ }
+ 
+-uint64_t ghc___aarch64_ldadd8_acq_rel(uint64_t v, uint64_t* p);
+-uint64_t ghc___aarch64_ldadd8_acq_rel(uint64_t v, uint64_t* p) {
++uint64_t ghc___aarch64_ldadd8_acq_rel(uint64_t v, _Atomic uint64_t* p);
++uint64_t ghc___aarch64_ldadd8_acq_rel(uint64_t v, _Atomic uint64_t* p) {
+   return atomic_fetch_add_explicit(p, v, memory_order_acq_rel);
+ }
+ 
+-uint64_t ghc___aarch64_ldadd8_sync(uint64_t v, uint64_t* p);
+-uint64_t ghc___aarch64_ldadd8_sync(uint64_t v, uint64_t* p) {
++uint64_t ghc___aarch64_ldadd8_sync(uint64_t v, _Atomic uint64_t* p);
++uint64_t ghc___aarch64_ldadd8_sync(uint64_t v, _Atomic uint64_t* p) {
+   return atomic_fetch_add_explicit(p, v, memory_order_seq_cst);
+ }
+ 
+-uint8_t ghc___aarch64_ldclr1_relax(uint8_t v, uint8_t* p);
+-uint8_t ghc___aarch64_ldclr1_relax(uint8_t v, uint8_t* p) {
++uint8_t ghc___aarch64_ldclr1_relax(uint8_t v, _Atomic uint8_t* p);
++uint8_t ghc___aarch64_ldclr1_relax(uint8_t v, _Atomic uint8_t* p) {
+   return atomic_fetch_and_explicit(p, v, memory_order_relaxed);
+ }
+ 
+-uint8_t ghc___aarch64_ldclr1_acq(uint8_t v, uint8_t* p);
+-uint8_t ghc___aarch64_ldclr1_acq(uint8_t v, uint8_t* p) {
++uint8_t ghc___aarch64_ldclr1_acq(uint8_t v, _Atomic uint8_t* p);
++uint8_t ghc___aarch64_ldclr1_acq(uint8_t v, _Atomic uint8_t* p) {
+   return atomic_fetch_and_explicit(p, v, memory_order_acquire);
+ }
+ 
+-uint8_t ghc___aarch64_ldclr1_rel(uint8_t v, uint8_t* p);
+-uint8_t ghc___aarch64_ldclr1_rel(uint8_t v, uint8_t* p) {
++uint8_t ghc___aarch64_ldclr1_rel(uint8_t v, _Atomic uint8_t* p);
++uint8_t ghc___aarch64_ldclr1_rel(uint8_t v, _Atomic uint8_t* p) {
+   return atomic_fetch_and_explicit(p, v, memory_order_release);
+ }
+ 
+-uint8_t ghc___aarch64_ldclr1_acq_rel(uint8_t v, uint8_t* p);
+-uint8_t ghc___aarch64_ldclr1_acq_rel(uint8_t v, uint8_t* p) {
++uint8_t ghc___aarch64_ldclr1_acq_rel(uint8_t v, _Atomic uint8_t* p);
++uint8_t ghc___aarch64_ldclr1_acq_rel(uint8_t v, _Atomic uint8_t* p) {
+   return atomic_fetch_and_explicit(p, v, memory_order_acq_rel);
+ }
+ 
+-uint8_t ghc___aarch64_ldclr1_sync(uint8_t v, uint8_t* p);
+-uint8_t ghc___aarch64_ldclr1_sync(uint8_t v, uint8_t* p) {
++uint8_t ghc___aarch64_ldclr1_sync(uint8_t v, _Atomic uint8_t* p);
++uint8_t ghc___aarch64_ldclr1_sync(uint8_t v, _Atomic uint8_t* p) {
+   return atomic_fetch_and_explicit(p, v, memory_order_seq_cst);
+ }
+ 
+-uint16_t ghc___aarch64_ldclr2_relax(uint16_t v, uint16_t* p);
+-uint16_t ghc___aarch64_ldclr2_relax(uint16_t v, uint16_t* p) {
++uint16_t ghc___aarch64_ldclr2_relax(uint16_t v, _Atomic uint16_t* p);
++uint16_t ghc___aarch64_ldclr2_relax(uint16_t v, _Atomic uint16_t* p) {
