[llvm-bugs] [Bug 31672] Assertion failed: ((TLI.getTypeAction(*DAG.getContext(), Op.getValueType()) == TargetLowering::TypeLegal || TLI.isTypeLegal(Op.getValueType()) || Op.getOpcode() == ISD::TargetC
https://llvm.org/bugs/show_bug.cgi?id=31672 Sanjay Patel changed: What|Removed |Added Status|NEW |RESOLVED Resolution|--- |FIXED --- Comment #6 from Sanjay Patel --- Should be fixed with: https://reviews.llvm.org/rL292758 -- You are receiving this mail because: You are on the CC list for the bug. ___ llvm-bugs mailing list llvm-bugs@lists.llvm.org http://lists.llvm.org/cgi-bin/mailman/listinfo/llvm-bugs
[llvm-bugs] [Bug 31718] New: clang crashes on valid code at -O2 on x86_64-linux-gnu while running pass 'Induction Variable Simplification': Assertion `L->isRecursivelyLCSSAForm(*DT, *LI) && "LCSSA req
https://llvm.org/bugs/show_bug.cgi?id=31718
Bug ID: 31718
Summary: clang crashes on valid code at -O2 on x86_64-linux-gnu
while running pass 'Induction Variable
Simplification': Assertion
`L->isRecursivelyLCSSAForm(*DT, *LI) && "LCSSA
required to run indvars!"' failed
Product: clang
Version: unspecified
Hardware: PC
OS: All
Status: NEW
Severity: normal
Priority: P
Component: -New Bugs
Assignee: unassignedclangb...@nondot.org
Reporter: s...@cs.ucdavis.edu
CC: llvm-bugs@lists.llvm.org
Classification: Unclassified
$ clang -v
clang version 5.0.0 (trunk 292714)
Target: x86_64-unknown-linux-gnu
Thread model: posix
InstalledDir: /usr/local/clang-trunk/bin
Found candidate GCC installation: /usr/lib/gcc/x86_64-linux-gnu/4.4
Found candidate GCC installation: /usr/lib/gcc/x86_64-linux-gnu/4.4.7
Found candidate GCC installation: /usr/lib/gcc/x86_64-linux-gnu/4.6
Found candidate GCC installation: /usr/lib/gcc/x86_64-linux-gnu/4.6.4
Found candidate GCC installation: /usr/lib/gcc/x86_64-linux-gnu/4.7
Found candidate GCC installation: /usr/lib/gcc/x86_64-linux-gnu/4.7.4
Found candidate GCC installation: /usr/lib/gcc/x86_64-linux-gnu/4.8
Found candidate GCC installation: /usr/lib/gcc/x86_64-linux-gnu/4.8.5
Found candidate GCC installation: /usr/lib/gcc/x86_64-linux-gnu/4.9
Found candidate GCC installation: /usr/lib/gcc/x86_64-linux-gnu/4.9.3
Found candidate GCC installation: /usr/lib/gcc/x86_64-linux-gnu/5
Found candidate GCC installation: /usr/lib/gcc/x86_64-linux-gnu/5.4.0
Found candidate GCC installation: /usr/lib/gcc/x86_64-linux-gnu/6
Found candidate GCC installation: /usr/lib/gcc/x86_64-linux-gnu/6.0.0
Selected GCC installation: /usr/lib/gcc/x86_64-linux-gnu/5.4.0
Candidate multilib: .;@m64
Candidate multilib: 32;@m32
Candidate multilib: x32;@mx32
Selected multilib: .;@m64
$
$ clang -Os small.c
$
$ clang -O2 small.c
clang-5.0:
/tmp/llvm-builder/llvm-source-trunk/lib/Transforms/Scalar/IndVarSimplify.cpp:2369:
bool {anonymous}::IndVarSimplify::run(llvm::Loop*): Assertion
`L->isRecursivelyLCSSAForm(*DT, *LI) && "LCSSA required to run indvars!"'
failed.
