From 351086591d938aaf884d475261ae96ec5da00384 Mon Sep 17 00:00:00 2001
From: dengjianbo <dengjianbo@loongson.cn>
Date: Wed, 13 Sep 2023 15:34:59 +0800
Subject: [PATCH 22/29] LoongArch: Add ifunc support for strcpy,
 stpcpy{aligned, unaligned, lsx, lasx}

According to glibc strcpy and stpcpy microbenchmark test results(changed
to use generic_strcpy and generic_stpcpy instead of strlen + memcpy),
comparing with the generic version, this implementation could reduce the
runtime as following:

Name              Percent of rutime reduced
strcpy-aligned    8%-45%
strcpy-unaligned  8%-48%, comparing with the aligned version, unaligned
                  version takes less instructions to copy the tail of data
		  which length is less than 8. it also has better performance
		  in case src and dest cannot be both aligned with 8bytes
strcpy-lsx        20%-80%
strcpy-lasx       15%-86%
stpcpy-aligned    6%-43%
stpcpy-unaligned  8%-48%
stpcpy-lsx        10%-80%
stpcpy-lasx       10%-87%

Signed-off-by: Peng Fan <fanpeng@loongson.cn>
Signed-off-by: ticat_fp <fanpeng@loongson.cn>
---
 sysdeps/loongarch/lp64/multiarch/Makefile     |   8 +
 .../lp64/multiarch/ifunc-impl-list.c          |  18 ++
 .../loongarch/lp64/multiarch/stpcpy-aligned.S |  27 +++
 .../loongarch/lp64/multiarch/stpcpy-lasx.S    |  22 ++
 sysdeps/loongarch/lp64/multiarch/stpcpy-lsx.S |  22 ++
 .../lp64/multiarch/stpcpy-unaligned.S         |  22 ++
 sysdeps/loongarch/lp64/multiarch/stpcpy.c     |  42 ++++
 .../loongarch/lp64/multiarch/strcpy-aligned.S | 202 ++++++++++++++++
 .../loongarch/lp64/multiarch/strcpy-lasx.S    | 215 ++++++++++++++++++
 sysdeps/loongarch/lp64/multiarch/strcpy-lsx.S | 212 +++++++++++++++++
 .../lp64/multiarch/strcpy-unaligned.S         | 138 +++++++++++
 sysdeps/loongarch/lp64/multiarch/strcpy.c     |  35 +++
 12 files changed, 963 insertions(+)
 create mode 100644 sysdeps/loongarch/lp64/multiarch/stpcpy-aligned.S
 create mode 100644 sysdeps/loongarch/lp64/multiarch/stpcpy-lasx.S
 create mode 100644 sysdeps/loongarch/lp64/multiarch/stpcpy-lsx.S
 create mode 100644 sysdeps/loongarch/lp64/multiarch/stpcpy-unaligned.S
 create mode 100644 sysdeps/loongarch/lp64/multiarch/stpcpy.c
 create mode 100644 sysdeps/loongarch/lp64/multiarch/strcpy-aligned.S
 create mode 100644 sysdeps/loongarch/lp64/multiarch/strcpy-lasx.S
 create mode 100644 sysdeps/loongarch/lp64/multiarch/strcpy-lsx.S
 create mode 100644 sysdeps/loongarch/lp64/multiarch/strcpy-unaligned.S
 create mode 100644 sysdeps/loongarch/lp64/multiarch/strcpy.c

diff --git a/sysdeps/loongarch/lp64/multiarch/Makefile b/sysdeps/loongarch/lp64/multiarch/Makefile
index 360a6718..39550bea 100644
--- a/sysdeps/loongarch/lp64/multiarch/Makefile
+++ b/sysdeps/loongarch/lp64/multiarch/Makefile
@@ -16,6 +16,14 @@ sysdep_routines += \
   strcmp-lsx \
   strncmp-aligned \
   strncmp-lsx \
+  strcpy-aligned \
+  strcpy-unaligned \
+  strcpy-lsx \
+  strcpy-lasx \
+  stpcpy-aligned \
+  stpcpy-unaligned \
+  stpcpy-lsx \
+  stpcpy-lasx \
   memcpy-aligned \
   memcpy-unaligned \
   memmove-unaligned \
