This is the parity generation/rebuild using 128-bits NEON for Aarch64.

cmd/raidz_test/raidz_test.h

@@ -34,6 +34,8 @@ static const char *raidz_impl_names[] = {
 	"sse2",
 	"ssse3",
 	"avx2",
+	"aarch64_neon",
+	"aarch64_neonx2",
 	NULL
 };
 

include/linux/Makefile.am

@@ -8,6 +8,7 @@ KERNEL_H = \
 	$(top_srcdir)/include/linux/utsname_compat.h \
 	$(top_srcdir)/include/linux/kmap_compat.h \
 	$(top_srcdir)/include/linux/simd_x86.h \
+	$(top_srcdir)/include/linux/simd_aarch64.h \
 	$(top_srcdir)/include/linux/mod_compat.h
 
 USER_H =

include/linux/simd_aarch64.h

@@ -0,0 +1,62 @@
+/*
+ * CDDL HEADER START
+ *
+ * The contents of this file are subject to the terms of the
+ * Common Development and Distribution License (the "License").
+ * You may not use this file except in compliance with the License.
+ *
+ * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
+ * or http://www.opensolaris.org/os/licensing.
+ * See the License for the specific language governing permissions
+ * and limitations under the License.
+ *
+ * When distributing Covered Code, include this CDDL HEADER in each
+ * file and include the License file at usr/src/OPENSOLARIS.LICENSE.
+ * If applicable, add the following below this CDDL HEADER, with the
+ * fields enclosed by brackets "[]" replaced with your own identifying
+ * information: Portions Copyright [yyyy] [name of copyright owner]
+ *
+ * CDDL HEADER END
+ */
+/*
+ * Copyright (C) 2016 Romain Dolbeau <[email protected]>.
+ */
+
+/*
+ * USER API:
+ *
+ * Kernel fpu methods:
+ * 	kfpu_begin()
+ * 	kfpu_end()
+ */
+
+#ifndef _SIMD_AARCH64_H
+#define	_SIMD_AARCH64_H
+
+#include <sys/isa_defs.h>
+
+#if defined(__aarch64__)
+
+#include <sys/types.h>
+
+#if defined(_KERNEL)
+#include <asm/neon.h>
+#define	kfpu_begin()		\
+{					\
+	kernel_neon_begin();		\
+}
+#define	kfpu_end()			\
+{					\
+	kernel_neon_end();		\
+}
+#else
+/*
+ * fpu dummy methods for userspace
+ */
+#define	kfpu_begin() 	do {} while (0)
+#define	kfpu_end() 		do {} while (0)
+#endif /* defined(_KERNEL) */
+
+#endif /* __aarch64__ */
+
+#endif /* _SIMD_AARCH64_H */

include/sys/vdev_raidz_impl.h

@@ -141,6 +141,10 @@ extern const raidz_impl_ops_t vdev_raidz_ssse3_impl;
 #if defined(__x86_64) && defined(HAVE_AVX2)	/* only x86_64 for now */
 extern const raidz_impl_ops_t vdev_raidz_avx2_impl;
 #endif
+#if defined(__aarch64__)
+extern const raidz_impl_ops_t vdev_raidz_aarch64_neon_impl;
+extern const raidz_impl_ops_t vdev_raidz_aarch64_neonx2_impl;
+#endif
 
 /*
  * Commonly used raidz_map helpers

lib/libzpool/Makefile.am

@@ -99,6 +99,8 @@ KERNEL_C = \
 	vdev_raidz_math_sse2.c \
 	vdev_raidz_math_ssse3.c \
 	vdev_raidz_math_avx2.c \
+	vdev_raidz_math_aarch64_neon.c \
+	vdev_raidz_math_aarch64_neonx2.c \
 	vdev_root.c \
 	zap.c \
 	zap_leaf.c \

