Assertion failure in VPlanTransforms.cpp

[ilovepi]

We're seeing an assertion failure trip when building Fuchsia w/ coverage instrumentation.

Here is a reproducer
llvm-profdata-a2a90a.zip

I'm working on reducing it, and @gulfemsavrun has started a bisect.

The toolchain we're qualifying is from Oct 2 (commit b147019).

clang++: /usr/local/google/home/paulkirth/llvm-project/llvm/lib/Transforms/Vectorize/VPlanTransforms.cpp:3917: static DenseMap<const SCEV *, Value *> llvm::VPlanTransforms::expandSCEVs(VPlan &, ScalarEvolution &): Assertion `none_of(*Entry, IsaPred<VPExpandSCEVRecipe>) && "VPExpandSCEVRecipes must be at the beginning of the entry block, " "after any VPIRInstructions"' failed.
PLEASE submit a bug report to https://github.com/llvm/llvm-project/issues/ and include the crash backtrace, preprocessed source, and associated run script.
Stack dump:
0.      Program arguments: /usr/local/google/home/paulkirth/llvm-project/build/bin/clang++ -cc1 -triple x86_64-unknown-windows-msvc19.33.0 -O2 -emit-obj -massembler-fatal-warnings -disable-free -clear-ast-before-backend -disable-llvm-verifier -discard-value-names -main-file-name llvm-profdata.cc -mrelocation-model pic -pic-level 2 -mframe-pointer=non-leaf -relaxed-aliasing -fmath-errno -ffp-contract=on -fno-rounding-math -mconstructor-aliases -fms-volatile -ffreestanding -D_HAS_STATIC_RTTI=0 -target-cpu x86-64-v2 -mllvm -x86-branches-within-32B-boundaries -tune-cpu generic -gno-column-info -gcodeview -debug-info-kind=constructor -fdebug-compilation-dir=. -object-file-name=kernel.efi_x64-coverage\\obj\\src\\lib\\llvm-profdata\\libllvm-profdata.llvm-profdata.cc.o -mllvm -crash-diagnostics-dir=clang-crashreports -ffunction-sections -fdata-sections -fcoverage-compilation-dir=. -sys-header-deps -D _LIBCPP_DISABLE_VISIBILITY_ANNOTATIONS -D _LIBCPP_ENABLE_THREAD_SAFETY_ANNOTATIONS=1 -D TOOLCHAIN_VERSION=mSdw4dJP68zDVUDZeMkK7B6_Qa4ra24Ty4wlu7xGETgC -D __Fuchsia_API_level__=4293918720 -D ZX_ASSERT_LEVEL=2 -D __PE_COFF__ -D WITH_FRAME_POINTERS=1 -D HAVE_LLVM_PROFDATA=1 -U _MSC_VER -U _MSC_FULL_VER -Wall -Wextra -Wconversion -Wextra-semi -Wimplicit-fallthrough -Wnewline-eof -Wstrict-prototypes -Wwrite-strings -Wno-sign-conversion -Wno-unused-parameter -Wnonportable-system-include-path -Wno-missing-field-initializers -Wno-extra-qualification -Wno-cast-function-type-mismatch -Wno-unknown-warning-option -Wno-missing-template-arg-list-after-template-kw -Wno-deprecated-pragma -Wno-nontrivial-memaccess -Wthread-safety -Wno-unknown-warning-option -Wno-thread-safety-reference-return -Werror -Wno-builtin-macro-redefined -Wformat=2 -Wmissing-declarations -Wvla -Wshadow -Wno-deprecated-this-capture -std=c++20 -fdeprecated-macro -fstack-size-section -ferror-limit 19 -ftemplate-backtrace-limit=0 -fvisibility=hidden -fvisibility-inlines-hidden -ftrivial-auto-var-init=pattern -fno-rtti -fno-use-cxa-atexit -fms-extensions -fms-compatibility -fms-compatibility-version=19.33 -fno-threadsafe-statics -fno-implicit-modules -fskip-odr-check-in-gmf -faligned-allocation -fnew-alignment=8 -fcolor-diagnostics -vectorize-loops -vectorize-slp -faddrsig -x c++ llvm-profdata-a2a90a.cpp
1.      <eof> parser at end of file
2.      Optimizer
