[VectorCombine] scalarizeLoadExtract - infinite loop if we scalarize a dead extract waiting to be erased

[RKSimon]

Reduced test case:

define void @infinite_loop(ptr %p0) {
  %load = load <4 x float>, ptr %p0, align 8
  %shuffle0 = shufflevector <4 x float> zeroinitializer, <4 x float> %load, <4 x i32> <i32 0, i32 4, i32 1, i32 5>
  %shuffle1 = shufflevector <4 x float> %shuffle0, <4 x float> zeroinitializer, <4 x i32> <i32 0, i32 4, i32 poison, i32 poison>
  %extract = extractelement <4 x float> %load, i64 1
  ret void
}

scalarizeLoadExtract iterates across all users of a load, and doesn't account for possible orders in the worklist that prevent unused instructions that use the load from being erased afterward.

Fix incoming

[llvmbot]

@llvm/issue-subscribers-bug

Author: Simon Pilgrim (RKSimon)

Reduced test case:

define void @<!-- -->infinite_loop(ptr %p0) {
  %load = load &lt;4 x float&gt;, ptr %p0, align 8
  %shuffle0 = shufflevector &lt;4 x float&gt; zeroinitializer, &lt;4 x float&gt; %load, &lt;4 x i32&gt; &lt;i32 0, i32 4, i32 1, i32 5&gt;
  %shuffle1 = shufflevector &lt;4 x float&gt; %shuffle0, &lt;4 x float&gt; zeroinitializer, &lt;4 x i32&gt; &lt;i32 0, i32 4, i32 poison, i32 poison&gt;
  %extract = extractelement &lt;4 x float&gt; %load, i64 1
  ret void
}

scalarizeLoadExtract iterates across all users of a load, and doesn't account for possible orders in the worklist that prevent unused instructions that use the load from being erased afterward.

Fix incoming