man/man5/zfs-module-parameters.5

@@ -1763,6 +1763,8 @@ Possible options are:
   sse2     - implementation using SSE2 instruction set (64bit x86 only)
   ssse3    - implementation using SSSE3 instruction set (64bit x86 only)
   avx2     - implementation using AVX2 instruction set (64bit x86 only)
+  aarch64_neon - implementation using NEON (Aarch64/64 bit ARMv8 only)
+  aarch64_neonx2 - implementation using NEON with more unrolling (Aarch64/64 bit ARMv8 only)
 .sp
 Default value: \fBfastest\fR.
 .RE

module/zfs/Makefile.in

@@ -115,3 +115,6 @@ $(MODULE)-objs += dsl_userhold.o
 $(MODULE)-$(CONFIG_X86) += vdev_raidz_math_sse2.o
 $(MODULE)-$(CONFIG_X86) += vdev_raidz_math_ssse3.o
 $(MODULE)-$(CONFIG_X86) += vdev_raidz_math_avx2.o
+
+$(MODULE)-$(CONFIG_ARM64) += vdev_raidz_math_aarch64_neon.o
+$(MODULE)-$(CONFIG_ARM64) += vdev_raidz_math_aarch64_neonx2.o

module/zfs/vdev_raidz_math.c

@@ -55,7 +55,11 @@ const raidz_impl_ops_t *raidz_all_maths[] = {
 	&vdev_raidz_ssse3_impl,
 #endif
 #if defined(__x86_64) && defined(HAVE_AVX2)	/* only x86_64 for now */
-	&vdev_raidz_avx2_impl
+	&vdev_raidz_avx2_impl,
+#endif
+#if defined(__aarch64__)
+	&vdev_raidz_aarch64_neon_impl,
+	&vdev_raidz_aarch64_neonx2_impl,
 #endif
 };
 
@@ -275,11 +279,11 @@ raidz_math_kstat_headers(char *buf, size_t size)
 	off = snprintf(buf, size, "%-17s", "implementation");
 
 	for (i = 0; i < ARRAY_SIZE(raidz_gen_name); i++)
-		off += snprintf(buf + off, size - off, "%-12s",
+		off += snprintf(buf + off, size - off, "%-16s",
 		    raidz_gen_name[i]);
 
 	for (i = 0; i < ARRAY_SIZE(raidz_rec_name); i++)
-		off += snprintf(buf + off, size - off, "%-12s",
+		off += snprintf(buf + off, size - off, "%-16s",
 		    raidz_rec_name[i]);
 
 	(void) snprintf(buf + off, size - off, "\n");
@@ -302,12 +306,12 @@ raidz_math_kstat_data(char *buf, size_t size, void *data)
 
 		for (i = 0; i < ARRAY_SIZE(raidz_gen_name); i++) {
 			int id = fstat->gen[i];
-			off += snprintf(buf + off, size - off, "%-12s",
+			off += snprintf(buf + off, size - off, "%-16s",
 			    raidz_supp_impl[id]->name);
 		}
 		for (i = 0; i < ARRAY_SIZE(raidz_rec_name); i++) {
 			int id = fstat->rec[i];
-			off += snprintf(buf + off, size - off, "%-12s",
+			off += snprintf(buf + off, size - off, "%-16s",
 			    raidz_supp_impl[id]->name);
 		}
 	} else {
@@ -317,11 +321,11 @@ raidz_math_kstat_data(char *buf, size_t size, void *data)
 		    raidz_supp_impl[id]->name);
 
 		for (i = 0; i < ARRAY_SIZE(raidz_gen_name); i++)
-			off += snprintf(buf + off, size - off, "%-12llu",
+			off += snprintf(buf + off, size - off, "%-16llu",
 			    (u_longlong_t) cstat->gen[i]);
 
 		for (i = 0; i < ARRAY_SIZE(raidz_rec_name); i++)
-			off += snprintf(buf + off, size - off, "%-12llu",
+			off += snprintf(buf + off, size - off, "%-16llu",
 			    (u_longlong_t) cstat->rec[i]);
 	}