Disable coalescing optimizations when using generic target.

The main issue is that they end up generating a number of smaller
vector ops (e.g. 4-wide and 8-wide on the 16-wide generic target,
which the examples/intrinsics implementations don't currently
support.

This fixes a number of failing tests for now; it may be worth
generalizing the stuff in examples/intrinsics at some point,
since as a general principle, e.g. if generating LLVM IR output,
the coalescing optimizations are still desirable.

Issue #175.

opt.cpp

@@ -392,7 +392,8 @@ Optimize(llvm::Module *module, int optLevel) {
             g->target.vectorWidth > 1) {
             optPM.add(CreateGSToLoadStorePass());
         
-            if (g->opt.disableCoalescing == false) {
+            if (g->opt.disableCoalescing == false &&
+                g->target.isa != Target::GENERIC) {
                 // It is important to run this here to make it easier to
                 // finding matching gathers we can coalesce..
                 optPM.add(llvm::createEarlyCSEPass());