diff --git a/sysdeps/loongarch/lp64/multiarch/ifunc-impl-list.c b/sysdeps/loongarch/lp64/multiarch/ifunc-impl-list.c
index e397d58c..39a14f1d 100644
--- a/sysdeps/loongarch/lp64/multiarch/ifunc-impl-list.c
+++ b/sysdeps/loongarch/lp64/multiarch/ifunc-impl-list.c
@@ -76,6 +76,24 @@ __libc_ifunc_impl_list (const char *name, struct libc_ifunc_impl *array,
 	      IFUNC_IMPL_ADD (array, i, strncmp, 1, __strncmp_aligned)
 	      )
 
+  IFUNC_IMPL (i, name, strcpy,
+#if !defined __loongarch_soft_float
+	      IFUNC_IMPL_ADD (array, i, strcpy, SUPPORT_LASX, __strcpy_lasx)
+	      IFUNC_IMPL_ADD (array, i, strcpy, SUPPORT_LSX, __strcpy_lsx)
+#endif
+	      IFUNC_IMPL_ADD (array, i, strcpy, SUPPORT_UAL, __strcpy_unaligned)
+	      IFUNC_IMPL_ADD (array, i, strcpy, 1, __strcpy_aligned)
+	      )
+
+  IFUNC_IMPL (i, name, stpcpy,
+#if !defined __loongarch_soft_float
+	      IFUNC_IMPL_ADD (array, i, stpcpy, SUPPORT_LASX, __stpcpy_lasx)
+	      IFUNC_IMPL_ADD (array, i, stpcpy, SUPPORT_LSX, __stpcpy_lsx)
+#endif
+	      IFUNC_IMPL_ADD (array, i, stpcpy, SUPPORT_UAL, __stpcpy_unaligned)
+	      IFUNC_IMPL_ADD (array, i, stpcpy, 1, __stpcpy_aligned)
+	      )
+
   IFUNC_IMPL (i, name, memcpy,
 #if !defined __loongarch_soft_float
               IFUNC_IMPL_ADD (array, i, memcpy, SUPPORT_LASX, __memcpy_lasx)
diff --git a/sysdeps/loongarch/lp64/multiarch/stpcpy-aligned.S b/sysdeps/loongarch/lp64/multiarch/stpcpy-aligned.S
new file mode 100644
index 00000000..1f763db6
--- /dev/null
+++ b/sysdeps/loongarch/lp64/multiarch/stpcpy-aligned.S
@@ -0,0 +1,27 @@
+/* stpcpy-aligned implementation is in strcpy-aligned.S.
+   Copyright (C) 2023 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 IS_IN (libc)
+# define STPCPY __stpcpy_aligned
+#else
+# define STPCPY __stpcpy
+#endif
+
+#define USE_AS_STPCPY
+#define STRCPY STPCPY
+#include "strcpy-aligned.S"
diff --git a/sysdeps/loongarch/lp64/multiarch/stpcpy-lasx.S b/sysdeps/loongarch/lp64/multiarch/stpcpy-lasx.S
new file mode 100644
index 00000000..13d6c953
--- /dev/null
+++ b/sysdeps/loongarch/lp64/multiarch/stpcpy-lasx.S
@@ -0,0 +1,22 @@
+/* stpcpy-lasx implementation is in strcpy-lasx.S.
+   Copyright (C) 2023 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/>.  */
+
+#define STPCPY __stpcpy_lasx
+#define USE_AS_STPCPY
+#define STRCPY STPCPY
+#include "strcpy-lasx.S"
diff --git a/sysdeps/loongarch/lp64/multiarch/stpcpy-lsx.S b/sysdeps/loongarch/lp64/multiarch/stpcpy-lsx.S
new file mode 100644
index 00000000..e0f17ab5
--- /dev/null
+++ b/sysdeps/loongarch/lp64/multiarch/stpcpy-lsx.S
@@ -0,0 +1,22 @@
+/* stpcpy-lsx implementation is in strcpy-lsx.S.
+   Copyright (C) 2023 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/>.  */
+
+#define STPCPY __stpcpy_lsx
+#define USE_AS_STPCPY
+#define STRCPY STPCPY
+#include "strcpy-lsx.S"
diff --git a/sysdeps/loongarch/lp64/multiarch/stpcpy-unaligned.S b/sysdeps/loongarch/lp64/multiarch/stpcpy-unaligned.S
new file mode 100644
index 00000000..cc2f9712
--- /dev/null
+++ b/sysdeps/loongarch/lp64/multiarch/stpcpy-unaligned.S
@@ -0,0 +1,22 @@
+/* stpcpy-unaligned implementation is in strcpy-unaligned.S.
