<li class="listitem">
Directly as initialized by the process described for the <a class="link" href="performance.html#poly_collection.performance.insertion_tests" title="Insertion tests">insertion
tests</a>. The sequence of types is complex enough that CPU's branch
- prediction mechanisms are not able to fully anticipate it <a href="#ftn.poly_collection.performance.processing_tests.f0" class="footnote" name="poly_collection.performance.processing_tests.f0"><sup class="footnote">[19]</sup></a>. As elements are ordered according to their construction
+ prediction mechanisms are not able to fully anticipate it <a href="#ftn.poly_collection.performance.processing_tests.f0" class="footnote" name="poly_collection.performance.processing_tests.f0"><sup class="footnote">[21]</sup></a>. As elements are ordered according to their construction
time, certain degree of memory contiguity is expected.
</li>
<li class="listitem">
</div>
<div class="footnotes">
<br><hr style="width:100; text-align:left;margin-left: 0">
-<div id="ftn.poly_collection.performance.processing_tests.f0" class="footnote"><p><a href="#poly_collection.performance.processing_tests.f0" class="para"><sup class="para">[19] </sup></a>
+<div id="ftn.poly_collection.performance.processing_tests.f0" class="footnote"><p><a href="#poly_collection.performance.processing_tests.f0" class="para"><sup class="para">[21] </sup></a>
This has been verified empirically: simpler cycles did indeed yield
better execution times.
</p></div>