Fix the error handling in ERR_get_state:

[oweals/openssl.git] / crypto / bn / asm / mips-mont.pl

diff --git a/crypto/bn/asm/mips-mont.pl b/crypto/bn/asm/mips-mont.pl
index e2395f4b5ddab09599717eb21f41f3a66454b5b1..a907571bec3f48af164c1401db8f4cc170c3b93b 100644 (file)
--- a/crypto/bn/asm/mips-mont.pl
+++ b/crypto/bn/asm/mips-mont.pl
@@ -1,19 +1,30 @@
-#!/usr/bin/env perl
+#! /usr/bin/env perl
+# Copyright 2010-2016 The OpenSSL Project Authors. All Rights Reserved.
+#
+# Licensed under the OpenSSL license (the "License").  You may not use
+# this file except in compliance with the License.  You can obtain a copy
+# in the file LICENSE in the source distribution or at
+# https://www.openssl.org/source/license.html
+
 #
 # ====================================================================
-# Written by Andy Polyakov <appro@fy.chalmers.se> for the OpenSSL
+# Written by Andy Polyakov <appro@openssl.org> for the OpenSSL
 # project. The module is, however, dual licensed under OpenSSL and
 # CRYPTOGAMS licenses depending on where you obtain it. For further
 # details see http://www.openssl.org/~appro/cryptogams/.
 # ====================================================================
 
 # This module doesn't present direct interest for OpenSSL, because it
-# doesn't provide better performance for longer keys. While 512-bit
-# RSA private key operations are 40% faster, 1024-bit ones are hardly
-# faster at all, while longer key operations are slower by up to 20%.
-# It might be of interest to embedded system developers though, as
-# it's smaller than 1KB, yet offers ~3x improvement over compiler
-# generated code.
+# doesn't provide better performance for longer keys, at least not on
+# in-order-execution cores. While 512-bit RSA sign operations can be
+# 65% faster in 64-bit mode, 1024-bit ones are only 15% faster, and
+# 4096-bit ones are up to 15% slower. In 32-bit mode it varies from
+# 16% improvement for 512-bit RSA sign to -33% for 4096-bit RSA
+# verify:-( All comparisons are against bn_mul_mont-free assembler.
+# The module might be of interest to embedded system developers, as
+# the code is smaller than 1KB, yet offers >3x improvement on MIPS64
+# and 75-30% [less for longer keys] on MIPS32 over compiler-generated
+# code.
 
 ######################################################################
 # There is a number of MIPS ABI in use, O32 and N32/64 are most
@@ -42,7 +53,7 @@
 # ($s0,$s1,$s2,$s3,$s4,$s5,$s6,$s7)=map("\$$_",(16..23));
 # ($gp,$sp,$fp,$ra)=map("\$$_",(28..31));
 #
-$flavour = shift; # supported flavours are o32,n32,64,nubi32,nubi64
+$flavour = shift || "o32"; # supported flavours are o32,n32,64,nubi32,nubi64
 
 if ($flavour =~ /64|n32/i) {
        $PTR_ADD="dadd";        # incidentally works even on n32
@@ -63,7 +74,7 @@ $SAVED_REGS_MASK = ($flavour =~ /nubi/i) ? 0x00fff000 : 0x00ff0000;
 #
 ######################################################################
 
-while (($output=shift) && ($output!~/^\w[\w\-]*\.\w+$/)) {}
+while (($output=shift) && ($output!~/\w[\w\-]*\.\w+$/)) {}
 open STDOUT,">$output";
 
 if ($flavour =~ /64|n32/i) {
@@ -126,9 +137,12 @@ $code.=<<___ if ($flavour =~ /o32/i);
 ___
 $code.=<<___;
        slt     $at,$num,4
-       beqzl   $at,bn_mul_mont_internal
+       bnez    $at,1f
        li      $t0,0
-       jr      $ra
+       slt     $at,$num,17     # on in-order CPU
+       bnez    $at,bn_mul_mont_internal
+       nop
+1:     jr      $ra
        li      $a0,0
 .end   bn_mul_mont