forked from OSchip/llvm-project
parent
ee6a4fc680
commit
fb619e9686
|
@ -291,7 +291,7 @@ bool WebAssemblyTargetLowering::allowsMisalignedMemoryAccesses(
|
|||
// WebAssembly supports unaligned accesses, though it should be declared
|
||||
// with the p2align attribute on loads and stores which do so, and there
|
||||
// may be a performance impact. We tell LLVM they're "fast" because
|
||||
// for the kinds of things that LLVM uses this for (merging agacent stores
|
||||
// for the kinds of things that LLVM uses this for (merging adjacent stores
|
||||
// of constants, etc.), WebAssembly implementations will either want the
|
||||
// unaligned access or they'll split anyway.
|
||||
if (Fast)
|
||||
|
|
Loading…
Reference in New Issue