@@ -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;
@@ -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__ */
@@ -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
@@ -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
@@ -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
@@ -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
@@ -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