On a lightly loaded 24Kc, as measured by 'openssl speed sha1', shows between
27% and 120% speedup depending on block size. SHA1 is notably used in
Transmission for piece verification.
>>> no-perlasm <<<
The 'numbers' are in 1000s of bytes per second processed.
type 16 bytes 64 bytes 256 bytes 1024 bytes 8192 bytes
sha1 1421.10k 4019.38k 8498.85k 11452.85k 13073.53k
>>> mips32_asm <<<
The 'numbers' are in 1000s of bytes per second processed.
type 16 bytes 64 bytes 256 bytes 1024 bytes 8192 bytes
sha1 1814.05k 6046.81k 14637.55k 23456.13k 28825.77k
relative:
27% 50% 72% 104% 120%
---
I haven't tested this too thoroughly, so at this point I'm just looking for
people willing to try this out and for feedback in general. If it doesn't cause
too much grief, I can clean it up to a point where it's merge-able.
package/openssl/Makefile | 2 +-
.../openssl/patches/110-optimize-for-size.patch | 2 +-
package/openssl/patches/210-fix_aes_fips.patch | 64 ++++++++++++++++++++
3 files changed, 66 insertions(+), 2 deletions(-)
create mode 100644 package/openssl/patches/210-fix_aes_fips.patch
diff --git a/package/openssl/Makefile b/package/openssl/Makefile
index d95783a..4dfcdf6 100644
--- a/package/openssl/Makefile
+++ b/package/openssl/Makefile
@@ -89,7 +89,7 @@ ifeq ($(CONFIG_x86_64),y)
OPENSSL_TARGET:=linux-x86_64
else
OPENSSL_TARGET:=linux-openwrt
- OPENSSL_OPTIONS+=no-perlasm no-sse2
+ OPENSSL_OPTIONS+=no-sse2
endif
define Build/Configure
diff --git a/package/openssl/patches/110-optimize-for-size.patch
b/package/openssl/patches/110-optimize-for-size.patch
index c0160a3..8895f4f 100644
--- a/package/openssl/patches/110-optimize-for-size.patch
+++ b/package/openssl/patches/110-optimize-for-size.patch
@@ -5,7 +5,7 @@
"linux-alpha-ccc","ccc:-fast -readonly_strings -DL_ENDIAN
-DTERMIO::-D_REENTRANT:::SIXTY_FOUR_BIT_LONG RC4_CHUNK DES_INT DES_PTR
DES_RISC1 DES_UNROLL:${alpha_asm}",
"linux-alpha+bwx-ccc","ccc:-fast -readonly_strings -DL_ENDIAN
-DTERMIO::-D_REENTRANT:::SIXTY_FOUR_BIT_LONG RC4_CHAR RC4_CHUNK DES_INT DES_PTR
DES_RISC1 DES_UNROLL:${alpha_asm}",
+# OpenWrt targets
-+"linux-openwrt","gcc:-DTERMIO \$(OPENWRT_OPTIMIZATION_FLAGS)
-fomit-frame-pointer -Wall::-D_REENTRANT::-ldl:BN_LLONG RC4_CHAR RC4_CHUNK
DES_INT DES_UNROLL
BF_PTR:${no_asm}:dlfcn:linux-shared:-fPIC::.so.\$(SHLIB_MAJOR).\$(SHLIB_MINOR)",
++"linux-openwrt","gcc:-DTERMIO \$(OPENWRT_OPTIMIZATION_FLAGS)
-fomit-frame-pointer -Wall::-D_REENTRANT::-ldl:BN_LLONG RC4_CHAR RC4_CHUNK
DES_INT DES_UNROLL
BF_PTR:${mips32_asm}:o32:dlfcn:linux-shared:-fPIC::.so.\$(SHLIB_MAJOR).\$(SHLIB_MINOR)",
