[llvm] [RegAlloc] Account for use availability when applying rematerializable weight discount (PR #159180)

[Matt Arsenault via llvm-commits]

================
@@ -36,24 +36,18 @@ body:             |
   ; CHECK-NEXT:   SI_SPILL_AV128_SAVE killed $vgpr0_vgpr1_vgpr2_vgpr3, %stack.1, $sgpr32, 0, implicit $exec :: (store (s128) into %stack.1, align 4, addrspace 5)
   ; CHECK-NEXT:   renamable $vgpr0_vgpr1_vgpr2_vgpr3 = SI_SPILL_AV128_RESTORE %stack.2, $sgpr32, 0, implicit $exec :: (load (s128) from %stack.2, align 4, addrspace 5)
   ; CHECK-NEXT:   renamable $vgpr0 = V_FMA_F32_e64 0, $vgpr6, 0, $vgpr6, 0, killed $vgpr2, 0, 0, implicit $mode, implicit $exec
-  ; CHECK-NEXT:   SI_SPILL_AV128_SAVE $vgpr4_vgpr5_vgpr6_vgpr7, %stack.4, $sgpr32, 0, implicit $exec :: (store (s128) into %stack.4, align 4, addrspace 5)
   ; CHECK-NEXT:   renamable $vgpr0 = V_TRUNC_F32_e32 killed $vgpr0, implicit $mode, implicit $exec
   ; CHECK-NEXT:   SI_SPILL_AV32_SAVE killed $vgpr0, %stack.3, $sgpr32, 0, implicit $exec :: (store (s32) into %stack.3, addrspace 5)
+  ; CHECK-NEXT:   renamable $vgpr0_vgpr1_vgpr2_vgpr3 = SI_SPILL_AV128_RESTORE %stack.2, $sgpr32, 0, implicit $exec :: (load (s128) from %stack.2, align 4, addrspace 5)
----------------
arsenm wrote:

This avoided a lot of spills. However there's a good chance this no longer is going down the last chance recoloring path, but that's inevitable with these sorts of tests 

https://github.com/llvm/llvm-project/pull/159180