[mathicgb] 84/393: Fixed bug in hash table hit-verification code.

[Doug Torrance]

This is an automated email from the git hooks/post-receive script.

dtorrance-guest pushed a commit to branch upstream
in repository mathicgb.

commit 928528a3f0aa49b90ce821e855d0f27a3c73d6ae
Author: Bjarke Hammersholt Roune <bjarkehr.code at gmail.com>
Date:   Fri Oct 26 18:41:18 2012 +0200

    Fixed bug in hash table hit-verification code.
---
 src/mathicgb/PolyRing.hpp | 8 ++++----
 1 file changed, 4 insertions(+), 4 deletions(-)

diff --git a/src/mathicgb/PolyRing.hpp b/src/mathicgb/PolyRing.hpp
index 9756f39..71668e7 100755
--- a/src/mathicgb/PolyRing.hpp
+++ b/src/mathicgb/PolyRing.hpp
@@ -660,11 +660,11 @@ inline bool PolyRing::monomialIsProductOfHintTrue(
     uint64 A, B, AB;
     // We have to use std::memcpy here because just casting to a int64 breaks
     // the strict aliasing rule which implies undefined behavior. Both MSVC and
-    // gcc don't actually call memcpy here. MSVC is a tiny big slower for this
+    // gcc don't actually call memcpy here. MSVC is a tiny bit slower for this
     // code than for casting while GCC seems to be exactly the same speed.
-    std::memcpy(&A, &a[i], 8);
-    std::memcpy(&B, &b[i], 8);
-    std::memcpy(&AB, &ab[i], 8);
+    std::memcpy(&A, &a[i*2], 8);
+    std::memcpy(&B, &b[i*2], 8);
+    std::memcpy(&AB, &ab[i*2], 8);
     orOfXor |= AB ^ (A + B);
   }
   MATHICGB_ASSERT((orOfXor == 0) == monomialIsProductOf(a, b, ab));

-- 
Alioth's /usr/local/bin/git-commit-notice on /srv/git.debian.org/git/debian-science/packages/mathicgb.git