# Android: linux-* but without -DTERMIO and pointers to headers and libs.
"android","gcc:-mandroid -I\$(ANDROID_DEV)/include -B\$(ANDROID_DEV)/lib -O3
-fomit-frame-pointer -Wall::-D_REENTRANT::-ldl:BN_LLONG RC4_CHAR RC4_CHUNK
DES_INT DES_UNROLL
BF_PTR:${no_asm}:dlfcn:linux-shared:-fPIC::.so.\$(SHLIB_MAJOR).\$(SHLIB_MINOR)",
diff --git a/package/openssl/patches/210-fix_aes_fips.patch
b/package/openssl/patches/210-fix_aes_fips.patch
new file mode 100644
index 0000000..bbfc63a
--- /dev/null
+++ b/package/openssl/patches/210-fix_aes_fips.patch
@@ -0,0 +1,64 @@
+--- a/crypto/aes/asm/aes-mips.pl
++++ b/crypto/aes/asm/aes-mips.pl
+@@ -1036,9 +1036,9 @@ _mips_AES_set_encrypt_key:
+ nop
+ .end _mips_AES_set_encrypt_key
+
+-.globl AES_set_encrypt_key
+-.ent AES_set_encrypt_key
+-AES_set_encrypt_key:
++.globl private_AES_set_encrypt_key
++.ent private_AES_set_encrypt_key
++private_AES_set_encrypt_key:
+ .frame $sp,$FRAMESIZE,$ra
+ .mask $SAVED_REGS_MASK,-$SZREG
+ .set noreorder
+@@ -1060,7 +1060,7 @@ $code.=<<___ if ($flavour =~ /nubi/i); #
+ ___
+ $code.=<<___ if ($flavour !~ /o32/i); # non-o32 PIC-ification
+ .cplocal $Tbl
+- .cpsetup $pf,$zero,AES_set_encrypt_key
++ .cpsetup $pf,$zero,private_AES_set_encrypt_key
+ ___
+ $code.=<<___;
+ .set reorder
+@@ -1083,7 +1083,7 @@ ___
+ $code.=<<___;
+ jr $ra
+ $PTR_ADD $sp,$FRAMESIZE
+-.end AES_set_encrypt_key
++.end private_AES_set_encrypt_key
+ ___
+
+ my ($head,$tail)=($inp,$bits);
+@@ -1091,9 +1091,9 @@ my ($tp1,$tp2,$tp4,$tp8,$tp9,$tpb,$tpd,$
+ my ($m,$x80808080,$x7f7f7f7f,$x1b1b1b1b)=($at,$t0,$t1,$t2);
+ $code.=<<___;
+ .align 5
+-.globl AES_set_decrypt_key
+-.ent AES_set_decrypt_key
+-AES_set_decrypt_key:
++.globl private_AES_set_decrypt_key
++.ent private_AES_set_decrypt_key
++private_AES_set_decrypt_key:
+ .frame $sp,$FRAMESIZE,$ra
+ .mask $SAVED_REGS_MASK,-$SZREG
+ .set noreorder
+@@ -1115,7 +1115,7 @@ $code.=<<___ if ($flavour =~ /nubi/i); #
+ ___
+ $code.=<<___ if ($flavour !~ /o32/i); # non-o32 PIC-ification
+ .cplocal $Tbl
+- .cpsetup $pf,$zero,AES_set_decrypt_key
++ .cpsetup $pf,$zero,private_AES_set_decrypt_key
+ ___
+ $code.=<<___;
+ .set reorder
+@@ -1226,7 +1226,7 @@ ___
+ $code.=<<___;
+ jr $ra
+ $PTR_ADD $sp,$FRAMESIZE
+-.end AES_set_decrypt_key
++.end private_AES_set_decrypt_key
+ ___
+ }}}
+
--
1.7.7.3
_______________________________________________
openwrt-devel mailing list
openwrt-devel@lists.openwrt.org
https://lists.openwrt.org/mailman/listinfo/openwrt-devel
