Segmentation fault during optimization pass with aarch64 simd

[calebzulawski]

The following code seg faults during optimization. I've produced it on both macOS and Linux.

Discussions:
https://rust-lang.zulipchat.com/#narrow/stream/257879-project-portable-simd/topic/2021-03-29.20Meeting
https://rust-lang.zulipchat.com/#narrow/stream/257879-project-portable-simd/topic/Aarch64.20LLVM.20bug
https://rust-lang.zulipchat.com/#narrow/stream/187780-t-compiler.2Fwg-llvm/topic/Debugging.20segfault.20in.20LLVM

Code

#![feature(const_generics, platform_intrinsics, repr_simd)]
#![allow(incomplete_features)]

extern "platform-intrinsic" {
    fn simd_eq<T, U>(a: T, b: T) -> U;
    fn simd_and<T>(a: T, b: T) -> T;
}

#[derive(Copy, Clone)]
#[repr(simd)]
pub struct F<const N: usize>([f32; N]);

impl<const N: usize> F<N> {
    pub fn splat(value: f32) -> Self {
        Self([value; N])
    }

    pub fn eq(self, other: Self) -> I<N> {
        unsafe { simd_eq(self, other) }
    }
}

#[repr(simd)]
pub struct I<const N: usize>([i32; N]);

impl<const N: usize> core::ops::BitAnd for I<N> {
    type Output = Self;
    fn bitand(self, other: Self) -> Self {
        unsafe { simd_and(self, other) }
    }
}

impl<const N: usize> I<N> {
    pub fn to_array(self) -> [bool; N] {
        let mut array = [false; N];
        let mut i = 0;
        while i < N {
            array[i] = self.0[i] == -1;
            i += 1;
        }
        array
    }
}

#[cfg(test)]
mod test {
    use super::*;
    #[test]
    fn bug() {
        let a = F::<64>::splat(0.);
        let b = F::<64>::splat(0.);
        let c = F::<64>::splat(0.);
        let m = a.eq(b) & a.eq(c);
        assert!(m.to_array()[0]);
    }
}

Meta

rustc --version --verbose:

rustc 1.53.0-nightly (138fd56cf 2021-04-02)
binary: rustc
commit-hash: 138fd56cf9598b4bf016634c768dca128a83a5d7
commit-date: 2021-04-02
host: x86_64-apple-darwin
release: 1.53.0-nightly
LLVM version: 12.0.0

Error output

   Compiling bug v0.1.0 (/Users/caleb/repos/bug)
error: could not compile `bug`

Caused by:
  process didn't exit successfully: `rustc --crate-name bug --edition=2018 src/lib.rs --error-format=json --json=diagnostic-rendered-ansi --emit=dep-info,link -C opt-level=3 -C embed-bitcode=no --test -C metadata=1974b6bb1d213572 -C extra-filename=-1974b6bb1d213572 --out-dir /Users/caleb/repos/bug/target/aarch64-apple-darwin/release/deps --target aarch64-apple-darwin -L dependency=/Users/caleb/repos/bug/target/aarch64-apple-darwin/release/deps -L dependency=/Users/caleb/repos/bug/target/release/deps` (signal: 11, SIGSEGV: invalid memory reference)

Backtrace

Backtrace from https://rust-lang.zulipchat.com/#narrow/stream/257879-project-portable-simd/topic/2021-03-29.20Meeting/near/232369790

