[Mlir-commits] [mlir] [mlir][gpu] Add `broadcast_lane` op (PR #152808)

[Ivan Butygin]

================
@@ -2511,6 +2511,43 @@ bool WarpExecuteOnLane0Op::areTypesCompatible(Type lhs, Type rhs) {
       verifyDistributedType(lhs, rhs, getWarpSize(), getOperation()));
 }
 
+//===----------------------------------------------------------------------===//
+// GPU_BroadcastLaneOp
+//===----------------------------------------------------------------------===//
+
+void gpu::BroadcastLaneOp::inferResultRanges(
+    ArrayRef<ConstantIntRanges> argRanges, SetIntRangeFn setResultRange) {
+  setResultRange(getResult(), argRanges.front());
+}
+
+Speculation::Speculatability gpu::BroadcastLaneOp::getSpeculatability() {
+  switch (getBroadcastType()) {
+  case BroadcastType::first_lane:
+    // Cannot speculate first_lane broadcast, because speculating it across
+    // control flow can change the active lanes.
+    return Speculation::NotSpeculatable;
+  case BroadcastType::any_lane:
+    LLVM_FALLTHROUGH;
+  case BroadcastType::lane:
----------------
Hardcode84 wrote:

technically, CSE can merge ops across blocks
```
%1 = gpu.broadcast %src, lane %idx
scf.if lane_id < 13 {
    %2 = gpu.broadcast %src, lane %idx
}
```
not sure if it's a problem, though

https://github.com/llvm/llvm-project/pull/152808