most fit proceed check dbnull following distribute variable?

this doubt comes adult occasionally, nonetheless i haven't seen tolerable answer.

a customary settlement (row datarow):

 (row["value"] != dbnull.value)
 {
      someobject.member = row["value"];
 }

my initial doubt some-more fit (i've flipped condition):

  row["value"] == dbnull.value; // or
  row["value"] dbnull; // or
  row["value"].gettype() == typeof(dbnull) // or... any suggestions?

indicates .gettype() should faster, nonetheless maybe compiler knows few tricks i don't?

second question, value caching value row["value"] does compiler optimize indexer divided anyway?

for example:

  vigilant valueholder;
  (dbnull.value == (valueholder = row["value"])) {}

notes:

i'm seeking personally checking dbnull following assignment (not beforehand optimization, etc.).

i benchmarked few scenarios (time seconds, 10,000,000 trials):

row["value"] == dbnull.value: 00:00:01.5478995
row["value"] dbnull: 00:00:01.6306578
row["value"].gettype() == typeof(dbnull): 00:00:02.0138757

object.referenceequals same opening "=="

the many engaging result? mismatch name buttress box (for example, "value" instead "value", takes roughly 10 times longer (for string):

row["value"] == dbnull.value: 00:00:12.2792374

the cool story seems can't feeling adult buttress the index, following pledge buttress name feed indexer matches datacolumn's name exactly.

caching value also appears perceptibly twice fast:

no caching: 00:00:03.0996622
with caching: 00:00:01.5659920

so many fit slight seems be:

 vigilant temp;
 twine variable;
 (dbnull.value != (temp = row["value"]))
 {
      non-static = temp.tostring();
 }

code1