diff options
Diffstat (limited to 'drivers/builtin_openssl2/crypto/modes/asm')
| -rw-r--r-- | drivers/builtin_openssl2/crypto/modes/asm/ghash-armv4.pl | 6 | ||||
| -rw-r--r-- | drivers/builtin_openssl2/crypto/modes/asm/ghash-x86.pl | 2 |
2 files changed, 4 insertions, 4 deletions
diff --git a/drivers/builtin_openssl2/crypto/modes/asm/ghash-armv4.pl b/drivers/builtin_openssl2/crypto/modes/asm/ghash-armv4.pl index d91586ee29..e46f8e34da 100644 --- a/drivers/builtin_openssl2/crypto/modes/asm/ghash-armv4.pl +++ b/drivers/builtin_openssl2/crypto/modes/asm/ghash-armv4.pl @@ -374,8 +374,8 @@ gcm_ghash_neon: vdup.8 $xi,`&Dlo("$IN")`[0] @ broadcast lowest byte .Linner_neon: subs $cnt,$cnt,#1 - vmull.p8 $Qlo,$Hlo,$xi @ H.lo·Xi[i] - vmull.p8 $Qhi,$Hhi,$xi @ H.hi·Xi[i] + vmull.p8 $Qlo,$Hlo,$xi @ H.lo·Xi[i] + vmull.p8 $Qhi,$Hhi,$xi @ H.hi·Xi[i] vext.8 $IN,$zero,#1 @ IN>>=8 veor $Z,$Qpost @ modulo-scheduled part @@ -388,7 +388,7 @@ gcm_ghash_neon: vsli.8 $Zo,$T,#1 @ compose the "carry" byte vext.8 $Z,$zero,#1 @ Z>>=8 - vmull.p8 $R,$Zo,$mod @ "carry"·0xe1 + vmull.p8 $R,$Zo,$mod @ "carry"·0xe1 vshr.u8 $Zo,$T,#7 @ save Z's bottom bit vext.8 $Qpost,$Qlo,$zero,#1 @ Qlo>>=8 veor $Z,$Qhi diff --git a/drivers/builtin_openssl2/crypto/modes/asm/ghash-x86.pl b/drivers/builtin_openssl2/crypto/modes/asm/ghash-x86.pl index 83c727e07f..2426cd0c8a 100644 --- a/drivers/builtin_openssl2/crypto/modes/asm/ghash-x86.pl +++ b/drivers/builtin_openssl2/crypto/modes/asm/ghash-x86.pl @@ -346,7 +346,7 @@ $S=12; # shift factor for rem_4bit # effective address calculation and finally merge of value to Z.hi. # Reference to rem_4bit is scheduled so late that I had to >>4 # rem_4bit elements. This resulted in 20-45% procent improvement -# on contemporary µ-archs. +# on contemporary µ-archs. { my $cnt; my $rem_4bit = "eax"; |