Add x86-64 memmove with unaligned load/store and rep movsb

Implement x86-64 memmove with unaligned load/store and rep movsb.
Support 16-byte, 32-byte and 64-byte vector register sizes.  When
size <= 8 times of vector register size, there is no check for
address overlap bewteen source and destination.  Since overhead for
overlap check is small when size > 8 times of vector register size,
memcpy is an alias of memmove.

A single file provides 2 implementations of memmove, one with rep movsb
and the other without rep movsb.  They share the same codes when size is
between 2 times of vector register size and REP_MOVSB_THRESHOLD which
is 2KB for 16-byte vector register size and scaled up by large vector
register size.

Key features:

1. Use overlapping load and store to avoid branch.
2. For size <= 8 times of vector register size, load  all sources into
registers and store them together.
3. If there is no address overlap bewteen source and destination, copy
from both ends with 4 times of vector register size at a time.
4. If address of destination > address of source, backward copy 8 times
of vector register size at a time.
5. Otherwise, forward copy 8 times of vector register size at a time.
6. Use rep movsb only for forward copy.  Avoid slow backward rep movsb
by fallbacking to backward copy 8 times of vector register size at a
time.
7. Skip when address of destination == address of source.

	[BZ #19776]
	* sysdeps/x86_64/multiarch/Makefile (sysdep_routines): Add
	memmove-sse2-unaligned-erms, memmove-avx-unaligned-erms and
	memmove-avx512-unaligned-erms.
	* sysdeps/x86_64/multiarch/ifunc-impl-list.c
	(__libc_ifunc_impl_list): Test
	__memmove_chk_avx512_unaligned_2,
	__memmove_chk_avx512_unaligned_erms,
	__memmove_chk_avx_unaligned_2, __memmove_chk_avx_unaligned_erms,
	__memmove_chk_sse2_unaligned_2,
	__memmove_chk_sse2_unaligned_erms, __memmove_avx_unaligned_2,
	__memmove_avx_unaligned_erms, __memmove_avx512_unaligned_2,
	__memmove_avx512_unaligned_erms, __memmove_erms,
	__memmove_sse2_unaligned_2, __memmove_sse2_unaligned_erms,
	__memcpy_chk_avx512_unaligned_2,
	__memcpy_chk_avx512_unaligned_erms,
	__memcpy_chk_avx_unaligned_2, __memcpy_chk_avx_unaligned_erms,
	__memcpy_chk_sse2_unaligned_2, __memcpy_chk_sse2_unaligned_erms,
	__memcpy_avx_unaligned_2, __memcpy_avx_unaligned_erms,
	__memcpy_avx512_unaligned_2, __memcpy_avx512_unaligned_erms,
	__memcpy_sse2_unaligned_2, __memcpy_sse2_unaligned_erms,
	__memcpy_erms, __mempcpy_chk_avx512_unaligned_2,
	__mempcpy_chk_avx512_unaligned_erms,
	__mempcpy_chk_avx_unaligned_2, __mempcpy_chk_avx_unaligned_erms,
	__mempcpy_chk_sse2_unaligned_2, __mempcpy_chk_sse2_unaligned_erms,
	__mempcpy_avx512_unaligned_2, __mempcpy_avx512_unaligned_erms,
	__mempcpy_avx_unaligned_2, __mempcpy_avx_unaligned_erms,
	__mempcpy_sse2_unaligned_2, __mempcpy_sse2_unaligned_erms and
	__mempcpy_erms.
	* sysdeps/x86_64/multiarch/memmove-avx-unaligned-erms.S: New
	file.
	* sysdeps/x86_64/multiarch/memmove-avx512-unaligned-erms.S:
	Likwise.
	* sysdeps/x86_64/multiarch/memmove-sse2-unaligned-erms.S:
	Likwise.
	* sysdeps/x86_64/multiarch/memmove-vec-unaligned-erms.S:
	Likwise.

