diff mbox series

[v1,2/2] x86: Add support for building {w}memmove{_chk} with explicit ISA level

Message ID 20220628152757.17922-2-goldstein.w.n@gmail.com
State New
Headers show
Series [v1,1/2] x86: Move mem{p}{mov|cpy}_{chk_}erms to its own file | expand

Commit Message

Noah Goldstein June 28, 2022, 3:27 p.m. UTC
1. Refactor files so that all implementations are in the multiarch
   directory
    - Moved the implementation portion of memmove sse2 from memmove.S
      to multiarch/memmove-sse2.S

    - The non-multiarch file now only includes one of the
      implementations in the multiarch directory based on the compiled
      ISA level (only used for non-multiarch builds.  Otherwise we go
      through the ifunc selector).

2. Add ISA level build guards to different implementations.
    - I.e memmove-avx2-unaligned-erms.S which is ISA level 3 will only
      build if compiled ISA level <= 3. Otherwise there is no reason
      to include it as we will always use one of the ISA level 4
      implementations (memmove-evex-unaligned-erms.S).

3. Add new multiarch/rtld-memmove.S that just include the
   non-multiarch memmove.S which will in turn select the best
   implementation based on the compiled ISA level.

4. Refactor the ifunc selector and ifunc implementation list to use
   the ISA level aware wrapper macros that allow functions below the
   compiled ISA level (with a guranteed replacement) to be skipped.

Tested with and without multiarch on x86_64 for ISA levels:
{generic, x86-64-v2, x86-64-v3, x86-64-v4}

And m32 with and without multiarch.
isa raising memmove
---
 sysdeps/x86_64/memmove.S                      |  34 +--
 sysdeps/x86_64/multiarch/Makefile             |   1 +
 sysdeps/x86_64/multiarch/ifunc-impl-list.c    | 275 ++++++++++--------
 sysdeps/x86_64/multiarch/ifunc-memmove.h      |  47 +--
 .../multiarch/memmove-avx-unaligned-erms.S    |  15 +-
 .../multiarch/memmove-avx512-no-vzeroupper.S  |   3 +-
 .../multiarch/memmove-avx512-unaligned-erms.S |  11 +-
 .../multiarch/memmove-evex-unaligned-erms.S   |  16 +-
 .../x86_64/multiarch/memmove-shlib-compat.h   |  26 ++
 .../multiarch/memmove-sse2-unaligned-erms.S   |  36 ++-
 sysdeps/x86_64/multiarch/memmove-ssse3.S      |  41 +--
 sysdeps/x86_64/multiarch/rtld-memmove.S       |  18 ++
 12 files changed, 327 insertions(+), 196 deletions(-)
 create mode 100644 sysdeps/x86_64/multiarch/memmove-shlib-compat.h
 create mode 100644 sysdeps/x86_64/multiarch/rtld-memmove.S
diff mbox series

Patch

diff --git a/sysdeps/x86_64/memmove.S b/sysdeps/x86_64/memmove.S
index 78e8d974d9..ef1b2496f4 100644
--- a/sysdeps/x86_64/memmove.S
+++ b/sysdeps/x86_64/memmove.S
@@ -16,17 +16,6 @@ 
    License along with the GNU C Library; if not, see
    <https://www.gnu.org/licenses/>.  */
 
-#include <sysdep.h>
-
-#define VEC_SIZE	16
-#define VEC(i)		xmm##i
-#define PREFETCHNT	prefetchnta
-#define VMOVNT		movntdq
-/* Use movups and movaps for smaller code sizes.  */
-#define VMOVU		movups
-#define VMOVA		movaps
-#define MOV_SIZE	3
-#define SECTION(p)		p
 
 #ifdef USE_MULTIARCH
 # if !IS_IN (libc)
@@ -42,12 +31,20 @@ 
 #if !defined USE_MULTIARCH || !IS_IN (libc)
 # define MEMPCPY_SYMBOL(p,s)		__mempcpy
 #endif
-#ifndef MEMMOVE_SYMBOL
-# define MEMMOVE_CHK_SYMBOL(p,s)	p
-# define MEMMOVE_SYMBOL(p,s)		memmove
-#endif
 
-#include "multiarch/memmove-vec-unaligned-erms.S"
+#define MEMMOVE_CHK_SYMBOL(p,s)	p
+#define MEMMOVE_SYMBOL(p,s)		memmove
+
+
+#define DEFAULT_IMPL_V1	"multiarch/memmove-sse2-unaligned-erms.S"
+#define DEFAULT_IMPL_V3	"multiarch/memmove-avx-unaligned-erms.S"
+#define DEFAULT_IMPL_V4	"multiarch/memmove-evex-unaligned-erms.S"
+
+#include "isa-default-impl.h"
+
+#if defined USE_MULTIARCH && !IS_IN (libc)
+weak_alias (__mempcpy, mempcpy)
+#endif
 
 #ifndef USE_MULTIARCH
 libc_hidden_builtin_def (memmove)
@@ -59,13 +56,10 @@  libc_hidden_def (__mempcpy)
 weak_alias (__mempcpy, mempcpy)
 libc_hidden_builtin_def (mempcpy)
 
