x86-64: Add ifunc-avx2.h functions with 256-bit EVEX

Update ifunc-avx2.h, strchr.c, strcmp.c, strncmp.c and wcsnlen.c to
select the function optimized with 256-bit EVEX instructions using
YMM16-YMM31 registers to avoid RTM abort with usable AVX512VL, AVX512BW
and BMI2 since VZEROUPPER isn't needed at function exit.

For strcmp/strncmp, prefer AVX2 strcmp/strncmp if Prefer_AVX2_STRCMP
is set.

(cherry picked from commit 1fd8c163a8)

sysdeps/x86_64/multiarch/Makefile

@@ -43,7 +43,17 @@ sysdep_routines += strncat-c stpncpy-c strncpy-c \
 		   memmove-avx512-unaligned-erms \
 		   memset-sse2-unaligned-erms \
 		   memset-avx2-unaligned-erms \
-		   memset-avx512-unaligned-erms
+		   memset-avx512-unaligned-erms \
+		   memchr-evex \
+		   memrchr-evex \
+		   rawmemchr-evex \
+		   strchr-evex \
+		   strchrnul-evex \
+		   strcmp-evex \
+		   strlen-evex \
+		   strncmp-evex \
+		   strnlen-evex \
+		   strrchr-evex
 CFLAGS-varshift.c += -msse4
 CFLAGS-strcspn-c.c += -msse4
 CFLAGS-strpbrk-c.c += -msse4
@@ -60,7 +70,14 @@ sysdep_routines += wmemcmp-sse4 wmemcmp-ssse3 wmemcmp-c \
 		   wcschr-sse2 wcschr-avx2 \
 		   wcsrchr-sse2 wcsrchr-avx2 \
 		   wcsnlen-sse4_1 wcsnlen-c \
-		   wcslen-sse2 wcslen-avx2 wcsnlen-avx2
+		   wcslen-sse2 wcslen-avx2 wcsnlen-avx2 \
+		   wcschr-evex \
+		   wcscmp-evex \
+		   wcslen-evex \
+		   wcsncmp-evex \
+		   wcsnlen-evex \
+		   wcsrchr-evex \
+		   wmemchr-evex
 endif
 
 ifeq ($(subdir),debug)

sysdeps/x86_64/multiarch/ifunc-avx2.h

@@ -21,16 +21,24 @@
 
 extern __typeof (REDIRECT_NAME) OPTIMIZE (sse2) attribute_hidden;
 extern __typeof (REDIRECT_NAME) OPTIMIZE (avx2) attribute_hidden;
+extern __typeof (REDIRECT_NAME) OPTIMIZE (evex) attribute_hidden;
 
 static inline void *
 IFUNC_SELECTOR (void)
 {
   const struct cpu_features* cpu_features = __get_cpu_features ();
 
-  if (!CPU_FEATURES_ARCH_P (cpu_features, Prefer_No_VZEROUPPER)
+  if (CPU_FEATURES_ARCH_P (cpu_features, AVX2_Usable)
-      && CPU_FEATURES_ARCH_P (cpu_features, AVX2_Usable)
       && CPU_FEATURES_ARCH_P (cpu_features, AVX_Fast_Unaligned_Load))
-    return OPTIMIZE (avx2);
+    {
+      if (CPU_FEATURES_ARCH_P (cpu_features, AVX512VL_Usable)
+	  && CPU_FEATURES_ARCH_P (cpu_features, AVX512BW_Usable)
+	  && CPU_FEATURES_CPU_P (cpu_features, BMI2))
+	return OPTIMIZE (evex);
+
+      if (!CPU_FEATURES_ARCH_P (cpu_features, Prefer_No_VZEROUPPER))
+	return OPTIMIZE (avx2);
+    }
 
   return OPTIMIZE (sse2);
 }

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

@@ -43,6 +43,11 @@ __libc_ifunc_impl_list (const char *name, struct libc_ifunc_impl *array,
 	      IFUNC_IMPL_ADD (array, i, memchr,
 			      HAS_ARCH_FEATURE (AVX2_Usable),
 			      __memchr_avx2)
+	      IFUNC_IMPL_ADD (array, i, memchr,
+			      (HAS_ARCH_FEATURE (AVX512VL_Usable)
+			       && HAS_ARCH_FEATURE (AVX512BW_Usable)
+			       && HAS_CPU_FEATURE (BMI2)),
+			      __memchr_evex)
 	      IFUNC_IMPL_ADD (array, i, memchr, 1, __memchr_sse2))
 
   /* Support sysdeps/x86_64/multiarch/memcmp.c.  */
@@ -121,6 +126,11 @@ __libc_ifunc_impl_list (const char *name, struct libc_ifunc_impl *array,
 	      IFUNC_IMPL_ADD (array, i, memrchr,
 			      HAS_ARCH_FEATURE (AVX2_Usable),
 			      __memrchr_avx2)
+	      IFUNC_IMPL_ADD (array, i, memrchr,
+			      (HAS_ARCH_FEATURE (AVX512VL_Usable)
+			       && HAS_ARCH_FEATURE (AVX512BW_Usable)),
+			      __memrchr_evex)
+
 	      IFUNC_IMPL_ADD (array, i, memrchr, 1, __memrchr_sse2))
 
 #ifdef SHARED
@@ -179,6 +189,11 @@ __libc_ifunc_impl_list (const char *name, struct libc_ifunc_impl *array,
 	      IFUNC_IMPL_ADD (array, i, rawmemchr,
 			      HAS_ARCH_FEATURE (AVX2_Usable),
 			      __rawmemchr_avx2)
+	      IFUNC_IMPL_ADD (array, i, rawmemchr,
+			      (HAS_ARCH_FEATURE (AVX512VL_Usable)
+			       && HAS_ARCH_FEATURE (AVX512BW_Usable)
+			       && HAS_CPU_FEATURE (BMI2)),
+			      __rawmemchr_evex)
 	      IFUNC_IMPL_ADD (array, i, rawmemchr, 1, __rawmemchr_sse2))
 
   /* Support sysdeps/x86_64/multiarch/strlen.c.  */
@@ -186,6 +201,10 @@ __libc_ifunc_impl_list (const char *name, struct libc_ifunc_impl *array,
 	      IFUNC_IMPL_ADD (array, i, strlen,
 			      HAS_ARCH_FEATURE (AVX2_Usable),
 			      __strlen_avx2)
+	      IFUNC_IMPL_ADD (array, i, strlen,
+			      (HAS_ARCH_FEATURE (AVX512VL_Usable)
+			       && HAS_ARCH_FEATURE (AVX512BW_Usable)),
+			      __strlen_evex)
 	      IFUNC_IMPL_ADD (array, i, strlen, 1, __strlen_sse2))
 
   /* Support sysdeps/x86_64/multiarch/strnlen.c.  */
@@ -193,6 +212,10 @@ __libc_ifunc_impl_list (const char *name, struct libc_ifunc_impl *array,
 	      IFUNC_IMPL_ADD (array, i, strnlen,
 			      HAS_ARCH_FEATURE (AVX2_Usable),
 			      __strnlen_avx2)
+	      IFUNC_IMPL_ADD (array, i, strnlen,
+			      (HAS_ARCH_FEATURE (AVX512VL_Usable)
+			       && HAS_ARCH_FEATURE (AVX512BW_Usable)),
+			      __strnlen_evex)
 	      IFUNC_IMPL_ADD (array, i, strnlen, 1, __strnlen_sse2))
 
   /* Support sysdeps/x86_64/multiarch/stpncpy.c.  */
@@ -255,6 +278,11 @@ __libc_ifunc_impl_list (const char *name, struct libc_ifunc_impl *array,
 	      IFUNC_IMPL_ADD (array, i, strchr,
 			      HAS_ARCH_FEATURE (AVX2_Usable),
 			      __strchr_avx2)
+	      IFUNC_IMPL_ADD (array, i, strchr,
+			      (HAS_ARCH_FEATURE (AVX512VL_Usable)
+			       && HAS_ARCH_FEATURE (AVX512BW_Usable)
+			       && HAS_CPU_FEATURE (BMI2)),
+			      __strchr_evex)
 	      IFUNC_IMPL_ADD (array, i, strchr, 1, __strchr_sse2_no_bsf)
 	      IFUNC_IMPL_ADD (array, i, strchr, 1, __strchr_sse2))
 
@@ -263,6 +291,11 @@ __libc_ifunc_impl_list (const char *name, struct libc_ifunc_impl *array,
 	      IFUNC_IMPL_ADD (array, i, strchrnul,
 			      HAS_ARCH_FEATURE (AVX2_Usable),
 			      __strchrnul_avx2)
+	      IFUNC_IMPL_ADD (array, i, strchrnul,
+			      (HAS_ARCH_FEATURE (AVX512VL_Usable)
+			       && HAS_ARCH_FEATURE (AVX512BW_Usable)
+			       && HAS_CPU_FEATURE (BMI2)),
+			      __strchrnul_evex)
 	      IFUNC_IMPL_ADD (array, i, strchrnul, 1, __strchrnul_sse2))
 
   /* Support sysdeps/x86_64/multiarch/strrchr.c.  */
@@ -270,6 +303,10 @@ __libc_ifunc_impl_list (const char *name, struct libc_ifunc_impl *array,
 	      IFUNC_IMPL_ADD (array, i, strrchr,
 			      HAS_ARCH_FEATURE (AVX2_Usable),
 			      __strrchr_avx2)
+	      IFUNC_IMPL_ADD (array, i, strrchr,
+			      (HAS_ARCH_FEATURE (AVX512VL_Usable)
+			       && HAS_ARCH_FEATURE (AVX512BW_Usable)),
+			      __strrchr_evex)
 	      IFUNC_IMPL_ADD (array, i, strrchr, 1, __strrchr_sse2))
 
   /* Support sysdeps/x86_64/multiarch/strcmp.c.  */
@@ -277,6 +314,11 @@ __libc_ifunc_impl_list (const char *name, struct libc_ifunc_impl *array,
 	      IFUNC_IMPL_ADD (array, i, strcmp,
 			      HAS_ARCH_FEATURE (AVX2_Usable),
 			      __strcmp_avx2)
+	      IFUNC_IMPL_ADD (array, i, strcmp,
+			      (HAS_ARCH_FEATURE (AVX512VL_Usable)
+			       && HAS_ARCH_FEATURE (AVX512BW_Usable)
+			       && HAS_CPU_FEATURE (BMI2)),
+			      __strcmp_evex)
 	      IFUNC_IMPL_ADD (array, i, strcmp, HAS_CPU_FEATURE (SSE4_2),
 			      __strcmp_sse42)
 	      IFUNC_IMPL_ADD (array, i, strcmp, HAS_CPU_FEATURE (SSSE3),
@@ -370,6 +412,11 @@ __libc_ifunc_impl_list (const char *name, struct libc_ifunc_impl *array,
 	      IFUNC_IMPL_ADD (array, i, wcschr,
 			      HAS_ARCH_FEATURE (AVX2_Usable),
 			      __wcschr_avx2)
+	      IFUNC_IMPL_ADD (array, i, wcschr,
+			      (HAS_ARCH_FEATURE (AVX512VL_Usable)
+			       && HAS_ARCH_FEATURE (AVX512BW_Usable)
+			       && HAS_CPU_FEATURE (BMI2)),
+			      __wcschr_evex)
 	      IFUNC_IMPL_ADD (array, i, wcschr, 1, __wcschr_sse2))
 