+   return atomic_fetch_and_explicit(p, v, memory_order_relaxed);
+ }
+ 
+-uint16_t ghc___aarch64_ldclr2_acq(uint16_t v, uint16_t* p);
+-uint16_t ghc___aarch64_ldclr2_acq(uint16_t v, uint16_t* p) {
++uint16_t ghc___aarch64_ldclr2_acq(uint16_t v, _Atomic uint16_t* p);
++uint16_t ghc___aarch64_ldclr2_acq(uint16_t v, _Atomic uint16_t* p) {
+   return atomic_fetch_and_explicit(p, v, memory_order_acquire);
+ }
+ 
+-uint16_t ghc___aarch64_ldclr2_rel(uint16_t v, uint16_t* p);
+-uint16_t ghc___aarch64_ldclr2_rel(uint16_t v, uint16_t* p) {
++uint16_t ghc___aarch64_ldclr2_rel(uint16_t v, _Atomic uint16_t* p);
++uint16_t ghc___aarch64_ldclr2_rel(uint16_t v, _Atomic uint16_t* p) {
+   return atomic_fetch_and_explicit(p, v, memory_order_release);
+ }
+ 
+-uint16_t ghc___aarch64_ldclr2_acq_rel(uint16_t v, uint16_t* p);
+-uint16_t ghc___aarch64_ldclr2_acq_rel(uint16_t v, uint16_t* p) {
++uint16_t ghc___aarch64_ldclr2_acq_rel(uint16_t v, _Atomic uint16_t* p);
++uint16_t ghc___aarch64_ldclr2_acq_rel(uint16_t v, _Atomic uint16_t* p) {
+   return atomic_fetch_and_explicit(p, v, memory_order_acq_rel);
+ }
+ 
+-uint16_t ghc___aarch64_ldclr2_sync(uint16_t v, uint16_t* p);
+-uint16_t ghc___aarch64_ldclr2_sync(uint16_t v, uint16_t* p) {
++uint16_t ghc___aarch64_ldclr2_sync(uint16_t v, _Atomic uint16_t* p);
++uint16_t ghc___aarch64_ldclr2_sync(uint16_t v, _Atomic uint16_t* p) {
+   return atomic_fetch_and_explicit(p, v, memory_order_seq_cst);
+ }
+ 
+-uint32_t ghc___aarch64_ldclr4_relax(uint32_t v, uint32_t* p);
+-uint32_t ghc___aarch64_ldclr4_relax(uint32_t v, uint32_t* p) {
++uint32_t ghc___aarch64_ldclr4_relax(uint32_t v, _Atomic uint32_t* p);
++uint32_t ghc___aarch64_ldclr4_relax(uint32_t v, _Atomic uint32_t* p) {
+   return atomic_fetch_and_explicit(p, v, memory_order_relaxed);
+ }
+ 
+-uint32_t ghc___aarch64_ldclr4_acq(uint32_t v, uint32_t* p);
+-uint32_t ghc___aarch64_ldclr4_acq(uint32_t v, uint32_t* p) {
++uint32_t ghc___aarch64_ldclr4_acq(uint32_t v, _Atomic uint32_t* p);
++uint32_t ghc___aarch64_ldclr4_acq(uint32_t v, _Atomic uint32_t* p) {
+   return atomic_fetch_and_explicit(p, v, memory_order_acquire);
+ }
+ 
+-uint32_t ghc___aarch64_ldclr4_rel(uint32_t v, uint32_t* p);
+-uint32_t ghc___aarch64_ldclr4_rel(uint32_t v, uint32_t* p) {
++uint32_t ghc___aarch64_ldclr4_rel(uint32_t v, _Atomic uint32_t* p);
++uint32_t ghc___aarch64_ldclr4_rel(uint32_t v, _Atomic uint32_t* p) {
+   return atomic_fetch_and_explicit(p, v, memory_order_release);
+ }
+ 
+-uint32_t ghc___aarch64_ldclr4_acq_rel(uint32_t v, uint32_t* p);
+-uint32_t ghc___aarch64_ldclr4_acq_rel(uint32_t v, uint32_t* p) {
++uint32_t ghc___aarch64_ldclr4_acq_rel(uint32_t v, _Atomic uint32_t* p);