+   Copyright (C) 2023 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/>.  */
+
+#define STPCPY __stpcpy_unaligned
+#define USE_AS_STPCPY
+#define STRCPY STPCPY
+#include "strcpy-unaligned.S"
diff --git a/sysdeps/loongarch/lp64/multiarch/stpcpy.c b/sysdeps/loongarch/lp64/multiarch/stpcpy.c
new file mode 100644
index 00000000..d4860d7a
--- /dev/null
+++ b/sysdeps/loongarch/lp64/multiarch/stpcpy.c
@@ -0,0 +1,42 @@
+/* Multiple versions of stpcpy.
+   All versions must be listed in ifunc-impl-list.c.
+   Copyright (C) 2017-2023 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/>.  */
+
+/* Define multiple versions only for the definition in libc.  */
+#if IS_IN (libc)
+# define stpcpy __redirect_stpcpy
+# define __stpcpy __redirect___stpcpy
+# define NO_MEMPCPY_STPCPY_REDIRECT
+# define __NO_STRING_INLINES
+# include <string.h>
+# undef stpcpy
+# undef __stpcpy
+
+# define SYMBOL_NAME stpcpy
+# include "ifunc-lasx.h"
+
+libc_ifunc_redirected (__redirect_stpcpy, __stpcpy, IFUNC_SELECTOR ());
+
+weak_alias (__stpcpy, stpcpy)
+# ifdef SHARED
+__hidden_ver1 (__stpcpy, __GI___stpcpy, __redirect___stpcpy)
+  __attribute__ ((visibility ("hidden"))) __attribute_copy__ (stpcpy);
+__hidden_ver1 (stpcpy, __GI_stpcpy, __redirect_stpcpy)
+  __attribute__ ((visibility ("hidden"))) __attribute_copy__ (stpcpy);
+# endif
+#endif
diff --git a/sysdeps/loongarch/lp64/multiarch/strcpy-aligned.S b/sysdeps/loongarch/lp64/multiarch/strcpy-aligned.S
new file mode 100644
index 00000000..4ed539fd
--- /dev/null
+++ b/sysdeps/loongarch/lp64/multiarch/strcpy-aligned.S
@@ -0,0 +1,202 @@
+/* Optimized strcpy stpcpy aligned implementation using basic LoongArch
+   instructions.
+   Copyright (C) 2023 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 <sysdep.h>
+#include <sys/regdef.h>
+#include <sys/asm.h>
+
+#if IS_IN (libc)
+# ifndef STRCPY
+#  define STRCPY __strcpy_aligned
+# endif
+#else
+# ifndef STRCPY
+#  define STRCPY strcpy
+# endif
+#endif
+
+LEAF(STRCPY, 6)
+    andi        a3, a0, 0x7
+    move        a2, a0
+    beqz        a3, L(dest_align)
+    sub.d       a5, a1, a3
+    addi.d      a5, a5, 8
+
+L(make_dest_align):
+    ld.b        t0, a1, 0
+    addi.d      a1, a1, 1
+    st.b        t0, a2, 0
+    addi.d      a2, a2, 1
+    beqz        t0, L(al_out)
+
+    bne         a1, a5, L(make_dest_align)
+
+L(dest_align):
+    andi        a4, a1, 7
+    bstrins.d   a1, zero, 2, 0
+
+    lu12i.w     t5, 0x1010
+    ld.d        t0, a1, 0
+    ori         t5, t5, 0x101
+    bstrins.d   t5, t5, 63, 32
+
+    slli.d      t6, t5, 0x7
+    bnez        a4, L(unalign)
+    sub.d       t1, t0, t5
+    andn        t2, t6, t0
+
+    and         t3, t1, t2
+    bnez        t3, L(al_end)
+
+L(al_loop):
+    st.d        t0, a2, 0
+    ld.d        t0, a1, 8
+
+    addi.d      a1, a1, 8
+    addi.d      a2, a2, 8
+    sub.d       t1, t0, t5
+    andn        t2, t6, t0
+
+    and         t3, t1, t2
+    beqz        t3, L(al_loop)
+
+L(al_end):
+    ctz.d       t1, t3
