Re-apply r113679, which was reverted in r113720, which added a paid of new instcombine transforms

to expose greater opportunities for store narrowing in codegen.  This patch fixes a potential
infinite loop in instcombine caused by one of the introduced transforms being overly aggressive.


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@113763 91177308-0d34-0410-b5e6-96231b3b80d8

test/CodeGen/ARM/fast-isel.ll

@@ -15,22 +15,6 @@ entry:
   ret i32 %add
 }
 
-define float @fp_ops(float %a, float %b) nounwind {
-entry:
-  %a.addr = alloca float, align 4
-  %b.addr = alloca float, align 4
-  store float %a, float* %a.addr
-  store float %b, float* %b.addr
-  %tmp = load float* %a.addr
-  %tmp1 = load float* %b.addr
-  %mul = fmul float %tmp, %tmp1
-  %tmp2 = load float* %b.addr
-  %tmp3 = load float* %a.addr
-  %mul2 = fmul float %tmp2, %tmp3
-  %add = fadd float %mul, %mul2
-  ret float %mul
-}
-
 define i32* @foo(i32* %p, i32* %q, i32** %z) nounwind {
 entry:
   %r = load i32* %p