+
 # if defined SHARED && IS_IN (libc)
 #  undef memcpy
 #  include <shlib-compat.h>
 versioned_symbol (libc, __memcpy, memcpy, GLIBC_2_14);
-
-#  if SHLIB_COMPAT (libc, GLIBC_2_2_5, GLIBC_2_14)
-compat_symbol (libc, memmove, memcpy, GLIBC_2_2_5);
-#  endif
 # endif
 #endif
diff --git a/sysdeps/x86_64/multiarch/Makefile b/sysdeps/x86_64/multiarch/Makefile
index 0525cef3fe..18cea04423 100644
--- a/sysdeps/x86_64/multiarch/Makefile
+++ b/sysdeps/x86_64/multiarch/Makefile
@@ -18,6 +18,7 @@  sysdep_routines += \
   memmove-avx-unaligned-erms-rtm \
   memmove-avx512-no-vzeroupper \
   memmove-avx512-unaligned-erms \
+  memmove-erms \
   memmove-evex-unaligned-erms \
   memmove-sse2-unaligned-erms \
   memmove-ssse3 \
diff --git a/sysdeps/x86_64/multiarch/ifunc-impl-list.c b/sysdeps/x86_64/multiarch/ifunc-impl-list.c
index 4450e2c593..5f7a0aa186 100644
--- a/sysdeps/x86_64/multiarch/ifunc-impl-list.c
+++ b/sysdeps/x86_64/multiarch/ifunc-impl-list.c
@@ -109,83 +109,93 @@  __libc_ifunc_impl_list (const char *name, struct libc_ifunc_impl *array,
 #ifdef SHARED
   /* Support sysdeps/x86_64/multiarch/memmove_chk.c.  */
   IFUNC_IMPL (i, name, __memmove_chk,
-	      IFUNC_IMPL_ADD (array, i, __memmove_chk,
+	      IFUNC_IMPL_ADD (array, i, __memmove_chk, 1,
+			      __memmove_chk_erms)
+	      X86_IFUNC_IMPL_ADD_V4 (array, i, __memmove_chk,
 			      CPU_FEATURE_USABLE (AVX512F),
 			      __memmove_chk_avx512_no_vzeroupper)
-	      IFUNC_IMPL_ADD (array, i, __memmove_chk,
+	      X86_IFUNC_IMPL_ADD_V4 (array, i, __memmove_chk,
 			      CPU_FEATURE_USABLE (AVX512VL),
 			      __memmove_chk_avx512_unaligned)
-	      IFUNC_IMPL_ADD (array, i, __memmove_chk,
+	      X86_IFUNC_IMPL_ADD_V4 (array, i, __memmove_chk,
 			      CPU_FEATURE_USABLE (AVX512VL),
 			      __memmove_chk_avx512_unaligned_erms)
-	      IFUNC_IMPL_ADD (array, i, __memmove_chk,
+	      X86_IFUNC_IMPL_ADD_V4 (array, i, __memmove_chk,
+			      CPU_FEATURE_USABLE (AVX512VL),
+			      __memmove_chk_evex_unaligned)
+	      X86_IFUNC_IMPL_ADD_V4 (array, i, __memmove_chk,
+			      CPU_FEATURE_USABLE (AVX512VL),
+			      __memmove_chk_evex_unaligned_erms)
+	      X86_IFUNC_IMPL_ADD_V3 (array, i, __memmove_chk,
 			      CPU_FEATURE_USABLE (AVX),
 			      __memmove_chk_avx_unaligned)
-	      IFUNC_IMPL_ADD (array, i, __memmove_chk,
+	      X86_IFUNC_IMPL_ADD_V3 (array, i, __memmove_chk,
 			      CPU_FEATURE_USABLE (AVX),
 			      __memmove_chk_avx_unaligned_erms)
-	      IFUNC_IMPL_ADD (array, i, __memmove_chk,
+	      X86_IFUNC_IMPL_ADD_V3 (array, i, __memmove_chk,
 			      (CPU_FEATURE_USABLE (AVX)
 			       && CPU_FEATURE_USABLE (RTM)),
 			      __memmove_chk_avx_unaligned_rtm)
-	      IFUNC_IMPL_ADD (array, i, __memmove_chk,
+	      X86_IFUNC_IMPL_ADD_V3 (array, i, __memmove_chk,
 			      (CPU_FEATURE_USABLE (AVX)
 			       && CPU_FEATURE_USABLE (RTM)),
 			      __memmove_chk_avx_unaligned_erms_rtm)
-	      IFUNC_IMPL_ADD (array, i, __memmove_chk,
-			      CPU_FEATURE_USABLE (AVX512VL),
-			      __memmove_chk_evex_unaligned)
-	      IFUNC_IMPL_ADD (array, i, __memmove_chk,
-			      CPU_FEATURE_USABLE (AVX512VL),
-			      __memmove_chk_evex_unaligned_erms)
-	      IFUNC_IMPL_ADD (array, i, __memmove_chk,
+	      /* By V3 we assume fast aligned copy.  */
+	      X86_IFUNC_IMPL_ADD_V2 (array, i, __memmove_chk,
 			      CPU_FEATURE_USABLE (SSSE3),
 			      __memmove_chk_ssse3)
-	      IFUNC_IMPL_ADD (array, i, __memmove_chk, 1,
+	      /* Can be lowered to V1 if a general V2 implementation is
+             added (ssse3 is too optimized around aligned copy to be
+             better as general purpose memmove).  */
+	      X86_IFUNC_IMPL_ADD_V2 (array, i, __memmove_chk, 1,
 			      __memmove_chk_sse2_unaligned)
-	      IFUNC_IMPL_ADD (array, i, __memmove_chk, 1,
-			      __memmove_chk_sse2_unaligned_erms)
-	      IFUNC_IMPL_ADD (array, i, __memmove_chk, 1,
-			      __memmove_chk_erms))
+	      X86_IFUNC_IMPL_ADD_V2 (array, i, __memmove_chk, 1,
+			      __memmove_chk_sse2_unaligned_erms))
 #endif
 
