[Local] Do not ignore zexts in salvageDebugInfo, PR45923

[vsk]

Summary

When salvaging a dead zext instruction, append a convert operation to
the DIExpressions of the debug uses of the instruction, to prevent the
salvaged value from being sign-extended.

I confirmed that lldb prints out the correct unsigned result for "f" in
the example from PR45923 with this changed applied.

[aprantl]

Great!

llvm/lib/Transforms/Utils/Local.cpp

1702

I'm pretty sure this condition predates DW_OP_convert.

[aprantl]

llvm/test/Transforms/InstCombine/cast-mul-select.ll

178

I have not thought about the exact semantics, but I'm wondering if it would be possible to contract a sequence on DW_OP_convert operations in the backend in DwarfExpression.cpp to have a smaller encoding? I guess this would be easier if we distinguished between extending and truncating a value in our IR. Perhaps we'd need to introduce DW_OP_LLVM_zext & friends?

[vsk]

llvm/test/Transforms/InstCombine/cast-mul-select.ll

178

Yeah, we could try to simplify dwarf expressions really late in the backend, and/or do something like DW_OP_LLVM_zext to avoid emitting silly expressions in the first place. The former might allow us to compact-ify dwarf in more cases where there's emergent complexity in the expression, but is also quite possibly overkill (cc @JDevlieghere). We should probably look at what kind of expressions actually pop out of a stage2 build, then figure out where (if anywhere) we ought to make things more compact.

llvm/lib/Transforms/Utils/Local.cpp

DIExpression *llvm::salvageDebugInfoImpl(Instruction &I,
 
   // initializer-list helper for applying operators to the source DIExpression.
   auto applyOps = [&](ArrayRef<uint64_t> Opcodes) -> DIExpression * {
     SmallVector<uint64_t, 8> Ops(Opcodes.begin(), Opcodes.end());
     return doSalvage(Ops);
   };
 
   if (auto *CI = dyn_cast<CastInst>(&I)) {
-    // No-op casts and zexts are irrelevant for debug info.
-    if (CI->isNoopCast(DL) || isa<ZExtInst>(&I))
+    // No-op casts are irrelevant for debug info.
+    if (CI->isNoopCast(DL))
       return SrcDIExpr;
 
     Type *Type = CI->getType();
-    // Casts other than Trunc or SExt to scalar types cannot be salvaged.
-    if (Type->isVectorTy() || (!isa<TruncInst>(&I) && !isa<SExtInst>(&I)))
+    // Casts other than Trunc, SExt, or ZExt to scalar types cannot be salvaged.
+    if (Type->isVectorTy() ||
+        !(isa<TruncInst>(&I) || isa<SExtInst>(&I) || isa<ZExtInst>(&I)))
       return nullptr;
 
     Value *FromValue = CI->getOperand(0);
     unsigned FromTypeBitSize = FromValue->getType()->getScalarSizeInBits();
     unsigned ToTypeBitSize = Type->getScalarSizeInBits();
 
     return applyOps(DIExpression::getExtOps(FromTypeBitSize, ToTypeBitSize,
                                             isa<SExtInst>(&I)));

llvm/test/Transforms/InstCombine/cast-mul-select.ll

 ; RUN: opt < %s -instcombine -S | FileCheck %s
 ; RUN: opt -debugify -instcombine -S < %s | FileCheck %s -check-prefix DBGINFO
 
 target datalayout = "e-p:64:64:64-i1:8:8-i8:8:8-i16:16:16-i32:32:32-i64:64:64-f32:32:32-f64:64:64-v64:64:64-v128:128:128-a0:0:64-s0:64:64-f80:128:128-n8:16:32"
 
 define i32 @mul(i32 %x, i32 %y) {
 ; CHECK-LABEL: @mul(
 ; CHECK-NEXT:    [[C:%.*]] = mul i32 [[X:%.*]], [[Y:%.*]]
 ; CHECK-NEXT:    [[D:%.*]] = and i32 [[C]], 255
 ; CHECK-NEXT:    ret i32 [[D]]
 
 ; Test that when zext is evaluated in different type
 ; we preserve the debug information in the resulting
 ; instruction.
 ; DBGINFO-LABEL: @mul(
-; DBGINFO-NEXT:    call void @llvm.dbg.value(metadata i32 %x
-; DBGINFO-NEXT:    call void @llvm.dbg.value(metadata i32 %y
+; DBGINFO-NEXT:    call void @llvm.dbg.value(metadata i32 %x, {{.*}} !DIExpression(DW_OP_LLVM_convert, 32, DW_ATE_unsigned, DW_OP_LLVM_convert, 8, DW_ATE_unsigned, DW_OP_stack_value))
+; DBGINFO-NEXT:    call void @llvm.dbg.value(metadata i32 %y, {{.*}} !DIExpression(DW_OP_LLVM_convert, 32, DW_ATE_unsigned, DW_OP_LLVM_convert, 8, DW_ATE_unsigned, DW_OP_stack_value))
 ; DBGINFO-NEXT:    [[C:%.*]] = mul i32 {{.*}}
 ; DBGINFO-NEXT:    [[D:%.*]] = and i32 {{.*}}
 ; DBGINFO-NEXT:    call void @llvm.dbg.value(metadata i32 [[C]]
 ; DBGINFO-NEXT:    call void @llvm.dbg.value(metadata i32 [[D]]
 
   %A = trunc i32 %x to i8
   %B = trunc i32 %y to i8
   %C = mul i8 %A, %B

 
 exit:
   unreachable
 }
 
 ; Check that we don't drop debug info when a zext is removed.
 define i1 @foo(i1 zeroext %b) {
 ; DBGINFO-LABEL: @foo(
-; DBGINFO-NEXT:  call void @llvm.dbg.value(metadata i1 %b
+; DBGINFO-NEXT:  call void @llvm.dbg.value(metadata i1 %b, {{.*}} !DIExpression(DW_OP_LLVM_convert, 1, DW_ATE_unsigned, DW_OP_LLVM_convert, 8, DW_ATE_unsigned, DW_OP_stack_value))
 ; DBGINFO-NEXT:  ret i1 %b
 
   %frombool = zext i1 %b to i8 
   ret i1 %b
 }