3.      Running pass "function<eager-inv>(drop-unnecessary-assumes,float2int,lower-constant-intrinsics,loop(loop-rotate<header-duplication;no-prepare-for-lto>,loop-deletion),loop-distribute,inject-tli-mappings,loop-vectorize<no-interleave-forced-only;no-vectorize-forced-only;>,infer-alignment,loop-load-elim,instcombine<max-iterations=1;no-verify-fixpoint>,simplifycfg<bonus-inst-threshold=1;forward-switch-cond;switch-range-to-icmp;switch-to-lookup;no-keep-loops;hoist-common-insts;no-hoist-loads-stores-with-cond-faulting;sink-common-insts;speculate-blocks;simplify-cond-branch;no-speculate-unpredictables>,slp-vectorizer,vector-combine,instcombine<max-iterations=1;no-verify-fixpoint>,loop-unroll<O2>,transform-warning,sroa<preserve-cfg>,infer-alignment,instcombine<max-iterations=1;no-verify-fixpoint>,loop-mssa(licm<allowspeculation>),alignment-from-assumptions,loop-sink,instsimplify,div-rem-pairs,tailcallelim,simplifycfg<bonus-inst-threshold=1;no-forward-switch-cond;switch-range-to-icmp;no-switch-to-lookup;keep-loops;no-hoist-common-insts;hoist-loads-stores-with-cond-faulting;no-sink-common-insts;speculate-blocks;simplify-cond-branch;speculate-unpredictables>)" on module "llvm-profdata-a2a90a.cpp"
4.      Running pass "loop-vectorize<no-interleave-forced-only;no-vectorize-forced-only;>" on function "?MergeLiveData@LlvmProfdata@@QEAAXULiveData@1@@Z"
#0 0x00005634a0128b28 llvm::sys::PrintStackTrace(llvm::raw_ostream&, int) /usr/local/google/home/paulkirth/llvm-project/llvm/lib/Support/Unix/Signals.inc:834:13
llvm-profdata-a2a90a.sh: line 4: 3184705 Aborted                 (core dumped) "$HOME/llvm-project/build/bin/clang++" "-cc1" "-triple" "x86_64-unknown-windows-msvc19.33.0" "-O2" "-emit-obj" "-massembler-fatal-warnings" "-disable-free" "-clear-ast-before-backend" "-disable-llvm-verifier" "-discard-value-names" "-main-file-name" "llvm-profdata.cc" "-mrelocation-model" "pic" "-pic-level" "2" "-mframe-pointer=non-leaf" "-relaxed-aliasing" "-fmath-errno" "-ffp-contract=on" "-fno-rounding-math" "-mconstructor-aliases" "-fms-volatile" "-ffreestanding" "-D_HAS_STATIC_RTTI=0" "-target-cpu" "x86-64-v2" "-mllvm" "-x86-branches-within-32B-boundaries" "-tune-cpu" "generic" "-gno-column-info" "-gcodeview" "-debug-info-kind=constructor" "-fdebug-compilation-dir=." "-object-file-name=kernel.efi_x64-coverage\\obj\\src\\lib\\llvm-profdata\\libllvm-profdata.llvm-profdata.cc.o" "-mllvm" "-crash-diagnostics-dir=clang-crashreports" "-ffunction-sections" "-fdata-sections" "-fcoverage-compilation-dir=." "-sys-header-deps" "-D" "_LIBCPP_DISABLE_VISIBILITY_ANNOTATIONS" "-D" "_LIBCPP_ENABLE_THREAD_SAFETY_ANNOTATIONS=1" "-D" "TOOLCHAIN_VERSION=mSdw4dJP68zDVUDZeMkK7B6_Qa4ra24Ty4wlu7xGETgC" "-D" "__Fuchsia_API_level__=4293918720" "-D" "ZX_ASSERT_LEVEL=2" "-D" "__PE_COFF__" "-D" "WITH_FRAME_POINTERS=1" "-D" "HAVE_LLVM_PROFDATA=1" "-U" "_MSC_VER" "-U" "_MSC_FULL_VER" "-Wall" "-Wextra" "-Wconversion" "-Wextra-semi" "-Wimplicit-fallthrough" "-Wnewline-eof" "-Wstrict-prototypes" "-Wwrite-strings" "-Wno-sign-conversion" "-Wno-unused-parameter" "-Wnonportable-system-include-path" "-Wno-missing-field-initializers" "-Wno-extra-qualification" "-Wno-cast-function-type-mismatch" "-Wno-unknown-warning-option" "-Wno-missing-template-arg-list-after-template-kw" "-Wno-deprecated-pragma" "-Wno-nontrivial-memaccess" "-Wthread-safety" "-Wno-unknown-warning-option" "-Wno-thread-safety-reference-return" "-Werror" "-Wno-builtin-macro-redefined" "-Wformat=2" "-Wmissing-declarations" "-Wvla" "-Wshadow" "-Wno-deprecated-this-capture" "-std=c++20" "-fdeprecated-macro" "-fstack-size-section" "-ferror-limit" "19" "-ftemplate-backtrace-limit=0" "-fvisibility=hidden" "-fvisibility-inlines-hidden" "-ftrivial-auto-var-init=pattern" "-fno-rtti" "-fno-use-cxa-atexit" "-fms-extensions" "-fms-compatibility" "-fms-compatibility-version=19.33" "-fno-threadsafe-statics" "-fno-implicit-modules" "-fskip-odr-check-in-gmf" "-faligned-allocation" "-fnew-alignment=8" "-fcolor-diagnostics" "-vectorize-loops" "-vectorize-slp" "-faddrsig" "-x" "c++" "llvm-profdata-a2a90a.cpp"