#0 0x01e1a2d8 llvm::sys::PrintStackTrace(llvm::raw_ostream&)
(/usr/local/clang-trunk/bin/clang-5.0+0x1e1a2d8)
#1 0x01e17f6e llvm::sys::RunSignalHandlers()
(/usr/local/clang-trunk/bin/clang-5.0+0x1e17f6e)
#2 0x01e180e2 SignalHandler(int)
(/usr/local/clang-trunk/bin/clang-5.0+0x1e180e2)
#3 0x7f45a22a5390 __restore_rt
(/lib/x86_64-linux-gnu/libpthread.so.0+0x11390)
#4 0x7f45a1018428 gsignal
/build/glibc-t3gR2i/glibc-2.23/signal/../sysdeps/unix/sysv/linux/raise.c:54:0
#5 0x7f45a101a02a abort /build/glibc-t3gR2i/glibc-2.23/stdlib/abort.c:91:0
#6 0x7f45a1010bd7 __assert_fail_base
/build/glibc-t3gR2i/glibc-2.23/assert/assert.c:92:0
#7 0x7f45a1010c82 (/lib/x86_64-linux-gnu/libc.so.6+0x2dc82)
#8 0x01c8447b (anonymous
namespace)::IndVarSimplifyLegacyPass::runOnLoop(llvm::Loop*,
llvm::LPPassManager&) [clone .part.440] [clone .constprop.452]
(/usr/local/clang-trunk/bin/clang-5.0+0x1c8447b)
#9 0x0281930b llvm::LPPassManager::runOnFunction(llvm::Function&)
(/usr/local/clang-trunk/bin/clang-5.0+0x281930b)
#10 0x019f3173 llvm::FPPassManager::runOnFunction(llvm::Function&)
(/usr/local/clang-trunk/bin/clang-5.0+0x19f3173)
#11 0x027dbbcf (anonymous
namespace)::CGPassManager::runOnModule(llvm::Module&)
(/usr/local/clang-trunk/bin/clang-5.0+0x27dbbcf)
#12 0x019f2cdd llvm::legacy::PassManagerImpl::run(llvm::Module&)
(/usr/local/clang-trunk/bin/clang-5.0+0x19f2cdd)
#13 0x01fbc8c2 clang::EmitBackendOutput(clang::DiagnosticsEngine&,
clang::HeaderSearchOptions const&, clang::CodeGenOptions const&,
clang::TargetOptions const&, clang::LangOptions const&, llvm::DataLayout
const&, llvm::Module*, clang::BackendAction,
std::unique_ptr >)
(/usr/local/clang-trunk/bin/clang-5.0+0x1fbc8c2)
#14 0x02625d00
clang::BackendConsumer::HandleTranslationUnit(clang::ASTContext&)
(/usr/local/clang-trunk/bin/clang-5.0+0x2625d00)
#15 0x02a0d4d8 clang::ParseAST(clang::Sema&, bool, bool)
(/usr/local/clang-trunk/bin/clang-5.0+0x2a0d4d8)
#16 0x02620739 clang::CodeGenAction::ExecuteAction()
(/usr/local/clang-trunk/bin/clang-5.0+0x2620739)
#17 0x02317f16 clang::FrontendAction::Execute()
(/usr/local/clang-trunk/bin/clang-5.0+0x2317f16)
#18 0x022ea856
clang::CompilerInstance::ExecuteAction(clang::FrontendAction&)
(/usr/local/clang-trunk/bin/clang-5.0+0x22ea856)
#19 0x0239f63a
clang::ExecuteCompilerInvocation(clang::CompilerInstance*)
(/usr/local/clang-trunk/bin/clang-5.0+0x239f63a)
#20 0x00a49ac8 cc1_main(llvm::ArrayRef, char const*,
void*) (/usr/local/clang-trunk/bin/clang-5.0+0xa49ac8)
#21 0x009dbd9c main (/usr/local/clang-trunk/bin/clang-5.0+
[llvm-bugs] [Bug 31719] New: LLVM maintain useless state when using adc
https://llvm.org/bugs/show_bug.cgi?id=31719
Bug ID: 31719
Summary: LLVM maintain useless state when using adc
Product: libraries
Version: trunk
Hardware: PC
OS: Windows NT
Status: NEW
Severity: normal
Priority: P
Component: Backend: X86