++uint32_t ghc___aarch64_ldclr4_acq_rel(uint32_t v, _Atomic uint32_t* p) {
+   return atomic_fetch_and_explicit(p, v, memory_order_acq_rel);
+ }
+ 
+-uint32_t ghc___aarch64_ldclr4_sync(uint32_t v, uint32_t* p);
+-uint32_t ghc___aarch64_ldclr4_sync(uint32_t v, uint32_t* p) {
++uint32_t ghc___aarch64_ldclr4_sync(uint32_t v, _Atomic uint32_t* p);
++uint32_t ghc___aarch64_ldclr4_sync(uint32_t v, _Atomic uint32_t* p) {
+   return atomic_fetch_and_explicit(p, v, memory_order_seq_cst);
+ }
+ 
+-uint64_t ghc___aarch64_ldclr8_relax(uint64_t v, uint64_t* p);
+-uint64_t ghc___aarch64_ldclr8_relax(uint64_t v, uint64_t* p) {
++uint64_t ghc___aarch64_ldclr8_relax(uint64_t v, _Atomic uint64_t* p);
++uint64_t ghc___aarch64_ldclr8_relax(uint64_t v, _Atomic uint64_t* p) {
+   return atomic_fetch_and_explicit(p, v, memory_order_relaxed);
+ }
+ 
+-uint64_t ghc___aarch64_ldclr8_acq(uint64_t v, uint64_t* p);
+-uint64_t ghc___aarch64_ldclr8_acq(uint64_t v, uint64_t* p) {
++uint64_t ghc___aarch64_ldclr8_acq(uint64_t v, _Atomic uint64_t* p);
++uint64_t ghc___aarch64_ldclr8_acq(uint64_t v, _Atomic uint64_t* p) {
+   return atomic_fetch_and_explicit(p, v, memory_order_acquire);
+ }
+ 
+-uint64_t ghc___aarch64_ldclr8_rel(uint64_t v, uint64_t* p);
+-uint64_t ghc___aarch64_ldclr8_rel(uint64_t v, uint64_t* p) {
++uint64_t ghc___aarch64_ldclr8_rel(uint64_t v, _Atomic uint64_t* p);
++uint64_t ghc___aarch64_ldclr8_rel(uint64_t v, _Atomic uint64_t* p) {
+   return atomic_fetch_and_explicit(p, v, memory_order_release);
+ }
+ 
+-uint64_t ghc___aarch64_ldclr8_acq_rel(uint64_t v, uint64_t* p);
+-uint64_t ghc___aarch64_ldclr8_acq_rel(uint64_t v, uint64_t* p) {
++uint64_t ghc___aarch64_ldclr8_acq_rel(uint64_t v, _Atomic uint64_t* p);
++uint64_t ghc___aarch64_ldclr8_acq_rel(uint64_t v, _Atomic uint64_t* p) {
+   return atomic_fetch_and_explicit(p, v, memory_order_acq_rel);
+ }
+ 
+-uint64_t ghc___aarch64_ldclr8_sync(uint64_t v, uint64_t* p);
+-uint64_t ghc___aarch64_ldclr8_sync(uint64_t v, uint64_t* p) {
++uint64_t ghc___aarch64_ldclr8_sync(uint64_t v, _Atomic uint64_t* p);
++uint64_t ghc___aarch64_ldclr8_sync(uint64_t v, _Atomic uint64_t* p) {
+   return atomic_fetch_and_explicit(p, v, memory_order_seq_cst);
+ }
+ 
+-uint8_t ghc___aarch64_ldeor1_relax(uint8_t v, uint8_t* p);
+-uint8_t ghc___aarch64_ldeor1_relax(uint8_t v, uint8_t* p) {
++uint8_t ghc___aarch64_ldeor1_relax(uint8_t v, _Atomic uint8_t* p);
++uint8_t ghc___aarch64_ldeor1_relax(uint8_t v, _Atomic uint8_t* p) {
+   return atomic_fetch_xor_explicit(p, v, memory_order_relaxed);
+ }
+ 
+-uint8_t ghc___aarch64_ldeor1_acq(uint8_t v, uint8_t* p);