   /* Support sysdeps/x86_64/multiarch/wcsrchr.c.  */
@@ -377,6 +424,11 @@ __libc_ifunc_impl_list (const char *name, struct libc_ifunc_impl *array,
 	      IFUNC_IMPL_ADD (array, i, wcsrchr,
 			      HAS_ARCH_FEATURE (AVX2_Usable),
 			      __wcsrchr_avx2)
+	      IFUNC_IMPL_ADD (array, i, wcsrchr,
+			      (HAS_ARCH_FEATURE (AVX512VL_Usable)
+			       && HAS_ARCH_FEATURE (AVX512BW_Usable)
+			       && HAS_CPU_FEATURE (BMI2)),
+			      __wcsrchr_evex)
 	      IFUNC_IMPL_ADD (array, i, wcsrchr, 1, __wcsrchr_sse2))
 
   /* Support sysdeps/x86_64/multiarch/wcscmp.c.  */
@@ -384,6 +436,11 @@ __libc_ifunc_impl_list (const char *name, struct libc_ifunc_impl *array,
 	      IFUNC_IMPL_ADD (array, i, wcscmp,
 			      HAS_ARCH_FEATURE (AVX2_Usable),
 			      __wcscmp_avx2)
+	      IFUNC_IMPL_ADD (array, i, wcscmp,
+			      (HAS_ARCH_FEATURE (AVX512VL_Usable)
+			       && HAS_ARCH_FEATURE (AVX512BW_Usable)
+			       && HAS_CPU_FEATURE (BMI2)),
+			      __wcscmp_evex)
 	      IFUNC_IMPL_ADD (array, i, wcscmp, 1, __wcscmp_sse2))
 
   /* Support sysdeps/x86_64/multiarch/wcsncmp.c.  */
@@ -391,6 +448,11 @@ __libc_ifunc_impl_list (const char *name, struct libc_ifunc_impl *array,
 	      IFUNC_IMPL_ADD (array, i, wcsncmp,
 			      HAS_ARCH_FEATURE (AVX2_Usable),
 			      __wcsncmp_avx2)
+	      IFUNC_IMPL_ADD (array, i, wcsncmp,
+			      (HAS_ARCH_FEATURE (AVX512VL_Usable)
+			       && HAS_ARCH_FEATURE (AVX512BW_Usable)
+			       && HAS_CPU_FEATURE (BMI2)),
+			      __wcsncmp_evex)
 	      IFUNC_IMPL_ADD (array, i, wcsncmp, 1, __wcsncmp_sse2))
 
   /* Support sysdeps/x86_64/multiarch/wcscpy.c.  */
@@ -404,6 +466,11 @@ __libc_ifunc_impl_list (const char *name, struct libc_ifunc_impl *array,
 	      IFUNC_IMPL_ADD (array, i, wcslen,
 			      HAS_ARCH_FEATURE (AVX2_Usable),
 			      __wcslen_avx2)
+	      IFUNC_IMPL_ADD (array, i, wcslen,
+			      (HAS_ARCH_FEATURE (AVX512VL_Usable)
+			       && HAS_ARCH_FEATURE (AVX512BW_Usable)
+			       && HAS_CPU_FEATURE (BMI2)),
+			      __wcslen_evex)
 	      IFUNC_IMPL_ADD (array, i, wcslen, 1, __wcslen_sse2))
 
   /* Support sysdeps/x86_64/multiarch/wcsnlen.c.  */
@@ -411,6 +478,11 @@ __libc_ifunc_impl_list (const char *name, struct libc_ifunc_impl *array,
 	      IFUNC_IMPL_ADD (array, i, wcsnlen,
 			      HAS_ARCH_FEATURE (AVX2_Usable),
 			      __wcsnlen_avx2)
+	      IFUNC_IMPL_ADD (array, i, wcsnlen,
+			      (HAS_ARCH_FEATURE (AVX512VL_Usable)
+			       && HAS_ARCH_FEATURE (AVX512BW_Usable)
+			       && HAS_CPU_FEATURE (BMI2)),
+			      __wcsnlen_evex)
 	      IFUNC_IMPL_ADD (array, i, wcsnlen,
 			      HAS_CPU_FEATURE (SSE4_1),
 			      __wcsnlen_sse4_1)
@@ -421,6 +493,11 @@ __libc_ifunc_impl_list (const char *name, struct libc_ifunc_impl *array,
 	      IFUNC_IMPL_ADD (array, i, wmemchr,
 			      HAS_ARCH_FEATURE (AVX2_Usable),
 			      __wmemchr_avx2)
+	      IFUNC_IMPL_ADD (array, i, wmemchr,
+			      (HAS_ARCH_FEATURE (AVX512VL_Usable)
+			       && HAS_ARCH_FEATURE (AVX512BW_Usable)
+			       && HAS_CPU_FEATURE (BMI2)),
+			      __wmemchr_evex)
 	      IFUNC_IMPL_ADD (array, i, wmemchr, 1, __wmemchr_sse2))
 
   /* Support sysdeps/x86_64/multiarch/wmemcmp.c.  */
@@ -568,6 +645,10 @@ __libc_ifunc_impl_list (const char *name, struct libc_ifunc_impl *array,
 	      IFUNC_IMPL_ADD (array, i, strncmp,
 			      HAS_ARCH_FEATURE (AVX2_Usable),
 			      __strncmp_avx2)
+	      IFUNC_IMPL_ADD (array, i, strncmp,
+			      (HAS_ARCH_FEATURE (AVX512VL_Usable)
+			       && HAS_ARCH_FEATURE (AVX512BW_Usable)),
+			      __strncmp_evex)
 	      IFUNC_IMPL_ADD (array, i, strncmp, HAS_CPU_FEATURE (SSE4_2),
 			      __strncmp_sse42)
 	      IFUNC_IMPL_ADD (array, i, strncmp, HAS_CPU_FEATURE (SSSE3),