   /* Support sysdeps/x86_64/multiarch/memmove.c.  */
   IFUNC_IMPL (i, name, memmove,
-	      IFUNC_IMPL_ADD (array, i, memmove,
+	      IFUNC_IMPL_ADD (array, i, memmove, 1,
+			      __memmove_erms)
+	      X86_IFUNC_IMPL_ADD_V4 (array, i, memmove,
+			      CPU_FEATURE_USABLE (AVX512F),
+			      __memmove_avx512_no_vzeroupper)
+	      X86_IFUNC_IMPL_ADD_V4 (array, i, memmove,
+			      CPU_FEATURE_USABLE (AVX512VL),
+			      __memmove_avx512_unaligned)
+	      X86_IFUNC_IMPL_ADD_V4 (array, i, memmove,
+			      CPU_FEATURE_USABLE (AVX512VL),
+			      __memmove_avx512_unaligned_erms)
+	      X86_IFUNC_IMPL_ADD_V4 (array, i, memmove,
+			      CPU_FEATURE_USABLE (AVX512VL),
+			      __memmove_evex_unaligned)
+	      X86_IFUNC_IMPL_ADD_V4 (array, i, memmove,
+			      CPU_FEATURE_USABLE (AVX512VL),
+			      __memmove_evex_unaligned_erms)
+	      X86_IFUNC_IMPL_ADD_V3 (array, i, memmove,
 			      CPU_FEATURE_USABLE (AVX),
 			      __memmove_avx_unaligned)
-	      IFUNC_IMPL_ADD (array, i, memmove,
+	      X86_IFUNC_IMPL_ADD_V3 (array, i, memmove,
 			      CPU_FEATURE_USABLE (AVX),
 			      __memmove_avx_unaligned_erms)
-	      IFUNC_IMPL_ADD (array, i, memmove,
+	      X86_IFUNC_IMPL_ADD_V3 (array, i, memmove,
 			      (CPU_FEATURE_USABLE (AVX)
 			       && CPU_FEATURE_USABLE (RTM)),
 			      __memmove_avx_unaligned_rtm)
-	      IFUNC_IMPL_ADD (array, i, memmove,
+	      X86_IFUNC_IMPL_ADD_V3 (array, i, memmove,
 			      (CPU_FEATURE_USABLE (AVX)
 			       && CPU_FEATURE_USABLE (RTM)),
 			      __memmove_avx_unaligned_erms_rtm)
-	      IFUNC_IMPL_ADD (array, i, memmove,
-			      CPU_FEATURE_USABLE (AVX512VL),
-			      __memmove_evex_unaligned)
-	      IFUNC_IMPL_ADD (array, i, memmove,
-			      CPU_FEATURE_USABLE (AVX512VL),
-			      __memmove_evex_unaligned_erms)
-	      IFUNC_IMPL_ADD (array, i, memmove,
-			      CPU_FEATURE_USABLE (AVX512F),
-			      __memmove_avx512_no_vzeroupper)
-	      IFUNC_IMPL_ADD (array, i, memmove,
-			      CPU_FEATURE_USABLE (AVX512VL),
-			      __memmove_avx512_unaligned)
-	      IFUNC_IMPL_ADD (array, i, memmove,
-			      CPU_FEATURE_USABLE (AVX512VL),
-			      __memmove_avx512_unaligned_erms)
-	      IFUNC_IMPL_ADD (array, i, memmove, CPU_FEATURE_USABLE (SSSE3),
+	      /* By V3 we assume fast aligned copy.  */
+	      X86_IFUNC_IMPL_ADD_V2 (array, i, memmove,
+			      CPU_FEATURE_USABLE (SSSE3),
 			      __memmove_ssse3)
-	      IFUNC_IMPL_ADD (array, i, memmove, 1, __memmove_erms)
-	      IFUNC_IMPL_ADD (array, i, memmove, 1,
+	      /* Can be lowered to V1 if a general V2 implementation is
+             added (ssse3 is too optimized around aligned copy to be
+             better as general purpose memmove).  */
+	      X86_IFUNC_IMPL_ADD_V2 (array, i, memmove, 1,
 			      __memmove_sse2_unaligned)
-	      IFUNC_IMPL_ADD (array, i, memmove, 1,
+	      X86_IFUNC_IMPL_ADD_V2 (array, i, memmove, 1,
 			      __memmove_sse2_unaligned_erms))
 
   /* Support sysdeps/x86_64/multiarch/memrchr.c.  */
@@ -847,165 +857,186 @@  __libc_ifunc_impl_list (const char *name, struct libc_ifunc_impl *array,
 #ifdef SHARED
   /* Support sysdeps/x86_64/multiarch/memcpy_chk.c.  */
   IFUNC_IMPL (i, name, __memcpy_chk,
-	      IFUNC_IMPL_ADD (array, i, __memcpy_chk,
+	      IFUNC_IMPL_ADD (array, i, __memcpy_chk, 1,
+			      __memcpy_chk_erms)
+	      X86_IFUNC_IMPL_ADD_V4 (array, i, __memcpy_chk,
 			      CPU_FEATURE_USABLE (AVX512F),
 			      __memcpy_chk_avx512_no_vzeroupper)
-	      IFUNC_IMPL_ADD (array, i, __memcpy_chk,
+	      X86_IFUNC_IMPL_ADD_V4 (array, i, __memcpy_chk,
 			      CPU_FEATURE_USABLE (AVX512VL),
 			      __memcpy_chk_avx512_unaligned)
-	      IFUNC_IMPL_ADD (array, i, __memcpy_chk,
+	      X86_IFUNC_IMPL_ADD_V4 (array, i, __memcpy_chk,
 			      CPU_FEATURE_USABLE (AVX512VL),
 			      __memcpy_chk_avx512_unaligned_erms)
-	      IFUNC_IMPL_ADD (array, i, __memcpy_chk,
+	      X86_IFUNC_IMPL_ADD_V4 (array, i, __memcpy_chk,
+			      CPU_FEATURE_USABLE (AVX512VL),
+			      __memcpy_chk_evex_unaligned)
+	      X86_IFUNC_IMPL_ADD_V4 (array, i, __memcpy_chk,
+			      CPU_FEATURE_USABLE (AVX512VL),
+			      __memcpy_chk_evex_unaligned_erms)
+	      X86_IFUNC_IMPL_ADD_V3 (array, i, __memcpy_chk,
 			      CPU_FEATURE_USABLE (AVX),
 			      __memcpy_chk_avx_unaligned)
-	      IFUNC_IMPL_ADD (array, i, __memcpy_chk,
+	      X86_IFUNC_IMPL_ADD_V3 (array, i, __memcpy_chk,
 			      CPU_FEATURE_USABLE (AVX),
 			      __memcpy_chk_avx_unaligned_erms)
-	      IFUNC_IMPL_ADD (array, i, __memcpy_chk,
+	      X86_IFUNC_IMPL_ADD_V3 (array, i, __memcpy_chk,
 			      (CPU_FEATURE_USABLE (AVX)
 			       && CPU_FEATURE_USABLE (RTM)),
 			      __memcpy_chk_avx_unaligned_rtm)
-	      IFUNC_IMPL_ADD (array, i, __memcpy_chk,
+	      X86_IFUNC_IMPL_ADD_V3 (array, i, __memcpy_chk,
 			      (CPU_FEATURE_USABLE (AVX)
 			       && CPU_FEATURE_USABLE (RTM)),
 			      __memcpy_chk_avx_unaligned_erms_rtm)
-	      IFUNC_IMPL_ADD (array, i, __memcpy_chk,
-			      CPU_FEATURE_USABLE (AVX512VL),
-			      __memcpy_chk_evex_unaligned)
-	      IFUNC_IMPL_ADD (array, i, __memcpy_chk,
-			      CPU_FEATURE_USABLE (AVX512VL),
-			      __memcpy_chk_evex_unaligned_erms)
-	      IFUNC_IMPL_ADD (array, i, __memcpy_chk,
+	      /* By V3 we assume fast aligned copy.  */
+	      X86_IFUNC_IMPL_ADD_V2 (array, i, __memcpy_chk,
 			      CPU_FEATURE_USABLE (SSSE3),
 			      __memcpy_chk_ssse3)
-	      IFUNC_IMPL_ADD (array, i, __memcpy_chk, 1,
+	      /* Can be lowered to V1 if a general V2 implementation is
+             added (ssse3 is too optimized around aligned copy to be
+             better as general purpose memmove).  */
+	      X86_IFUNC_IMPL_ADD_V2 (array, i, __memcpy_chk, 1,
 			      __memcpy_chk_sse2_unaligned)
-	      IFUNC_IMPL_ADD (array, i, __memcpy_chk, 1,
-			      __memcpy_chk_sse2_unaligned_erms)
-	      IFUNC_IMPL_ADD (array, i, __memcpy_chk, 1,
-			      __memcpy_chk_erms))
+	      X86_IFUNC_IMPL_ADD_V2 (array, i, __memcpy_chk, 1,
+			      __memcpy_chk_sse2_unaligned_erms))
 #endif
 