+-uint8_t ghc___aarch64_ldeor1_acq(uint8_t v, uint8_t* p) {
++uint8_t ghc___aarch64_ldeor1_acq(uint8_t v, _Atomic uint8_t* p);
++uint8_t ghc___aarch64_ldeor1_acq(uint8_t v, _Atomic uint8_t* p) {
+   return atomic_fetch_xor_explicit(p, v, memory_order_acquire);
+ }
+ 
+-uint8_t ghc___aarch64_ldeor1_rel(uint8_t v, uint8_t* p);
+-uint8_t ghc___aarch64_ldeor1_rel(uint8_t v, uint8_t* p) {
++uint8_t ghc___aarch64_ldeor1_rel(uint8_t v, _Atomic uint8_t* p);
++uint8_t ghc___aarch64_ldeor1_rel(uint8_t v, _Atomic uint8_t* p) {
+   return atomic_fetch_xor_explicit(p, v, memory_order_release);
+ }
+ 
+-uint8_t ghc___aarch64_ldeor1_acq_rel(uint8_t v, uint8_t* p);
+-uint8_t ghc___aarch64_ldeor1_acq_rel(uint8_t v, uint8_t* p) {
++uint8_t ghc___aarch64_ldeor1_acq_rel(uint8_t v, _Atomic uint8_t* p);
++uint8_t ghc___aarch64_ldeor1_acq_rel(uint8_t v, _Atomic uint8_t* p) {
+   return atomic_fetch_xor_explicit(p, v, memory_order_acq_rel);
+ }
+ 
+-uint8_t ghc___aarch64_ldeor1_sync(uint8_t v, uint8_t* p);
+-uint8_t ghc___aarch64_ldeor1_sync(uint8_t v, uint8_t* p) {
++uint8_t ghc___aarch64_ldeor1_sync(uint8_t v, _Atomic uint8_t* p);
++uint8_t ghc___aarch64_ldeor1_sync(uint8_t v, _Atomic uint8_t* p) {
+   return atomic_fetch_xor_explicit(p, v, memory_order_seq_cst);
+ }
+ 
+-uint16_t ghc___aarch64_ldeor2_relax(uint16_t v, uint16_t* p);
+-uint16_t ghc___aarch64_ldeor2_relax(uint16_t v, uint16_t* p) {
++uint16_t ghc___aarch64_ldeor2_relax(uint16_t v, _Atomic uint16_t* p);
++uint16_t ghc___aarch64_ldeor2_relax(uint16_t v, _Atomic uint16_t* p) {
+   return atomic_fetch_xor_explicit(p, v, memory_order_relaxed);
+ }
+ 
+-uint16_t ghc___aarch64_ldeor2_acq(uint16_t v, uint16_t* p);
+-uint16_t ghc___aarch64_ldeor2_acq(uint16_t v, uint16_t* p) {
++uint16_t ghc___aarch64_ldeor2_acq(uint16_t v, _Atomic uint16_t* p);
++uint16_t ghc___aarch64_ldeor2_acq(uint16_t v, _Atomic uint16_t* p) {
+   return atomic_fetch_xor_explicit(p, v, memory_order_acquire);
+ }
+ 
+-uint16_t ghc___aarch64_ldeor2_rel(uint16_t v, uint16_t* p);
+-uint16_t ghc___aarch64_ldeor2_rel(uint16_t v, uint16_t* p) {
++uint16_t ghc___aarch64_ldeor2_rel(uint16_t v, _Atomic uint16_t* p);
++uint16_t ghc___aarch64_ldeor2_rel(uint16_t v, _Atomic uint16_t* p) {
+   return atomic_fetch_xor_explicit(p, v, memory_order_release);
+ }
+ 
+-uint16_t ghc___aarch64_ldeor2_acq_rel(uint16_t v, uint16_t* p);
+-uint16_t ghc___aarch64_ldeor2_acq_rel(uint16_t v, uint16_t* p) {
++uint16_t ghc___aarch64_ldeor2_acq_rel(uint16_t v, _Atomic uint16_t* p);
++uint16_t ghc___aarch64_ldeor2_acq_rel(uint16_t v, _Atomic uint16_t* p) {
+   return atomic_fetch_xor_explicit(p, v, memory_order_acq_rel);
+ }
+ 