ChangeLog

@@ -1,3 +1,43 @@
+2016-03-31   H.J. Lu  <hongjiu.lu@intel.com>
+
+	[BZ #19776]
+	* sysdeps/x86_64/multiarch/Makefile (sysdep_routines): Add
+	memmove-sse2-unaligned-erms, memmove-avx-unaligned-erms and
+	memmove-avx512-unaligned-erms.
+	* sysdeps/x86_64/multiarch/ifunc-impl-list.c
+	(__libc_ifunc_impl_list): Test
+	__memmove_chk_avx512_unaligned_2,
+	__memmove_chk_avx512_unaligned_erms,
+	__memmove_chk_avx_unaligned_2, __memmove_chk_avx_unaligned_erms,
+	__memmove_chk_sse2_unaligned_2,
+	__memmove_chk_sse2_unaligned_erms, __memmove_avx_unaligned_2,
+	__memmove_avx_unaligned_erms, __memmove_avx512_unaligned_2,
+	__memmove_avx512_unaligned_erms, __memmove_erms,
+	__memmove_sse2_unaligned_2, __memmove_sse2_unaligned_erms,
+	__memcpy_chk_avx512_unaligned_2,
+	__memcpy_chk_avx512_unaligned_erms,
+	__memcpy_chk_avx_unaligned_2, __memcpy_chk_avx_unaligned_erms,
+	__memcpy_chk_sse2_unaligned_2, __memcpy_chk_sse2_unaligned_erms,
+	__memcpy_avx_unaligned_2, __memcpy_avx_unaligned_erms,
+	__memcpy_avx512_unaligned_2, __memcpy_avx512_unaligned_erms,
+	__memcpy_sse2_unaligned_2, __memcpy_sse2_unaligned_erms,
+	__memcpy_erms, __mempcpy_chk_avx512_unaligned_2,
+	__mempcpy_chk_avx512_unaligned_erms,
+	__mempcpy_chk_avx_unaligned_2, __mempcpy_chk_avx_unaligned_erms,
+	__mempcpy_chk_sse2_unaligned_2, __mempcpy_chk_sse2_unaligned_erms,
+	__mempcpy_avx512_unaligned_2, __mempcpy_avx512_unaligned_erms,
+	__mempcpy_avx_unaligned_2, __mempcpy_avx_unaligned_erms,
+	__mempcpy_sse2_unaligned_2, __mempcpy_sse2_unaligned_erms and
+	__mempcpy_erms.
+	* sysdeps/x86_64/multiarch/memmove-avx-unaligned-erms.S: New
+	file.
+	* sysdeps/x86_64/multiarch/memmove-avx512-unaligned-erms.S:
+	Likwise.
+	* sysdeps/x86_64/multiarch/memmove-sse2-unaligned-erms.S:
+	Likwise.
+	* sysdeps/x86_64/multiarch/memmove-vec-unaligned-erms.S:
+	Likwise.
+
 2016-03-31  Stefan Liebler  <stli@linux.vnet.ibm.com>
 
 	* sysdeps/s390/bits/link.h: (La_s390_vr) New typedef.

sysdeps/x86_64/multiarch/Makefile

@@ -20,7 +20,10 @@ sysdep_routines += strncat-c stpncpy-c strncpy-c strcmp-ssse3 \
 		   strcat-sse2-unaligned strncat-sse2-unaligned \
 		   strchr-sse2-no-bsf memcmp-ssse3 strstr-sse2-unaligned \
 		   strcspn-c strpbrk-c strspn-c varshift memset-avx2 \
-		   memset-avx512-no-vzeroupper
+		   memset-avx512-no-vzeroupper \
+		   memmove-sse2-unaligned-erms \
+		   memmove-avx-unaligned-erms \
+		   memmove-avx512-unaligned-erms
 CFLAGS-varshift.c += -msse4
 CFLAGS-strcspn-c.c += -msse4
 CFLAGS-strpbrk-c.c += -msse4

sysdeps/x86_64/multiarch/ifunc-impl-list.c

@@ -52,16 +52,32 @@ __libc_ifunc_impl_list (const char *name, struct libc_ifunc_impl *array,
 	      IFUNC_IMPL_ADD (array, i, __memmove_chk,
 			      HAS_ARCH_FEATURE (AVX512F_Usable),
 			      __memmove_chk_avx512_no_vzeroupper)
+	      IFUNC_IMPL_ADD (array, i, __memmove_chk,
+			      HAS_ARCH_FEATURE (AVX512F_Usable),
+			      __memmove_chk_avx512_unaligned_2)
+	      IFUNC_IMPL_ADD (array, i, __memmove_chk,
+			      HAS_ARCH_FEATURE (AVX512F_Usable),
+			      __memmove_chk_avx512_unaligned_erms)
 #endif
 	      IFUNC_IMPL_ADD (array, i, __memmove_chk,
 			      HAS_ARCH_FEATURE (AVX_Usable),
 			      __memmove_chk_avx_unaligned)
+	      IFUNC_IMPL_ADD (array, i, __memmove_chk,
+			      HAS_ARCH_FEATURE (AVX_Usable),
+			      __memmove_chk_avx_unaligned_2)
+	      IFUNC_IMPL_ADD (array, i, __memmove_chk,
+			      HAS_ARCH_FEATURE (AVX_Usable),
+			      __memmove_chk_avx_unaligned_erms)
 	      IFUNC_IMPL_ADD (array, i, __memmove_chk,
 			      HAS_CPU_FEATURE (SSSE3),
 			      __memmove_chk_ssse3_back)
 	      IFUNC_IMPL_ADD (array, i, __memmove_chk,
 			      HAS_CPU_FEATURE (SSSE3),
 			      __memmove_chk_ssse3)
+	      IFUNC_IMPL_ADD (array, i, __memmove_chk, 1,
+			      __memmove_chk_sse2_unaligned_2)
+	      IFUNC_IMPL_ADD (array, i, __memmove_chk, 1,
+			      __memmove_chk_sse2_unaligned_erms)
 	      IFUNC_IMPL_ADD (array, i, __memmove_chk, 1,
 			      __memmove_chk_sse2))
 