sysdeps/x86_64/multiarch/memchr-evex.S

@@ -0,0 +1,381 @@
+/* memchr/wmemchr optimized with 256-bit EVEX instructions.
+   Copyright (C) 2021 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)
+
+# include <sysdep.h>
+
+# ifndef MEMCHR
+#  define MEMCHR	__memchr_evex
+# endif
+
+# ifdef USE_AS_WMEMCHR
+#  define VPBROADCAST	vpbroadcastd
+#  define VPCMP		vpcmpd
+#  define SHIFT_REG	r8d
+# else
+#  define VPBROADCAST	vpbroadcastb
+#  define VPCMP		vpcmpb
+#  define SHIFT_REG	ecx
+# endif
+
+# define XMMMATCH	xmm16
+# define YMMMATCH	ymm16
+# define YMM1		ymm17
+# define YMM2		ymm18
+# define YMM3		ymm19
+# define YMM4		ymm20
+# define YMM5		ymm21
+# define YMM6		ymm22
+
+# define VEC_SIZE 32
+
+	.section .text.evex,"ax",@progbits
+ENTRY (MEMCHR)
+# ifndef USE_AS_RAWMEMCHR
+	/* Check for zero length.  */
+	test	%RDX_LP, %RDX_LP
+	jz	L(zero)
+# endif
+	movl	%edi, %ecx
+# ifdef USE_AS_WMEMCHR
+	shl	$2, %RDX_LP
+# else
+#  ifdef __ILP32__
+	/* Clear the upper 32 bits.  */
+	movl	%edx, %edx
+#  endif
+# endif
+	/* Broadcast CHAR to YMMMATCH.  */
+	VPBROADCAST %esi, %YMMMATCH
+	/* Check if we may cross page boundary with one vector load.  */
+	andl	$(2 * VEC_SIZE - 1), %ecx
+	cmpl	$VEC_SIZE, %ecx
+	ja	L(cros_page_boundary)
+
+	/* Check the first VEC_SIZE bytes.  */
+	VPCMP	$0, (%rdi), %YMMMATCH, %k1
+	kmovd	%k1, %eax
+	testl	%eax, %eax
+
+# ifndef USE_AS_RAWMEMCHR
+	jnz	L(first_vec_x0_check)
+	/* Adjust length and check the end of data.  */
+	subq	$VEC_SIZE, %rdx
+	jbe	L(zero)
+# else
+	jnz	L(first_vec_x0)
+# endif
+
+	/* Align data for aligned loads in the loop.  */
+	addq	$VEC_SIZE, %rdi
+	andl	$(VEC_SIZE - 1), %ecx
+	andq	$-VEC_SIZE, %rdi
+
+# ifndef USE_AS_RAWMEMCHR
+	/* Adjust length.  */
+	addq	%rcx, %rdx
+
+	subq	$(VEC_SIZE * 4), %rdx
+	jbe	L(last_4x_vec_or_less)
+# endif
+	jmp	L(more_4x_vec)
+
+	.p2align 4
+L(cros_page_boundary):
+	andl	$(VEC_SIZE - 1), %ecx
+# ifdef USE_AS_WMEMCHR
+	/* NB: Divide shift count by 4 since each bit in K1 represent 4
+	   bytes.  */
+	movl	%ecx, %SHIFT_REG
+	sarl	$2, %SHIFT_REG
+# endif
+	andq	$-VEC_SIZE, %rdi
+	VPCMP	$0, (%rdi), %YMMMATCH, %k1
+	kmovd	%k1, %eax
+	/* Remove the leading bytes.  */
+	sarxl	%SHIFT_REG, %eax, %eax
+	testl	%eax, %eax
+	jz	L(aligned_more)
+	tzcntl	%eax, %eax
+# ifdef USE_AS_WMEMCHR
+	/* NB: Multiply wchar_t count by 4 to get the number of bytes.  */
+	sall	$2, %eax
+# endif
+# ifndef USE_AS_RAWMEMCHR
+	/* Check the end of data.  */
+	cmpq	%rax, %rdx
+	jbe	L(zero)
+# endif
+	addq	%rdi, %rax
+	addq	%rcx, %rax
+	ret
+
+	.p2align 4
+L(aligned_more):
+# ifndef USE_AS_RAWMEMCHR
+        /* Calculate "rdx + rcx - VEC_SIZE" with "rdx - (VEC_SIZE - rcx)"
+	   instead of "(rdx + rcx) - VEC_SIZE" to void possible addition
+	   overflow.  */
+	negq	%rcx
+	addq	$VEC_SIZE, %rcx
+
+	/* Check the end of data.  */
+	subq	%rcx, %rdx
+	jbe	L(zero)
+# endif
+
+	addq	$VEC_SIZE, %rdi
+
+# ifndef USE_AS_RAWMEMCHR
+	subq	$(VEC_SIZE * 4), %rdx
+	jbe	L(last_4x_vec_or_less)
+# endif
+
+L(more_4x_vec):
+	/* Check the first 4 * VEC_SIZE.  Only one VEC_SIZE at a time
+	   since data is only aligned to VEC_SIZE.  */
+	VPCMP	$0, (%rdi), %YMMMATCH, %k1
+	kmovd	%k1, %eax
+	testl	%eax, %eax
+	jnz	L(first_vec_x0)
+
+	VPCMP	$0, VEC_SIZE(%rdi), %YMMMATCH, %k1
+	kmovd	%k1, %eax
+	testl	%eax, %eax
+	jnz	L(first_vec_x1)
+
+	VPCMP	$0, (VEC_SIZE * 2)(%rdi), %YMMMATCH, %k1
+	kmovd	%k1, %eax
+	testl	%eax, %eax
+	jnz	L(first_vec_x2)
+
+	VPCMP	$0, (VEC_SIZE * 3)(%rdi), %YMMMATCH, %k1
+	kmovd	%k1, %eax
+	testl	%eax, %eax
+	jnz	L(first_vec_x3)
+
+	addq	$(VEC_SIZE * 4), %rdi
+
+# ifndef USE_AS_RAWMEMCHR
+	subq	$(VEC_SIZE * 4), %rdx
+	jbe	L(last_4x_vec_or_less)
+# endif
+
+	/* Align data to 4 * VEC_SIZE.  */
+	movq	%rdi, %rcx
+	andl	$(4 * VEC_SIZE - 1), %ecx
+	andq	$-(4 * VEC_SIZE), %rdi
+
+# ifndef USE_AS_RAWMEMCHR
+	/* Adjust length.  */
+	addq	%rcx, %rdx
+# endif
+
+	.p2align 4
+L(loop_4x_vec):
+	/* Compare 4 * VEC at a time forward.  */
+	VPCMP	$0, (%rdi), %YMMMATCH, %k1
+	VPCMP	$0, VEC_SIZE(%rdi), %YMMMATCH, %k2
+	kord	%k1, %k2, %k5
+	VPCMP	$0, (VEC_SIZE * 2)(%rdi), %YMMMATCH, %k3
+	VPCMP	$0, (VEC_SIZE * 3)(%rdi), %YMMMATCH, %k4
+
+	kord	%k3, %k4, %k6
+	kortestd %k5, %k6
+	jnz	L(4x_vec_end)
+
+	addq	$(VEC_SIZE * 4), %rdi
+
+# ifdef USE_AS_RAWMEMCHR
+	jmp	L(loop_4x_vec)
+# else
+	subq	$(VEC_SIZE * 4), %rdx
+	ja	L(loop_4x_vec)
+
+L(last_4x_vec_or_less):
+	/* Less than 4 * VEC and aligned to VEC_SIZE.  */
+	addl	$(VEC_SIZE * 2), %edx
+	jle	L(last_2x_vec)
+
+	VPCMP	$0, (%rdi), %YMMMATCH, %k1
+	kmovd	%k1, %eax
+	testl	%eax, %eax
+	jnz	L(first_vec_x0)
+
+	VPCMP	$0, VEC_SIZE(%rdi), %YMMMATCH, %k1
+	kmovd	%k1, %eax
+	testl	%eax, %eax
+	jnz	L(first_vec_x1)
+
+	VPCMP	$0, (VEC_SIZE * 2)(%rdi), %YMMMATCH, %k1
+	kmovd	%k1, %eax
+	testl	%eax, %eax
+
+	jnz	L(first_vec_x2_check)
+	subl	$VEC_SIZE, %edx
+	jle	L(zero)
+
+	VPCMP	$0, (VEC_SIZE * 3)(%rdi), %YMMMATCH, %k1
+	kmovd	%k1, %eax
+	testl	%eax, %eax
+
+	jnz	L(first_vec_x3_check)
+	xorl	%eax, %eax
+	ret
+
+	.p2align 4
+L(last_2x_vec):
+	addl	$(VEC_SIZE * 2), %edx
+	VPCMP	$0, (%rdi), %YMMMATCH, %k1
+	kmovd	%k1, %eax
+	testl	%eax, %eax
+
+	jnz	L(first_vec_x0_check)
+	subl	$VEC_SIZE, %edx
+	jle	L(zero)
+
+	VPCMP	$0, VEC_SIZE(%rdi), %YMMMATCH, %k1
+	kmovd	%k1, %eax
+	testl	%eax, %eax
+	jnz	L(first_vec_x1_check)
+	xorl	%eax, %eax
+	ret
+
+	.p2align 4
+L(first_vec_x0_check):
+	tzcntl	%eax, %eax
+# ifdef USE_AS_WMEMCHR
+	/* NB: Multiply wchar_t count by 4 to get the number of bytes.  */
+	sall	$2, %eax
+# endif
+	/* Check the end of data.  */
+	cmpq	%rax, %rdx
+	jbe	L(zero)
+	addq	%rdi, %rax
+	ret
+
+	.p2align 4
+L(first_vec_x1_check):
+	tzcntl	%eax, %eax
+# ifdef USE_AS_WMEMCHR
+	/* NB: Multiply wchar_t count by 4 to get the number of bytes.  */
+	sall	$2, %eax
+# endif
+	/* Check the end of data.  */
+	cmpq	%rax, %rdx
+	jbe	L(zero)
+	addq	$VEC_SIZE, %rax
+	addq	%rdi, %rax
+	ret
+
+	.p2align 4
+L(first_vec_x2_check):
+	tzcntl	%eax, %eax
+# ifdef USE_AS_WMEMCHR
+	/* NB: Multiply wchar_t count by 4 to get the number of bytes.  */
+	sall	$2, %eax
+# endif
+	/* Check the end of data.  */
+	cmpq	%rax, %rdx
+	jbe	L(zero)
+	addq	$(VEC_SIZE * 2), %rax
+	addq	%rdi, %rax
+	ret
+
+	.p2align 4
+L(first_vec_x3_check):
+	tzcntl	%eax, %eax
+# ifdef USE_AS_WMEMCHR
+	/* NB: Multiply wchar_t count by 4 to get the number of bytes.  */
+	sall	$2, %eax
+# endif
+	/* Check the end of data.  */
+	cmpq	%rax, %rdx
+	jbe	L(zero)
+	addq	$(VEC_SIZE * 3), %rax
+	addq	%rdi, %rax
+	ret
+
+	.p2align 4
+L(zero):
+	xorl	%eax, %eax
+	ret
+# endif
+
+	.p2align 4
+L(first_vec_x0):
+	tzcntl	%eax, %eax
+# ifdef USE_AS_WMEMCHR
+	/* NB: Multiply wchar_t count by 4 to get the number of bytes.  */
+	leaq	(%rdi, %rax, 4), %rax
+# else
+	addq	%rdi, %rax
+# endif
+	ret
+
+	.p2align 4
+L(first_vec_x1):
+	tzcntl	%eax, %eax
+# ifdef USE_AS_WMEMCHR
+	/* NB: Multiply wchar_t count by 4 to get the number of bytes.  */
+	leaq	VEC_SIZE(%rdi, %rax, 4), %rax
+# else
+	addq	$VEC_SIZE, %rax
+	addq	%rdi, %rax
+# endif
+	ret
+
+	.p2align 4
+L(first_vec_x2):
+	tzcntl	%eax, %eax
+# ifdef USE_AS_WMEMCHR
+	/* NB: Multiply wchar_t count by 4 to get the number of bytes.  */
+	leaq	(VEC_SIZE * 2)(%rdi, %rax, 4), %rax
+# else
+	addq	$(VEC_SIZE * 2), %rax
+	addq	%rdi, %rax
+# endif
+	ret
+
+	.p2align 4
+L(4x_vec_end):
+	kmovd	%k1, %eax
+	testl	%eax, %eax
+	jnz	L(first_vec_x0)
+	kmovd	%k2, %eax
+	testl	%eax, %eax
+	jnz	L(first_vec_x1)
+	kmovd	%k3, %eax
+	testl	%eax, %eax
+	jnz	L(first_vec_x2)
+	kmovd	%k4, %eax
+	testl	%eax, %eax
+L(first_vec_x3):
+	tzcntl	%eax, %eax
+# ifdef USE_AS_WMEMCHR
+	/* NB: Multiply wchar_t count by 4 to get the number of bytes.  */
+	leaq	(VEC_SIZE * 3)(%rdi, %rax, 4), %rax
+# else
+	addq	$(VEC_SIZE * 3), %rax
+	addq	%rdi, %rax
+# endif
+	ret
+
+END (MEMCHR)
+#endif

