[Mlir-commits] [mlir] [mlir][emitc] Add 'emitc.switch' op to the dialect (PR #102331)

[Gil Rapaport]

================
@@ -1302,5 +1304,195 @@ def EmitC_SubscriptOp : EmitC_Op<"subscript", []> {
   let assemblyFormat = "$value `[` $indices `]` attr-dict `:` functional-type(operands, results)";
 }
 
+def EmitC_SwitchOp : EmitC_Op<"switch", [RecursiveMemoryEffects,
+    SingleBlockImplicitTerminator<"emitc::YieldOp">,
+    DeclareOpInterfaceMethods<RegionBranchOpInterface,
+                              ["getRegionInvocationBounds",
+                               "getEntrySuccessorRegions"]>]> {
+  let summary = "Switch operation";
+  let description = [{
+    The `emitc.switch` is a control-flow operation that branches to one of
+    the given regions based on the values of the argument and the cases.
+    The operand to a switch operation is a opaque or integral value.
+
+    The operation always has a "default" region and any number of case regions
+    denoted by integer constants. Control-flow transfers to the case region
+    whose constant value equals the value of the argument. If the argument does
+    not equal any of the case values, control-flow transfer to the "default"
+    region.
+
+    The operation does not return any value. Moreover, case regions and 
+    default region must be explicitly terminated using the `emitc.yield` operation.
+
+    Example:
+
+    ```mlir
+    // Example for i32:
----------------
aniragil wrote:

Examples section seems overly detailed (cf. to examples in other EmitC ops):

- Selector's type doesn't imply significantly different semantics/output, so demonstrating a single type seems enough
- The `--declare-variables-at-top` flag affects other ops in a similar manner (and isn't demonstrated for them)
- Showing the op itself (without the function context) seems sufficient
- A single variable + call in the default section seems enough

Interested users can find more detailed examples in the dialect's tests.

https://github.com/llvm/llvm-project/pull/102331