@@ -70,15 +86,32 @@ __libc_ifunc_impl_list (const char *name, struct libc_ifunc_impl *array,
 	      IFUNC_IMPL_ADD (array, i, memmove,
 			      HAS_ARCH_FEATURE (AVX_Usable),
 			      __memmove_avx_unaligned)
+	      IFUNC_IMPL_ADD (array, i, memmove,
+			      HAS_ARCH_FEATURE (AVX_Usable),
+			      __memmove_avx_unaligned_2)
+	      IFUNC_IMPL_ADD (array, i, memmove,
+			      HAS_ARCH_FEATURE (AVX_Usable),
+			      __memmove_avx_unaligned_erms)
 #ifdef HAVE_AVX512_ASM_SUPPORT
 	      IFUNC_IMPL_ADD (array, i, memmove,
 			      HAS_ARCH_FEATURE (AVX512F_Usable),
 			      __memmove_avx512_no_vzeroupper)
+	      IFUNC_IMPL_ADD (array, i, memmove,
+			      HAS_ARCH_FEATURE (AVX512F_Usable),
+			      __memmove_avx512_unaligned_2)
+	      IFUNC_IMPL_ADD (array, i, memmove,
+			      HAS_ARCH_FEATURE (AVX512F_Usable),
+			      __memmove_avx512_unaligned_erms)
 #endif
 	      IFUNC_IMPL_ADD (array, i, memmove, HAS_CPU_FEATURE (SSSE3),
 			      __memmove_ssse3_back)
 	      IFUNC_IMPL_ADD (array, i, memmove, HAS_CPU_FEATURE (SSSE3),
 			      __memmove_ssse3)
+	      IFUNC_IMPL_ADD (array, i, memmove, 1, __memmove_erms)
+	      IFUNC_IMPL_ADD (array, i, memmove, 1,
+			      __memmove_sse2_unaligned_2)
+	      IFUNC_IMPL_ADD (array, i, memmove, 1,
+			      __memmove_sse2_unaligned_erms)
 	      IFUNC_IMPL_ADD (array, i, memmove, 1, __memmove_sse2))
 
   /* Support sysdeps/x86_64/multiarch/memset_chk.S.  */
@@ -267,16 +300,32 @@ __libc_ifunc_impl_list (const char *name, struct libc_ifunc_impl *array,
 	      IFUNC_IMPL_ADD (array, i, __memcpy_chk,
 			      HAS_ARCH_FEATURE (AVX512F_Usable),
 			      __memcpy_chk_avx512_no_vzeroupper)
+	      IFUNC_IMPL_ADD (array, i, __memcpy_chk,
+			      HAS_ARCH_FEATURE (AVX512F_Usable),
+			      __memcpy_chk_avx512_unaligned_2)
+	      IFUNC_IMPL_ADD (array, i, __memcpy_chk,
+			      HAS_ARCH_FEATURE (AVX512F_Usable),
+			      __memcpy_chk_avx512_unaligned_erms)
 #endif
 	      IFUNC_IMPL_ADD (array, i, __memcpy_chk,
 			      HAS_ARCH_FEATURE (AVX_Usable),
 			      __memcpy_chk_avx_unaligned)
+	      IFUNC_IMPL_ADD (array, i, __memcpy_chk,
+			      HAS_ARCH_FEATURE (AVX_Usable),
+			      __memcpy_chk_avx_unaligned_2)
+	      IFUNC_IMPL_ADD (array, i, __memcpy_chk,
+			      HAS_ARCH_FEATURE (AVX_Usable),
+			      __memcpy_chk_avx_unaligned_erms)
 	      IFUNC_IMPL_ADD (array, i, __memcpy_chk,
 			      HAS_CPU_FEATURE (SSSE3),
 			      __memcpy_chk_ssse3_back)
 	      IFUNC_IMPL_ADD (array, i, __memcpy_chk,
 			      HAS_CPU_FEATURE (SSSE3),
 			      __memcpy_chk_ssse3)
+	      IFUNC_IMPL_ADD (array, i, __memcpy_chk, 1,
+			      __memcpy_chk_sse2_unaligned_2)
+	      IFUNC_IMPL_ADD (array, i, __memcpy_chk, 1,
+			      __memcpy_chk_sse2_unaligned_erms)
 	      IFUNC_IMPL_ADD (array, i, __memcpy_chk, 1,
 			      __memcpy_chk_sse2))
 