   /* Support sysdeps/x86_64/multiarch/memcpy.c.  */
   IFUNC_IMPL (i, name, memcpy,
-	      IFUNC_IMPL_ADD (array, i, memcpy,
+	      IFUNC_IMPL_ADD (array, i, memcpy, 1,
+			      __memcpy_erms)
+	      X86_IFUNC_IMPL_ADD_V4 (array, i, memcpy,
+			      CPU_FEATURE_USABLE (AVX512F),
+			      __memcpy_avx512_no_vzeroupper)
+	      X86_IFUNC_IMPL_ADD_V4 (array, i, memcpy,
+			      CPU_FEATURE_USABLE (AVX512VL),
+			      __memcpy_avx512_unaligned)
+	      X86_IFUNC_IMPL_ADD_V4 (array, i, memcpy,
+			      CPU_FEATURE_USABLE (AVX512VL),
+			      __memcpy_avx512_unaligned_erms)
+	      X86_IFUNC_IMPL_ADD_V4 (array, i, memcpy,
+			      CPU_FEATURE_USABLE (AVX512VL),
+			      __memcpy_evex_unaligned)
+	      X86_IFUNC_IMPL_ADD_V4 (array, i, memcpy,
+			      CPU_FEATURE_USABLE (AVX512VL),
+			      __memcpy_evex_unaligned_erms)
+	      X86_IFUNC_IMPL_ADD_V3 (array, i, memcpy,
 			      CPU_FEATURE_USABLE (AVX),
 			      __memcpy_avx_unaligned)
-	      IFUNC_IMPL_ADD (array, i, memcpy,
+	      X86_IFUNC_IMPL_ADD_V3 (array, i, memcpy,
 			      CPU_FEATURE_USABLE (AVX),
 			      __memcpy_avx_unaligned_erms)
-	      IFUNC_IMPL_ADD (array, i, memcpy,
+	      X86_IFUNC_IMPL_ADD_V3 (array, i, memcpy,
 			      (CPU_FEATURE_USABLE (AVX)
 			       && CPU_FEATURE_USABLE (RTM)),
 			      __memcpy_avx_unaligned_rtm)
-	      IFUNC_IMPL_ADD (array, i, memcpy,
+	      X86_IFUNC_IMPL_ADD_V3 (array, i, memcpy,
 			      (CPU_FEATURE_USABLE (AVX)
 			       && CPU_FEATURE_USABLE (RTM)),
 			      __memcpy_avx_unaligned_erms_rtm)
-	      IFUNC_IMPL_ADD (array, i, memcpy,
-			      CPU_FEATURE_USABLE (AVX512VL),
-			      __memcpy_evex_unaligned)
-	      IFUNC_IMPL_ADD (array, i, memcpy,
-			      CPU_FEATURE_USABLE (AVX512VL),
-			      __memcpy_evex_unaligned_erms)
-	      IFUNC_IMPL_ADD (array, i, memcpy, CPU_FEATURE_USABLE (SSSE3),
+	      /* By V3 we assume fast aligned copy.  */
+	      X86_IFUNC_IMPL_ADD_V2 (array, i, memcpy,
+			      CPU_FEATURE_USABLE (SSSE3),
 			      __memcpy_ssse3)
-	      IFUNC_IMPL_ADD (array, i, memcpy,
-			      CPU_FEATURE_USABLE (AVX512F),
-			      __memcpy_avx512_no_vzeroupper)
-	      IFUNC_IMPL_ADD (array, i, memcpy,
-			      CPU_FEATURE_USABLE (AVX512VL),
-			      __memcpy_avx512_unaligned)
-	      IFUNC_IMPL_ADD (array, i, memcpy,
-			      CPU_FEATURE_USABLE (AVX512VL),
-			      __memcpy_avx512_unaligned_erms)
-	      IFUNC_IMPL_ADD (array, i, memcpy, 1, __memcpy_sse2_unaligned)
-	      IFUNC_IMPL_ADD (array, i, memcpy, 1,
-			      __memcpy_sse2_unaligned_erms)
-	      IFUNC_IMPL_ADD (array, i, memcpy, 1, __memcpy_erms))
+	      /* Can be lowered to V1 if a general V2 implementation is
+             added (ssse3 is too optimized around aligned copy to be
+             better as general purpose memmove).  */
+	      X86_IFUNC_IMPL_ADD_V2 (array, i, memcpy, 1,
+			      __memcpy_sse2_unaligned)
+	      X86_IFUNC_IMPL_ADD_V2 (array, i, memcpy, 1,
+			      __memcpy_sse2_unaligned_erms))
 