[ilovepi]

bad.ll

; *** IR Dump Before LoopVectorizePass on ?MergeLiveData@LlvmProfdata@@QEAAXULiveData@1@@Z ***
; ModuleID = 'reduced.ll'
source_filename = "reduced.ll"
target datalayout = "e-m:w-p270:32:32-p271:32:32-p272:64:64-i64:64-i128:128-f80:128-n8:16:32:64-S128"
target triple = "x86_64-unknown-windows-msvc19.33.0"

@"?CountersBegin_placeholder@?A0xE34CF3CB@@3PADA" = external global [1 x i8]

; Function Attrs: mustprogress nofree norecurse nosync nounwind willreturn memory(write, inaccessiblemem: none)
define noalias noundef ptr @"?ProfCountersData@?A0xE34CF3CB@@YA?AV?$span@D$0?0@__ktl@std@@XZ"() local_unnamed_addr #0 {
  store i64 sub (i64 0, i64 ptrtoint (ptr getelementptr inbounds nuw (i8, ptr @"?CountersBegin_placeholder@?A0xE34CF3CB@@3PADA", i64 1) to i64)), ptr inttoptr (i64 8 to ptr), align 8
  ret ptr null
}

; Function Attrs: nofree norecurse nosync nounwind memory(write, inaccessiblemem: none)
define void @"?MergeLiveData@LlvmProfdata@@QEAAXULiveData@1@@Z"() local_unnamed_addr #1 {
  store i64 sub (i64 0, i64 ptrtoint (ptr getelementptr inbounds nuw (i8, ptr @"?CountersBegin_placeholder@?A0xE34CF3CB@@3PADA", i64 1) to i64)), ptr inttoptr (i64 8 to ptr), align 8
  %.not.i = icmp eq i64 sub (i64 0, i64 ptrtoint (ptr getelementptr inbounds nuw (i8, ptr @"?CountersBegin_placeholder@?A0xE34CF3CB@@3PADA", i64 1) to i64)), 0
  br i1 %.not.i, label %"??$MergeData@EUlogical_and@__ktl@std@@@?A0xE34CF3CB@@YAXV?$span@W4byte@std@@$0?0@__ktl@std@@V?$span@$$CBW4byte@std@@$0?0@12@@Z.exit", label %.lr.ph.i.preheader

.lr.ph.i.preheader:                               ; preds = %0
  br label %.lr.ph.i

