bn/asm/vis3-mont.pl: fix intermittent EC failures on SPARC T3.

BLKINIT optimization worked on T4, but for some reason appears "too
aggressive" for T3 triggering intermiitent EC failures. It's not clear
why only EC is affected...

Reviewed-by: Tim Hudson <tjh@openssl.org>

diff --git a/crypto/bn/asm/vis3-mont.pl b/crypto/bn/asm/vis3-mont.pl
index 8da52c638c09e8280d95d023dbdd6efe5af63533..1d8dc9e08ea20609b2d9ad5c5ee03ca6cad7f27b 100644 (file)
--- a/crypto/bn/asm/vis3-mont.pl
+++ b/crypto/bn/asm/vis3-mont.pl
@@ -100,7 +100,7 @@ $code.=<<___;
        ld      [$ap+12],       $t3
        or      $t0,    $aj,    $aj
        add     $ap,    16,     $ap
        ld      [$ap+12],       $t3
        or      $t0,    $aj,    $aj
        add     $ap,    16,     $ap

-       stxa    $aj,    [$anp]0xe2      ! converted ap[0]
+       stx     $aj,    [$anp]          ! converted ap[0]

 
        mulx    $aj,    $m0,    $lo0    ! ap[0]*bp[0]
        umulxhi $aj,    $m0,    $hi0
 
        mulx    $aj,    $m0,    $lo0    ! ap[0]*bp[0]
        umulxhi $aj,    $m0,    $hi0


@@ -150,7 +150,7 @@ $code.=<<___;
        sllx    $t1,    32,     $aj
        add     $ap,    8,      $ap
        or      $t0,    $aj,    $aj
        sllx    $t1,    32,     $aj
        add     $ap,    8,      $ap
        or      $t0,    $aj,    $aj

-       stxa    $aj,    [$anp]0xe2      ! converted ap[j]
+       stx     $aj,    [$anp]          ! converted ap[j]

 
        ld      [$np+0],        $t2     ! np[j]
        addcc   $nlo,   $hi1,   $lo1
 
        ld      [$np+0],        $t2     ! np[j]
        addcc   $nlo,   $hi1,   $lo1


@@ -169,7 +169,7 @@ $code.=<<___;
        addcc   $lo0,   $lo1,   $lo1    ! np[j]*m1+ap[j]*bp[0]
        umulxhi $nj,    $m1,    $nj     ! nhi=nj
        addxc   %g0,    $hi1,   $hi1
        addcc   $lo0,   $lo1,   $lo1    ! np[j]*m1+ap[j]*bp[0]
        umulxhi $nj,    $m1,    $nj     ! nhi=nj
        addxc   %g0,    $hi1,   $hi1

-       stxa    $lo1,   [$tp]0xe2       ! tp[j-1]
+       stx     $lo1,   [$tp]           ! tp[j-1]

        add     $tp,    8,      $tp     ! tp++
 
        brnz,pt $cnt,   .L1st
        add     $tp,    8,      $tp     ! tp++
 
        brnz,pt $cnt,   .L1st


@@ -182,12 +182,12 @@ $code.=<<___;
        addxc   $nj,    %g0,    $hi1
        addcc   $lo0,   $lo1,   $lo1    ! np[j]*m1+ap[j]*bp[0]
        addxc   %g0,    $hi1,   $hi1
        addxc   $nj,    %g0,    $hi1
        addcc   $lo0,   $lo1,   $lo1    ! np[j]*m1+ap[j]*bp[0]
        addxc   %g0,    $hi1,   $hi1

-       stxa    $lo1,   [$tp]0xe2       ! tp[j-1]
+       stx     $lo1,   [$tp]           ! tp[j-1]

        add     $tp,    8,      $tp
 
        addcc   $hi0,   $hi1,   $hi1
        addxc   %g0,    %g0,    $ovf    ! upmost overflow bit
        add     $tp,    8,      $tp
 
        addcc   $hi0,   $hi1,   $hi1
        addxc   %g0,    %g0,    $ovf    ! upmost overflow bit

-       stxa    $hi1,   [$tp]0xe2
+       stx     $hi1,   [$tp]

        add     $tp,    8,      $tp
 \f
        ba      .Louter
        add     $tp,    8,      $tp
 \f
        ba      .Louter