[llvm] [AMDGPU] Use V_FMAC_F64 in "if (cond) a -= c" (PR #168710)

[Matt Arsenault via llvm-commits]

================
@@ -260,6 +281,15 @@ bool GCNPreRAOptimizationsImpl::run(MachineFunction &MF) {
     Changed |= processReg(Reg);
   }
 
+  if (ST.shouldUseConditionalSubToFMAF64()) {
+    for (MachineBasicBlock &MBB : MF) {
----------------
arsenm wrote:

Can you integrate this better so it's not doing another full loop over the function?

https://github.com/llvm/llvm-project/pull/168710