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

test/mjsunit/mjsunit.status

 # Copyright 2012 the V8 project authors. All rights reserved.
 # Redistribution and use in source and binary forms, with or without
 # modification, are permitted provided that the following conditions are
 # met:
 #
 #     * Redistributions of source code must retain the above copyright
 #       notice, this list of conditions and the following disclaimer.
 #     * Redistributions in binary form must reproduce the above
 #       copyright notice, this list of conditions and the following
 #       disclaimer in the documentation and/or other materials provided
@@ skipping 291 matching lines @@
   'copy-on-write-assert': [PASS, SLOW],
   'debug-scopes': [PASS, SLOW],
   'es7/object-observe': [PASS, ['mode == debug', SLOW]],
   'numops-fuzz-part*': [PASS, ['mode == debug', SLOW]],
   'readonly': [PASS, SLOW],
   'regress/regress-1200351': [PASS, ['mode == debug', SLOW]],
   'regress/regress-crbug-474297': [PASS, ['mode == debug', SLOW]],
 
   # 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]],
 
   # TODO(bradnelson) Fix and re-enable.
   'wasm/embenchen/copy': [FAIL],  # wrong answer
   'wasm/embenchen/box2d': [SKIP],   # hang
   'wasm/embenchen/lua_binarytrees': [SKIP],  # fails decode
   'wasm/embenchen/zlib': [SKIP],  # fails gc-stress
 
@@ skipping 580 matching lines @@
   # Tests taking too long.
   'array-functions-prototype-misc': [SKIP],
   'strong/implicit-conversions': [SKIP],
   'strong/load-element-mutate-backing-store': [SKIP],
 
   # Stack overflow.
   'big-array-literal': [SKIP],
 }],  # 'gcov_coverage'
 
 ]