sysdeps/x86_64/multiarch/memrchr-evex.S

@@ -0,0 +1,337 @@
+/* memrchr optimized with 256-bit EVEX instructions.
+   Copyright (C) 2021 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)
+
+# include <sysdep.h>
+
+# define VMOVA		vmovdqa64
+
+# define YMMMATCH	ymm16
+
+# define VEC_SIZE 32
+
+	.section .text.evex,"ax",@progbits
+ENTRY (__memrchr_evex)
+	/* Broadcast CHAR to YMMMATCH.  */
+	vpbroadcastb %esi, %YMMMATCH
+
+	sub	$VEC_SIZE, %RDX_LP
+	jbe	L(last_vec_or_less)
+
+	add	%RDX_LP, %RDI_LP
+
+	/* Check the last VEC_SIZE bytes.  */
+	vpcmpb	$0, (%rdi), %YMMMATCH, %k1
+	kmovd	%k1, %eax
+	testl	%eax, %eax
+	jnz	L(last_vec_x0)
+
+	subq	$(VEC_SIZE * 4), %rdi
+	movl	%edi, %ecx
+	andl	$(VEC_SIZE - 1), %ecx
+	jz	L(aligned_more)
+
+	/* Align data for aligned loads in the loop.  */
+	addq	$VEC_SIZE, %rdi
+	addq	$VEC_SIZE, %rdx
+	andq	$-VEC_SIZE, %rdi
+	subq	%rcx, %rdx
+
+	.p2align 4
+L(aligned_more):
+	subq	$(VEC_SIZE * 4), %rdx
+	jbe	L(last_4x_vec_or_less)
+
+	/* Check the last 4 * VEC_SIZE.  Only one VEC_SIZE at a time
+	   since data is only aligned to VEC_SIZE.  */
+	vpcmpb	$0, (VEC_SIZE * 3)(%rdi), %YMMMATCH, %k1
+	kmovd	%k1, %eax
+	testl	%eax, %eax
+	jnz	L(last_vec_x3)
+
+	vpcmpb	$0, (VEC_SIZE * 2)(%rdi), %YMMMATCH, %k2
+	kmovd	%k2, %eax
+	testl	%eax, %eax
+	jnz	L(last_vec_x2)
+
+	vpcmpb	$0, VEC_SIZE(%rdi), %YMMMATCH, %k3
+	kmovd	%k3, %eax
+	testl	%eax, %eax
+	jnz	L(last_vec_x1)
+
+	vpcmpb	$0, (%rdi), %YMMMATCH, %k4
+	kmovd	%k4, %eax
+	testl	%eax, %eax
+	jnz	L(last_vec_x0)
+
+	/* Align data to 4 * VEC_SIZE for loop with fewer branches.
+	   There are some overlaps with above if data isn't aligned
+	   to 4 * VEC_SIZE.  */
+	movl	%edi, %ecx
+	andl	$(VEC_SIZE * 4 - 1), %ecx
+	jz	L(loop_4x_vec)
+
+	addq	$(VEC_SIZE * 4), %rdi
+	addq	$(VEC_SIZE * 4), %rdx
+	andq	$-(VEC_SIZE * 4), %rdi
+	subq	%rcx, %rdx
+
+	.p2align 4
+L(loop_4x_vec):
+	/* Compare 4 * VEC at a time forward.  */
+	subq	$(VEC_SIZE * 4), %rdi
+	subq	$(VEC_SIZE * 4), %rdx
+	jbe	L(last_4x_vec_or_less)
+
+	vpcmpb	$0, (%rdi), %YMMMATCH, %k1
+	vpcmpb	$0, VEC_SIZE(%rdi), %YMMMATCH, %k2
+	kord	%k1, %k2, %k5
+	vpcmpb	$0, (VEC_SIZE * 2)(%rdi), %YMMMATCH, %k3
+	vpcmpb	$0, (VEC_SIZE * 3)(%rdi), %YMMMATCH, %k4
+
+	kord	%k3, %k4, %k6
+	kortestd %k5, %k6
+	jz	L(loop_4x_vec)
+
+	/* There is a match.  */
+	kmovd	%k4, %eax
+	testl	%eax, %eax
+	jnz	L(last_vec_x3)
+
+	kmovd	%k3, %eax
+	testl	%eax, %eax
+	jnz	L(last_vec_x2)
+
+	kmovd	%k2, %eax
+	testl	%eax, %eax
+	jnz	L(last_vec_x1)
+
+	kmovd	%k1, %eax
+	bsrl	%eax, %eax
+	addq	%rdi, %rax
+	ret
+
+	.p2align 4
+L(last_4x_vec_or_less):
+	addl	$(VEC_SIZE * 4), %edx
+	cmpl	$(VEC_SIZE * 2), %edx
+	jbe	L(last_2x_vec)
+
+	vpcmpb	$0, (VEC_SIZE * 3)(%rdi), %YMMMATCH, %k1
+	kmovd	%k1, %eax
+	testl	%eax, %eax
+	jnz	L(last_vec_x3)
+
+	vpcmpb	$0, (VEC_SIZE * 2)(%rdi), %YMMMATCH, %k2
+	kmovd	%k2, %eax
+	testl	%eax, %eax
+	jnz	L(last_vec_x2)
+
+	vpcmpb	$0, VEC_SIZE(%rdi), %YMMMATCH, %k3
+	kmovd	%k3, %eax
+	testl	%eax, %eax
+	jnz	L(last_vec_x1_check)
+	cmpl	$(VEC_SIZE * 3), %edx
+	jbe	L(zero)
+
+	vpcmpb	$0, (%rdi), %YMMMATCH, %k4
+	kmovd	%k4, %eax
+	testl	%eax, %eax
+	jz	L(zero)
+	bsrl	%eax, %eax
+	subq	$(VEC_SIZE * 4), %rdx
+	addq	%rax, %rdx
+	jl	L(zero)
+	addq	%rdi, %rax
+	ret
+
+	.p2align 4
+L(last_2x_vec):
+	vpcmpb	$0, (VEC_SIZE * 3)(%rdi), %YMMMATCH, %k1
+	kmovd	%k1, %eax
+	testl	%eax, %eax
+	jnz	L(last_vec_x3_check)
+	cmpl	$VEC_SIZE, %edx
+	jbe	L(zero)
+
+	vpcmpb	$0, (VEC_SIZE * 2)(%rdi), %YMMMATCH, %k1
+	kmovd	%k1, %eax
+	testl	%eax, %eax
+	jz	L(zero)
+	bsrl	%eax, %eax
+	subq	$(VEC_SIZE * 2), %rdx
+	addq	%rax, %rdx
+	jl	L(zero)
+	addl	$(VEC_SIZE * 2), %eax
+	addq	%rdi, %rax
+	ret
+
+	.p2align 4
+L(last_vec_x0):
+	bsrl	%eax, %eax
+	addq	%rdi, %rax
+	ret
+
+	.p2align 4
+L(last_vec_x1):
+	bsrl	%eax, %eax
+	addl	$VEC_SIZE, %eax
+	addq	%rdi, %rax
+	ret
+
+	.p2align 4
+L(last_vec_x2):
+	bsrl	%eax, %eax
+	addl	$(VEC_SIZE * 2), %eax
+	addq	%rdi, %rax
+	ret
+
+	.p2align 4
+L(last_vec_x3):
+	bsrl	%eax, %eax
+	addl	$(VEC_SIZE * 3), %eax
+	addq	%rdi, %rax
+	ret
+
+	.p2align 4
+L(last_vec_x1_check):
+	bsrl	%eax, %eax
+	subq	$(VEC_SIZE * 3), %rdx
+	addq	%rax, %rdx
+	jl	L(zero)
+	addl	$VEC_SIZE, %eax
+	addq	%rdi, %rax
+	ret
+
+	.p2align 4
+L(last_vec_x3_check):
+	bsrl	%eax, %eax
+	subq	$VEC_SIZE, %rdx
+	addq	%rax, %rdx
+	jl	L(zero)
+	addl	$(VEC_SIZE * 3), %eax
+	addq	%rdi, %rax
+	ret
+
+	.p2align 4
+L(zero):
+	xorl	%eax, %eax
+	ret
+
+	.p2align 4
+L(last_vec_or_less_aligned):
+	movl	%edx, %ecx
+
+	vpcmpb	$0, (%rdi), %YMMMATCH, %k1
+
+	movl	$1, %edx
+	/* Support rdx << 32.  */
+	salq	%cl, %rdx
+	subq	$1, %rdx
+
+	kmovd	%k1, %eax
+
+	/* Remove the trailing bytes.  */
+	andl	%edx, %eax
+	testl	%eax, %eax
+	jz	L(zero)
+
+	bsrl	%eax, %eax
+	addq	%rdi, %rax
+	ret
+
+	.p2align 4
+L(last_vec_or_less):
+	addl	$VEC_SIZE, %edx
+
+	/* Check for zero length.  */
+	testl	%edx, %edx
+	jz	L(zero)
+
+	movl	%edi, %ecx
+	andl	$(VEC_SIZE - 1), %ecx
+	jz	L(last_vec_or_less_aligned)
+
+	movl	%ecx, %esi
+	movl	%ecx, %r8d
+	addl	%edx, %esi
+	andq	$-VEC_SIZE, %rdi
+
+	subl	$VEC_SIZE, %esi
+	ja	L(last_vec_2x_aligned)
+
+	/* Check the last VEC.  */
+	vpcmpb	$0, (%rdi), %YMMMATCH, %k1
+	kmovd	%k1, %eax
+
+	/* Remove the leading and trailing bytes.  */
+	sarl	%cl, %eax
+	movl	%edx, %ecx
+
+	movl	$1, %edx
+	sall	%cl, %edx
+	subl	$1, %edx
+
+	andl	%edx, %eax
+	testl	%eax, %eax
+	jz	L(zero)
+
+	bsrl	%eax, %eax
+	addq	%rdi, %rax
+	addq	%r8, %rax
+	ret
+
+	.p2align 4
+L(last_vec_2x_aligned):
+	movl	%esi, %ecx
+
+	/* Check the last VEC.  */
+	vpcmpb	$0, VEC_SIZE(%rdi), %YMMMATCH, %k1
+
+	movl	$1, %edx
+	sall	%cl, %edx
+	subl	$1, %edx
+
+	kmovd	%k1, %eax
+
+	/* Remove the trailing bytes.  */
+	andl	%edx, %eax
+
+	testl	%eax, %eax
+	jnz	L(last_vec_x1)
+
+	/* Check the second last VEC.  */
+	vpcmpb	$0, (%rdi), %YMMMATCH, %k1
+
+	movl	%r8d, %ecx
+
+	kmovd	%k1, %eax
+
+	/* Remove the leading bytes.  Must use unsigned right shift for
+	   bsrl below.  */
+	shrl	%cl, %eax
+	testl	%eax, %eax
+	jz	L(zero)
+
+	bsrl	%eax, %eax
+	addq	%rdi, %rax
+	addq	%r8, %rax
+	ret
+END (__memrchr_evex)
+#endif

sysdeps/x86_64/multiarch/rawmemchr-evex.S

@@ -0,0 +1,4 @@
+#define MEMCHR __rawmemchr_evex
+#define USE_AS_RAWMEMCHR 1
+
+#include "memchr-evex.S"

