[LoopInterchange] Ignore debug intrinsics during legality checks.

[aprantl]

This is generally great, I think the helper function would be useful outside of this source file, too.

lib/Transforms/Scalar/LoopInterchange.cpp

82 ↗

(On Diff #141379)

This adapter should really be available in BasicBlock.h or somewhere more accessible if it doesn't exist already. Could you move it there (or elsewhere appropriate) and add a doxygen comment?

1000 ↗

(On Diff #141379)

I'm assuming that dbg intrinsic calls don't read memory, so this change may be a noop. Feel free to keep it for readability though.

[vsk]

lib/Transforms/Scalar/LoopInterchange.cpp

82 ↗

(On Diff #141379)

+ 1, I think having iterator_range<iterator> instructions() and ... instructions_nodbg() in BasicBlock would be nice.

[fhahn]

lib/Transforms/Scalar/LoopInterchange.cpp

82 ↗

(On Diff #141379)

Ok great, I will move it there. Do you think we should just have a member function instructionsNodbg() or a function taking an instruction iterator and returning a filtered iterator? It's just a question how we want to compose it with things like reverse

[vsk]

lib/Transforms/Scalar/LoopInterchange.cpp

82 ↗

(On Diff #141379)

I think we should have both. We'd need 'skipDebugIntrinsics' to implement 'instructionsNodbg', and 'instructionsNodbg' would make it easier to write correct code going forwards.

Wdyt of placing 'skipDebugIntrinsics' in IR/DebugInfo.h?

[mattd]

lib/Transforms/Scalar/LoopInterchange.cpp

82 ↗

(On Diff #141379)

I'm a bit late to the party here, but have been keeping an eye on this discussion. I really like this idea. What about exposing this in IR/Intrinisics.h instead? For instance, what if there were other intrinsics that we would like to skip over as well (e.g., lifetime.start, lifetime.end). In that case I can imagine (in the future) an iterator that takes a list of types to skip-over, like a parameter pack or something, although I'm not sure if that would work with make_filter_range, just a thought.

[fhahn]

lib/Transforms/Scalar/LoopInterchange.cpp

82 ↗

(On Diff #141379)

Having an easy way to skip all kinds of instructions sounds like a great idea. I will try to put initial patches together in the next few days.

[aprantl]

LGTM with whatever new API for skipping we come up with.

[fhahn]

Rebased after rL330875 went in

llvm/trunk/lib/Transforms/Scalar/LoopInterchange.cpp

if (!InnerIndexVarInc) {
     return true;
   }
 
   // Since we split the inner loop latch on this induction variable. Make sure
   // we do not have any instruction between the induction variable and branch
   // instruction.
 
   bool FoundInduction = false;
-  for (const Instruction &I : llvm::reverse(*InnerLoopLatch)) {
+  for (const Instruction &I :
+       llvm::reverse(InnerLoopLatch->instructionsWithoutDebug())) {
     if (isa<BranchInst>(I) || isa<CmpInst>(I) || isa<TruncInst>(I) ||
         isa<ZExtInst>(I))
       continue;
 
     // We found an instruction. If this is not induction variable then it is not
     // safe to split this loop latch.
     if (!I.isIdenticalTo(InnerIndexVarInc)) {
       DEBUG(dbgs() << "Found unsupported instructions between induction "
if (!isLegalToInterChangeLoops(DepMatrix, InnerLoopId, OuterLoopId)) {
     ORE->emit([&]() {
       return OptimizationRemarkMissed(DEBUG_TYPE, "Dependence",
                                       InnerLoop->getStartLoc(),
                                       InnerLoop->getHeader())
              << "Cannot interchange loops due to dependences.";
     });
     return false;
   }
-
   // Check if outer and inner loop contain legal instructions only.
   for (auto *BB : OuterLoop->blocks())
-    for (Instruction &I : *BB)
+    for (Instruction &I : BB->instructionsWithoutDebug())
       if (CallInst *CI = dyn_cast<CallInst>(&I)) {
         // readnone functions do not prevent interchanging.
         if (CI->doesNotReadMemory())
           continue;
         DEBUG(dbgs() << "Loops with call instructions cannot be interchanged "
                      << "safely.");
         ORE->emit([&]() {
           return OptimizationRemarkMissed(DEBUG_TYPE, "CallInst",

