diff options
| author | Andy Polyakov <appro@openssl.org> | 2007-09-13 07:27:10 +0000 |
|---|---|---|
| committer | Andy Polyakov <appro@openssl.org> | 2007-09-13 07:27:10 +0000 |
| commit | 399f94bfb45c993d7e08d604532d1dc72dc944a4 (patch) | |
| tree | 6367064780fabe04693545dc0f4d9e66297a0fdd /crypto/sha | |
| parent | 1fa29843fa2f23d01b7896b667e44fd60ed6d9ce (diff) | |
Commentary updates.
Diffstat (limited to 'crypto/sha')
| -rwxr-xr-x | crypto/sha/asm/sha1-ppc.pl | 1 | ||||
| -rwxr-xr-x | crypto/sha/asm/sha1-x86_64.pl | 2 | ||||
| -rw-r--r-- | crypto/sha/asm/sha256-armv4.pl | 8 | ||||
| -rwxr-xr-x | crypto/sha/asm/sha512-ppc.pl | 1 | ||||
| -rw-r--r-- | crypto/sha/asm/sha512-s390x.pl | 2 |
5 files changed, 9 insertions, 5 deletions
diff --git a/crypto/sha/asm/sha1-ppc.pl b/crypto/sha/asm/sha1-ppc.pl index fbd051652c..2c84d59149 100755 --- a/crypto/sha/asm/sha1-ppc.pl +++ b/crypto/sha/asm/sha1-ppc.pl @@ -18,6 +18,7 @@ # -m64 -m32 # ---------------------------------- # PPC970,gcc-4.0.0 +76% +59% +# Power6,xlc-7 +68% +33% $output = shift; diff --git a/crypto/sha/asm/sha1-x86_64.pl b/crypto/sha/asm/sha1-x86_64.pl index 93a4f01796..f7ed67a726 100755 --- a/crypto/sha/asm/sha1-x86_64.pl +++ b/crypto/sha/asm/sha1-x86_64.pl @@ -27,7 +27,7 @@ # gcc 3.4 32-bit asm cycles/byte # Opteron +45% +20% 6.8 # Xeon P4 +65% +0% 9.9 -# Core2 +60% +10% 8.8 +# Core2 +60% +10% 7.0 $output=shift; diff --git a/crypto/sha/asm/sha256-armv4.pl b/crypto/sha/asm/sha256-armv4.pl index d2465e3aa8..4dd09619a6 100644 --- a/crypto/sha/asm/sha256-armv4.pl +++ b/crypto/sha/asm/sha256-armv4.pl @@ -7,9 +7,11 @@ # details see http://www.openssl.org/~appro/cryptogams/. # ==================================================================== -# SHA256 block procedure for ARMv4. +# SHA256 block procedure for ARMv4. May 2007. -# Performance is ~2x better than gcc 3.4 generated code. +# Performance is ~2x better than gcc 3.4 generated code and in "abso- +# lute" terms is ~2250 cycles per 64-byte block or ~35 cycles per +# byte. $ctx="r0"; $t0="r0"; $inp="r1"; @@ -165,7 +167,7 @@ $code.=<<___; tst lr,#1 moveq pc,lr @ be binary compatible with V4, yet bx lr @ interoperable with Thumb ISA:-) -.size sha256_block_data_order,.-sha1_block_data_order +.size sha256_block_data_order,.-sha256_block_data_order .asciz "SHA256 block transform for ARMv4, CRYPTOGAMS by <appro\@openssl.org>" ___ diff --git a/crypto/sha/asm/sha512-ppc.pl b/crypto/sha/asm/sha512-ppc.pl index 087faab866..0ec4b0bcc3 100755 --- a/crypto/sha/asm/sha512-ppc.pl +++ b/crypto/sha/asm/sha512-ppc.pl @@ -16,6 +16,7 @@ # -m64 -m32 | -m64 -m32 # --------------------------------------+----------------------- # PPC970,gcc-4.0.0 +50% +38% | +40% +410%(*) +# Power6,xlc-7 +150% +90% | +100% +430%(*) # # (*) 64-bit code in 32-bit application context, which actually is # on TODO list. It should be noted that for safe deployment in diff --git a/crypto/sha/asm/sha512-s390x.pl b/crypto/sha/asm/sha512-s390x.pl index 5dd17473fa..d2aceec68c 100644 --- a/crypto/sha/asm/sha512-s390x.pl +++ b/crypto/sha/asm/sha512-s390x.pl @@ -12,7 +12,7 @@ # April 2007. # # sha256_block_data_order is reportedly >3 times faster than gcc 3.3 -# generated code (must to be a bug in compiler, as improvement is +# generated code (must be a bug in compiler, as improvement is # "pathologically" high, in particular in comparison to other SHA # modules). But the real twist is that it detects if hardware support # for SHA256 is available and in such case utilizes it. Then the |