This is the mail archive of the
libc-alpha@sourceware.org
mailing list for the glibc project.
[PATCH] Check ERMS in memmove/memcpy/mempcpy/memset
- From: "H.J. Lu" <hongjiu dot lu at intel dot com>
- To: GNU C Library <libc-alpha at sourceware dot org>
- Date: Wed, 29 Jun 2016 11:42:53 -0700
- Subject: [PATCH] Check ERMS in memmove/memcpy/mempcpy/memset
- Authentication-results: sourceware.org; auth=none
- Reply-to: "H.J. Lu" <hjl dot tools at gmail dot com>
Although the Enhanced REP MOVSB/STOSB (ERMS) implementations of memmove,
memcpy, mempcpy and memset aren't used by the current processors, this
patch adds Prefer_ERMS check in memmove, memcpy, mempcpy and memset so
that glibc developers can experiment with it using GLIBC_IFUNC.
OK for master?
H.J.
---
* sysdeps/x86/cpu-features.c (init_cpu_features): Also
check Prefer_ERMS.
* sysdeps/x86/cpu-features.h (bit_arch_Prefer_ERMS): New.
(index_arch_Prefer_ERMS): Likewise.
* sysdeps/x86_64/multiarch/memcpy.S (__new_memcpy): Return
__memcpy_erms for Prefer_ERMS.
* sysdeps/x86_64/multiarch/memmove-vec-unaligned-erms.S
(__memmove_erms): Enabled for libc.a.
* ysdeps/x86_64/multiarch/memmove.S (__libc_memmove): Return
__memmove_erms or Prefer_ERMS.
* sysdeps/x86_64/multiarch/mempcpy.S (__mempcpy): Return
__mempcpy_erms for Prefer_ERMS.
* sysdeps/x86_64/multiarch/memset.S (memset): Return
__memset_erms for Prefer_ERMS.
---
sysdeps/x86/cpu-features.c | 1 +
sysdeps/x86/cpu-features.h | 3 +++
sysdeps/x86_64/multiarch/memcpy.S | 3 +++
sysdeps/x86_64/multiarch/memmove-vec-unaligned-erms.S | 6 +++++-
sysdeps/x86_64/multiarch/memmove.S | 3 +++
sysdeps/x86_64/multiarch/mempcpy.S | 3 +++
sysdeps/x86_64/multiarch/memset.S | 3 +++
7 files changed, 21 insertions(+), 1 deletion(-)
diff --git a/sysdeps/x86/cpu-features.c b/sysdeps/x86/cpu-features.c
index 0e5765e..d506aa5 100644
--- a/sysdeps/x86/cpu-features.c
+++ b/sysdeps/x86/cpu-features.c
@@ -473,6 +473,7 @@ no_cpuid:
case 13:
CHECK_GLIBC_IFUNC_ARCH_OFF (AVX2_Usable);
CHECK_GLIBC_IFUNC_ARCH_OFF (FMA4_Usable);
+ CHECK_GLIBC_IFUNC_ARCH_BOTH (Prefer_ERMS);
CHECK_GLIBC_IFUNC_ARCH_NEED_CPU_BOTH (Slow_SSE4_2,
SSE4_2);
break;
diff --git a/sysdeps/x86/cpu-features.h b/sysdeps/x86/cpu-features.h
index 2bd9371..97ffe76 100644
--- a/sysdeps/x86/cpu-features.h
+++ b/sysdeps/x86/cpu-features.h
@@ -36,6 +36,7 @@
#define bit_arch_Prefer_MAP_32BIT_EXEC (1 << 16)
#define bit_arch_Prefer_No_VZEROUPPER (1 << 17)
#define bit_arch_Fast_Unaligned_Copy (1 << 18)
+#define bit_arch_Prefer_ERMS (1 << 19)
/* CPUID Feature flags. */
@@ -105,6 +106,7 @@
# define index_arch_Prefer_MAP_32BIT_EXEC FEATURE_INDEX_1*FEATURE_SIZE
# define index_arch_Prefer_No_VZEROUPPER FEATURE_INDEX_1*FEATURE_SIZE
# define index_arch_Fast_Unaligned_Copy FEATURE_INDEX_1*FEATURE_SIZE
+# define index_arch_Prefer_ERMS FEATURE_INDEX_1*FEATURE_SIZE
# if defined (_LIBC) && !IS_IN (nonlib)
@@ -274,6 +276,7 @@ extern const struct cpu_features *__get_cpu_features (void)
