Fixes buggy FPU tests for MIPS hardware.

Fixes buggy FPU tests for MIPS hardware.

[zra, 2013-06-01 00:15:44]

For the MIPS hardware the results need to be in D0, for the MIPS simulator the
results need to be in V1:V0.

[regis, 2013-06-01 00:38:15]

Is this only a temporary fix?
The right way to fix this would be to change the test macros to use the same
registers in the simulator as on the hardware.

[zra, 2013-06-03 15:26:17]

On 2013/06/01 00:38:15, regis wrote:
> Is this only a temporary fix?
> The right way to fix this would be to change the test macros to use the same
> registers in the simulator as on the hardware.

Yes, it's just a temporary fix to make sure there were no other problems with
the floating point instructions on the hardware. But I think the right way isn't
quite as simple as fixing the test macros since Simulator::Call would have to
know which registers to return -- V1:V0 or F1:F0. Maybe I could just add a flag
to Simulator::Call to select the right ones, which would default to V1:V0.

[zra, 2013-06-10 16:56:40]

On 2013/06/03 15:26:17, zra wrote:
> Yes, it's just a temporary fix to make sure there were no other problems with
> the floating point instructions on the hardware. But I think the right way
isn't
> quite as simple as fixing the test macros since Simulator::Call would have to
> know which registers to return -- V1:V0 or F1:F0. Maybe I could just add a
flag
> to Simulator::Call to select the right ones, which would default to V1:V0.

I added this flag to Simulator::Call and cleaned up the tests to match.

Cheers,
Zach

[zra, 2013-06-17 21:44:17]

On 2013/06/10 16:56:40, zra wrote:
> I added this flag to Simulator::Call and cleaned up the tests to match.
> 
> Cheers,
> Zach

ping

[regis, 2013-06-18 08:42:12]

LGTM

https://codereview.chromium.org/15874005/diff/9001/runtime/vm/simulator_mips.h
File runtime/vm/simulator_mips.h (right):

https://codereview.chromium.org/15874005/diff/9001/runtime/vm/simulator_mips....
runtime/vm/simulator_mips.h:54: void set_dregister_double(DRegister freg, double
value);
I would expect set_dregister to take a double. The other variant taking an
int64_t could be called set_dregister_bits.

https://codereview.chromium.org/15874005/diff/9001/runtime/vm/simulator_mips....
runtime/vm/simulator_mips.h:57: double get_dregister_double(DRegister freg)
const;
ditto with return types.

https://codereview.chromium.org/15874005/diff/9001/runtime/vm/simulator_mips....
runtime/vm/simulator_mips.h:103: // result on return.
The comment should explain the fp_return parameter.

[zra, 2013-06-18 16:47:25]

Thanks!

https://codereview.chromium.org/15874005/diff/9001/runtime/vm/simulator_mips.h
File runtime/vm/simulator_mips.h (right):

https://codereview.chromium.org/15874005/diff/9001/runtime/vm/simulator_mips....
runtime/vm/simulator_mips.h:54: void set_dregister_double(DRegister freg, double
value);
On 2013/06/18 08:42:12, regis wrote:
> I would expect set_dregister to take a double. The other variant taking an
> int64_t could be called set_dregister_bits.

Done.

https://codereview.chromium.org/15874005/diff/9001/runtime/vm/simulator_mips....
runtime/vm/simulator_mips.h:57: double get_dregister_double(DRegister freg)
const;
On 2013/06/18 08:42:12, regis wrote:
> ditto with return types.

Done.

https://codereview.chromium.org/15874005/diff/9001/runtime/vm/simulator_mips....
runtime/vm/simulator_mips.h:103: // result on return.
On 2013/06/18 08:42:12, regis wrote:
> The comment should explain the fp_return parameter.

Done.