 #ifdef SHARED
   /* Support sysdeps/x86_64/multiarch/mempcpy_chk.c.  */
   IFUNC_IMPL (i, name, __mempcpy_chk,
-	      IFUNC_IMPL_ADD (array, i, __mempcpy_chk,
+	      IFUNC_IMPL_ADD (array, i, __mempcpy_chk, 1,
+			      __mempcpy_chk_erms)
+	      X86_IFUNC_IMPL_ADD_V4 (array, i, __mempcpy_chk,
 			      CPU_FEATURE_USABLE (AVX512F),
 			      __mempcpy_chk_avx512_no_vzeroupper)
-	      IFUNC_IMPL_ADD (array, i, __mempcpy_chk,
+	      X86_IFUNC_IMPL_ADD_V4 (array, i, __mempcpy_chk,
 			      CPU_FEATURE_USABLE (AVX512VL),
 			      __mempcpy_chk_avx512_unaligned)
-	      IFUNC_IMPL_ADD (array, i, __mempcpy_chk,
+	      X86_IFUNC_IMPL_ADD_V4 (array, i, __mempcpy_chk,
 			      CPU_FEATURE_USABLE (AVX512VL),
 			      __mempcpy_chk_avx512_unaligned_erms)
-	      IFUNC_IMPL_ADD (array, i, __mempcpy_chk,
+	      X86_IFUNC_IMPL_ADD_V4 (array, i, __mempcpy_chk,
+			      CPU_FEATURE_USABLE (AVX512VL),
+			      __mempcpy_chk_evex_unaligned)
+	      X86_IFUNC_IMPL_ADD_V4 (array, i, __mempcpy_chk,
+			      CPU_FEATURE_USABLE (AVX512VL),
+			      __mempcpy_chk_evex_unaligned_erms)
+	      X86_IFUNC_IMPL_ADD_V3 (array, i, __mempcpy_chk,
 			      CPU_FEATURE_USABLE (AVX),
 			      __mempcpy_chk_avx_unaligned)
-	      IFUNC_IMPL_ADD (array, i, __mempcpy_chk,
+	      X86_IFUNC_IMPL_ADD_V3 (array, i, __mempcpy_chk,
 			      CPU_FEATURE_USABLE (AVX),
 			      __mempcpy_chk_avx_unaligned_erms)
-	      IFUNC_IMPL_ADD (array, i, __mempcpy_chk,
+	      X86_IFUNC_IMPL_ADD_V3 (array, i, __mempcpy_chk,
 			      (CPU_FEATURE_USABLE (AVX)
 			       && CPU_FEATURE_USABLE (RTM)),
 			      __mempcpy_chk_avx_unaligned_rtm)
-	      IFUNC_IMPL_ADD (array, i, __mempcpy_chk,
+	      X86_IFUNC_IMPL_ADD_V3 (array, i, __mempcpy_chk,
 			      (CPU_FEATURE_USABLE (AVX)
 			       && CPU_FEATURE_USABLE (RTM)),
 			      __mempcpy_chk_avx_unaligned_erms_rtm)
-	      IFUNC_IMPL_ADD (array, i, __mempcpy_chk,
-			      CPU_FEATURE_USABLE (AVX512VL),
-			      __mempcpy_chk_evex_unaligned)
-	      IFUNC_IMPL_ADD (array, i, __mempcpy_chk,
-			      CPU_FEATURE_USABLE (AVX512VL),
-			      __mempcpy_chk_evex_unaligned_erms)
-	      IFUNC_IMPL_ADD (array, i, __mempcpy_chk,
+	      /* By V3 we assume fast aligned copy.  */
+	      X86_IFUNC_IMPL_ADD_V2 (array, i, __mempcpy_chk,
 			      CPU_FEATURE_USABLE (SSSE3),
 			      __mempcpy_chk_ssse3)
-	      IFUNC_IMPL_ADD (array, i, __mempcpy_chk, 1,
+	      /* Can be lowered to V1 if a general V2 implementation is
+             added (ssse3 is too optimized around aligned copy to be
+             better as general purpose memmove).  */
+	      X86_IFUNC_IMPL_ADD_V2 (array, i, __mempcpy_chk, 1,
 			      __mempcpy_chk_sse2_unaligned)
-	      IFUNC_IMPL_ADD (array, i, __mempcpy_chk, 1,
-			      __mempcpy_chk_sse2_unaligned_erms)
-	      IFUNC_IMPL_ADD (array, i, __mempcpy_chk, 1,
-			      __mempcpy_chk_erms))
+	      X86_IFUNC_IMPL_ADD_V2 (array, i, __mempcpy_chk, 1,
+			      __mempcpy_chk_sse2_unaligned_erms))
 #endif
 
   /* Support sysdeps/x86_64/multiarch/mempcpy.c.  */
   IFUNC_IMPL (i, name, mempcpy,
-	      IFUNC_IMPL_ADD (array, i, mempcpy,
+	      IFUNC_IMPL_ADD (array, i, mempcpy, 1,
+			      __mempcpy_erms)
+	      X86_IFUNC_IMPL_ADD_V4 (array, i, mempcpy,
 			      CPU_FEATURE_USABLE (AVX512F),
 			      __mempcpy_avx512_no_vzeroupper)
-	      IFUNC_IMPL_ADD (array, i, mempcpy,
+	      X86_IFUNC_IMPL_ADD_V4 (array, i, mempcpy,
 			      CPU_FEATURE_USABLE (AVX512VL),
 			      __mempcpy_avx512_unaligned)
-	      IFUNC_IMPL_ADD (array, i, mempcpy,
+	      X86_IFUNC_IMPL_ADD_V4 (array, i, mempcpy,
 			      CPU_FEATURE_USABLE (AVX512VL),
 			      __mempcpy_avx512_unaligned_erms)
-	      IFUNC_IMPL_ADD (array, i, mempcpy,
+	      X86_IFUNC_IMPL_ADD_V4 (array, i, mempcpy,
+			      CPU_FEATURE_USABLE (AVX512VL),
+			      __mempcpy_evex_unaligned)
+	      X86_IFUNC_IMPL_ADD_V4 (array, i, mempcpy,
+			      CPU_FEATURE_USABLE (AVX512VL),
+			      __mempcpy_evex_unaligned_erms)
+	      X86_IFUNC_IMPL_ADD_V3 (array, i, mempcpy,
 			      CPU_FEATURE_USABLE (AVX),
 			      __mempcpy_avx_unaligned)
-	      IFUNC_IMPL_ADD (array, i, mempcpy,
+	      X86_IFUNC_IMPL_ADD_V3 (array, i, mempcpy,
 			      CPU_FEATURE_USABLE (AVX),
 			      __mempcpy_avx_unaligned_erms)
-	      IFUNC_IMPL_ADD (array, i, mempcpy,
+	      X86_IFUNC_IMPL_ADD_V3 (array, i, mempcpy,
 			      (CPU_FEATURE_USABLE (AVX)
 			       && CPU_FEATURE_USABLE (RTM)),
 			      __mempcpy_avx_unaligned_rtm)
-	      IFUNC_IMPL_ADD (array, i, mempcpy,
+	      X86_IFUNC_IMPL_ADD_V3 (array, i, mempcpy,
 			      (CPU_FEATURE_USABLE (AVX)
 			       && CPU_FEATURE_USABLE (RTM)),
 			      __mempcpy_avx_unaligned_erms_rtm)
-	      IFUNC_IMPL_ADD (array, i, mempcpy,
-			      CPU_FEATURE_USABLE (AVX512VL),
-			      __mempcpy_evex_unaligned)
-	      IFUNC_IMPL_ADD (array, i, mempcpy,
-			      CPU_FEATURE_USABLE (AVX512VL),
-			      __mempcpy_evex_unaligned_erms)
-	      IFUNC_IMPL_ADD (array, i, mempcpy, CPU_FEATURE_USABLE (SSSE3),
+	      /* By V3 we assume fast aligned copy.  */
+	      X86_IFUNC_IMPL_ADD_V2 (array, i, mempcpy,
+			      CPU_FEATURE_USABLE (SSSE3),
 			      __mempcpy_ssse3)
-	      IFUNC_IMPL_ADD (array, i, mempcpy, 1,
+	      /* Can be lowered to V1 if a general V2 implementation is
+             added (ssse3 is too optimized around aligned copy to be
+             better as general purpose memmove).  */
+	      X86_IFUNC_IMPL_ADD_V2 (array, i, mempcpy, 1,
 			      __mempcpy_sse2_unaligned)
-	      IFUNC_IMPL_ADD (array, i, mempcpy, 1,
-			      __mempcpy_sse2_unaligned_erms)
-	      IFUNC_IMPL_ADD (array, i, mempcpy, 1, __mempcpy_erms))
+	      X86_IFUNC_IMPL_ADD_V2 (array, i, mempcpy, 1,
+			      __mempcpy_sse2_unaligned_erms))
 