+-uint16_t ghc___aarch64_ldeor2_sync(uint16_t v, uint16_t* p);
+-uint16_t ghc___aarch64_ldeor2_sync(uint16_t v, uint16_t* p) {
++uint16_t ghc___aarch64_ldeor2_sync(uint16_t v, _Atomic uint16_t* p);
++uint16_t ghc___aarch64_ldeor2_sync(uint16_t v, _Atomic uint16_t* p) {
+   return atomic_fetch_xor_explicit(p, v, memory_order_seq_cst);
+ }
+ 
+-uint32_t ghc___aarch64_ldeor4_relax(uint32_t v, uint32_t* p);
+-uint32_t ghc___aarch64_ldeor4_relax(uint32_t v, uint32_t* p) {
++uint32_t ghc___aarch64_ldeor4_relax(uint32_t v, _Atomic uint32_t* p);
++uint32_t ghc___aarch64_ldeor4_relax(uint32_t v, _Atomic uint32_t* p) {
+   return atomic_fetch_xor_explicit(p, v, memory_order_relaxed);
+ }
+ 
+-uint32_t ghc___aarch64_ldeor4_acq(uint32_t v, uint32_t* p);
+-uint32_t ghc___aarch64_ldeor4_acq(uint32_t v, uint32_t* p) {
++uint32_t ghc___aarch64_ldeor4_acq(uint32_t v, _Atomic uint32_t* p);
++uint32_t ghc___aarch64_ldeor4_acq(uint32_t v, _Atomic uint32_t* p) {
+   return atomic_fetch_xor_explicit(p, v, memory_order_acquire);
+ }
+ 
+-uint32_t ghc___aarch64_ldeor4_rel(uint32_t v, uint32_t* p);
+-uint32_t ghc___aarch64_ldeor4_rel(uint32_t v, uint32_t* p) {
++uint32_t ghc___aarch64_ldeor4_rel(uint32_t v, _Atomic uint32_t* p);
++uint32_t ghc___aarch64_ldeor4_rel(uint32_t v, _Atomic uint32_t* p) {
+   return atomic_fetch_xor_explicit(p, v, memory_order_release);
+ }
+ 
+-uint32_t ghc___aarch64_ldeor4_acq_rel(uint32_t v, uint32_t* p);
+-uint32_t ghc___aarch64_ldeor4_acq_rel(uint32_t v, uint32_t* p) {
++uint32_t ghc___aarch64_ldeor4_acq_rel(uint32_t v, _Atomic uint32_t* p);
++uint32_t ghc___aarch64_ldeor4_acq_rel(uint32_t v, _Atomic uint32_t* p) {
+   return atomic_fetch_xor_explicit(p, v, memory_order_acq_rel);
+ }
+ 
+-uint32_t ghc___aarch64_ldeor4_sync(uint32_t v, uint32_t* p);
+-uint32_t ghc___aarch64_ldeor4_sync(uint32_t v, uint32_t* p) {
++uint32_t ghc___aarch64_ldeor4_sync(uint32_t v, _Atomic uint32_t* p);
++uint32_t ghc___aarch64_ldeor4_sync(uint32_t v, _Atomic uint32_t* p) {
+   return atomic_fetch_xor_explicit(p, v, memory_order_seq_cst);
+ }
+ 
+-uint64_t ghc___aarch64_ldeor8_relax(uint64_t v, uint64_t* p);
+-uint64_t ghc___aarch64_ldeor8_relax(uint64_t v, uint64_t* p) {
++uint64_t ghc___aarch64_ldeor8_relax(uint64_t v, _Atomic uint64_t* p);
++uint64_t ghc___aarch64_ldeor8_relax(uint64_t v, _Atomic uint64_t* p) {
+   return atomic_fetch_xor_explicit(p, v, memory_order_relaxed);
+ }
+ 
+-uint64_t ghc___aarch64_ldeor8_acq(uint64_t v, uint64_t* p);
+-uint64_t ghc___aarch64_ldeor8_acq(uint64_t v, uint64_t* p) {
++uint64_t ghc___aarch64_ldeor8_acq(uint64_t v, _Atomic uint64_t* p);
++uint64_t ghc___aarch64_ldeor8_acq(uint64_t v, _Atomic uint64_t* p) {
