[AMDGPU] Fix not rescheduling without clustering

Regions are sometimes skipped which should be rescheduled without memory op
clustering. RegionIdx is not incremented when iterating over regions that
are flagged to be skipped, causing the index to be incorrect.

Thanks to Vang Thao for discovering this bug!

Reviewed By: rampitec

Differential Revision: https://reviews.llvm.org/D85498

Author: vangthao95

llvm/lib/Target/AMDGPU/GCNSchedStrategy.cpp

@@ -567,8 +567,10 @@ void GCNScheduleDAGMILive::finalizeSchedule() {
       SavedMutations.swap(Mutations);
 
     for (auto Region : Regions) {
-      if (Stage == UnclusteredReschedule && !RescheduleRegions[RegionIdx])
+      if (Stage == UnclusteredReschedule && !RescheduleRegions[RegionIdx]) {
+        ++RegionIdx;
         continue;
+      }
 
       RegionBegin = Region.first;
       RegionEnd = Region.second;

llvm/test/CodeGen/AMDGPU/schedule-regpressure-limit-clustering.ll

@@ -3,6 +3,9 @@
 ; Interleave loads and stores to fit into 9 VGPR limit.
 ; This requires to avoid load/store clustering.
 
+; Reschedule the second scheduling region without clustering while
+; the first region is skipped.
+
 ; GCN: global_load_dwordx4
 ; GCN: global_store_dwordx4
 ; GCN: global_load_dwordx4
@@ -12,10 +15,13 @@
 ; GCN: NumVgprs: {{[0-9]$}}
 ; GCN: ScratchSize: 0{{$}}
 
-define amdgpu_kernel void @load_store_max_9vgprs(<4 x i32> addrspace(1)* nocapture noalias readonly %arg, <4 x i32> addrspace(1)* nocapture noalias %arg1) #1 {
+define amdgpu_kernel void @load_store_max_9vgprs(<4 x i32> addrspace(1)* nocapture noalias readonly %arg, <4 x i32> addrspace(1)* nocapture noalias %arg1, i1 %cnd) #1 {
 bb:
   %id = call i32 @llvm.amdgcn.workitem.id.x()
   %base = getelementptr inbounds <4 x i32>, <4 x i32> addrspace(1)* %arg, i32 %id
+  br i1 %cnd, label %bb1, label %bb2
+
+bb1:
   %tmp = getelementptr inbounds <4 x i32>, <4 x i32> addrspace(1)* %base, i32 1
   %tmp2 = load <4 x i32>, <4 x i32> addrspace(1)* %tmp, align 4
   %tmp3 = getelementptr inbounds <4 x i32>, <4 x i32> addrspace(1)* %base, i32 3
@@ -27,6 +33,9 @@ bb:
   store <4 x i32> %tmp4, <4 x i32> addrspace(1)* %tmp7, align 4
   %tmp8 = getelementptr inbounds <4 x i32>, <4 x i32> addrspace(1)* %arg1, i64 5
   store <4 x i32> %tmp6, <4 x i32> addrspace(1)* %tmp8, align 4
+  br label %bb2
+
+bb2:
   ret void
 }