[Webkit-unassigned] [Bug 30144] MIPS JIT Supports

[bugzilla-daemon at webkit.org]

https://bugs.webkit.org/show_bug.cgi?id=30144





--- Comment #15 from Chao-ying Fu <fu at mips.com>  2009-10-09 16:14:58 PDT ---
(In reply to comment #14)
> (In reply to comment #13)
> > (In reply to comment #10)
> > > > > +		    fprintf (stderr, "Error: Cannot create MIPS jump (J) due to the top four bits are different.\n");
> > > > > +		}
> > > > > +		*insn = 0x08000000 | ((new_target_address >> 2) & 0x3ffffff);
> > > > > +	    }
> > > > 
> > > > What happens if a jump cannot be created? Crash?
> > > 
> > >   Yes, if we cannot use "J", it will crash.
> > > But this type of jump can cover a 28-bit range of targets.  From running
> > > sunspider and v8 tests, 28-bit jumps seem ok.
> > 
> > The JavaScriptCore tests are essential. Did you test with
> > ./WebKitTools/Scripts/run-javascriptcore-tests?
> > 
> > In additional, you should test at least with
> > ./WebKitTools/Scripts/run-webkit-tests LayoutTests/fast/js as well! The new JS
> > regression tests were landed there.
> 
> Thank you for the testing information!  I will run these two tests and verify
> the correctness.

I finished ./WebKitTools/Scripts/run-javascriptcore-tests .
The results are as follows.
        ecma/GlobalObject/15.1.2.3-2.js
        ecma_3/RegExp/regress-72964.js
        js1_5/Regress/regress-159334.js

3 regressions found.
0 tests fixed.

I debugged the first two tests and fixed them.
The first test failed due to that we need to relocate MIPS jumps after the
buffer grows.
The second test failed due to that we need to use zero-extended load to load
16-bit data.
The third test failed under both non-JIT and JIT versions.  I will debug it
next week.

I will attached a new patch that fixed the first two tests.  Thanks!

-- 
Configure bugmail: https://bugs.webkit.org/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are the assignee for the bug.