[SIMD/test] Do not access indeterminate simd vectors with lane().

Depending on implementation, this may require bitmask operations on the indeterminate value, which requires an lvalue-to-rvalue conversion, which results in undefined behaviour except in some very special cases.

dune/common/simd/test.hh

@@ -157,7 +157,10 @@ namespace Dune {
       template<class V>
       static V make123()
       {
-        V vec;
+        // initialize to avoid undefined behaviour if assigning to lane()
+        // involves lvalue-to-rvalue conversions, e.g. due to bitmask
+        // operations
+        V vec{ Scalar<V>{0} };
         for(std::size_t l = 0; l < lanes(vec); ++l)
           lane(l, vec) = l + 1;
         return vec;
@@ -179,7 +182,7 @@ namespace Dune {
       template<class V>
       static V leftVector()
       {
-        V res;
+        V res{ Scalar<V>{0} };
         for(std::size_t l = 0; l < lanes(res); ++l)
           lane(l, res) = Scalar<V>(l+1);
         return res;
@@ -188,7 +191,7 @@ namespace Dune {
       template<class V>
       static V rightVector()
       {
-        V res;
+        V res{ Scalar<V>{0} };
         for(std::size_t l = 0; l < lanes(res); ++l)
           // do not exceed number of bits in char (for shifts)
           // avoid 0 (for / and %)
@@ -307,7 +310,7 @@ namespace Dune {
       template<class V>
       void checkLane()
       {
-        V vec;
+        V vec{ Scalar<V>{0} };
         // check lane() on mutable lvalues
         for(std::size_t l = 0; l < lanes(vec); ++l)
           lane(l, vec) = l + 1;
@@ -1326,7 +1329,7 @@ namespace Dune {
         DUNE_SIMD_CHECK(allFalse(M(false)) == true);
         DUNE_SIMD_CHECK(anyFalse(M(false)) == true);
 
-        M mixedVec;
+        M mixedVec{ Scalar<M>{0} };
         for(std::size_t l = 0; l < lanes(mixedVec); ++l)
           lane(l, mixedVec) = (l % 2);
 
@@ -1385,8 +1388,8 @@ namespace Dune {
         DUNE_SIMD_CHECK(allTrue(cond(M(true),  vec1, vec2) == vec1));
         DUNE_SIMD_CHECK(allTrue(cond(M(false), vec1, vec2) == vec2));
 
-        V mixedResult;
-        M mixedMask;
+        V mixedResult{ Scalar<V>{0} };
+        M mixedMask{ Scalar<M>{0} };
         for(std::size_t l = 0; l < lanes(mixedMask); ++l)
         {
           lane(l, mixedMask  ) = (l % 2);