# define index_arch_Prefer_MAP_32BIT_EXEC FEATURE_INDEX_1
# define index_arch_Prefer_No_VZEROUPPER FEATURE_INDEX_1
# define index_arch_Fast_Unaligned_Copy FEATURE_INDEX_1
+# define index_arch_Prefer_ERMS FEATURE_INDEX_1
#endif /* !__ASSEMBLER__ */
diff --git a/sysdeps/x86_64/multiarch/memcpy.S b/sysdeps/x86_64/multiarch/memcpy.S
index f6771a4..df7fbac 100644
--- a/sysdeps/x86_64/multiarch/memcpy.S
+++ b/sysdeps/x86_64/multiarch/memcpy.S
@@ -29,6 +29,9 @@
ENTRY(__new_memcpy)
.type __new_memcpy, @gnu_indirect_function
LOAD_RTLD_GLOBAL_RO_RDX
+ lea __memcpy_erms(%rip), %RAX_LP
+ HAS_ARCH_FEATURE (Prefer_ERMS)
+ jnz 2f
# ifdef HAVE_AVX512_ASM_SUPPORT
HAS_ARCH_FEATURE (AVX512F_Usable)
jz 1f
diff --git a/sysdeps/x86_64/multiarch/memmove-vec-unaligned-erms.S b/sysdeps/x86_64/multiarch/memmove-vec-unaligned-erms.S
index a2cce39..4893ea4 100644
--- a/sysdeps/x86_64/multiarch/memmove-vec-unaligned-erms.S
+++ b/sysdeps/x86_64/multiarch/memmove-vec-unaligned-erms.S
@@ -150,13 +150,15 @@ L(nop):
#if defined USE_MULTIARCH && IS_IN (libc)
END (MEMMOVE_SYMBOL (__memmove, unaligned))
-# if VEC_SIZE == 16 && defined SHARED
+# if VEC_SIZE == 16
+# if defined SHARED
/* Only used to measure performance of REP MOVSB. */
ENTRY (__mempcpy_erms)
movq %rdi, %rax
addq %rdx, %rax
jmp L(start_movsb)
END (__mempcpy_erms)
+# endif
ENTRY (__memmove_erms)
movq %rdi, %rax
@@ -181,7 +183,9 @@ L(movsb_backward):
cld
ret
END (__memmove_erms)
+# if defined SHARED
strong_alias (__memmove_erms, __memcpy_erms)
+# endif
# endif
# ifdef SHARED
diff --git a/sysdeps/x86_64/multiarch/memmove.S b/sysdeps/x86_64/multiarch/memmove.S
index 25c3586..8e1c6ac 100644
--- a/sysdeps/x86_64/multiarch/memmove.S
+++ b/sysdeps/x86_64/multiarch/memmove.S
@@ -27,6 +27,9 @@
ENTRY(__libc_memmove)
.type __libc_memmove, @gnu_indirect_function
LOAD_RTLD_GLOBAL_RO_RDX
+ lea __memmove_erms(%rip), %RAX_LP
+ HAS_ARCH_FEATURE (Prefer_ERMS)
+ jnz 2f
# ifdef HAVE_AVX512_ASM_SUPPORT
HAS_ARCH_FEATURE (AVX512F_Usable)
jz 1f
diff --git a/sysdeps/x86_64/multiarch/mempcpy.S b/sysdeps/x86_64/multiarch/mempcpy.S
index f9c6df3..4011a1a 100644
--- a/sysdeps/x86_64/multiarch/mempcpy.S
+++ b/sysdeps/x86_64/multiarch/mempcpy.S
@@ -29,6 +29,9 @@
ENTRY(__mempcpy)
.type __mempcpy, @gnu_indirect_function
LOAD_RTLD_GLOBAL_RO_RDX
+ lea __mempcpy_erms(%rip), %RAX_LP
+ HAS_ARCH_FEATURE (Prefer_ERMS)
+ jnz 2f
# ifdef HAVE_AVX512_ASM_SUPPORT
HAS_ARCH_FEATURE (AVX512F_Usable)
jz 1f
diff --git a/sysdeps/x86_64/multiarch/memset.S b/sysdeps/x86_64/multiarch/memset.S
index 4e52d8f..2b964a0 100644
--- a/sysdeps/x86_64/multiarch/memset.S
+++ b/sysdeps/x86_64/multiarch/memset.S
@@ -26,6 +26,9 @@
ENTRY(memset)
.type memset, @gnu_indirect_function
LOAD_RTLD_GLOBAL_RO_RDX
+ lea __memset_erms(%rip), %RAX_LP
+ HAS_ARCH_FEATURE (Prefer_ERMS)
+ jnz 2f
lea __memset_sse2_unaligned_erms(%rip), %RAX_LP
HAS_CPU_FEATURE (ERMS)
jnz 1f
--
2.7.4