svn commit: r320554 - in head: lib/libmd sys/modules/crypto

[Allan Jude]

Author: allanjude
Date: Sat Jul  1 21:18:06 2017
New Revision: 320554
URL: https://svnweb.freebsd.org/changeset/base/320554

Log:
  Increase loop unrolling for skein hashes
  
  This patch was inspired by an opposite change made to shrink the code
  for the boot loader.
  
  On my i7-4770, it increases the skein1024 speed from 470 to 550 MB/s
  
  Reviewed by:	sbruno
  MFC after:	1 month
  Sponsored by:	ScaleEngine Inc.
  Differential Revision:	https://reviews.freebsd.org/D7824

Modified:
  head/lib/libmd/Makefile
  head/sys/modules/crypto/Makefile

Modified: head/lib/libmd/Makefile
==============================================================================
--- head/lib/libmd/Makefile	Sat Jul  1 20:25:22 2017	(r320553)
+++ head/lib/libmd/Makefile	Sat Jul  1 21:18:06 2017	(r320554)
@@ -88,6 +88,8 @@ sys/md5.h: ${SRCTOP}/sys/${.TARGET} .NOMETA
 CFLAGS+= -I${.CURDIR} -I${SRCTOP}/sys/crypto/sha2
 CFLAGS+= -I${SRCTOP}/sys/crypto/skein
 CFLAGS+= -DWEAK_REFS
+# unroll the 256 and 512 loops, half unroll the 1024
+CFLAGS+= -DSKEIN_LOOP=995
 .PATH: ${.CURDIR}/${MACHINE_ARCH} ${SRCTOP}/sys/crypto/sha2
 .PATH: ${SRCTOP}/sys/crypto/skein ${SRCTOP}/sys/crypto/skein/${MACHINE_ARCH}
 
@@ -101,6 +103,8 @@ CFLAGS+= -DRMD160_ASM
 .endif
 .if exists(${MACHINE_ARCH}/skein_block_asm.s)
 AFLAGS += --strip-local-absolute
+# Fully unroll all loops in the assembly optimized version
+AFLAGS+= --defsym SKEIN_LOOP=0
 SRCS+= skein_block_asm.s
 CFLAGS+= -DSKEIN_ASM -DSKEIN_USE_ASM=1792 # list of block functions to replace with assembly: 256+512+1024 = 1792
 .endif

Modified: head/sys/modules/crypto/Makefile
==============================================================================
--- head/sys/modules/crypto/Makefile	Sat Jul  1 20:25:22 2017	(r320553)
+++ head/sys/modules/crypto/Makefile	Sat Jul  1 21:18:06 2017	(r320554)
@@ -19,11 +19,15 @@ SRCS	+= camellia.c camellia-api.c
 SRCS	+= des_ecb.c des_enc.c des_setkey.c
 SRCS	+= sha1.c sha256c.c sha512c.c
 SRCS	+= skein.c skein_block.c
+# unroll the 256 and 512 loops, half unroll the 1024
+CFLAGS+= -DSKEIN_LOOP=995
 .if exists(${MACHINE_ARCH}/skein_block_asm.s)
 .PATH:	${SRCTOP}/sys/crypto/skein/${MACHINE_ARCH}
 SRCS	+= skein_block_asm.s
 CFLAGS	+= -DSKEIN_ASM -DSKEIN_USE_ASM=1792 # list of block functions to replace with assembly: 256+512+1024 = 1792
 ACFLAGS	+= -DELF -Wa,--noexecstack
+# Fully unroll all loops in the assembly optimized version
+AFLAGS+= --defsym SKEIN_LOOP=0
 .endif
 SRCS	+= siphash.c
 SRCS	+= gmac.c gfmult.c