   /* Support sysdeps/x86_64/multiarch/strncmp.c.  */
   IFUNC_IMPL (i, name, strncmp,
diff --git a/sysdeps/x86_64/multiarch/ifunc-memmove.h b/sysdeps/x86_64/multiarch/ifunc-memmove.h
index fb01fbb301..1643d32887 100644
--- a/sysdeps/x86_64/multiarch/ifunc-memmove.h
+++ b/sysdeps/x86_64/multiarch/ifunc-memmove.h
@@ -20,11 +20,19 @@ 
 #include <init-arch.h>
 
 extern __typeof (REDIRECT_NAME) OPTIMIZE (erms) attribute_hidden;
-extern __typeof (REDIRECT_NAME) OPTIMIZE (sse2_unaligned)
+
+extern __typeof (REDIRECT_NAME) OPTIMIZE (avx512_unaligned)
   attribute_hidden;
-extern __typeof (REDIRECT_NAME) OPTIMIZE (sse2_unaligned_erms)
+extern __typeof (REDIRECT_NAME) OPTIMIZE (avx512_unaligned_erms)
   attribute_hidden;
-extern __typeof (REDIRECT_NAME) OPTIMIZE (ssse3) attribute_hidden;
+extern __typeof (REDIRECT_NAME) OPTIMIZE (avx512_no_vzeroupper)
+  attribute_hidden;
+
+extern __typeof (REDIRECT_NAME) OPTIMIZE (evex_unaligned)
+  attribute_hidden;
+extern __typeof (REDIRECT_NAME) OPTIMIZE (evex_unaligned_erms)
+  attribute_hidden;
+
 extern __typeof (REDIRECT_NAME) OPTIMIZE (avx_unaligned) attribute_hidden;
 extern __typeof (REDIRECT_NAME) OPTIMIZE (avx_unaligned_erms)
   attribute_hidden;
@@ -32,30 +40,27 @@  extern __typeof (REDIRECT_NAME) OPTIMIZE (avx_unaligned_rtm)
   attribute_hidden;
 extern __typeof (REDIRECT_NAME) OPTIMIZE (avx_unaligned_erms_rtm)
   attribute_hidden;
-extern __typeof (REDIRECT_NAME) OPTIMIZE (evex_unaligned)
-  attribute_hidden;
-extern __typeof (REDIRECT_NAME) OPTIMIZE (evex_unaligned_erms)
-  attribute_hidden;
-extern __typeof (REDIRECT_NAME) OPTIMIZE (avx512_unaligned)
-  attribute_hidden;
-extern __typeof (REDIRECT_NAME) OPTIMIZE (avx512_unaligned_erms)
+
+extern __typeof (REDIRECT_NAME) OPTIMIZE (ssse3) attribute_hidden;
+
+extern __typeof (REDIRECT_NAME) OPTIMIZE (sse2_unaligned)
   attribute_hidden;
-extern __typeof (REDIRECT_NAME) OPTIMIZE (avx512_no_vzeroupper)
+extern __typeof (REDIRECT_NAME) OPTIMIZE (sse2_unaligned_erms)
   attribute_hidden;
 
 static inline void *
 IFUNC_SELECTOR (void)
 {
-  const struct cpu_features* cpu_features = __get_cpu_features ();
+  const struct cpu_features *cpu_features = __get_cpu_features ();
 
   if (CPU_FEATURES_ARCH_P (cpu_features, Prefer_ERMS)
       || CPU_FEATURES_ARCH_P (cpu_features, Prefer_FSRM))
     return OPTIMIZE (erms);
 
-  if (CPU_FEATURE_USABLE_P (cpu_features, AVX512F)
+  if (X86_ISA_CPU_FEATURE_USABLE_P (cpu_features, AVX512F)
       && !CPU_FEATURES_ARCH_P (cpu_features, Prefer_No_AVX512))
     {
-      if (CPU_FEATURE_USABLE_P (cpu_features, AVX512VL))
+      if (X86_ISA_CPU_FEATURE_USABLE_P (cpu_features, AVX512VL))
 	{
 	  if (CPU_FEATURE_USABLE_P (cpu_features, ERMS))
 	    return OPTIMIZE (avx512_unaligned_erms);
@@ -66,9 +71,10 @@  IFUNC_SELECTOR (void)
       return OPTIMIZE (avx512_no_vzeroupper);
     }
 
-  if (CPU_FEATURES_ARCH_P (cpu_features, AVX_Fast_Unaligned_Load))
+  if (X86_ISA_CPU_FEATURES_ARCH_P (cpu_features,
+				   AVX_Fast_Unaligned_Load, ))
     {
-      if (CPU_FEATURE_USABLE_P (cpu_features, AVX512VL))
+      if (X86_ISA_CPU_FEATURE_USABLE_P (cpu_features, AVX512VL))
 	{
 	  if (CPU_FEATURE_USABLE_P (cpu_features, ERMS))
 	    return OPTIMIZE (evex_unaligned_erms);
@@ -84,7 +90,8 @@  IFUNC_SELECTOR (void)
 	  return OPTIMIZE (avx_unaligned_rtm);
 	}
 
-      if (!CPU_FEATURES_ARCH_P (cpu_features, Prefer_No_VZEROUPPER))
+      if (X86_ISA_CPU_FEATURES_ARCH_P (cpu_features,
+				       Prefer_No_VZEROUPPER, !))
 	{
 	  if (CPU_FEATURE_USABLE_P (cpu_features, ERMS))
 	    return OPTIMIZE (avx_unaligned_erms);
@@ -93,7 +100,11 @@  IFUNC_SELECTOR (void)
 	}
     }
 