sysdeps/x86_64/multiarch/strchr-evex.S

@@ -0,0 +1,335 @@
+/* strchr/strchrnul optimized with 256-bit EVEX instructions.
+   Copyright (C) 2021 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)
+
+# include <sysdep.h>
+
+# ifndef STRCHR
+#  define STRCHR	__strchr_evex
+# endif
+
+# define VMOVU		vmovdqu64
+# define VMOVA		vmovdqa64
+
+# ifdef USE_AS_WCSCHR
+#  define VPBROADCAST	vpbroadcastd
+#  define VPCMP		vpcmpd
+#  define VPMINU	vpminud
+#  define CHAR_REG	esi
+#  define SHIFT_REG	r8d
+# else
+#  define VPBROADCAST	vpbroadcastb
+#  define VPCMP		vpcmpb
+#  define VPMINU	vpminub
+#  define CHAR_REG	sil
+#  define SHIFT_REG	ecx
+# endif
+
+# define XMMZERO	xmm16
+
+# define YMMZERO	ymm16
+# define YMM0		ymm17
+# define YMM1		ymm18
+# define YMM2		ymm19
+# define YMM3		ymm20
+# define YMM4		ymm21
+# define YMM5		ymm22
+# define YMM6		ymm23
+# define YMM7		ymm24
+# define YMM8		ymm25
+
+# define VEC_SIZE 32
+# define PAGE_SIZE 4096
+
+	.section .text.evex,"ax",@progbits
+ENTRY (STRCHR)
+	movl	%edi, %ecx
+# ifndef USE_AS_STRCHRNUL
+	xorl	%edx, %edx
+# endif
+
+	/* Broadcast CHAR to YMM0.	*/
+	VPBROADCAST %esi, %YMM0
+
+	vpxorq	%XMMZERO, %XMMZERO, %XMMZERO
+
+	/* Check if we cross page boundary with one vector load.  */
+	andl	$(PAGE_SIZE - 1), %ecx
+	cmpl	$(PAGE_SIZE - VEC_SIZE), %ecx
+	ja  L(cross_page_boundary)
+
+	/* Check the first VEC_SIZE bytes. Search for both CHAR and the
+	   null bytes.  */
+	VMOVU	(%rdi), %YMM1
+
+	/* Leaves only CHARS matching esi as 0.  */
+	vpxorq	%YMM1, %YMM0, %YMM2
+	VPMINU	%YMM2, %YMM1, %YMM2
+	/* Each bit in K0 represents a CHAR or a null byte in YMM1.  */
+	VPCMP	$0, %YMMZERO, %YMM2, %k0
+	ktestd	%k0, %k0
+	jz	L(more_vecs)
+	kmovd	%k0, %eax
+	tzcntl	%eax, %eax
+	/* Found CHAR or the null byte.	 */
+# ifdef USE_AS_WCSCHR
+	/* NB: Multiply wchar_t count by 4 to get the number of bytes.  */
+	leaq	(%rdi, %rax, 4), %rax
+# else
+	addq	%rdi, %rax
+# endif
+# ifndef USE_AS_STRCHRNUL
+	cmp (%rax), %CHAR_REG
+	cmovne	%rdx, %rax
+# endif
+	ret
+
+	.p2align 4
+L(more_vecs):
+	/* Align data for aligned loads in the loop.  */
+	andq	$-VEC_SIZE, %rdi
+L(aligned_more):
+
+	/* Check the next 4 * VEC_SIZE.	 Only one VEC_SIZE at a time
+	   since data is only aligned to VEC_SIZE.	*/
+	VMOVA	VEC_SIZE(%rdi), %YMM1
+	addq	$VEC_SIZE, %rdi
+
+	/* Leaves only CHARS matching esi as 0.  */
+	vpxorq	%YMM1, %YMM0, %YMM2
+	VPMINU	%YMM2, %YMM1, %YMM2
+	/* Each bit in K0 represents a CHAR or a null byte in YMM1.  */
+	VPCMP	$0, %YMMZERO, %YMM2, %k0
+	kmovd	%k0, %eax
+	testl	%eax, %eax
+	jnz	L(first_vec_x0)
+
+	VMOVA	VEC_SIZE(%rdi), %YMM1
+	/* Leaves only CHARS matching esi as 0.  */
+	vpxorq	%YMM1, %YMM0, %YMM2
+	VPMINU	%YMM2, %YMM1, %YMM2
+	/* Each bit in K0 represents a CHAR or a null byte in YMM1.  */
+	VPCMP	$0, %YMMZERO, %YMM2, %k0
+	kmovd	%k0, %eax
+	testl	%eax, %eax
+	jnz	L(first_vec_x1)
+
+	VMOVA	(VEC_SIZE * 2)(%rdi), %YMM1
+	/* Leaves only CHARS matching esi as 0.  */
+	vpxorq	%YMM1, %YMM0, %YMM2
+	VPMINU	%YMM2, %YMM1, %YMM2
+	/* Each bit in K0 represents a CHAR or a null byte in YMM1.  */
+	VPCMP	$0, %YMMZERO, %YMM2, %k0
+	kmovd	%k0, %eax
+	testl	%eax, %eax
+	jnz	L(first_vec_x2)
+
+	VMOVA	(VEC_SIZE * 3)(%rdi), %YMM1
+	/* Leaves only CHARS matching esi as 0.  */
+	vpxorq	%YMM1, %YMM0, %YMM2
+	VPMINU	%YMM2, %YMM1, %YMM2
+	/* Each bit in K0 represents a CHAR or a null byte in YMM1.  */
+	VPCMP	$0, %YMMZERO, %YMM2, %k0
+	ktestd	%k0, %k0
+	jz	L(prep_loop_4x)
+
+	kmovd	%k0, %eax
+	tzcntl	%eax, %eax
+	/* Found CHAR or the null byte.	 */
+# ifdef USE_AS_WCSCHR
+	/* NB: Multiply wchar_t count by 4 to get the number of bytes.  */
+	leaq	(VEC_SIZE * 3)(%rdi, %rax, 4), %rax
+# else
+	leaq	(VEC_SIZE * 3)(%rdi, %rax), %rax
+# endif
+# ifndef USE_AS_STRCHRNUL
+	cmp (%rax), %CHAR_REG
+	cmovne	%rdx, %rax
+# endif
+	ret
+
+	.p2align 4
+L(first_vec_x0):
+	tzcntl	%eax, %eax
+	/* Found CHAR or the null byte.	 */
+# ifdef USE_AS_WCSCHR
+	/* NB: Multiply wchar_t count by 4 to get the number of bytes.  */
+	leaq	(%rdi, %rax, 4), %rax
+# else
+	addq	%rdi, %rax
+# endif
+# ifndef USE_AS_STRCHRNUL
+	cmp (%rax), %CHAR_REG
+	cmovne	%rdx, %rax
+# endif
+	ret
+
+	.p2align 4
+L(first_vec_x1):
+	tzcntl	%eax, %eax
+	/* Found CHAR or the null byte.	 */
+# ifdef USE_AS_WCSCHR
+	/* NB: Multiply wchar_t count by 4 to get the number of bytes.  */
+	leaq	VEC_SIZE(%rdi, %rax, 4), %rax
+# else
+	leaq	VEC_SIZE(%rdi, %rax), %rax
+# endif
+# ifndef USE_AS_STRCHRNUL
+	cmp (%rax), %CHAR_REG
+	cmovne	%rdx, %rax
+# endif
+	ret
+
+	.p2align 4
+L(first_vec_x2):
+	tzcntl	%eax, %eax
+	/* Found CHAR or the null byte.	 */
+# ifdef USE_AS_WCSCHR
+	/* NB: Multiply wchar_t count by 4 to get the number of bytes.  */
+	leaq	(VEC_SIZE * 2)(%rdi, %rax, 4), %rax
+# else
+	leaq	(VEC_SIZE * 2)(%rdi, %rax), %rax
+# endif
+# ifndef USE_AS_STRCHRNUL
+	cmp (%rax), %CHAR_REG
+	cmovne	%rdx, %rax
+# endif
+	ret
+
+L(prep_loop_4x):
+	/* Align data to 4 * VEC_SIZE.	*/
+	andq	$-(VEC_SIZE * 4), %rdi
+
+	.p2align 4
+L(loop_4x_vec):
+	/* Compare 4 * VEC at a time forward.  */
+	VMOVA	(VEC_SIZE * 4)(%rdi), %YMM1
+	VMOVA	(VEC_SIZE * 5)(%rdi), %YMM2
+	VMOVA	(VEC_SIZE * 6)(%rdi), %YMM3
+	VMOVA	(VEC_SIZE * 7)(%rdi), %YMM4
+
+	/* Leaves only CHARS matching esi as 0.  */
+	vpxorq	%YMM1, %YMM0, %YMM5
+	vpxorq	%YMM2, %YMM0, %YMM6
+	vpxorq	%YMM3, %YMM0, %YMM7
+	vpxorq	%YMM4, %YMM0, %YMM8
+
+	VPMINU	%YMM5, %YMM1, %YMM5
+	VPMINU	%YMM6, %YMM2, %YMM6
+	VPMINU	%YMM7, %YMM3, %YMM7
+	VPMINU	%YMM8, %YMM4, %YMM8
+
+	VPMINU	%YMM5, %YMM6, %YMM1
+	VPMINU	%YMM7, %YMM8, %YMM2
+
+	VPMINU	%YMM1, %YMM2, %YMM1
+
+	/* Each bit in K0 represents a CHAR or a null byte.  */
+	VPCMP	$0, %YMMZERO, %YMM1, %k0
+
+	addq	$(VEC_SIZE * 4), %rdi
+
+	ktestd	%k0, %k0
+	jz	L(loop_4x_vec)
+
+	/* Each bit in K0 represents a CHAR or a null byte in YMM1.  */
+	VPCMP	$0, %YMMZERO, %YMM5, %k0
+	kmovd	%k0, %eax
+	testl	%eax, %eax
+	jnz	L(first_vec_x0)
+
+	/* Each bit in K1 represents a CHAR or a null byte in YMM2.  */
+	VPCMP	$0, %YMMZERO, %YMM6, %k1
+	kmovd	%k1, %eax
+	testl	%eax, %eax
+	jnz	L(first_vec_x1)
+
+	/* Each bit in K2 represents a CHAR or a null byte in YMM3.  */
+	VPCMP	$0, %YMMZERO, %YMM7, %k2
+	/* Each bit in K3 represents a CHAR or a null byte in YMM4.  */
+	VPCMP	$0, %YMMZERO, %YMM8, %k3
+
+# ifdef USE_AS_WCSCHR
+	/* NB: Each bit in K2/K3 represents 4-byte element.  */
+	kshiftlw $8, %k3, %k1
+# else
+	kshiftlq $32, %k3, %k1
+# endif
+
+	/* Each bit in K1 represents a NULL or a mismatch.  */
+	korq	%k1, %k2, %k1
+	kmovq	%k1, %rax
+
+	tzcntq  %rax, %rax
+# ifdef USE_AS_WCSCHR
+	/* NB: Multiply wchar_t count by 4 to get the number of bytes.  */
+	leaq	(VEC_SIZE * 2)(%rdi, %rax, 4), %rax
+# else
+	leaq	(VEC_SIZE * 2)(%rdi, %rax), %rax
+# endif
+# ifndef USE_AS_STRCHRNUL
+	cmp (%rax), %CHAR_REG
+	cmovne	%rdx, %rax
+# endif
+	ret
+
+	/* Cold case for crossing page with first load.	 */
+	.p2align 4
+L(cross_page_boundary):
+	andq	$-VEC_SIZE, %rdi
+	andl	$(VEC_SIZE - 1), %ecx
+
+	VMOVA	(%rdi), %YMM1
+
+	/* Leaves only CHARS matching esi as 0.  */
+	vpxorq	%YMM1, %YMM0, %YMM2
+	VPMINU	%YMM2, %YMM1, %YMM2
+	/* Each bit in K0 represents a CHAR or a null byte in YMM1.  */
+	VPCMP	$0, %YMMZERO, %YMM2, %k0
+	kmovd	%k0, %eax
+	testl	%eax, %eax
+
+# ifdef USE_AS_WCSCHR
+	/* NB: Divide shift count by 4 since each bit in K1 represent 4
+	   bytes.  */
+	movl	%ecx, %SHIFT_REG
+	sarl    $2, %SHIFT_REG
+# endif
+
+	/* Remove the leading bits.	 */
+	sarxl	%SHIFT_REG, %eax, %eax
+	testl	%eax, %eax
+
+	jz	L(aligned_more)
+	tzcntl	%eax, %eax
+	addq	%rcx, %rdi
+# ifdef USE_AS_WCSCHR
+	/* NB: Multiply wchar_t count by 4 to get the number of bytes.  */
+	leaq	(%rdi, %rax, 4), %rax
+# else
+	addq	%rdi, %rax
+# endif
+# ifndef USE_AS_STRCHRNUL
+	cmp (%rax), %CHAR_REG
+	cmovne	%rdx, %rax
+# endif
+	ret
+
+END (STRCHR)
+# endif

sysdeps/x86_64/multiarch/strchr.c

@@ -29,16 +29,24 @@
 extern __typeof (REDIRECT_NAME) OPTIMIZE (sse2) attribute_hidden;
 extern __typeof (REDIRECT_NAME) OPTIMIZE (sse2_no_bsf) attribute_hidden;
 extern __typeof (REDIRECT_NAME) OPTIMIZE (avx2) attribute_hidden;
+extern __typeof (REDIRECT_NAME) OPTIMIZE (evex) attribute_hidden;
 
 static inline void *
 IFUNC_SELECTOR (void)
 {
   const struct cpu_features* cpu_features = __get_cpu_features ();
 
