[turbofan] more regalloc fixes

test/mjsunit/regress/regress-5911.js

+// Copyright 2017 the V8 project authors. All rights reserved.
+// Use of this source code is governed by a BSD-style license that can be
+// found in the LICENSE file.
+
+load("test/mjsunit/wasm/wasm-constants.js");
+load("test/mjsunit/wasm/wasm-module-builder.js");
+
+(function() {
+    var builder = new WasmModuleBuilder();
+    builder.addMemory(32, 32, false);
+    builder.addFunction("test", kSig_i_iii)
+        .addBodyWithEnd([
+            // body:
+            kExprI64Const, 0x42,
+            kExprI64Const, 0x7a,
+            kExprI64RemU,
+            kExprI64Const, 0x42,
+            kExprI64Const, 0x37,
+            kExprI64Mul,
+            kExprI64Const, 0x36,
+            kExprI64Mul,
+            kExprI64Const, 0x42,
+            kExprI64Ctz,
+            kExprI64Ctz,
+            kExprI64Shl,
+            kExprF32SConvertI64,
+            kExprUnreachable,

[Eric Holk, 2017/02/01 19:04:25]

It looks like this test unconditionally hits this unreachable instruction. If
so, how does this test detect the bug?

[ahaas, 2017/02/02 20:21:42]

This test case crashed during compilation, not during execution. The unreachable
instruction can be helpful for compilation because with the unreachable it is
more likely that a WebAssembly program validates.

+            kExprEnd,   // @21
+        ])
+        .exportFunc();
+    var module = new WebAssembly.Module(builder.toBuffer());
+})();