+    srli.d      t1, t1, 3
+    addi.d      t1, t1, 1
+
+    andi        a3, t1, 8
+    andi        a4, t1, 4
+    andi        a5, t1, 2
+    andi        a6, t1, 1
+
+L(al_end_8):
+    beqz        a3, L(al_end_4)
+    st.d        t0, a2, 0
+#ifdef USE_AS_STPCPY
+    addi.d      a0, a2, 7
+#endif
+    jr          ra
+L(al_end_4):
+    beqz        a4, L(al_end_2)
+    st.w        t0, a2, 0
+    addi.d      a2, a2, 4
+    srli.d      t0, t0, 32
+L(al_end_2):
+    beqz        a5, L(al_end_1)
+    st.h        t0, a2, 0
+    addi.d      a2, a2, 2
+    srli.d      t0, t0, 16
+L(al_end_1):
+    beqz        a6, L(al_out)
+    st.b        t0, a2, 0
+    addi.d      a2, a2, 1
+L(al_out):
+#ifdef USE_AS_STPCPY
+    addi.d      a0, a2, -1
+#endif
+    jr          ra
+
+    .align      4
+L(unalign):
+    slli.d      a5, a4, 3
+    li.d        t1, -1
+    sub.d       a6, zero, a5
+
+    srl.d       a7, t0, a5
+    sll.d       t7, t1, a6
+
+    or          t0, a7, t7
+    sub.d       t1, t0, t5
+    andn        t2, t6, t0
+    and         t3, t1, t2
+
+    bnez        t3, L(un_end)
+
+    ld.d        t4, a1, 8
+
+    sub.d       t1, t4, t5
+    andn        t2, t6, t4
+    sll.d       t0, t4, a6
+    and         t3, t1, t2
+
+    or          t0, t0, a7
+    bnez        t3, L(un_end_with_remaining)
+
+L(un_loop):
+    srl.d       a7, t4, a5
+
+    ld.d        t4, a1, 16
+    addi.d      a1, a1, 8
+
+    st.d        t0, a2, 0
+    addi.d      a2, a2, 8
+
+    sub.d       t1, t4, t5
+    andn        t2, t6, t4
+    sll.d       t0, t4, a6
+    and         t3, t1, t2
+
+    or          t0, t0, a7
+    beqz        t3, L(un_loop)
+
+L(un_end_with_remaining):
+    ctz.d       t1, t3
+    srli.d      t1, t1, 3
+    addi.d      t1, t1, 1
+    sub.d       t1, t1, a4
+
+    blt         t1, zero, L(un_end_less_8)
+    st.d        t0, a2, 0
+    addi.d      a2, a2, 8
+    beqz        t1, L(un_out)
+    srl.d       t0, t4, a5
+    b           L(un_end_less_8)
+
+L(un_end):
+    ctz.d       t1, t3
+    srli.d      t1, t1, 3
+    addi.d      t1, t1, 1
+
+L(un_end_less_8):
+    andi        a4, t1, 4
+    andi        a5, t1, 2
+    andi        a6, t1, 1
+L(un_end_4):
+    beqz        a4, L(un_end_2)
+    st.w        t0, a2, 0
+    addi.d      a2, a2, 4
+    srli.d      t0, t0, 32
+L(un_end_2):
+    beqz        a5, L(un_end_1)
+    st.h        t0, a2, 0
+    addi.d      a2, a2, 2
+    srli.d      t0, t0, 16
+L(un_end_1):
+    beqz        a6, L(un_out)
+    st.b        t0, a2, 0
+    addi.d      a2, a2, 1
+L(un_out):
+#ifdef USE_AS_STPCPY
+    addi.d      a0, a2, -1
+#endif
+    jr          ra
+END(STRCPY)
+
+libc_hidden_builtin_def (STRCPY)
diff --git a/sysdeps/loongarch/lp64/multiarch/strcpy-lasx.S b/sysdeps/loongarch/lp64/multiarch/strcpy-lasx.S
new file mode 100644
index 00000000..c2825612
--- /dev/null
+++ b/sysdeps/loongarch/lp64/multiarch/strcpy-lasx.S
@@ -0,0 +1,215 @@
+/* Optimized strcpy stpcpy implementation using LoongArch LASX instructions.