-  if (!CPU_FEATURES_ARCH_P (cpu_features, Prefer_No_VZEROUPPER)
+  if (CPU_FEATURES_ARCH_P (cpu_features, AVX2_Usable)
-      && CPU_FEATURES_ARCH_P (cpu_features, AVX2_Usable)
       && CPU_FEATURES_ARCH_P (cpu_features, AVX_Fast_Unaligned_Load))
-    return OPTIMIZE (avx2);
+    {
+      if (CPU_FEATURES_ARCH_P (cpu_features, AVX512VL_Usable)
+	  && CPU_FEATURES_ARCH_P (cpu_features, AVX512BW_Usable)
+	  && CPU_FEATURES_CPU_P (cpu_features, BMI2))
+	return OPTIMIZE (evex);
+
+      if (!CPU_FEATURES_ARCH_P (cpu_features, Prefer_No_VZEROUPPER))
+	return OPTIMIZE (avx2);
+    }
 
   if (CPU_FEATURES_ARCH_P (cpu_features, Slow_BSF))
     return OPTIMIZE (sse2_no_bsf);

sysdeps/x86_64/multiarch/strchrnul-evex.S

@@ -0,0 +1,3 @@
+#define STRCHR __strchrnul_evex
+#define USE_AS_STRCHRNUL 1
+#include "strchr-evex.S"

sysdeps/x86_64/multiarch/strcmp.c

@@ -30,16 +30,25 @@ extern __typeof (REDIRECT_NAME) OPTIMIZE (sse2) attribute_hidden;
 extern __typeof (REDIRECT_NAME) OPTIMIZE (sse2_unaligned) attribute_hidden;
 extern __typeof (REDIRECT_NAME) OPTIMIZE (ssse3) attribute_hidden;
 extern __typeof (REDIRECT_NAME) OPTIMIZE (avx2) attribute_hidden;
+extern __typeof (REDIRECT_NAME) OPTIMIZE (evex) attribute_hidden;
 
 static inline void *
 IFUNC_SELECTOR (void)
 {
   const struct cpu_features* cpu_features = __get_cpu_features ();
 
-  if (!CPU_FEATURES_ARCH_P (cpu_features, Prefer_No_VZEROUPPER)
+  if (CPU_FEATURES_ARCH_P (cpu_features, AVX2_Usable)
-      && CPU_FEATURES_ARCH_P (cpu_features, AVX2_Usable)
       && CPU_FEATURES_ARCH_P (cpu_features, AVX_Fast_Unaligned_Load))
-    return OPTIMIZE (avx2);
+    {
+      if (CPU_FEATURES_ARCH_P (cpu_features, AVX512VL_Usable)
+	  && CPU_FEATURES_ARCH_P (cpu_features, AVX512BW_Usable)
+	  && CPU_FEATURES_CPU_P (cpu_features, BMI2)
+	  && !CPU_FEATURES_ARCH_P (cpu_features, Prefer_AVX2_STRCMP))
+	return OPTIMIZE (evex);
+
+      if (!CPU_FEATURES_ARCH_P (cpu_features, Prefer_No_VZEROUPPER))
+	return OPTIMIZE (avx2);
+    }
 
   if (CPU_FEATURES_ARCH_P (cpu_features, Fast_Unaligned_Load))
     return OPTIMIZE (sse2_unaligned);

sysdeps/x86_64/multiarch/strlen-evex.S

@@ -0,0 +1,436 @@
+/* strlen/strnlen/wcslen/wcsnlen optimized with 256-bit EVEX instructions.
+   Copyright (C) 2021 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)
+
+# include <sysdep.h>
+
+# ifndef STRLEN
+#  define STRLEN	__strlen_evex
+# endif
+
+# define VMOVA		vmovdqa64
+
+# ifdef USE_AS_WCSLEN
+#  define VPCMP		vpcmpd
+#  define VPMINU	vpminud
+#  define SHIFT_REG	r9d
+# else
+#  define VPCMP		vpcmpb
+#  define VPMINU	vpminub
+#  define SHIFT_REG	ecx
+# endif
+
+# define XMMZERO	xmm16
+# define YMMZERO	ymm16
+# define YMM1		ymm17
+# define YMM2		ymm18
+# define YMM3		ymm19
+# define YMM4		ymm20
+# define YMM5		ymm21
+# define YMM6		ymm22
+
+# define VEC_SIZE 32
+
+	.section .text.evex,"ax",@progbits
+ENTRY (STRLEN)
+# ifdef USE_AS_STRNLEN
+	/* Check for zero length.  */
+	test	%RSI_LP, %RSI_LP
+	jz	L(zero)
+#  ifdef USE_AS_WCSLEN
+	shl	$2, %RSI_LP
+#  elif defined __ILP32__
+	/* Clear the upper 32 bits.  */
+	movl	%esi, %esi
+#  endif
+	mov	%RSI_LP, %R8_LP
+# endif
+	movl	%edi, %ecx
+	movq	%rdi, %rdx
+	vpxorq	%XMMZERO, %XMMZERO, %XMMZERO
+
+	/* Check if we may cross page boundary with one vector load.  */
+	andl	$(2 * VEC_SIZE - 1), %ecx
+	cmpl	$VEC_SIZE, %ecx
+	ja	L(cros_page_boundary)
+
+	/* Check the first VEC_SIZE bytes.  Each bit in K0 represents a
+	   null byte.  */
+	VPCMP	$0, (%rdi), %YMMZERO, %k0
+	kmovd	%k0, %eax
+	testl	%eax, %eax
+
+# ifdef USE_AS_STRNLEN
+	jnz	L(first_vec_x0_check)
+	/* Adjust length and check the end of data.  */
+	subq	$VEC_SIZE, %rsi
+	jbe	L(max)
+# else
+	jnz	L(first_vec_x0)
+# endif
+
+	/* Align data for aligned loads in the loop.  */
+	addq	$VEC_SIZE, %rdi
+	andl	$(VEC_SIZE - 1), %ecx
+	andq	$-VEC_SIZE, %rdi
+
+# ifdef USE_AS_STRNLEN
+	/* Adjust length.  */
+	addq	%rcx, %rsi
+
+	subq	$(VEC_SIZE * 4), %rsi
+	jbe	L(last_4x_vec_or_less)
+# endif
+	jmp	L(more_4x_vec)
+
+	.p2align 4
+L(cros_page_boundary):
+	andl	$(VEC_SIZE - 1), %ecx
+	andq	$-VEC_SIZE, %rdi
+
+# ifdef USE_AS_WCSLEN
+	/* NB: Divide shift count by 4 since each bit in K0 represent 4
+	   bytes.  */
+	movl	%ecx, %SHIFT_REG
+	sarl	$2, %SHIFT_REG
+# endif
+	VPCMP	$0, (%rdi), %YMMZERO, %k0
+	kmovd	%k0, %eax
+
+	/* Remove the leading bytes.  */
+	sarxl	%SHIFT_REG, %eax, %eax
+	testl	%eax, %eax
+	jz	L(aligned_more)
+	tzcntl	%eax, %eax
+# ifdef USE_AS_WCSLEN
+	/* NB: Multiply wchar_t count by 4 to get the number of bytes.  */
+	sall	$2, %eax
+# endif
+# ifdef USE_AS_STRNLEN
+	/* Check the end of data.  */
+	cmpq	%rax, %rsi
+	jbe	L(max)
+# endif
+	addq	%rdi, %rax
+	addq	%rcx, %rax
+	subq	%rdx, %rax
+# ifdef USE_AS_WCSLEN
+	shrq	$2, %rax
+# endif
+	ret
+
+	.p2align 4
+L(aligned_more):
+# ifdef USE_AS_STRNLEN
+        /* "rcx" is less than VEC_SIZE.  Calculate "rdx + rcx - VEC_SIZE"
+	    with "rdx - (VEC_SIZE - rcx)" instead of "(rdx + rcx) - VEC_SIZE"
+	    to void possible addition overflow.  */
+	negq	%rcx
+	addq	$VEC_SIZE, %rcx
+
+	/* Check the end of data.  */
+	subq	%rcx, %rsi
+	jbe	L(max)
+# endif
+
+	addq	$VEC_SIZE, %rdi
+
+# ifdef USE_AS_STRNLEN
+	subq	$(VEC_SIZE * 4), %rsi
+	jbe	L(last_4x_vec_or_less)
+# endif
+
+L(more_4x_vec):
+	/* Check the first 4 * VEC_SIZE.  Only one VEC_SIZE at a time
+	   since data is only aligned to VEC_SIZE.  */
+	VPCMP	$0, (%rdi), %YMMZERO, %k0
+	kmovd	%k0, %eax
+	testl	%eax, %eax
+	jnz	L(first_vec_x0)
+
+	VPCMP	$0, VEC_SIZE(%rdi), %YMMZERO, %k0
+	kmovd	%k0, %eax
+	testl	%eax, %eax
+	jnz	L(first_vec_x1)
+
+	VPCMP	$0, (VEC_SIZE * 2)(%rdi), %YMMZERO, %k0
+	kmovd	%k0, %eax
+	testl	%eax, %eax
+	jnz	L(first_vec_x2)
+
+	VPCMP	$0, (VEC_SIZE * 3)(%rdi), %YMMZERO, %k0
+	kmovd	%k0, %eax
+	testl	%eax, %eax
+	jnz	L(first_vec_x3)
+
+	addq	$(VEC_SIZE * 4), %rdi
+
+# ifdef USE_AS_STRNLEN
+	subq	$(VEC_SIZE * 4), %rsi
+	jbe	L(last_4x_vec_or_less)
+# endif
+
+	/* Align data to 4 * VEC_SIZE.  */
+	movq	%rdi, %rcx
+	andl	$(4 * VEC_SIZE - 1), %ecx
+	andq	$-(4 * VEC_SIZE), %rdi
+
+# ifdef USE_AS_STRNLEN
+	/* Adjust length.  */
+	addq	%rcx, %rsi
+# endif
+
+	.p2align 4
+L(loop_4x_vec):
+	/* Compare 4 * VEC at a time forward.  */
+	VMOVA	(%rdi), %YMM1
+	VMOVA	VEC_SIZE(%rdi), %YMM2
+	VMOVA	(VEC_SIZE * 2)(%rdi), %YMM3
+	VMOVA	(VEC_SIZE * 3)(%rdi), %YMM4
+
+	VPMINU	%YMM1, %YMM2, %YMM5
+	VPMINU	%YMM3, %YMM4, %YMM6
+
+	VPMINU	%YMM5, %YMM6, %YMM5
+	VPCMP	$0, %YMM5, %YMMZERO, %k0
+	ktestd	%k0, %k0
+	jnz	L(4x_vec_end)
+
+	addq	$(VEC_SIZE * 4), %rdi
+
+# ifndef USE_AS_STRNLEN
+	jmp	L(loop_4x_vec)
+# else
+	subq	$(VEC_SIZE * 4), %rsi
+	ja	L(loop_4x_vec)
+
+L(last_4x_vec_or_less):
+	/* Less than 4 * VEC and aligned to VEC_SIZE.  */
+	addl	$(VEC_SIZE * 2), %esi
+	jle	L(last_2x_vec)
+
+	VPCMP	$0, (%rdi), %YMMZERO, %k0
+	kmovd	%k0, %eax
+	testl	%eax, %eax
+	jnz	L(first_vec_x0)
+
+	VPCMP	$0, VEC_SIZE(%rdi), %YMMZERO, %k0
+	kmovd	%k0, %eax
+	testl	%eax, %eax
+	jnz	L(first_vec_x1)
+
+	VPCMP	$0, (VEC_SIZE * 2)(%rdi), %YMMZERO, %k0
+	kmovd	%k0, %eax
+	testl	%eax, %eax
+	jnz	L(first_vec_x2_check)
+	subl	$VEC_SIZE, %esi
+	jle	L(max)
+
+	VPCMP	$0, (VEC_SIZE * 3)(%rdi), %YMMZERO, %k0
+	kmovd	%k0, %eax
+	testl	%eax, %eax
+	jnz	L(first_vec_x3_check)
+	movq	%r8, %rax
+#  ifdef USE_AS_WCSLEN
+	shrq	$2, %rax
+#  endif
+	ret
+
+	.p2align 4
+L(last_2x_vec):
+	addl	$(VEC_SIZE * 2), %esi
+
+	VPCMP	$0, (%rdi), %YMMZERO, %k0
+	kmovd	%k0, %eax
+	testl	%eax, %eax
+	jnz	L(first_vec_x0_check)
+	subl	$VEC_SIZE, %esi
+	jle	L(max)
+
+	VPCMP	$0, VEC_SIZE(%rdi), %YMMZERO, %k0
+	kmovd	%k0, %eax
+	testl	%eax, %eax
+	jnz	L(first_vec_x1_check)
+	movq	%r8, %rax
+#  ifdef USE_AS_WCSLEN
+	shrq	$2, %rax
+#  endif
+	ret
+
+	.p2align 4
+L(first_vec_x0_check):
+	tzcntl	%eax, %eax
+# ifdef USE_AS_WCSLEN
+	/* NB: Multiply wchar_t count by 4 to get the number of bytes.  */
+	sall	$2, %eax
+# endif
+	/* Check the end of data.  */
+	cmpq	%rax, %rsi
+	jbe	L(max)
+	addq	%rdi, %rax
+	subq	%rdx, %rax
+#  ifdef USE_AS_WCSLEN
+	shrq	$2, %rax
+#  endif
+	ret
+
+	.p2align 4
+L(first_vec_x1_check):
+	tzcntl	%eax, %eax
+# ifdef USE_AS_WCSLEN
+	/* NB: Multiply wchar_t count by 4 to get the number of bytes.  */
+	sall	$2, %eax
+# endif
+	/* Check the end of data.  */
+	cmpq	%rax, %rsi
+	jbe	L(max)
+	addq	$VEC_SIZE, %rax
+	addq	%rdi, %rax
+	subq	%rdx, %rax
+#  ifdef USE_AS_WCSLEN
+	shrq	$2, %rax
+#  endif
+	ret
+
+	.p2align 4
+L(first_vec_x2_check):
+	tzcntl	%eax, %eax
+# ifdef USE_AS_WCSLEN
+	/* NB: Multiply wchar_t count by 4 to get the number of bytes.  */
+	sall	$2, %eax
+# endif
+	/* Check the end of data.  */
+	cmpq	%rax, %rsi
+	jbe	L(max)
+	addq	$(VEC_SIZE * 2), %rax
+	addq	%rdi, %rax
+	subq	%rdx, %rax
+#  ifdef USE_AS_WCSLEN
+	shrq	$2, %rax
+#  endif
+	ret
+
+	.p2align 4
+L(first_vec_x3_check):
+	tzcntl	%eax, %eax
+# ifdef USE_AS_WCSLEN
+	/* NB: Multiply wchar_t count by 4 to get the number of bytes.  */
+	sall	$2, %eax
+# endif
+	/* Check the end of data.  */
+	cmpq	%rax, %rsi
+	jbe	L(max)
+	addq	$(VEC_SIZE * 3), %rax
+	addq	%rdi, %rax
+	subq	%rdx, %rax
+#  ifdef USE_AS_WCSLEN
+	shrq	$2, %rax
+#  endif
+	ret
+
+	.p2align 4
+L(max):
+	movq	%r8, %rax
+#  ifdef USE_AS_WCSLEN
+	shrq	$2, %rax
+#  endif
+	ret
+
+	.p2align 4
+L(zero):
+	xorl	%eax, %eax
+	ret
+# endif
+
+	.p2align 4
+L(first_vec_x0):
+	tzcntl	%eax, %eax
+# ifdef USE_AS_WCSLEN
+	/* NB: Multiply wchar_t count by 4 to get the number of bytes.  */
+	sall	$2, %eax
+# endif
+	addq	%rdi, %rax
+	subq	%rdx, %rax
+# ifdef USE_AS_WCSLEN
+	shrq	$2, %rax
+# endif
+	ret
+
+	.p2align 4
+L(first_vec_x1):
+	tzcntl	%eax, %eax
+# ifdef USE_AS_WCSLEN
+	/* NB: Multiply wchar_t count by 4 to get the number of bytes.  */
+	sall	$2, %eax
+# endif
+	addq	$VEC_SIZE, %rax
+	addq	%rdi, %rax
+	subq	%rdx, %rax
+# ifdef USE_AS_WCSLEN
+	shrq	$2, %rax
+# endif
+	ret
+
+	.p2align 4
+L(first_vec_x2):
+	tzcntl	%eax, %eax
+# ifdef USE_AS_WCSLEN
+	/* NB: Multiply wchar_t count by 4 to get the number of bytes.  */
+	sall	$2, %eax
+# endif
+	addq	$(VEC_SIZE * 2), %rax
+	addq	%rdi, %rax
+	subq	%rdx, %rax
+# ifdef USE_AS_WCSLEN
+	shrq	$2, %rax
+# endif
+	ret
+
+	.p2align 4
+L(4x_vec_end):
+	VPCMP	$0, %YMM1, %YMMZERO, %k0
+	kmovd	%k0, %eax
+	testl	%eax, %eax
+	jnz	L(first_vec_x0)
+	VPCMP	$0, %YMM2, %YMMZERO, %k1
+	kmovd	%k1, %eax
+	testl	%eax, %eax
+	jnz	L(first_vec_x1)
+	VPCMP	$0, %YMM3, %YMMZERO, %k2
+	kmovd	%k2, %eax
+	testl	%eax, %eax
+	jnz	L(first_vec_x2)
+	VPCMP	$0, %YMM4, %YMMZERO, %k3
+	kmovd	%k3, %eax
+L(first_vec_x3):
+	tzcntl	%eax, %eax
+# ifdef USE_AS_WCSLEN
+	/* NB: Multiply wchar_t count by 4 to get the number of bytes.  */
+	sall	$2, %eax
+# endif
+	addq	$(VEC_SIZE * 3), %rax
+	addq	%rdi, %rax
+	subq	%rdx, %rax
+# ifdef USE_AS_WCSLEN
+	shrq	$2, %rax
+# endif
+	ret
+
+END (STRLEN)
+#endif

