[llvm] [WIP] [AMDGPU] [GlobalIsel] Combine Fmul with Select into ldexp instruction. (PR #120104)

[Vikash Gupta via llvm-commits]

================
@@ -445,3 +445,75 @@ void AMDGPUCombinerHelper::applyExpandPromotedF16FMed3(MachineInstr &MI,
   Builder.buildFMinNumIEEE(MI.getOperand(0), B1, C1);
   MI.eraseFromParent();
 }
+
+bool AMDGPUCombinerHelper::matchCombineFmulWithSelectToLdexp(
+    MachineInstr &MI, MachineInstr &Sel,
+    std::function<void(MachineIRBuilder &)> &MatchInfo) {
+  assert(MI.getOpcode() == TargetOpcode::G_FMUL);
+  assert(Sel.getOpcode() == TargetOpcode::G_SELECT);
+
+  Register Dst = MI.getOperand(0).getReg();
+  LLT DestTy = MRI.getType(Dst);
+  LLT ScalarDestTy = DestTy.getScalarType();
+
+  if ((ScalarDestTy == LLT::float64() || ScalarDestTy == LLT::float32() ||
----------------
vg0204 wrote:

> First case looks better, the bfloat case is too far removed from the final instructions do to all the promotion handling

You asked to go without this combine for bfloat as mentioned in the #104900, with the above comment after seeing assmebly with and without dagCombine.

https://github.com/llvm/llvm-project/pull/120104