-  if (CPU_FEATURE_USABLE_P (cpu_features, SSSE3)
+  if (X86_ISA_CPU_FEATURE_USABLE_P (cpu_features, SSSE3)
+      /* Leave this as runtime check.  The SSSE3 is optimized almost
+         exclusively for avoiding unaligned memory access during the
+         copy and by and large is not better than the sse2
+         implementation as a general purpose memmove.  */
       && !CPU_FEATURES_ARCH_P (cpu_features, Fast_Unaligned_Copy))
     {
       return OPTIMIZE (ssse3);
diff --git a/sysdeps/x86_64/multiarch/memmove-avx-unaligned-erms.S b/sysdeps/x86_64/multiarch/memmove-avx-unaligned-erms.S
index 975ae6c051..a14b155667 100644
--- a/sysdeps/x86_64/multiarch/memmove-avx-unaligned-erms.S
+++ b/sysdeps/x86_64/multiarch/memmove-avx-unaligned-erms.S
@@ -1,12 +1,23 @@ 
-#if IS_IN (libc)
+#include <isa-level.h>
+
+#if ISA_SHOULD_BUILD (3)
+
 # define VEC_SIZE	32
 # define VEC(i)		ymm##i
 # define VMOVNT		vmovntdq
 # define VMOVU		vmovdqu
 # define VMOVA		vmovdqa
 # define MOV_SIZE	4
+
 # define SECTION(p)		p##.avx
-# define MEMMOVE_SYMBOL(p,s)	p##_avx_##s
+
+# ifndef MEMMOVE_SYMBOL
+#  define MEMMOVE_SYMBOL(p,s)	p##_avx_##s
+# endif
 
 # include "memmove-vec-unaligned-erms.S"
+
+# if MINIMUM_X86_ISA_LEVEL == 3
+#  include "memmove-shlib-compat.h"
+# endif
 #endif
diff --git a/sysdeps/x86_64/multiarch/memmove-avx512-no-vzeroupper.S b/sysdeps/x86_64/multiarch/memmove-avx512-no-vzeroupper.S
index 42d15a142a..9c090d368b 100644
--- a/sysdeps/x86_64/multiarch/memmove-avx512-no-vzeroupper.S
+++ b/sysdeps/x86_64/multiarch/memmove-avx512-no-vzeroupper.S
@@ -17,8 +17,9 @@ 
    <https://www.gnu.org/licenses/>.  */
 
 #include <sysdep.h>
+#include <isa-level.h>
 
-#if IS_IN (libc)
+#if ISA_SHOULD_BUILD (4)
 
 # include "asm-syntax.h"
 
diff --git a/sysdeps/x86_64/multiarch/memmove-avx512-unaligned-erms.S b/sysdeps/x86_64/multiarch/memmove-avx512-unaligned-erms.S
index 0fa7126830..8d1568a7ba 100644
--- a/sysdeps/x86_64/multiarch/memmove-avx512-unaligned-erms.S
+++ b/sysdeps/x86_64/multiarch/memmove-avx512-unaligned-erms.S
@@ -1,4 +1,7 @@ 
-#if IS_IN (libc)
+#include <isa-level.h>
+
+#if ISA_SHOULD_BUILD (4)
+
 # define VEC_SIZE	64
 # define XMM0		xmm16
 # define XMM1		xmm17
@@ -26,8 +29,12 @@ 
 # define VMOVA		vmovdqa64
 # define VZEROUPPER
 # define MOV_SIZE	6
+
 # define SECTION(p)		p##.evex512
-# define MEMMOVE_SYMBOL(p,s)	p##_avx512_##s
+
+# ifndef MEMMOVE_SYMBOL
+#  define MEMMOVE_SYMBOL(p,s)	p##_avx512_##s
+# endif
 
 # include "memmove-vec-unaligned-erms.S"
 #endif
diff --git a/sysdeps/x86_64/multiarch/memmove-evex-unaligned-erms.S b/sysdeps/x86_64/multiarch/memmove-evex-unaligned-erms.S
index 88715441fe..2373017358 100644
--- a/sysdeps/x86_64/multiarch/memmove-evex-unaligned-erms.S
+++ b/sysdeps/x86_64/multiarch/memmove-evex-unaligned-erms.S
@@ -1,4 +1,7 @@ 
-#if IS_IN (libc)
+#include <isa-level.h>
+
+#if ISA_SHOULD_BUILD (4)
+
 # define VEC_SIZE	32
 # define XMM0		xmm16
 # define XMM1		xmm17
@@ -26,8 +29,17 @@ 
 # define VMOVA		vmovdqa64
 # define VZEROUPPER
 # define MOV_SIZE	6
+
 # define SECTION(p)		p##.evex
-# define MEMMOVE_SYMBOL(p,s)	p##_evex_##s
+
+# ifndef MEMMOVE_SYMBOL
+#  define MEMMOVE_SYMBOL(p,s)	p##_evex_##s
+# endif
 
 # include "memmove-vec-unaligned-erms.S"
+
+
+# if MINIMUM_X86_ISA_LEVEL == 4
+#  include "memmove-shlib-compat.h"
+# endif
 #endif
diff --git a/sysdeps/x86_64/multiarch/memmove-shlib-compat.h b/sysdeps/x86_64/multiarch/memmove-shlib-compat.h
new file mode 100644
index 0000000000..c0793d6eef
--- /dev/null
+++ b/sysdeps/x86_64/multiarch/memmove-shlib-compat.h
@@ -0,0 +1,26 @@ 
+/* Copyright (C) 2016-2022 Free Software Foundation, Inc.
+   This file is part of the GNU C Library.
+
+   The GNU C Library is free software; you can redistribute it and/or
+   modify it under the terms of the GNU Lesser General Public
+   License as published by the Free Software Foundation; either
+   version 2.1 of the License, or (at your option) any later version.
+
+   The GNU C Library 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
+   Lesser General Public License for more details.
+
+   You should have received a copy of the GNU Lesser General Public
+   License along with the GNU C Library; if not, see
+   <https://www.gnu.org/licenses/>.  */
+
+#if defined SHARED && IS_IN(libc)
+# include <shlib-compat.h>
+# if SHLIB_COMPAT(libc, GLIBC_2_2_5, GLIBC_2_14)
+/* Use __memmove_{isa_level}_unaligned to support overlapping
+   addresses.  */
+compat_symbol (libc, MEMMOVE_SYMBOL (__memmove, unaligned), memcpy,
+	       GLIBC_2_2_5);
+# endif
+#endif
diff --git a/sysdeps/x86_64/multiarch/memmove-sse2-unaligned-erms.S b/sysdeps/x86_64/multiarch/memmove-sse2-unaligned-erms.S
index 09e7c1d6cd..422a079902 100644
--- a/sysdeps/x86_64/multiarch/memmove-sse2-unaligned-erms.S
+++ b/sysdeps/x86_64/multiarch/memmove-sse2-unaligned-erms.S
@@ -16,18 +16,32 @@ 
    License along with the GNU C Library; if not, see
    <https://www.gnu.org/licenses/>.  */
 