sysdeps/x86_64/multiarch/strncmp-evex.S

@@ -0,0 +1,3 @@
+#define STRCMP	__strncmp_evex
+#define USE_AS_STRNCMP 1
+#include "strcmp-evex.S"

sysdeps/x86_64/multiarch/strncmp.c

@@ -30,16 +30,25 @@ extern __typeof (REDIRECT_NAME) OPTIMIZE (sse2) attribute_hidden;
 extern __typeof (REDIRECT_NAME) OPTIMIZE (ssse3) attribute_hidden;
 extern __typeof (REDIRECT_NAME) OPTIMIZE (sse42) attribute_hidden;
 extern __typeof (REDIRECT_NAME) OPTIMIZE (avx2) attribute_hidden;
+extern __typeof (REDIRECT_NAME) OPTIMIZE (evex) attribute_hidden;
 
 static inline void *
 IFUNC_SELECTOR (void)
 {
   const struct cpu_features* cpu_features = __get_cpu_features ();
 
-  if (!CPU_FEATURES_ARCH_P (cpu_features, Prefer_No_VZEROUPPER)
+  if (CPU_FEATURES_ARCH_P (cpu_features, AVX2_Usable)
-      && CPU_FEATURES_ARCH_P (cpu_features, AVX2_Usable)
       && CPU_FEATURES_ARCH_P (cpu_features, AVX_Fast_Unaligned_Load))
-    return OPTIMIZE (avx2);
+    {
+      if (CPU_FEATURES_ARCH_P (cpu_features, AVX512VL_Usable)
+	  && CPU_FEATURES_ARCH_P (cpu_features, AVX512BW_Usable)
+	  && CPU_FEATURES_CPU_P (cpu_features, BMI2)
+	  && !CPU_FEATURES_ARCH_P (cpu_features, Prefer_AVX2_STRCMP))
+	return OPTIMIZE (evex);
+
+      if (!CPU_FEATURES_ARCH_P (cpu_features, Prefer_No_VZEROUPPER))
+	return OPTIMIZE (avx2);
+    }
 
   if (CPU_FEATURES_CPU_P (cpu_features, SSE4_2)
       && !CPU_FEATURES_ARCH_P (cpu_features, Slow_SSE4_2))

sysdeps/x86_64/multiarch/strnlen-evex.S

@@ -0,0 +1,4 @@
+#define STRLEN __strnlen_evex
+#define USE_AS_STRNLEN 1
+
+#include "strlen-evex.S"