+   return atomic_fetch_xor_explicit(p, v, memory_order_acquire);
+ }
+ 
+-uint64_t ghc___aarch64_ldeor8_rel(uint64_t v, uint64_t* p);
+-uint64_t ghc___aarch64_ldeor8_rel(uint64_t v, uint64_t* p) {
++uint64_t ghc___aarch64_ldeor8_rel(uint64_t v, _Atomic uint64_t* p);
++uint64_t ghc___aarch64_ldeor8_rel(uint64_t v, _Atomic uint64_t* p) {
+   return atomic_fetch_xor_explicit(p, v, memory_order_release);
+ }
+ 
+-uint64_t ghc___aarch64_ldeor8_acq_rel(uint64_t v, uint64_t* p);
+-uint64_t ghc___aarch64_ldeor8_acq_rel(uint64_t v, uint64_t* p) {
++uint64_t ghc___aarch64_ldeor8_acq_rel(uint64_t v, _Atomic uint64_t* p);
++uint64_t ghc___aarch64_ldeor8_acq_rel(uint64_t v, _Atomic uint64_t* p) {
+   return atomic_fetch_xor_explicit(p, v, memory_order_acq_rel);
+ }
+ 
+-uint64_t ghc___aarch64_ldeor8_sync(uint64_t v, uint64_t* p);
+-uint64_t ghc___aarch64_ldeor8_sync(uint64_t v, uint64_t* p) {
++uint64_t ghc___aarch64_ldeor8_sync(uint64_t v, _Atomic uint64_t* p);
++uint64_t ghc___aarch64_ldeor8_sync(uint64_t v, _Atomic uint64_t* p) {
+   return atomic_fetch_xor_explicit(p, v, memory_order_seq_cst);
+ }
+ 
+-uint8_t ghc___aarch64_ldset1_relax(uint8_t v, uint8_t* p);
+-uint8_t ghc___aarch64_ldset1_relax(uint8_t v, uint8_t* p) {
++uint8_t ghc___aarch64_ldset1_relax(uint8_t v, _Atomic uint8_t* p);
++uint8_t ghc___aarch64_ldset1_relax(uint8_t v, _Atomic uint8_t* p) {
+   return atomic_fetch_or_explicit(p, v, memory_order_relaxed);
+ }
+ 
+-uint8_t ghc___aarch64_ldset1_acq(uint8_t v, uint8_t* p);
+-uint8_t ghc___aarch64_ldset1_acq(uint8_t v, uint8_t* p) {
++uint8_t ghc___aarch64_ldset1_acq(uint8_t v, _Atomic uint8_t* p);
++uint8_t ghc___aarch64_ldset1_acq(uint8_t v, _Atomic uint8_t* p) {
+   return atomic_fetch_or_explicit(p, v, memory_order_acquire);
+ }
+ 
+-uint8_t ghc___aarch64_ldset1_rel(uint8_t v, uint8_t* p);
+-uint8_t ghc___aarch64_ldset1_rel(uint8_t v, uint8_t* p) {
++uint8_t ghc___aarch64_ldset1_rel(uint8_t v, _Atomic uint8_t* p);
++uint8_t ghc___aarch64_ldset1_rel(uint8_t v, _Atomic uint8_t* p) {
+   return atomic_fetch_or_explicit(p, v, memory_order_release);
+ }
+ 
+-uint8_t ghc___aarch64_ldset1_acq_rel(uint8_t v, uint8_t* p);
+-uint8_t ghc___aarch64_ldset1_acq_rel(uint8_t v, uint8_t* p) {
++uint8_t ghc___aarch64_ldset1_acq_rel(uint8_t v, _Atomic uint8_t* p);
++uint8_t ghc___aarch64_ldset1_acq_rel(uint8_t v, _Atomic uint8_t* p) {
+   return atomic_fetch_or_explicit(p, v, memory_order_acq_rel);
+ }
+ 
+-uint8_t ghc___aarch64_ldset1_sync(uint8_t v, uint8_t* p);
+-uint8_t ghc___aarch64_ldset1_sync(uint8_t v, uint8_t* p) {
++uint8_t ghc___aarch64_ldset1_sync(uint8_t v, _Atomic uint8_t* p);
