[llvm-bugs] [Bug 25422] New: Incorrect liveness in DeadStoreElimination

[via llvm-bugs]

https://llvm.org/bugs/show_bug.cgi?id=25422

            Bug ID: 25422
           Summary: Incorrect liveness in DeadStoreElimination
           Product: libraries
           Version: trunk
          Hardware: All
                OS: All
            Status: NEW
          Severity: normal
          Priority: P
         Component: Scalar Optimizations
          Assignee: unassignedbugs at nondot.org
          Reporter: dan433584 at gmail.com
                CC: llvm-bugs at lists.llvm.org
    Classification: Unclassified

In this code:

declare void @foo()

@g = external global i8

define void @f() {
  store i8 1, i8* @g
  call void @foo() readnone
  store i8 0, i8* @g
  ret void
}

Dead-store elimination eliminates the first store, however this is incorrect if
the call to @foo can unwind, such that callers of @f can observe the value of
@g after the first store and before the second, even if @foo itself can't.

MemoryDependenceAnalysis currently says that the second store here has a direct
local dependence on the first store. There's a sense in which that is a correct
answer, but DeadStoreElimination inteprets this answer to mean that the first
store is dead. It's not immediately obvious which of these two should be fixed.

This is a reduced form of the issue reported in Rust here:
https://github.com/rust-lang/rust/issues/29485

-- 
You are receiving this mail because:
You are on the CC list for the bug.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.llvm.org/pipermail/llvm-bugs/attachments/20151105/9c58c8a4/attachment.html>