+   Copyright (C) 2023 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 <sysdep.h>
+#include <sys/regdef.h>
+#include <sys/asm.h>
+
+#if IS_IN (libc) && !defined __loongarch_soft_float
+
+# ifndef STRCPY
+#  define STRCPY __strcpy_lasx
+# endif
+
+# ifdef USE_AS_STPCPY
+#  define dstend a0
+# else
+#  define dstend a4
+# endif
+
+LEAF(STRCPY, 6)
+    ori             t8, zero, 0xfe0
+    andi            t0, a1, 0xfff
+    li.d            t7, -1
+    move            a2, a0
+
+    bltu            t8, t0, L(page_cross_start)
+L(start_entry):
+    xvld            xr0, a1, 0
+    li.d            t0, 32
+    andi            t1, a2, 0x1f
+
+    xvsetanyeqz.b   fcc0, xr0
+    sub.d           t0, t0, t1
+    bcnez           fcc0, L(end)
+    add.d           a1, a1, t0
+
+    xvst            xr0, a2, 0
+    andi            a3, a1, 0x1f
+    add.d           a2, a2, t0
+    bnez            a3, L(unaligned)
+
+
+    xvld            xr0, a1, 0
+    xvsetanyeqz.b   fcc0, xr0
+    bcnez           fcc0, L(al_end)
+L(al_loop):
+    xvst            xr0, a2, 0
+
+    xvld            xr0, a1, 32
+    addi.d          a2, a2, 32
+    addi.d          a1, a1, 32
+    xvsetanyeqz.b   fcc0, xr0
+
+    bceqz           fcc0, L(al_loop)
+L(al_end):
+    xvmsknz.b       xr0, xr0
+    xvpickve.w      xr1, xr0, 4
+    vilvl.h         vr0, vr1, vr0
+
+    movfr2gr.s      t0, fa0
+    cto.w           t0, t0
+    add.d           a1, a1, t0
+    xvld            xr0, a1, -31
+
+
+    add.d           dstend, a2, t0
+    xvst            xr0, dstend, -31
+    jr              ra
+    nop
+
+L(page_cross_start):
+    move            a4, a1
+    bstrins.d       a4, zero, 4, 0
+    xvld            xr0, a4, 0
+    xvmsknz.b       xr0, xr0
+
+    xvpickve.w      xr1, xr0, 4
+    vilvl.h         vr0, vr1, vr0
+    movfr2gr.s      t0, fa0
+    sra.w           t0, t0, a1
+
+    beq             t0, t7, L(start_entry)
+    b               L(tail)
+L(unaligned):
+    andi            t0, a1, 0xfff
+    bltu            t8, t0, L(un_page_cross)
+
+
+L(un_start_entry):
+    xvld            xr0, a1, 0
+    xvsetanyeqz.b   fcc0, xr0
+    bcnez           fcc0, L(un_end)
+    addi.d          a1, a1, 32
+
+L(un_loop):
+    xvst            xr0, a2, 0
+    andi            t0, a1, 0xfff
+    addi.d          a2, a2, 32
+    bltu            t8, t0, L(page_cross_loop)
+
+L(un_loop_entry):
+    xvld            xr0, a1, 0
+    addi.d          a1, a1, 32
+    xvsetanyeqz.b   fcc0, xr0
+    bceqz           fcc0, L(un_loop)
+
+    addi.d          a1, a1, -32
+L(un_end):
+    xvmsknz.b       xr0, xr0
+    xvpickve.w      xr1, xr0, 4
+    vilvl.h         vr0, vr1, vr0
+
+
+    movfr2gr.s      t0, fa0
+L(un_tail):
+    cto.w           t0, t0
+    add.d           a1, a1, t0
+    xvld            xr0, a1, -31
+
+    add.d           dstend, a2, t0
+    xvst            xr0, dstend, -31
+    jr              ra
+L(un_page_cross):
+    sub.d           a4, a1, a3
+
+    xvld            xr0, a4, 0
+    xvmsknz.b       xr0, xr0
+    xvpickve.w      xr1, xr0, 4
+    vilvl.h         vr0, vr1, vr0
+
+    movfr2gr.s      t0, fa0
+    sra.w           t0, t0, a1
+    beq             t0, t7, L(un_start_entry)
+    b               L(un_tail)
+
+
+L(page_cross_loop):
+    sub.d           a4, a1, a3
+    xvld            xr0, a4, 0
+    xvmsknz.b       xr0, xr0
+    xvpickve.w      xr1, xr0, 4
+