llvm/trunk/test/Transforms/LoopInterchange/debuginfo.ll

+; RUN: opt < %s -basicaa -loop-interchange -pass-remarks='loop-interchange' -pass-remarks-output=%t -S \
+; RUN:     -verify-dom-info -verify-loop-info | FileCheck %s
+; RUN: FileCheck -check-prefix=REMARK --input-file=%t %s
+
+
+target datalayout = "e-m:e-i64:64-f80:128-n8:16:32:64-S128"
+target triple = "x86_64-unknown-linux-gnu"
+
+@A = common global [100 x [100 x i64]] zeroinitializer
+
+;;  for(int i=0;i<100;i++)
+;;    for(int j=0;j<100;j++)
+;;      A[j][i] = A[j][i]+k;
+
+; REMARK:      Name:            Interchanged
+; REMARK-NEXT: Function:        interchange_01
+; CHECK: split
+
+define void @interchange_01(i64 %k, i64 %N) !dbg !5 {
+entry:
+  br label %for1.header
+
+for1.header:
+  %j23 = phi i64 [ 0, %entry ], [ %j.next24, %for1.inc10 ]
+  call void @llvm.dbg.value(metadata i64 %j, metadata !13, metadata !DIExpression()), !dbg !14
+  br label %for2
+
+for2:
+  %j = phi i64 [ %j.next, %for2 ], [ 0, %for1.header ]
+  call void @llvm.dbg.value(metadata i64 %j, metadata !13, metadata !DIExpression()), !dbg !14
+  %arrayidx5 = getelementptr inbounds [100 x [100 x i64]], [100 x [100 x i64]]* @A, i64 0, i64 %j, i64 %j23
+  %lv = load i64, i64* %arrayidx5
+  %add = add nsw i64 %lv, %k
+  store i64 %add, i64* %arrayidx5
+  %j.next = add nuw nsw i64 %j, 1
+  %exitcond = icmp eq i64 %j, 99
+  call void @llvm.dbg.value(metadata i64 %j, metadata !13, metadata !DIExpression()), !dbg !14
+  br i1 %exitcond, label %for1.inc10, label %for2
+
+for1.inc10:
+  %j.next24 = add nuw nsw i64 %j23, 1
+  call void @llvm.dbg.value(metadata i64 %j, metadata !13, metadata !DIExpression()), !dbg !14
+  %exitcond26 = icmp eq i64 %j23, 99
+  br i1 %exitcond26, label %for.end12, label %for1.header
+
+for.end12:
+  ret void
+}
+
+; Function Attrs: nounwind readnone speculatable
+declare void @llvm.dbg.value(metadata, metadata, metadata)
+
+!llvm.dbg.cu = !{!0}
+!llvm.module.flags = !{!3, !4}
+
+!0 = distinct !DICompileUnit(language: DW_LANG_C99, file: !1, producer: "clang", isOptimized: true, runtimeVersion: 0, emissionKind: FullDebug, enums: !2)
+!1 = !DIFile(filename: "test.c", directory: "/test")
+!2 = !{}
+!3 = !{i32 2, !"Dwarf Version", i32 4}
+!4 = !{i32 2, !"Debug Info Version", i32 3}
+!5 = distinct !DISubprogram(name: "foo", scope: !1, file: !1, line: 1, type: !6, isLocal: false, isDefinition: true, scopeLine: 1, flags: DIFlagPrototyped, isOptimized: true, unit: !0, variables: !12)
+!6 = !DISubroutineType(types: !7)
+!7 = !{null, !8, !8, !11}
+!8 = !DIDerivedType(tag: DW_TAG_restrict_type, baseType: !9)
+!9 = !DIDerivedType(tag: DW_TAG_pointer_type, baseType: !10, size: 32, align: 32)
+!10 = !DIBasicType(name: "float", size: 32, align: 32, encoding: DW_ATE_float)
+!11 = !DIBasicType(name: "int", size: 32, align: 32, encoding: DW_ATE_signed)
+!12 = !{!13}
+!13 = !DILocalVariable(name: "a", arg: 1, scope: !5, file: !1, line: 1, type: !8)
+!14 = !DILocation(line: 1, column: 27, scope: !5)