Assignee: unassignedb...@nondot.org
Reporter: deadal...@gmail.com
CC: llvm-bugs@lists.llvm.org
Classification: Unclassified
Sample IR (optimized):
; Function Attrs: norecurse nounwind readonly
define %scalar @foo(%scalar* nocapture readonly %this, %scalar %arg.b)
local_unnamed_addr #1 {
entry:
%0 = extractvalue %scalar %arg.b, 0
%.elt = extractvalue [4 x i64] %0, 0
%.elt24 = extractvalue [4 x i64] %0, 1
%.elt26 = extractvalue [4 x i64] %0, 2
%.elt28 = extractvalue [4 x i64] %0, 3
%1 = getelementptr inbounds %scalar , %scalar* %this, i64 0, i32 0, i64 0
%2 = load i64, i64* %1, align 8
%3 = zext i64 %2 to i128
%4 = zext i64 %.elt to i128
%5 = add nuw nsw i128 %3, %4
%6 = trunc i128 %5 to i64
%7 = lshr i128 %5, 64
%8 = getelementptr inbounds %scalar , %scalar * %this, i64 0, i32 0, i64 1
%9 = load i64, i64* %8, align 8
%10 = zext i64 %9 to i128
%11 = zext i64 %.elt24 to i128
%12 = add nuw nsw i128 %10, %11
%13 = add nuw nsw i128 %12, %7
%14 = trunc i128 %13 to i64
%15 = lshr i128 %13, 64
%16 = getelementptr inbounds %scalar , %scalar* %this, i64 0, i32 0, i64 2
%17 = load i64, i64* %16, align 8
%18 = zext i64 %17 to i128
%19 = zext i64 %.elt26 to i128
%20 = add nuw nsw i128 %18, %19
%21 = add nuw nsw i128 %20, %15
%22 = trunc i128 %21 to i64
%23 = lshr i128 %21, 64
%24 = getelementptr inbounds %scalar , %scalar* %this, i64 0, i32 0, i64 3
%25 = load i64, i64* %24, align 8
%26 = zext i64 %25 to i128
%27 = zext i64 %.elt28 to i128
%28 = add nuw nsw i128 %26, %27
%29 = add nuw nsw i128 %28, %23
%30 = trunc i128 %29 to i64
%31 = insertvalue [4 x i64] undef, i64 %6, 0
%32 = insertvalue [4 x i64] %31, i64 %14, 1
%33 = insertvalue [4 x i64] %32, i64 %22, 2
%34 = insertvalue [4 x i64] %33, i64 %30, 3
%35 = insertvalue %S6crypto5field6Scalar undef, [4 x i64] %34, 0
ret %scalar%35
}
attributes #0 = { norecurse nounwind readnone }
attributes #1 = { norecurse nounwind readonly }
Codegen:
foo:
addq(%rsi), %rdx
sbbq%r10, %r10
andl$1, %r10d
addq8(%rsi), %rcx
sbbq%r11, %r11
andl$1, %r11d
addq%r10, %rcx
adcq$0, %r11
addq16(%rsi), %r8
sbbq%rax, %rax
andl$1, %eax
addq%r11, %r8
adcq$0, %rax
addq24(%rsi), %r9
addq%rax, %r9
movq%rdx, (%rdi)
movq%rcx, 8(%rdi)
movq%r8, 16(%rdi)
movq%r9, 24(%rdi)
movq%rdi, %rax
retq
While LLVM is able to leverage the use of the adc instruction (good) it is
unclear why it is doing so in RAX and then adding RAX rather than using ADC
right away. See for instance:
adcq$0, %rax
addq24(%rsi), %r9
addq%rax, %r9
Uses adc to store the carry in RAX and then perform 2 additions, when it could
simply do
adcq24(%rsi), %r9
These routine are at the core of various crypto libraries and need to be fast.
Any chance to get better codegen here ?
--
You are receiving this mail because:
You are on the CC list for the bug.
___
llvm-bugs mailing list
llvm-bugs@lists.llvm.org
http://lists.llvm.org/cgi-bin/mailman/listinfo/llvm-bugs