++uint8_t ghc___aarch64_ldset1_sync(uint8_t v, _Atomic uint8_t* p) {
+   return atomic_fetch_or_explicit(p, v, memory_order_seq_cst);
+ }
+ 
+-uint16_t ghc___aarch64_ldset2_relax(uint16_t v, uint16_t* p);
+-uint16_t ghc___aarch64_ldset2_relax(uint16_t v, uint16_t* p) {
++uint16_t ghc___aarch64_ldset2_relax(uint16_t v, _Atomic uint16_t* p);
++uint16_t ghc___aarch64_ldset2_relax(uint16_t v, _Atomic uint16_t* p) {
+   return atomic_fetch_or_explicit(p, v, memory_order_relaxed);
+ }
+ 
+-uint16_t ghc___aarch64_ldset2_acq(uint16_t v, uint16_t* p);
+-uint16_t ghc___aarch64_ldset2_acq(uint16_t v, uint16_t* p) {
++uint16_t ghc___aarch64_ldset2_acq(uint16_t v, _Atomic uint16_t* p);
++uint16_t ghc___aarch64_ldset2_acq(uint16_t v, _Atomic uint16_t* p) {
+   return atomic_fetch_or_explicit(p, v, memory_order_acquire);
+ }
+ 
+-uint16_t ghc___aarch64_ldset2_rel(uint16_t v, uint16_t* p);
+-uint16_t ghc___aarch64_ldset2_rel(uint16_t v, uint16_t* p) {
++uint16_t ghc___aarch64_ldset2_rel(uint16_t v, _Atomic uint16_t* p);
++uint16_t ghc___aarch64_ldset2_rel(uint16_t v, _Atomic uint16_t* p) {
+   return atomic_fetch_or_explicit(p, v, memory_order_release);
+ }
+ 
+-uint16_t ghc___aarch64_ldset2_acq_rel(uint16_t v, uint16_t* p);
+-uint16_t ghc___aarch64_ldset2_acq_rel(uint16_t v, uint16_t* p) {
++uint16_t ghc___aarch64_ldset2_acq_rel(uint16_t v, _Atomic uint16_t* p);
++uint16_t ghc___aarch64_ldset2_acq_rel(uint16_t v, _Atomic uint16_t* p) {
+   return atomic_fetch_or_explicit(p, v, memory_order_acq_rel);
+ }
+ 
+-uint16_t ghc___aarch64_ldset2_sync(uint16_t v, uint16_t* p);
+-uint16_t ghc___aarch64_ldset2_sync(uint16_t v, uint16_t* p) {
++uint16_t ghc___aarch64_ldset2_sync(uint16_t v, _Atomic uint16_t* p);
++uint16_t ghc___aarch64_ldset2_sync(uint16_t v, _Atomic uint16_t* p) {
+   return atomic_fetch_or_explicit(p, v, memory_order_seq_cst);
+ }
+ 
+-uint32_t ghc___aarch64_ldset4_relax(uint32_t v, uint32_t* p);
+-uint32_t ghc___aarch64_ldset4_relax(uint32_t v, uint32_t* p) {
++uint32_t ghc___aarch64_ldset4_relax(uint32_t v, _Atomic uint32_t* p);
++uint32_t ghc___aarch64_ldset4_relax(uint32_t v, _Atomic uint32_t* p) {
+   return atomic_fetch_or_explicit(p, v, memory_order_relaxed);
+ }
+ 
+-uint32_t ghc___aarch64_ldset4_acq(uint32_t v, uint32_t* p);
+-uint32_t ghc___aarch64_ldset4_acq(uint32_t v, uint32_t* p) {
++uint32_t ghc___aarch64_ldset4_acq(uint32_t v, _Atomic uint32_t* p);
++uint32_t ghc___aarch64_ldset4_acq(uint32_t v, _Atomic uint32_t* p) {
+   return atomic_fetch_or_explicit(p, v, memory_order_acquire);
+ }
+ 
+-uint32_t ghc___aarch64_ldset4_rel(uint32_t v, uint32_t* p);
+-uint32_t ghc___aarch64_ldset4_rel(uint32_t v, uint32_t* p) {