* thread #9, name = 'LTO f64_ops.4vffyy4v-cgu.7', stop reason = EXC_BAD_ACCESS (code=1, address=0x8)
    frame #0: 0x0000000100473f7c librustc_driver-6566963bd0589bc0.dylib`llvm::Type::getTypeID() const + 12
    frame #1: 0x0000000100474468 librustc_driver-6566963bd0589bc0.dylib`llvm::Type::isVectorTy() const + 32
    frame #2: 0x000000010047453c librustc_driver-6566963bd0589bc0.dylib`llvm::Type::getScalarType() const + 32
    frame #3: 0x00000001007e4b68 librustc_driver-6566963bd0589bc0.dylib`llvm::Type::isIntOrIntVectorTy() const + 24
    frame #4: 0x000000010311a8b0 librustc_driver-6566963bd0589bc0.dylib`llvm::EVT::isExtendedInteger() const + 116
    frame #5: 0x000000010048cdb8 librustc_driver-6566963bd0589bc0.dylib`llvm::EVT::isInteger() const + 64
    frame #6: 0x000000010278d710 librustc_driver-6566963bd0589bc0.dylib`llvm::SelectionDAG::getNode(unsigned int, llvm::SDLoc const&, llvm::EVT, llvm::SDValue, llvm::SDNodeFlags) + 5776
    frame #7: 0x0000000102778264 librustc_driver-6566963bd0589bc0.dylib`llvm::SelectionDAG::getNode(unsigned int, llvm::SDLoc const&, llvm::EVT, llvm::SDValue) + 156
    frame #8: 0x000000010231db08 librustc_driver-6566963bd0589bc0.dylib`performExtendCombine(llvm::SDNode*, llvm::TargetLowering::DAGCombinerInfo&, llvm::SelectionDAG&) + 832
    frame #9: 0x0000000102317e48 librustc_driver-6566963bd0589bc0.dylib`llvm::AArch64TargetLowering::PerformDAGCombine(llvm::SDNode*, llvm::TargetLowering::DAGCombinerInfo&) const + 1128
    frame #10: 0x0000000102541298 librustc_driver-6566963bd0589bc0.dylib`(anonymous namespace)::DAGCombiner::combine(llvm::SDNode*) + 364
    frame #11: 0x000000010253a094 librustc_driver-6566963bd0589bc0.dylib`(anonymous namespace)::DAGCombiner::Run(llvm::CombineLevel) + 820
    frame #12: 0x0000000102539ce0 librustc_driver-6566963bd0589bc0.dylib`llvm::SelectionDAG::Combine(llvm::CombineLevel, llvm::AAResults*, llvm::CodeGenOpt::Level) + 92
    frame #13: 0x00000001027df038 librustc_driver-6566963bd0589bc0.dylib`llvm::SelectionDAGISel::CodeGenAndEmitDAG() + 1084
    frame #14: 0x00000001027deb8c librustc_driver-6566963bd0589bc0.dylib`llvm::SelectionDAGISel::SelectBasicBlock(llvm::ilist_iterator<llvm::ilist_detail::node_options<llvm::Instruction, true, false, void>, false, true>, llvm::ilist_iterator<llvm::ilist_detail::node_options<llvm::Instruction, true, false, void>, false, true>, bool&) + 352
    frame #15: 0x00000001027de010 librustc_driver-6566963bd0589bc0.dylib`llvm::SelectionDAGISel::SelectAllBasicBlocks(llvm::Function const&) + 4444
    frame #16: 0x00000001027dbb98 librustc_driver-6566963bd0589bc0.dylib`llvm::SelectionDAGISel::runOnMachineFunction(llvm::MachineFunction&) + 1568
    frame #17: 0x000000010229e0e8 librustc_driver-6566963bd0589bc0.dylib`(anonymous namespace)::AArch64DAGToDAGISel::runOnMachineFunction(llvm::MachineFunction&) + 56
    frame #18: 0x0000000102d18d00 librustc_driver-6566963bd0589bc0.dylib`llvm::MachineFunctionPass::runOnFunction(llvm::Function&) + 460
    frame #19: 0x000000010426b76c librustc_driver-6566963bd0589bc0.dylib`llvm::FPPassManager::runOnFunction(llvm::Function&) + 548
    frame #20: 0x0000000104272588 librustc_driver-6566963bd0589bc0.dylib`llvm::FPPassManager::runOnModule(llvm::Module&) + 116