@@ -285,6 +334,12 @@ __libc_ifunc_impl_list (const char *name, struct libc_ifunc_impl *array,
 	      IFUNC_IMPL_ADD (array, i, memcpy,
 			      HAS_ARCH_FEATURE (AVX_Usable),
 			      __memcpy_avx_unaligned)
+	      IFUNC_IMPL_ADD (array, i, memcpy,
+			      HAS_ARCH_FEATURE (AVX_Usable),
+			      __memcpy_avx_unaligned_2)
+	      IFUNC_IMPL_ADD (array, i, memcpy,
+			      HAS_ARCH_FEATURE (AVX_Usable),
+			      __memcpy_avx_unaligned_erms)
 	      IFUNC_IMPL_ADD (array, i, memcpy, HAS_CPU_FEATURE (SSSE3),
 			      __memcpy_ssse3_back)
 	      IFUNC_IMPL_ADD (array, i, memcpy, HAS_CPU_FEATURE (SSSE3),
@@ -293,8 +348,19 @@ __libc_ifunc_impl_list (const char *name, struct libc_ifunc_impl *array,
 	      IFUNC_IMPL_ADD (array, i, memcpy,
 			      HAS_ARCH_FEATURE (AVX512F_Usable),
 			      __memcpy_avx512_no_vzeroupper)
+	      IFUNC_IMPL_ADD (array, i, memcpy,
+			      HAS_ARCH_FEATURE (AVX512F_Usable),
+			      __memcpy_avx512_unaligned_2)
+	      IFUNC_IMPL_ADD (array, i, memcpy,
+			      HAS_ARCH_FEATURE (AVX512F_Usable),
+			      __memcpy_avx512_unaligned_erms)
 #endif
 	      IFUNC_IMPL_ADD (array, i, memcpy, 1, __memcpy_sse2_unaligned)
+	      IFUNC_IMPL_ADD (array, i, memcpy, 1,
+			      __memcpy_sse2_unaligned_2)
+	      IFUNC_IMPL_ADD (array, i, memcpy, 1,
+			      __memcpy_sse2_unaligned_erms)
+	      IFUNC_IMPL_ADD (array, i, memcpy, 1, __memcpy_erms)
 	      IFUNC_IMPL_ADD (array, i, memcpy, 1, __memcpy_sse2))
 
   /* Support sysdeps/x86_64/multiarch/mempcpy_chk.S.  */
@@ -303,16 +369,32 @@ __libc_ifunc_impl_list (const char *name, struct libc_ifunc_impl *array,
 	      IFUNC_IMPL_ADD (array, i, __mempcpy_chk,
 			      HAS_ARCH_FEATURE (AVX512F_Usable),
 			      __mempcpy_chk_avx512_no_vzeroupper)
+	      IFUNC_IMPL_ADD (array, i, __mempcpy_chk,
+			      HAS_ARCH_FEATURE (AVX512F_Usable),
+			      __mempcpy_chk_avx512_unaligned_2)
+	      IFUNC_IMPL_ADD (array, i, __mempcpy_chk,
+			      HAS_ARCH_FEATURE (AVX512F_Usable),
+			      __mempcpy_chk_avx512_unaligned_erms)
 #endif
 	      IFUNC_IMPL_ADD (array, i, __mempcpy_chk,
 			      HAS_ARCH_FEATURE (AVX_Usable),
 			      __mempcpy_chk_avx_unaligned)
+	      IFUNC_IMPL_ADD (array, i, __mempcpy_chk,
+			      HAS_ARCH_FEATURE (AVX_Usable),
+			      __mempcpy_chk_avx_unaligned_2)
+	      IFUNC_IMPL_ADD (array, i, __mempcpy_chk,
+			      HAS_ARCH_FEATURE (AVX_Usable),
+			      __mempcpy_chk_avx_unaligned_erms)
 	      IFUNC_IMPL_ADD (array, i, __mempcpy_chk,
 			      HAS_CPU_FEATURE (SSSE3),
 			      __mempcpy_chk_ssse3_back)
 	      IFUNC_IMPL_ADD (array, i, __mempcpy_chk,
 			      HAS_CPU_FEATURE (SSSE3),
 			      __mempcpy_chk_ssse3)
+	      IFUNC_IMPL_ADD (array, i, __mempcpy_chk, 1,
+			      __mempcpy_chk_sse2_unaligned_2)
+	      IFUNC_IMPL_ADD (array, i, __mempcpy_chk, 1,
+			      __mempcpy_chk_sse2_unaligned_erms)
 	      IFUNC_IMPL_ADD (array, i, __mempcpy_chk, 1,
 			      __mempcpy_chk_sse2))
 