+    vilvl.h         vr0, vr1, vr0
+    movfr2gr.s      t0, fa0
+    sra.w           t0, t0, a1
+    beq             t0, t7, L(un_loop_entry)
+
+    b               L(un_tail)
+L(end):
+    xvmsknz.b       xr0, xr0
+    xvpickve.w      xr1, xr0, 4
+    vilvl.h         vr0, vr1, vr0
+
+    movfr2gr.s      t0, fa0
+L(tail):
+    cto.w           t0, t0
+    add.d           dstend, a2, t0
+    add.d           a5, a1, t0
+
+L(less_32):
+    srli.d          t1, t0, 4
+    beqz            t1, L(less_16)
+    vld             vr0, a1, 0
+    vld             vr1, a5, -15
+
+    vst             vr0, a2, 0
+    vst             vr1, dstend, -15
+    jr              ra
+L(less_16):
+    srli.d          t1, t0, 3
+
+    beqz            t1, L(less_8)
+    ld.d            t2, a1, 0
+    ld.d            t3, a5, -7
+    st.d            t2, a2, 0
+
+    st.d            t3, dstend, -7
+    jr              ra
+L(less_8):
+    li.d            t1, 3
+    bltu            t0, t1, L(less_3)
+
+    ld.w            t2, a1, 0
+    ld.w            t3, a5, -3
+    st.w            t2, a2, 0
+    st.w            t3, dstend, -3
+
+    jr              ra
+L(less_3):
+    beqz            t0, L(zero_byte)
+    ld.h            t2, a1, 0
+
+    st.h            t2, a2, 0
+L(zero_byte):
+    st.b            zero, dstend, 0
+    jr              ra
+END(STRCPY)
+
+libc_hidden_builtin_def (STRCPY)
+#endif
diff --git a/sysdeps/loongarch/lp64/multiarch/strcpy-lsx.S b/sysdeps/loongarch/lp64/multiarch/strcpy-lsx.S
new file mode 100644
index 00000000..fc2498f7
--- /dev/null
+++ b/sysdeps/loongarch/lp64/multiarch/strcpy-lsx.S
@@ -0,0 +1,212 @@
+/* Optimized strcpy stpcpy implementation using LoongArch LSX instructions.
+   Copyright (C) 2023 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 <sysdep.h>
+#include <sys/regdef.h>
+#include <sys/asm.h>
+
+#if IS_IN (libc) && !defined __loongarch_soft_float
+
+# ifndef STRCPY
+#  define STRCPY __strcpy_lsx
+# endif
+
+LEAF(STRCPY, 6)
+    pcalau12i       t0, %pc_hi20(L(INDEX))
+    andi            a4, a1, 0xf
+    vld             vr1, t0, %pc_lo12(L(INDEX))
+    move            a2, a0
+
+    beqz            a4, L(load_start)
+    xor             t0, a1, a4
+    vld             vr0, t0, 0
+    vreplgr2vr.b    vr2, a4
+
+    vadd.b          vr2, vr2, vr1
+    vshuf.b         vr0, vr2, vr0, vr2
+    vsetanyeqz.b    fcc0, vr0
+    bcnez           fcc0, L(end)
+
+L(load_start):
+    vld             vr0, a1, 0
+    li.d            t1, 16
+    andi            a3, a2, 0xf
+    vsetanyeqz.b    fcc0, vr0
+
+
+    sub.d           t0, t1, a3
+    bcnez           fcc0, L(end)
+    add.d           a1, a1, t0
+    vst             vr0, a2, 0
+
+    andi            a3, a1, 0xf
+    add.d           a2, a2, t0
+    bnez            a3, L(unaligned)
+    vld             vr0, a1, 0
+
+    vsetanyeqz.b    fcc0, vr0
+    bcnez           fcc0, L(al_end)
+L(al_loop):
+    vst             vr0, a2, 0
+    vld             vr0, a1, 16
+
+    addi.d          a2, a2, 16
+    addi.d          a1, a1, 16
+    vsetanyeqz.b    fcc0, vr0
+    bceqz           fcc0, L(al_loop)
+
+
+L(al_end):
+    vmsknz.b        vr1, vr0
+    movfr2gr.s      t0, fa1
+    cto.w           t0, t0
+    add.d           a1, a1, t0
+
+    vld             vr0, a1, -15
+# ifdef USE_AS_STPCPY
+    add.d           a0, a2, t0
+    vst             vr0, a0, -15
+# else
+    add.d           a2, a2, t0
