[WebAssembly] Enable atomic expansion for unsupported atomicrmws
Summary:
Wasm does not have direct counterparts to some of LLVM IR's atomicrmw
instructions (min, max, umin, umax, and nand). This enables atomic
expansion using cmpxchg instruction within a loop for those atomicrmw
instructions.
Reviewers: dschuff
Subscribers: sbc100, jgravelle-google, sunfish, llvm-commits
Differential Revision: https://reviews.llvm.org/D49440
llvm-svn: 339084
diff --git a/llvm/lib/Target/WebAssembly/WebAssemblyInstrAtomics.td b/llvm/lib/Target/WebAssembly/WebAssemblyInstrAtomics.td
index ba6800e..2c66efe 100644
--- a/llvm/lib/Target/WebAssembly/WebAssemblyInstrAtomics.td
+++ b/llvm/lib/Target/WebAssembly/WebAssemblyInstrAtomics.td
@@ -660,10 +660,11 @@
// Atomic ternary read-modify-writes
//===----------------------------------------------------------------------===//
-// TODO LLVM IR's cmpxchg instruction returns a pair of {loaded value,
-// success flag}. When we use a success flag or both values, we can't make use
-// of truncate/extend versions of instructions for now, which is suboptimal. Add
-// selection rules for those cases too.
+// TODO LLVM IR's cmpxchg instruction returns a pair of {loaded value, success
+// flag}. When we use the success flag or both values, we can't make use of i64
+// truncate/extend versions of instructions for now, which is suboptimal.
+// Consider adding a pass after instruction selection that optimizes this case
+// if it is frequent.
let Defs = [ARGUMENTS] in {