.lr.ph.i:                                         ; preds = %.lr.ph.i.preheader, %.lr.ph.i
  %.01.i = phi i64 [ %2, %.lr.ph.i ], [ 0, %.lr.ph.i.preheader ]
  %1 = getelementptr i8, ptr null, i64 %.01.i
  store i8 poison, ptr %1, align 1
  %2 = add nuw i64 %.01.i, 1
  %exitcond.not.i = icmp eq i64 %2, sub (i64 0, i64 ptrtoint (ptr getelementptr inbounds nuw (i8, ptr @"?CountersBegin_placeholder@?A0xE34CF3CB@@3PADA", i64 1) to i64))
  br i1 %exitcond.not.i, label %"??$MergeData@EUlogical_and@__ktl@std@@@?A0xE34CF3CB@@YAXV?$span@W4byte@std@@$0?0@__ktl@std@@V?$span@$$CBW4byte@std@@$0?0@12@@Z.exit.loopexit", label %.lr.ph.i

"??$MergeData@EUlogical_and@__ktl@std@@@?A0xE34CF3CB@@YAXV?$span@W4byte@std@@$0?0@__ktl@std@@V?$span@$$CBW4byte@std@@$0?0@12@@Z.exit.loopexit": ; preds = %.lr.ph.i
  br label %"??$MergeData@EUlogical_and@__ktl@std@@@?A0xE34CF3CB@@YAXV?$span@W4byte@std@@$0?0@__ktl@std@@V?$span@$$CBW4byte@std@@$0?0@12@@Z.exit"

"??$MergeData@EUlogical_and@__ktl@std@@@?A0xE34CF3CB@@YAXV?$span@W4byte@std@@$0?0@__ktl@std@@V?$span@$$CBW4byte@std@@$0?0@12@@Z.exit": ; preds = %"??$MergeData@EUlogical_and@__ktl@std@@@?A0xE34CF3CB@@YAXV?$span@W4byte@std@@$0?0@__ktl@std@@V?$span@$$CBW4byte@std@@$0?0@12@@Z.exit.loopexit", %0
  ret void
}

; Function Attrs: nofree norecurse nosync nounwind memory(readwrite, inaccessiblemem: none)
define void @"?MergeCounters@?A0xE34CF3CB@@YAXV?$span@W4byte@std@@$0?0@__ktl@std@@V?$span@$$CBW4byte@std@@$0?0@23@@Z"() local_unnamed_addr #2 {
  tail call void @"??$MergeData@EUlogical_and@__ktl@std@@@?A0xE34CF3CB@@YAXV?$span@W4byte@std@@$0?0@__ktl@std@@V?$span@$$CBW4byte@std@@$0?0@12@@Z"()
  ret void
}

; Function Attrs: mustprogress nofree norecurse nosync nounwind willreturn memory(argmem: write)
define noalias noundef ptr @"??$?0PEAD@?$span@D$0?0@__ktl@std@@QEAA@PEAD_K@Z"(ptr writeonly captures(none) initializes((8, 16)) %0, i64 %1) local_unnamed_addr #3 {
  %3 = getelementptr i8, ptr %0, i64 8
  store i64 %1, ptr %3, align 8
  ret ptr null
}

; Function Attrs: nofree norecurse nosync nounwind memory(readwrite, inaccessiblemem: none)
define void @"??$MergeData@EUlogical_and@__ktl@std@@@?A0xE34CF3CB@@YAXV?$span@W4byte@std@@$0?0@__ktl@std@@V?$span@$$CBW4byte@std@@$0?0@12@@Z"() local_unnamed_addr #4 {
  %1 = load i64, ptr inttoptr (i64 8 to ptr), align 8
  %.not = icmp eq i64 %1, 0
  br i1 %.not, label %._crit_edge, label %.lr.ph

._crit_edge:                                      ; preds = %.lr.ph, %0
  ret void

.lr.ph:                                           ; preds = %0, %.lr.ph
  %.01 = phi i64 [ %3, %.lr.ph ], [ 0, %0 ]
  %2 = getelementptr i8, ptr null, i64 %.01
  store i8 poison, ptr %2, align 1
  %3 = add nuw i64 %.01, 1
  %exitcond.not = icmp eq i64 %3, %1
  br i1 %exitcond.not, label %._crit_edge, label %.lr.ph
}