@@ -322,14 +404,31 @@ __libc_ifunc_impl_list (const char *name, struct libc_ifunc_impl *array,
 	      IFUNC_IMPL_ADD (array, i, mempcpy,
 			      HAS_ARCH_FEATURE (AVX512F_Usable),
 			      __mempcpy_avx512_no_vzeroupper)
+	      IFUNC_IMPL_ADD (array, i, mempcpy,
+			      HAS_ARCH_FEATURE (AVX512F_Usable),
+			      __mempcpy_avx512_unaligned_2)
+	      IFUNC_IMPL_ADD (array, i, mempcpy,
+			      HAS_ARCH_FEATURE (AVX512F_Usable),
+			      __mempcpy_avx512_unaligned_erms)
 #endif
 	      IFUNC_IMPL_ADD (array, i, mempcpy,
 			      HAS_ARCH_FEATURE (AVX_Usable),
 			      __mempcpy_avx_unaligned)
+	      IFUNC_IMPL_ADD (array, i, mempcpy,
+			      HAS_ARCH_FEATURE (AVX_Usable),
+			      __mempcpy_avx_unaligned_2)
+	      IFUNC_IMPL_ADD (array, i, mempcpy,
+			      HAS_ARCH_FEATURE (AVX_Usable),
+			      __mempcpy_avx_unaligned_erms)
 	      IFUNC_IMPL_ADD (array, i, mempcpy, HAS_CPU_FEATURE (SSSE3),
 			      __mempcpy_ssse3_back)
 	      IFUNC_IMPL_ADD (array, i, mempcpy, HAS_CPU_FEATURE (SSSE3),
 			      __mempcpy_ssse3)
+	      IFUNC_IMPL_ADD (array, i, mempcpy, 1,
+			      __mempcpy_sse2_unaligned_2)
+	      IFUNC_IMPL_ADD (array, i, mempcpy, 1,
+			      __mempcpy_sse2_unaligned_erms)
+	      IFUNC_IMPL_ADD (array, i, mempcpy, 1, __mempcpy_erms)
 	      IFUNC_IMPL_ADD (array, i, mempcpy, 1, __mempcpy_sse2))
 
   /* Support sysdeps/x86_64/multiarch/strncmp.S.  */

sysdeps/x86_64/multiarch/memmove-avx-unaligned-erms.S

@@ -0,0 +1,9 @@
+#define VEC_SIZE	32
+#define VEC(i)		ymm##i
+#define VMOVU		vmovdqu
+#define VMOVA		vmovdqa
+
+#define SECTION(p)		p##.avx
+#define MEMMOVE_SYMBOL(p,s)	p##_avx_##s
+
+#include "memmove-vec-unaligned-erms.S"

sysdeps/x86_64/multiarch/memmove-avx512-unaligned-erms.S

@@ -0,0 +1,11 @@
+#ifdef HAVE_AVX512_ASM_SUPPORT
+# define VEC_SIZE	64
+# define VEC(i)		zmm##i
+# define VMOVU		vmovdqu64
+# define VMOVA		vmovdqa64
+
+# define SECTION(p)		p##.avx512
+# define MEMMOVE_SYMBOL(p,s)	p##_avx512_##s
+
+# include "memmove-vec-unaligned-erms.S"
+#endif

sysdeps/x86_64/multiarch/memmove-sse2-unaligned-erms.S

@@ -0,0 +1,9 @@
+#define VEC_SIZE	16
+#define VEC(i)		xmm##i
+#define VMOVU		movdqu
+#define VMOVA		movdqa
+
+#define SECTION(p)		p
+#define MEMMOVE_SYMBOL(p,s)	p##_sse2_##s
+
+#include "memmove-vec-unaligned-erms.S"