++uint32_t ghc___aarch64_ldset4_rel(uint32_t v, _Atomic uint32_t* p);
++uint32_t ghc___aarch64_ldset4_rel(uint32_t v, _Atomic uint32_t* p) {
+   return atomic_fetch_or_explicit(p, v, memory_order_release);
+ }
+ 
+-uint32_t ghc___aarch64_ldset4_acq_rel(uint32_t v, uint32_t* p);
+-uint32_t ghc___aarch64_ldset4_acq_rel(uint32_t v, uint32_t* p) {
++uint32_t ghc___aarch64_ldset4_acq_rel(uint32_t v, _Atomic uint32_t* p);
++uint32_t ghc___aarch64_ldset4_acq_rel(uint32_t v, _Atomic uint32_t* p) {
+   return atomic_fetch_or_explicit(p, v, memory_order_acq_rel);
+ }
+ 
+-uint32_t ghc___aarch64_ldset4_sync(uint32_t v, uint32_t* p);
+-uint32_t ghc___aarch64_ldset4_sync(uint32_t v, uint32_t* p) {
++uint32_t ghc___aarch64_ldset4_sync(uint32_t v, _Atomic uint32_t* p);
++uint32_t ghc___aarch64_ldset4_sync(uint32_t v, _Atomic uint32_t* p) {
+   return atomic_fetch_or_explicit(p, v, memory_order_seq_cst);
+ }
+ 
+-uint64_t ghc___aarch64_ldset8_relax(uint64_t v, uint64_t* p);
+-uint64_t ghc___aarch64_ldset8_relax(uint64_t v, uint64_t* p) {
++uint64_t ghc___aarch64_ldset8_relax(uint64_t v, _Atomic uint64_t* p);
++uint64_t ghc___aarch64_ldset8_relax(uint64_t v, _Atomic uint64_t* p) {
+   return atomic_fetch_or_explicit(p, v, memory_order_relaxed);
+ }
+ 
+-uint64_t ghc___aarch64_ldset8_acq(uint64_t v, uint64_t* p);
+-uint64_t ghc___aarch64_ldset8_acq(uint64_t v, uint64_t* p) {
++uint64_t ghc___aarch64_ldset8_acq(uint64_t v, _Atomic uint64_t* p);
++uint64_t ghc___aarch64_ldset8_acq(uint64_t v, _Atomic uint64_t* p) {
+   return atomic_fetch_or_explicit(p, v, memory_order_acquire);
+ }
+ 
+-uint64_t ghc___aarch64_ldset8_rel(uint64_t v, uint64_t* p);
+-uint64_t ghc___aarch64_ldset8_rel(uint64_t v, uint64_t* p) {
++uint64_t ghc___aarch64_ldset8_rel(uint64_t v, _Atomic uint64_t* p);
++uint64_t ghc___aarch64_ldset8_rel(uint64_t v, _Atomic uint64_t* p) {
+   return atomic_fetch_or_explicit(p, v, memory_order_release);
+ }
+ 
+-uint64_t ghc___aarch64_ldset8_acq_rel(uint64_t v, uint64_t* p);
+-uint64_t ghc___aarch64_ldset8_acq_rel(uint64_t v, uint64_t* p) {
++uint64_t ghc___aarch64_ldset8_acq_rel(uint64_t v, _Atomic uint64_t* p);
++uint64_t ghc___aarch64_ldset8_acq_rel(uint64_t v, _Atomic uint64_t* p) {
+   return atomic_fetch_or_explicit(p, v, memory_order_acq_rel);
+ }
+ 
+-uint64_t ghc___aarch64_ldset8_sync(uint64_t v, uint64_t* p);
+-uint64_t ghc___aarch64_ldset8_sync(uint64_t v, uint64_t* p) {
++uint64_t ghc___aarch64_ldset8_sync(uint64_t v, _Atomic uint64_t* p);
++uint64_t ghc___aarch64_ldset8_sync(uint64_t v, _Atomic uint64_t* p) {
+   return atomic_fetch_or_explicit(p, v, memory_order_seq_cst);
+ }
+