; Function Attrs: mustprogress nofree norecurse nosync nounwind willreturn memory(read, inaccessiblemem: none)
define i64 @"?size@?$span@E$0?0@__ktl@std@@QEBA_KXZ"() local_unnamed_addr #5 {
  %1 = load i64, ptr inttoptr (i64 8 to ptr), align 8
  ret i64 %1
}

; Function Attrs: mustprogress nofree norecurse nosync nounwind willreturn memory(none)
define noalias ptr @"??A?$span@E$0?0@__ktl@std@@QEBAAEAE_K@Z"(i64 %0) local_unnamed_addr #6 {
  %2 = getelementptr i8, ptr null, i64 %0
  ret ptr %2
}

attributes #0 = { mustprogress nofree norecurse nosync nounwind willreturn memory(write, inaccessiblemem: none) }
attributes #1 = { nofree norecurse nosync nounwind memory(write, inaccessiblemem: none) "no-builtins" }
attributes #2 = { nofree norecurse nosync nounwind memory(readwrite, inaccessiblemem: none) }
attributes #3 = { mustprogress nofree norecurse nosync nounwind willreturn memory(argmem: write) }
attributes #4 = { nofree norecurse nosync nounwind memory(readwrite, inaccessiblemem: none) "no-builtins" }
attributes #5 = { mustprogress nofree norecurse nosync nounwind willreturn memory(read, inaccessiblemem: none) }
attributes #6 = { mustprogress nofree norecurse nosync nounwind willreturn memory(none) }

opt bad.ll -passes="loop-vectorize<no-interleave-forced-only;no-vectorize-forced-only;>" -S

[ilovepi]

got it down a bit more

; RUN: opt %s -passes="loop-vectorize<no-interleave-forced-only;no-vectorize-forced-only;>" -S
target datalayout = "e-m:w-p270:32:32-p271:32:32-p272:64:64-i64:64-i128:128-f80:128-n8:16:32:64-S128"
target triple = "x86_64-unknown-windows-msvc19.33.0"

@"?CountersBegin_placeholder@?A0xE34CF3CB@@3PADA" = external global [1 x i8]

define void @"?MergeLiveData@LlvmProfdata@@QEAAXULiveData@1@@Z"() {
.lr.ph.i.preheader:
  br label %.lr.ph.i

.lr.ph.i:                                         ; preds = %.lr.ph.i, %.lr.ph.i.preheader
  %.01.i = phi i64 [ %1, %.lr.ph.i ], [ 0, %.lr.ph.i.preheader ]
  %0 = getelementptr i8, ptr null, i64 %.01.i
  store i8 0, ptr %0, align 1
  %1 = add i64 %.01.i, 1
  %exitcond.not.i = icmp eq i64 %.01.i, sub (i64 0, i64 ptrtoint (ptr getelementptr inbounds nuw (i8, ptr @"?CountersBegin_placeholder@?A0xE34CF3CB@@3PADA", i64 1) to i64))
  br i1 %exitcond.not.i, label %"??$MergeData@EUlogical_and@__ktl@std@@@?A0xE34CF3CB@@YAXV?$span@W4byte@std@@$0?0@__ktl@std@@V?$span@$$CBW4byte@std@@$0?0@12@@Z.exit.loopexit", label %.lr.ph.i

"??$MergeData@EUlogical_and@__ktl@std@@@?A0xE34CF3CB@@YAXV?$span@W4byte@std@@$0?0@__ktl@std@@V?$span@$$CBW4byte@std@@$0?0@12@@Z.exit.loopexit": ; preds = %.lr.ph.i
  ret void
}

[ilovepi]