-#if IS_IN (libc)
-# define MEMMOVE_SYMBOL(p,s)	p##_sse2_##s
-#else
-weak_alias (__mempcpy, mempcpy)
-#endif
+#include <isa-level.h>
+
+/* MINIMUM_X86_ISA_LEVEL <= 2 because there is no V2 implementation
+   so we need this to build for ISA V2 builds. */
+#if ISA_SHOULD_BUILD (2)
+
+# include <sysdep.h>
+
+# define VEC_SIZE	16
+# define VEC(i)		xmm##i
+# define PREFETCHNT	prefetchnta
+# define VMOVNT		movntdq
+/* Use movups and movaps for smaller code sizes.  */
+# define VMOVU		movups
+# define VMOVA		movaps
+# define MOV_SIZE	3
+
+# define SECTION(p)		p
+
+# ifndef MEMMOVE_SYMBOL
+#  define MEMMOVE_SYMBOL(p,s)	p##_sse2_##s
+# endif
 
-#include <sysdeps/x86_64/memmove.S>
+# include "multiarch/memmove-vec-unaligned-erms.S"
 
-#if defined SHARED && IS_IN (libc)
-# include <shlib-compat.h>
-# if SHLIB_COMPAT (libc, GLIBC_2_2_5, GLIBC_2_14)
-/* Use __memmove_sse2_unaligned to support overlapping addresses.  */
-compat_symbol (libc, __memmove_sse2_unaligned, memcpy, GLIBC_2_2_5);
+# if MINIMUM_X86_ISA_LEVEL <= 2
+#  include "memmove-shlib-compat.h"
 # endif
 #endif
diff --git a/sysdeps/x86_64/multiarch/memmove-ssse3.S b/sysdeps/x86_64/multiarch/memmove-ssse3.S
index 310ff62b86..60b2ba84e7 100644
--- a/sysdeps/x86_64/multiarch/memmove-ssse3.S
+++ b/sysdeps/x86_64/multiarch/memmove-ssse3.S
@@ -1,13 +1,17 @@ 
-#include <sysdep.h>
-
-#ifndef MEMMOVE
-# define MEMMOVE	__memmove_ssse3
-# define MEMMOVE_CHK	__memmove_chk_ssse3
-# define MEMCPY	__memcpy_ssse3
-# define MEMCPY_CHK	__memcpy_chk_ssse3
-# define MEMPCPY	__mempcpy_ssse3
-# define MEMPCPY_CHK	__mempcpy_chk_ssse3
-#endif
+#include <isa-level.h>
+
+#if ISA_SHOULD_BUILD (2)
+
+# include <sysdep.h>
+
+# ifndef MEMMOVE
+#  define MEMMOVE	__memmove_ssse3
+#  define MEMMOVE_CHK	__memmove_chk_ssse3
+#  define MEMCPY	__memcpy_ssse3
+#  define MEMCPY_CHK	__memcpy_chk_ssse3
+#  define MEMPCPY	__mempcpy_ssse3
+#  define MEMPCPY_CHK	__mempcpy_chk_ssse3
+# endif
 
 	.section .text.ssse3, "ax", @progbits
 ENTRY(MEMPCPY_CHK)
@@ -27,10 +31,10 @@  ENTRY(MEMMOVE_CHK)
 END(MEMMOVE_CHK)
 
 ENTRY_P2ALIGN(MEMMOVE, 6)
-# ifdef __ILP32__
+#  ifdef __ILP32__
 	/* Clear the upper 32 bits.  */
 	movl	%edx, %edx
-# endif
+#  endif
 	movq	%rdi, %rax
 L(start):
 	cmpq	$16, %rdx
@@ -124,11 +128,11 @@  L(more_2x_vec):
 	   loop.  */
 	movups	%xmm0, (%rdi)
 
-#ifdef SHARED_CACHE_SIZE_HALF
+# ifdef SHARED_CACHE_SIZE_HALF
 	cmp	$SHARED_CACHE_SIZE_HALF, %RDX_LP
-#else
+# else
 	cmp	__x86_shared_cache_size_half(%rip), %rdx
-#endif
+# endif
 	ja	L(large_memcpy)
 
 	leaq	-64(%rdi, %rdx), %r8
@@ -206,7 +210,7 @@  L(end_loop_fwd):
 
 	/* Extactly 64 bytes if `jmp L(end_loop_fwd)` is long encoding.
 	   60 bytes otherwise.  */
-#define ALIGNED_LOOP_FWD(align_by);	\
+# define ALIGNED_LOOP_FWD(align_by);	\
 	.p2align 6;	\
 L(loop_fwd_ ## align_by):	\
 	movaps	16(%rsi), %xmm0;	\
@@ -275,7 +279,7 @@  L(end_large_loop_fwd):
 
 	/* Size > 64 bytes and <= 96 bytes. 32-byte align between ensure
 	   96-byte spacing between each.  */
-#define ALIGNED_LARGE_LOOP_FWD(align_by);	\
+# define ALIGNED_LARGE_LOOP_FWD(align_by);	\
 	.p2align 5;	\
 L(large_loop_fwd_ ## align_by):	\
 	movaps	16(%rsi), %xmm0;	\
@@ -343,7 +347,7 @@  L(end_loop_bkwd):
 
 	/* Extactly 64 bytes if `jmp L(end_loop_bkwd)` is long encoding.
 	   60 bytes otherwise.  */
-#define ALIGNED_LOOP_BKWD(align_by);	\
+# define ALIGNED_LOOP_BKWD(align_by);	\
 	.p2align 6;	\
 L(loop_bkwd_ ## align_by):	\
 	movaps	32(%rsi), %xmm1;	\
@@ -382,3 +386,4 @@  END(MEMMOVE)
 
 strong_alias (MEMMOVE, MEMCPY)
 strong_alias (MEMMOVE_CHK, MEMCPY_CHK)
+#endif
diff --git a/sysdeps/x86_64/multiarch/rtld-memmove.S b/sysdeps/x86_64/multiarch/rtld-memmove.S
new file mode 100644
index 0000000000..1f3ad6433b
--- /dev/null
+++ b/sysdeps/x86_64/multiarch/rtld-memmove.S
@@ -0,0 +1,18 @@ 
+/* Copyright (C) 2022 Free Software Foundation, Inc.
+   This file is part of the GNU C Library.
+
+   The GNU C Library is free software; you can redistribute it and/or
+   modify it under the terms of the GNU Lesser General Public
+   License as published by the Free Software Foundation; either
+   version 2.1 of the License, or (at your option) any later version.
+
+   The GNU C Library 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
+   Lesser General Public License for more details.
+
+   You should have received a copy of the GNU Lesser General Public
+   License along with the GNU C Library; if not, see
+   <https://www.gnu.org/licenses/>.  */
+
+#include "../memmove.S"