// rather than starting all in a batch and then waiting once
		// afterwards. By running one goroutine at a time, we can take
		// advantage of runnext to bounce back and forth between
		// workers and this goroutine. In an overloaded application,
		// this can reduce GC start latency by prioritizing these
		// goroutines rather than waiting on the end of the run queue.
		<-ready

	return true
}

// noteRule is an easy way to track if a rule is matched when writing
// new ones.  Make the rule of interest also conditional on
//
//	noteRule("note to self: rule of interest matched")
//
// and that message will print when the rule matches.
func noteRule(s string) bool {
	fmt.Println(s)
	return true
}

  // optionally followed by a '-' and a colon-separated list of
  // negative patterns (tests to exclude).  A test is run if it
  // matches one of the positive patterns and does not match any of
  // the negative patterns.
  //
  // For example, *A*:Foo.* is a filter that matches any string that
  // contains the character 'A' or starts with "Foo.".
  bool should_run() const { return should_run_; }

    }) {is_stateless = false} : (tensor<i1>) -> tensor<*xf32>
```
### `-tf-functional-to-executor-conversion`

_Transform from func op to TF executor dialect._

### `-tf-fused-kernel-matcher`

_Matches computations corresponding to optimized fused kernels_

### `-tf-gpu-op-fusion`

_Fusion optimization for GPU targets_

This pass is performing fusion specific to GPU targets. This is an ad-hoc