// RUN: clang -cc1 -triple x86_64-unknown-windows-msvc19.33.0 -O2 -ffreestanding -std=c++20 -ftrivial-auto-var-init=pattern -fms-compatibility -vectorize-loops -emit-llvm %s
namespace std {
enum byte;
template <bool> using __enable_if_t = int;
template <bool __v> struct integral_constant {
  static const bool value = __v;
};
template <bool _Val> using _BoolConstant = integral_constant<_Val>;
template <class _Tp>
using __libcpp_remove_reference_t = __remove_reference_t(_Tp);
template <class _Tp>
using remove_reference_t = __libcpp_remove_reference_t<_Tp>;
template <class _Tp, class _Up>
using _IsSame = _BoolConstant<__is_same(_Tp, _Up)>;
template <class _Tp, class _Up>
concept __same_as_impl = _IsSame<_Tp, _Up>::value;
template <class _Tp, class _Up>
concept same_as = __same_as_impl<_Up, _Tp>;
template <bool, class _IfRes, class> using conditional_t = _IfRes;
template <class _Tp> using __remove_cvref_t = __remove_cvref(_Tp);
template <class _Tp> using remove_cvref_t = __remove_cvref_t<_Tp>;
template <class> class numeric_limits;
template <> struct numeric_limits<unsigned long long> {
  static constexpr long max() { return 0; }
};
constexpr size_t dynamic_extent = numeric_limits<size_t>::max();
template <typename, size_t = dynamic_extent> class span;
template <typename _Tp> struct span<_Tp> {
  using size_type = size_t;
  span subspan(size_type, size_type __count) { return {0, __count}; }
  size_type size() { return __size_; }
  size_type size_bytes() { return __size_; }
  _Tp &operator[](size_type __idx) { return __data_[__idx]; }
  _Tp *__data_;
  size_type __size_;
};
template <class, class _Rhs>
concept assignable_from = requires(_Rhs __rhs) { __rhs; };
template <class _Tp> constexpr bool is_object_v = assignable_from<_Tp, _Tp>;
template <template <class> class, class>
integral_constant<true> __sfinae_test_impl(int);
template <template <class> class _Templ, class... _Args>
using _IsValidExpansion = decltype(__sfinae_test_impl<_Templ, _Args...>(0));
template <class _Tp>
using __test_for_primary_template = __enable_if_t<_IsSame<_Tp, _Tp>::value>;
template <class _Tp>
using __is_primary_template =
    _IsValidExpansion<__test_for_primary_template, _Tp>;
template <class> struct __cond_value_type;
template <class _Tp>
  requires is_object_v<_Tp>
struct __cond_value_type<_Tp> {
  using value_type = _Tp;
};
template <class> struct indirectly_readable_traits;
template <class _Tp>
struct indirectly_readable_traits<_Tp *> : __cond_value_type<_Tp> {};
template <class _Tp>
concept __dereferenceable = requires(_Tp __t) { __t; };
template <__dereferenceable _Tp> using iter_reference_t = decltype(*_Tp());
template <class _Ip>
using iter_value_t =
    conditional_t<__is_primary_template<int>::value,
                  indirectly_readable_traits<_Ip>, int>::value_type;
template <class _Ip>
concept contiguous_iterator =
    same_as<iter_value_t<_Ip>, remove_cvref_t<iter_reference_t<_Ip>>>;
template <contiguous_iterator _It, class _EndOrSize>
span(_It, _EndOrSize) -> span<remove_reference_t<iter_reference_t<_It>>>;
} // namespace std
struct LiveData {
  std::span<std::byte> bitmap;
};
namespace std {
template <class> struct bit_or;
}
char BitmapBegin, BitmapBegin_0, BitmapEnd;
std::span<char> ProfBitmapData() {
  return std::span(&BitmapBegin_0, &BitmapEnd - &BitmapBegin);
}
template <typename T, template <typename> class>
void MergeData(std::span<std::byte> to) {
  std::span to_data{reinterpret_cast<T *>(0), to.size_bytes()};
  for (size_t i; i < to_data.size(); ++i)
    to_data[i] = 0;
}
template <typename T, template <typename> class Op, typename FromT>
void MergeSelfData(std::span<std::byte> to, std::span<FromT> from, char *) {
  MergeData<T, Op>(to.subspan(0, from.size_bytes()));
}
void MergeLiveData() {
  LiveData data;
  MergeSelfData<char, std::bit_or>(data.bitmap, ProfBitmapData(), "");
}

In case it helps, heres a reproducer from the frontend (e.g. incase there's a frontend change that caused bad IR to be emitted)

[ilovepi]

Also, this is still an issue at ToT (8889377 from today).