    frame #21: 0x000000010426bfd8 librustc_driver-6566963bd0589bc0.dylib`(anonymous namespace)::MPPassManager::runOnModule(llvm::Module&) + 688
    frame #22: 0x000000010426bb70 librustc_driver-6566963bd0589bc0.dylib`llvm::legacy::PassManagerImpl::run(llvm::Module&) + 272
    frame #23: 0x00000001042729a4 librustc_driver-6566963bd0589bc0.dylib`llvm::legacy::PassManager::run(llvm::Module&) + 36
    frame #24: 0x00000001003f9a58 librustc_driver-6566963bd0589bc0.dylib`LLVMRustWriteOutputFile + 696
    frame #25: 0x00000001003465a4 librustc_driver-6566963bd0589bc0.dylib`rustc_codegen_llvm::back::write::write_output_file::hc09c4bfb3a246709 + 356
  * frame #26: 0x0000000100344f60 librustc_driver-6566963bd0589bc0.dylib`rustc_codegen_llvm::back::write::codegen::with_codegen::h9282ea3405df7f19 + 140
    frame #27: 0x000000010034964c librustc_driver-6566963bd0589bc0.dylib`rustc_codegen_llvm::back::write::codegen::h18b6d47a71c14803 + 2616
    frame #28: 0x0000000100384f44 librustc_driver-6566963bd0589bc0.dylib`rustc_codegen_ssa::back::write::finish_intra_module_work::ha36a9bd9f4a0717d + 180
    frame #29: 0x00000001003802a8 librustc_driver-6566963bd0589bc0.dylib`rustc_codegen_ssa::back::write::execute_work_item::ha652a44c2011cc51 + 1228
    frame #30: 0x000000010038b80c librustc_driver-6566963bd0589bc0.dylib`std::sys_common::backtrace::__rust_begin_short_backtrace::he5bf95f074532528 + 180
    frame #31: 0x00000001003289e4 librustc_driver-6566963bd0589bc0.dylib`std::panicking::try::hc21045c482577b07 + 48
    frame #32: 0x000000010038e33c librustc_driver-6566963bd0589bc0.dylib`core::ops::function::FnOnce::call_once$u7b$$u7b$vtable.shim$u7d$$u7d$::hd4c96fd994a769dd + 144
    frame #33: 0x0000000115c4ee34 libstd-405150871270d457.dylib`_$LT$alloc..boxed..Box$LT$F$C$A$GT$$u20$as$u20$core..ops..function..FnOnce$LT$Args$GT$$GT$::call_once::hc929de8808b588b8 [inlined] _$LT$alloc..boxed..Box$LT$F$C$A$GT$$u20$as$u20$core..ops..function..FnOnce$LT$Args$GT$$GT$::call_once::h2b5dd4f23298a6af at boxed.rs:1546:9 [opt]
    frame #34: 0x0000000115c4ee28 libstd-405150871270d457.dylib`_$LT$alloc..boxed..Box$LT$F$C$A$GT$$u20$as$u20$core..ops..function..FnOnce$LT$Args$GT$$GT$::call_once::hc929de8808b588b8 at boxed.rs:1546 [opt]
    frame #35: 0x0000000115c5073c libstd-405150871270d457.dylib`std::sys::unix::thread::Thread::new::thread_start::hd74574d62a5c9da7 at thread.rs:71:17 [opt]

[calebzulawski]

@nagisa filed https://bugs.llvm.org/show_bug.cgi?id=49867 with the reduced LLVM IR from rustc, but I was only able to reproduce the segfault from rust with SIMD intrinsics with 64-length vectors, so I wonder if we have two separate bugs in the optimizer.

[nagisa]

Yeah, those are two distinct issues. The reproduction steps provided in the thread here resulted in a different crash in the first place.

I was not able to reproduce this. In order for that to happen the reproducer probably should be made much more self contained – no cargo, no --test machinery, just a single rustc invocation (which may use libstd, but ideally doesn't)

[calebzulawski]

Unfortunately I was not able to reproduce it without cargo test. I'm sure there's a way, but anything I wrote that looked like this example but outside of a test function did not crash. Perhaps tests have some subtle difference from say a simple executable that affects codegen (or maybe linking/LTO) in some way?