diff options
| author | Andy Polyakov <appro@openssl.org> | 2011-09-15 20:22:59 +0000 |
|---|---|---|
| committer | Andy Polyakov <appro@openssl.org> | 2011-09-15 20:22:59 +0000 |
| commit | 8ca28da0a798d2289b71b750dad191066600c166 (patch) | |
| tree | f99b2ae7ab9bbee9c5f98cc928d0a799d969eac5 /crypto/aes/asm/vpaes-x86.pl | |
| parent | b889a6046be029cb9874a78ab262471b184b5467 (diff) | |
Integrate Vector Permutation AES into build system.
Diffstat (limited to 'crypto/aes/asm/vpaes-x86.pl')
| -rw-r--r-- | crypto/aes/asm/vpaes-x86.pl | 7 |
1 files changed, 4 insertions, 3 deletions
diff --git a/crypto/aes/asm/vpaes-x86.pl b/crypto/aes/asm/vpaes-x86.pl index efe68dff41..f2414ccd23 100644 --- a/crypto/aes/asm/vpaes-x86.pl +++ b/crypto/aes/asm/vpaes-x86.pl @@ -21,7 +21,7 @@ # about its alignment... # # Performance summary. aes-586.pl column lists large-block CBC -# encrypt/decrypt/with-hypert-hreading-off(*) results in cycles per +# encrypt/decrypt/with-hyper-threading-off(*) results in cycles per # byte processed with 128-bit key, and vpaes-x86.pl column - # encrypt/decrypt. # @@ -41,7 +41,8 @@ # # (***) Less impressive improvement on Core 2 and Atom is due to slow # pshufb, yet it's respectable +32%/65% improvement on Core 2 -# and +58%/40% on Atom. +# and +58%/40% on Atom (as implied, over "hyper-threading-safe" +# code path). # # <appro@openssl.org> @@ -51,7 +52,7 @@ require "x86asm.pl"; &asm_init($ARGV[0],"vpaes-x86.pl",$x86only = $ARGV[$#ARGV] eq "386"); -$PREFIX="AES"; +$PREFIX="vpaes"; my ($round, $base, $magic, $key, $const, $inp, $out)= ("eax", "ebx", "ecx", "edx","ebp", "esi","edi"); |