sysdeps/x86_64/multiarch/strrchr-evex.S

@@ -0,0 +1,265 @@
+/* strrchr/wcsrchr optimized with 256-bit EVEX instructions.
+   Copyright (C) 2021 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)
+
+# include <sysdep.h>
+
+# ifndef STRRCHR
+#  define STRRCHR	__strrchr_evex
+# endif
+
+# define VMOVU		vmovdqu64
+# define VMOVA		vmovdqa64
+
+# ifdef USE_AS_WCSRCHR
+#  define VPBROADCAST	vpbroadcastd
+#  define VPCMP		vpcmpd
+#  define SHIFT_REG	r8d
+# else
+#  define VPBROADCAST	vpbroadcastb
+#  define VPCMP		vpcmpb
+#  define SHIFT_REG	ecx
+# endif
+
+# define XMMZERO	xmm16
+# define YMMZERO	ymm16
+# define YMMMATCH	ymm17
+# define YMM1		ymm18
+
+# define VEC_SIZE	32
+
+	.section .text.evex,"ax",@progbits
+ENTRY (STRRCHR)
+	movl	%edi, %ecx
+	/* Broadcast CHAR to YMMMATCH.  */
+	VPBROADCAST %esi, %YMMMATCH
+
+	vpxorq	%XMMZERO, %XMMZERO, %XMMZERO
+
+	/* Check if we may cross page boundary with one vector load.  */
+	andl	$(2 * VEC_SIZE - 1), %ecx
+	cmpl	$VEC_SIZE, %ecx
+	ja	L(cros_page_boundary)
+
+	VMOVU	(%rdi), %YMM1
+
+	/* Each bit in K0 represents a null byte in YMM1.  */
+	VPCMP	$0, %YMMZERO, %YMM1, %k0
+	/* Each bit in K1 represents a CHAR in YMM1.  */
+	VPCMP	$0, %YMMMATCH, %YMM1, %k1
+	kmovd	%k0, %ecx
+	kmovd	%k1, %eax
+
+	addq	$VEC_SIZE, %rdi
+
+	testl	%eax, %eax
+	jnz	L(first_vec)
+
+	testl	%ecx, %ecx
+	jnz	L(return_null)
+
+	andq	$-VEC_SIZE, %rdi
+	xorl	%edx, %edx
+	jmp	L(aligned_loop)
+
+	.p2align 4
+L(first_vec):
+	/* Check if there is a null byte.  */
+	testl	%ecx, %ecx
+	jnz	L(char_and_nul_in_first_vec)
+
+	/* Remember the match and keep searching.  */
+	movl	%eax, %edx
+	movq	%rdi, %rsi
+	andq	$-VEC_SIZE, %rdi
+	jmp	L(aligned_loop)
+
+	.p2align 4
+L(cros_page_boundary):
+	andl	$(VEC_SIZE - 1), %ecx
+	andq	$-VEC_SIZE, %rdi
+
+# ifdef USE_AS_WCSRCHR
+	/* NB: Divide shift count by 4 since each bit in K1 represent 4
+	   bytes.  */
+	movl	%ecx, %SHIFT_REG
+	sarl	$2, %SHIFT_REG
+# endif
+
+	VMOVA	(%rdi), %YMM1
+
+	/* Each bit in K0 represents a null byte in YMM1.  */
+	VPCMP	$0, %YMMZERO, %YMM1, %k0
+	/* Each bit in K1 represents a CHAR in YMM1.  */
+	VPCMP	$0, %YMMMATCH, %YMM1, %k1
+	kmovd	%k0, %edx
+	kmovd	%k1, %eax
+
+	shrxl	%SHIFT_REG, %edx, %edx
+	shrxl	%SHIFT_REG, %eax, %eax
+	addq	$VEC_SIZE, %rdi
+
+	/* Check if there is a CHAR.  */
+	testl	%eax, %eax
+	jnz	L(found_char)
+
+	testl	%edx, %edx
+	jnz	L(return_null)
+
+	jmp	L(aligned_loop)
+
+	.p2align 4
+L(found_char):
+	testl	%edx, %edx
+	jnz	L(char_and_nul)
+
+	/* Remember the match and keep searching.  */
+	movl	%eax, %edx
+	leaq	(%rdi, %rcx), %rsi
+
+	.p2align 4
+L(aligned_loop):
+	VMOVA	(%rdi), %YMM1
+	addq	$VEC_SIZE, %rdi
+
+	/* Each bit in K0 represents a null byte in YMM1.  */
+	VPCMP	$0, %YMMZERO, %YMM1, %k0
+	/* Each bit in K1 represents a CHAR in YMM1.  */
+	VPCMP	$0, %YMMMATCH, %YMM1, %k1
+	kmovd	%k0, %ecx
+	kmovd	%k1, %eax
+	orl	%eax, %ecx
+	jnz	L(char_nor_null)
+
+	VMOVA	(%rdi), %YMM1
+	add	$VEC_SIZE, %rdi
+
+	/* Each bit in K0 represents a null byte in YMM1.  */
+	VPCMP	$0, %YMMZERO, %YMM1, %k0
+	/* Each bit in K1 represents a CHAR in YMM1.  */
+	VPCMP	$0, %YMMMATCH, %YMM1, %k1
+	kmovd	%k0, %ecx
+	kmovd	%k1, %eax
+	orl	%eax, %ecx
+	jnz	L(char_nor_null)
+
+	VMOVA	(%rdi), %YMM1
+	addq	$VEC_SIZE, %rdi
+
+	/* Each bit in K0 represents a null byte in YMM1.  */
+	VPCMP	$0, %YMMZERO, %YMM1, %k0
+	/* Each bit in K1 represents a CHAR in YMM1.  */
+	VPCMP	$0, %YMMMATCH, %YMM1, %k1
+	kmovd	%k0, %ecx
+	kmovd	%k1, %eax
+	orl	%eax, %ecx
+	jnz	L(char_nor_null)
+
+	VMOVA	(%rdi), %YMM1
+	addq	$VEC_SIZE, %rdi
+
+	/* Each bit in K0 represents a null byte in YMM1.  */
+	VPCMP	$0, %YMMZERO, %YMM1, %k0
+	/* Each bit in K1 represents a CHAR in YMM1.  */
+	VPCMP	$0, %YMMMATCH, %YMM1, %k1
+	kmovd	%k0, %ecx
+	kmovd	%k1, %eax
+	orl	%eax, %ecx
+	jz	L(aligned_loop)
+
+	.p2align 4
+L(char_nor_null):
+	/* Find a CHAR or a null byte in a loop.  */
+	testl	%eax, %eax
+	jnz	L(match)
+L(return_value):
+	testl	%edx, %edx
+	jz	L(return_null)
+	movl	%edx, %eax
+	movq	%rsi, %rdi
+	bsrl	%eax, %eax
+# ifdef USE_AS_WCSRCHR
+	/* NB: Multiply wchar_t count by 4 to get the number of bytes.  */
+	leaq	-VEC_SIZE(%rdi, %rax, 4), %rax
+# else
+	leaq	-VEC_SIZE(%rdi, %rax), %rax
+# endif
+	ret
+
+	.p2align 4
+L(match):
+	/* Find a CHAR.  Check if there is a null byte.  */
+	kmovd	%k0, %ecx
+	testl	%ecx, %ecx
+	jnz	L(find_nul)
+
+	/* Remember the match and keep searching.  */
+	movl	%eax, %edx
+	movq	%rdi, %rsi
+	jmp	L(aligned_loop)
+
+	.p2align 4
+L(find_nul):
+	/* Mask out any matching bits after the null byte.  */
+	movl	%ecx, %r8d
+	subl	$1, %r8d
+	xorl	%ecx, %r8d
+	andl	%r8d, %eax
+	testl	%eax, %eax
+	/* If there is no CHAR here, return the remembered one.  */
+	jz	L(return_value)
+	bsrl	%eax, %eax
+# ifdef USE_AS_WCSRCHR
+	/* NB: Multiply wchar_t count by 4 to get the number of bytes.  */
+	leaq	-VEC_SIZE(%rdi, %rax, 4), %rax
+# else
+	leaq	-VEC_SIZE(%rdi, %rax), %rax
+# endif
+	ret
+
+	.p2align 4
+L(char_and_nul):
+	/* Find both a CHAR and a null byte.  */
+	addq	%rcx, %rdi
+	movl	%edx, %ecx
+L(char_and_nul_in_first_vec):
+	/* Mask out any matching bits after the null byte.  */
+	movl	%ecx, %r8d
+	subl	$1, %r8d
+	xorl	%ecx, %r8d
+	andl	%r8d, %eax
+	testl	%eax, %eax
+	/* Return null pointer if the null byte comes first.  */
+	jz	L(return_null)
+	bsrl	%eax, %eax
+# ifdef USE_AS_WCSRCHR
+	/* NB: Multiply wchar_t count by 4 to get the number of bytes.  */
+	leaq	-VEC_SIZE(%rdi, %rax, 4), %rax
+# else
+	leaq	-VEC_SIZE(%rdi, %rax), %rax
+# endif
+	ret
+
+	.p2align 4
+L(return_null):
+	xorl	%eax, %eax
+	ret
+
+END (STRRCHR)
+#endif

sysdeps/x86_64/multiarch/wcschr-evex.S

@@ -0,0 +1,3 @@
+#define STRCHR __wcschr_evex
+#define USE_AS_WCSCHR 1
+#include "strchr-evex.S"

sysdeps/x86_64/multiarch/wcscmp-evex.S

@@ -0,0 +1,4 @@
+#define STRCMP __wcscmp_evex
+#define USE_AS_WCSCMP 1
+
+#include "strcmp-evex.S"

sysdeps/x86_64/multiarch/wcslen-evex.S

@@ -0,0 +1,4 @@
+#define STRLEN __wcslen_evex
+#define USE_AS_WCSLEN 1
+
+#include "strlen-evex.S"

sysdeps/x86_64/multiarch/wcsncmp-evex.S

@@ -0,0 +1,5 @@
+#define STRCMP __wcsncmp_evex
+#define USE_AS_STRNCMP 1
+#define USE_AS_WCSCMP 1
+
+#include "strcmp-evex.S"

sysdeps/x86_64/multiarch/wcsnlen-evex.S

@@ -0,0 +1,5 @@
+#define STRLEN __wcsnlen_evex
+#define USE_AS_WCSLEN 1
+#define USE_AS_STRNLEN 1
+
+#include "strlen-evex.S"

sysdeps/x86_64/multiarch/wcsnlen.c

@@ -29,16 +29,24 @@
 extern __typeof (REDIRECT_NAME) OPTIMIZE (sse2) attribute_hidden;
 extern __typeof (REDIRECT_NAME) OPTIMIZE (sse4_1) attribute_hidden;
 extern __typeof (REDIRECT_NAME) OPTIMIZE (avx2) attribute_hidden;
+extern __typeof (REDIRECT_NAME) OPTIMIZE (evex) attribute_hidden;
 
 static inline void *
 IFUNC_SELECTOR (void)
 {
   const struct cpu_features* cpu_features = __get_cpu_features ();
 
-  if (!CPU_FEATURES_ARCH_P (cpu_features, Prefer_No_VZEROUPPER)
+  if (CPU_FEATURES_ARCH_P (cpu_features, AVX2_Usable)
-      && CPU_FEATURES_ARCH_P (cpu_features, AVX2_Usable)
       && CPU_FEATURES_ARCH_P (cpu_features, AVX_Fast_Unaligned_Load))
-    return OPTIMIZE (avx2);
+    {
+      if (CPU_FEATURES_ARCH_P (cpu_features, AVX512VL_Usable)
+	  && CPU_FEATURES_ARCH_P (cpu_features, AVX512BW_Usable)
+	  && CPU_FEATURES_CPU_P (cpu_features, BMI2))
+	return OPTIMIZE (evex);
+
+      if (!CPU_FEATURES_ARCH_P (cpu_features, Prefer_No_VZEROUPPER))
+	return OPTIMIZE (avx2);
+    }
 
   if (CPU_FEATURES_CPU_P (cpu_features, SSE4_1))
     return OPTIMIZE (sse4_1);

sysdeps/x86_64/multiarch/wcsrchr-evex.S

@@ -0,0 +1,3 @@
+#define STRRCHR __wcsrchr_evex
+#define USE_AS_WCSRCHR 1
+#include "strrchr-evex.S"

sysdeps/x86_64/multiarch/wmemchr-evex.S

@@ -0,0 +1,4 @@
+#define MEMCHR __wmemchr_evex
+#define USE_AS_WMEMCHR 1
+
+#include "memchr-evex.S"