MIPS: Enable asm-wasm mjsunit tests (skip on arm and mips).

test/mjsunit/mjsunit.status

Index: test/mjsunit/mjsunit.status
diff --git a/test/mjsunit/mjsunit.status b/test/mjsunit/mjsunit.status
index 07311899609545243d55ede5ea415ac8d5f31e8d..bdadb3c54f78619a09326ddbee27718927bde721 100644
--- a/test/mjsunit/mjsunit.status
+++ b/test/mjsunit/mjsunit.status
@@ -309,7 +309,7 @@
 
   # TODO(titzer): correct WASM adapter frame alignment on arm64
   'wasm/*': [PASS, ['arch == arm64', SKIP]],
-  'wasm/asm-wasm': [PASS, ['arch == arm or arch == arm64', SKIP]],
+  'wasm/asm-wasm': [PASS, ['arch in [arm, arm64, mips, mipsel, mips64, mips64el]', SKIP]],

[bradn, 2016/03/09 18:11:17]

I'm confused, doesn't this disable asm-wasm for mips?

[balazs.kilvady, 2016/03/10 13:44:30]

It does, that is our purpose. On MIPS trunc.d.w converts a double value which is
higher then MAX_INT to MAX_INT so asm-wasm test fails. ( Failing part is:
function TestConvertF64FromUnsigned() {
  "use asm";

  function caller() {
    var a = 0xffffffff;
    if ((+(a>>>0)) > 0.0) {
      if((+a) < 0.0) {
        return 26;

      }
    }
    return 0;
  }

  return {caller:caller};
}

The test fails on a target board and passes on simulator because of different
conversion, I think.
The static_cast<>() of the following code produces different results on Intel
and on MIPS:
            int val = static_cast<int>(expr->raw_value()->AsNumber());
            PrintF("double val: %f, int converted: 0x%x\n",
                   expr->raw_value()->AsNumber(), val);

The 'good' result of Intel (sim) is:
double val: 4294967295.000000, int converted: 0x80000000
the 'bad' result on MIPS board:
double val: 4294967295.000000, int converted: 0x7fffffff
Do you think we should do some workaround solution for MIPS (and perhaps for
ARM, too) or we can just skip the test like ARM guys did?

 
   # TODO(branelson): Figure out why ignition + asm->wasm fails embenchen.
   'wasm/embenchen/*': [PASS, ['ignition == True', SKIP]],