fix test/nilptr3.go

func f4(x *[10]int) {
  ...
  _ = x[9] // ERROR "generated nil check"
	for {
		if x[9] != 0 { // ERROR "removed nil check"
			break
		}
	}
}

Loop rotation duplicates conditional test Values into loop gaurd block:

...
loopGuard:
v6 (+159) = NilCheck <*[10]int> v5 v1
v15 (159) = OffPtr <*int> [72] v6
v9 (159) = Load <int> v15 v1
v7 (159) = Neq64 <bool> v9 v13
If v7 → b5 loopHeader

loopHeader:
v8 (+159) = NilCheck <*[10]int> v5 v1
v11 (159) = OffPtr <*int> [72] v8
v12 (159) = Load <int> v11 v1
v14 (159) = Neq64 <bool> v12 v13
Plain → b3

So we need to add a loop nilcheckelim pass after all other loop opts to
remove the v6 NilCheck in the loop guard, but that's not the whole
story.

The code _ = x[9] is expected to generate a nilcheck, buttighten pass
will move the LoweredNilCheck from the loop entry b1 to the loop guard.
At that point, there happens to be a CMPQconstload in the loop guard
that meets the conditions for late nilcheckelim, resulting in the
elimination of the LoweredNilCheck as well. Therefore, we also need to
modify the test to no longer expect the generated code to contain a
nilcheck.

Author: y1yang0

src/cmd/compile/internal/ssa/compile.go

@@ -492,6 +492,7 @@ var passes = [...]pass{
 	{name: "loop sccp", fn: sccp, disabled: !EnableLoopOpts},                  // optimize loop guard conditional test
 	{name: "loop opt", fn: opt, disabled: !EnableLoopOpts},                    // further optimize loop guard conditional test
 	{name: "loop deadcode late", fn: deadcode, disabled: !EnableLoopOpts},     // remove dead loop guard to simplify cfg
+	{name: "loop nilcheckelim", fn: nilcheckelim, disabled: !EnableLoopOpts},  // remove duplicated nil check in loop guard
 	{name: "writebarrier", fn: writebarrier, required: true},                  // expand write barrier ops
 	{name: "insert resched checks", fn: insertLoopReschedChecks,
 		disabled: !buildcfg.Experiment.PreemptibleLoops}, // insert resched checks in loops.

test/nilptr3.go

@@ -155,7 +155,7 @@ func f4(x *[10]int) {
 	// and the offset is small enough that if x is nil, the address will still be
 	// in the first unmapped page of memory.
 
-	_ = x[9] // ERROR "generated nil check" // bug: would like to remove this check (but nilcheck and load are in different blocks)
+	_ = x[9] // ERROR "removed nil check"
 
 	for {
 		if x[9] != 0 { // ERROR "removed nil check"