[llvm] AMDGPU: Fix assert when multi operands to update after folding imm (PR #148205)

[Matt Arsenault via llvm-commits]

================
@@ -0,0 +1,55 @@
+# NOTE: Assertions have been autogenerated by utils/update_mir_test_checks.py UTC_ARGS: --version 5
+# RUN: llc -mtriple=amdgcn-amd-hsa -mcpu=gfx1031 -run-pass=si-fold-operands -o - %s | FileCheck %s
+--- |
+  @global = external addrspace(3) global i32
+
+  define void @snork() {
+  bb:
+    br label %bb3
+
+  bb3:
+    br label %bb5
+
+  bb5:
+    ret void
+  }
+...
+---
+name:            snork
+body:             |
+  ; CHECK-LABEL: name: snork
+  ; CHECK: bb.0.bb:
+  ; CHECK-NEXT:   successors: %bb.1(0x40000000), %bb.2(0x40000000)
+  ; CHECK-NEXT: {{  $}}
+  ; CHECK-NEXT:   [[S_MOV_B32_:%[0-9]+]]:sreg_32 = S_MOV_B32 0
+  ; CHECK-NEXT:   [[REG_SEQUENCE:%[0-9]+]]:sgpr_128 = REG_SEQUENCE [[S_MOV_B32_]], %subreg.sub0, [[S_MOV_B32_]], %subreg.sub1, [[S_MOV_B32_]], %subreg.sub2, [[S_MOV_B32_]], %subreg.sub3
+  ; CHECK-NEXT:   [[V_MOV_B32_e32_:%[0-9]+]]:vgpr_32 = V_MOV_B32_e32 target-flags(amdgpu-abs32-lo) @global, implicit $exec
+  ; CHECK-NEXT:   [[COPY:%[0-9]+]]:vreg_128 = COPY [[REG_SEQUENCE]]
+  ; CHECK-NEXT:   DS_WRITE_B128_gfx9 killed [[V_MOV_B32_e32_]], [[COPY]], 32, 0, implicit $exec :: (store (s128) into `ptr addrspace(3) getelementptr inbounds nuw (i8, ptr addrspace(3) @global, i32 32)`, addrspace 3)
+  ; CHECK-NEXT:   S_CMP_LG_U32 0, 0, implicit-def $scc
+  ; CHECK-NEXT:   S_CBRANCH_SCC1 %bb.2, implicit $scc
+  ; CHECK-NEXT:   S_BRANCH %bb.1
+  ; CHECK-NEXT: {{  $}}
+  ; CHECK-NEXT: bb.1.bb3:
+  ; CHECK-NEXT:   successors: %bb.2(0x80000000)
+  ; CHECK-NEXT: {{  $}}
+  ; CHECK-NEXT: bb.2.bb5:
+  ; CHECK-NEXT:   SI_RETURN
+  bb.0.bb:
+    successors: %bb.1, %bb.2
+
+    %9:sreg_32 = S_MOV_B32 0
----------------
arsenm wrote:

Should also run this through -run-pass=none to compact the register numbers 

https://github.com/llvm/llvm-project/pull/148205