+    vst             vr0, a2, -15
+# endif
+    jr              ra
+
+L(end):
+    vmsknz.b        vr1, vr0
+    movfr2gr.s      t0, fa1
+    cto.w           t0, t0
+    addi.d          t0, t0, 1
+
+L(end_16):
+    andi            t1, t0, 16
+    beqz            t1, L(end_8)
+    vst             vr0, a2, 0
+# ifdef USE_AS_STPCPY
+    addi.d          a0, a2, 15
+# endif
+    jr              ra
+
+L(end_8):
+    andi            t2, t0, 8
+    andi            t3, t0, 4
+    andi            t4, t0, 2
+    andi            t5, t0, 1
+
+    beqz            t2, L(end_4)
+    vstelm.d        vr0, a2, 0, 0
+    addi.d          a2, a2, 8
+    vbsrl.v         vr0, vr0, 8
+
+L(end_4):
+    beqz            t3, L(end_2)
+    vstelm.w        vr0, a2, 0, 0
+    addi.d          a2, a2, 4
+    vbsrl.v         vr0, vr0, 4
+
+L(end_2):
+    beqz            t4, L(end_1)
+    vstelm.h        vr0, a2, 0, 0
+    addi.d          a2, a2, 2
+    vbsrl.v         vr0, vr0, 2
+
+
+L(end_1):
+    beqz            t5, L(out)
+    vstelm.b        vr0, a2, 0, 0
+    addi.d          a2, a2, 1
+L(out):
+# ifdef USE_AS_STPCPY
+    addi.d          a0, a2, -1
+# endif
+    jr              ra
+
+    .align          4
+L(unaligned):
+    bstrins.d       a1, zero, 3, 0
+    vld             vr2, a1, 0
+    vreplgr2vr.b    vr3, a3
+    vslt.b          vr4, vr1, vr3
+
+    vor.v           vr0, vr2, vr4
+    vsetanyeqz.b    fcc0, vr0
+    bcnez           fcc0, L(un_first_end)
+    vld             vr0, a1, 16
+
+    vadd.b          vr3, vr3, vr1
+    vshuf.b         vr4, vr0, vr2, vr3
+    vsetanyeqz.b    fcc0, vr0
+    bcnez           fcc0, L(un_end)
+
+
+    vor.v           vr2, vr0, vr0
+    addi.d          a1, a1, 16
+L(un_loop):
+    vld             vr0, a1, 16
+    vst             vr4, a2, 0
+
+    addi.d          a2, a2, 16
+    vshuf.b         vr4, vr0, vr2, vr3
+    vsetanyeqz.b    fcc0, vr0
+    bcnez           fcc0, L(un_end)
+
+    vld             vr2, a1, 32
+    vst             vr4, a2, 0
+    addi.d          a1, a1, 32
+    addi.d          a2, a2, 16
+
+    vshuf.b         vr4, vr2, vr0, vr3
+    vsetanyeqz.b    fcc0, vr2
+    bceqz           fcc0, L(un_loop)
+    vor.v           vr0, vr2, vr2
+
+
+    addi.d          a1, a1, -16
+L(un_end):
+    vsetanyeqz.b    fcc0, vr4
+    bcnez           fcc0, 1f
+    vst             vr4, a2, 0
+
+1:
+    vmsknz.b        vr1, vr0
+    movfr2gr.s      t0, fa1
+    cto.w           t0, t0
+    add.d           a1, a1, t0
+
+    vld             vr0, a1, 1
+    add.d           a2, a2, t0
+    sub.d           a2, a2, a3
+    vst             vr0, a2, 1
+# ifdef USE_AS_STPCPY
+    addi.d          a0, a2, 16
+# endif
+    jr              ra
+L(un_first_end):
+    addi.d          a2, a2, -16
+    addi.d          a1, a1, -16
+    b               1b
+END(STRCPY)
+
+    .section        .rodata.cst16,"M",@progbits,16
+    .align          4
+L(INDEX):
+    .dword          0x0706050403020100
+    .dword          0x0f0e0d0c0b0a0908
+
+libc_hidden_builtin_def (STRCPY)
+#endif
diff --git a/sysdeps/loongarch/lp64/multiarch/strcpy-unaligned.S b/sysdeps/loongarch/lp64/multiarch/strcpy-unaligned.S
new file mode 100644
index 00000000..9e31883b
--- /dev/null
+++ b/sysdeps/loongarch/lp64/multiarch/strcpy-unaligned.S
@@ -0,0 +1,138 @@
+/* Optimized strcpy unaligned implementation using basic LoongArch
+   instructions.
