[llvm] [AMDGPU] Don't optimize agpr phis if the operand doesn't have subreg use (PR #91267)

[Matt Arsenault via llvm-commits]

================
@@ -0,0 +1,58 @@
+# RUN: llc -mtriple=amdgcn-amd-amdhsa -mcpu=gfx908 -verify-machineinstrs -run-pass si-fold-operands -o - %s | FileCheck %s
+
+# CHECK-NOT: V_ACCVGPR_READ_B32_e64
+
+---
+name:            skip_optimize_agpr_phi_without_subreg_use
+tracksRegLiveness: true
+machineFunctionInfo:
+  isEntryFunction: true
+  scratchRSrcReg:  '$sgpr0_sgpr1_sgpr2_sgpr3'
+body:             |
+  bb.0:
+    successors: %bb.1(0x80000000); %bb.1(100.00%)
+
+    %10:vgpr_32 = V_MOV_B32_e32 0, implicit $exec
----------------
arsenm wrote:

Should use -run-pass=none to compact the register numbers 

https://github.com/llvm/llvm-project/pull/91267