[clang] [llvm] [X86] Reduce -ffixed-r compile-time overhead (PR #184606)

[Alexis Engelke via llvm-commits]

================
@@ -517,13 +517,15 @@ BitVector X86RegisterInfo::getReservedRegs(const MachineFunction &MF) const {
 
   auto &ST = MF.getSubtarget<X86Subtarget>();
   if (ST.is64Bit()) {
-    for (size_t Reg = 0; Reg < getNumRegs(); Reg++) {
-      // Set r# as reserved register if user required
-      if (ST.isRegisterReservedByUser(Reg)) {
+    // Set r# as reserved register if user required
----------------
aengelke wrote:

Fast path for ReservedRReg being empty?

https://github.com/llvm/llvm-project/pull/184606