+   Copyright (C) 2023 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 <sysdep.h>
+#include <sys/regdef.h>
+#include <sys/asm.h>
+
+#if IS_IN (libc)
+
+# ifndef STRCPY
+#  define STRCPY __strcpy_unaligned
+# endif
+
+# ifdef USE_AS_STPCPY
+#  define dstend a0
+# else
+#  define dstend a4
+# endif
+
+LEAF(STRCPY, 6)
+    lu12i.w     t5, 0x01010
+    li.w        t0, 0xff8
+    ori         t5, t5, 0x101
+    andi        t1, a1, 0xfff
+
+    bstrins.d   t5, t5, 63, 32
+    move        a2, a0
+    slli.d      t6, t5, 7
+    bltu        t0, t1, L(page_cross)
+
+L(start_entry):
+    ld.d        t0, a1, 0
+    li.d        t3, 8
+    andi        a3, a1, 0x7
+    sub.d       t1, t0, t5
+
+    andn        t2, t6, t0
+    sub.d       t3, t3, a3
+    and         t1, t1, t2
+    bnez        t1, L(end)
+
+
+    add.d       a1, a1, t3
+    st.d        t0, a2, 0
+    add.d       a2, a2, t3
+    ld.d        t0, a1, 0
+
+    sub.d       t1, t0, t5
+    andn        t2, t6, t0
+    and         t1, t1, t2
+    bnez        t1, L(long_end)
+
+L(loop):
+    st.d        t0, a2, 0
+    ld.d        t0, a1, 8
+    addi.d      a2, a2, 8
+    addi.d      a1, a1, 8
+
+    sub.d       t1, t0, t5
+    andn        t2, t6, t0
+    and         t1, t1, t2
+    beqz        t1, L(loop)
+
+
+L(long_end):
+    ctz.d       t1, t1
+    srli.d      t1, t1, 3
+    add.d       a1, a1, t1
+    ld.d        t0, a1, -7
+
+    add.d       dstend, a2, t1
+    st.d        t0, dstend, -7
+    jr          ra
+    nop
+
+L(end):
+    ctz.d       t1, t1
+    srli.d      t1, t1, 3
+    add.d       a3, a1, t1
+    add.d       dstend, a2, t1
+
+L(less_8):
+    li.d        t0, 3
+    bltu        t1, t0, L(less_3)
+    ld.w        t1, a1, 0
+    ld.w        t2, a3, -3
+
+
+    st.w        t1, a2, 0
+    st.w        t2, dstend, -3
+    jr          ra
+L(less_3):
+    beqz        t1, L(zero_bytes)
+
+    ld.h        t1, a1, 0
+    st.h        t1, a2, 0
+L(zero_bytes):
+    st.b        zero, dstend, 0
+    jr          ra
+
+L(page_cross):
+    move        a4, a1
+    bstrins.d   a4, zero, 2, 0
+    ld.d        t0, a4, 0
+    li.d        t3, -1
+
+    slli.d      t4, a1, 3
+    srl.d       t3, t3, t4
+    srl.d       t0, t0, t4
+    orn         t0, t0, t3
+
+
+    sub.d       t1, t0, t5
+    andn        t2, t6, t0
+    and         t1, t1, t2
+    beqz        t1, L(start_entry)
+
+    b           L(end)
+END(STRCPY)
+
+libc_hidden_builtin_def (STRCPY)
+#endif
diff --git a/sysdeps/loongarch/lp64/multiarch/strcpy.c b/sysdeps/loongarch/lp64/multiarch/strcpy.c
new file mode 100644
index 00000000..46afd068
--- /dev/null
+++ b/sysdeps/loongarch/lp64/multiarch/strcpy.c
@@ -0,0 +1,35 @@
+/* Multiple versions of strcpy.
+   All versions must be listed in ifunc-impl-list.c.
+   Copyright (C) 2023 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/>.  */
+
+/* Define multiple versions only for the definition in libc.  */
+#if IS_IN (libc)
+# define strcpy __redirect_strcpy
+# include <string.h>
+# undef strcpy
+
+# define SYMBOL_NAME strcpy
+# include "ifunc-lasx.h"
+
+libc_ifunc_redirected (__redirect_strcpy, strcpy, IFUNC_SELECTOR ());
+
+# ifdef SHARED
+__hidden_ver1 (strcpy, __GI_strcpy, __redirect_strcpy)
+  __attribute__ ((visibility ("hidden"))) __attribute_copy__ (strcpy);
+# endif
+#endif
-- 
2.33.0