View mode: basic / threaded / horizontal-split · Log in · Help
July 25, 2012
Comparison between GDC test suite results on x86 and ARM
Hi,

I just ran the GDC test suite on x86 and ARM. I have attached the 
results to this message.

Quick summary:

x86:

# of expected passes		1287
# of unexpected failures	21
# of unresolved testcases	17

ARM:

# of expected passes		1280
# of unexpected failures	26
# of unresolved testcases	18

Not bad - the ARM port is progressing nicely.

As for specific files...

These tests fail on x86 but *not* on ARM:

FAIL: fail_compilation/fail6451.d output-exists-not fail6451.s
FAIL: runnable/test42.d (internal compiler error)

These tests fail on ARM but *not* on x86:

FAIL: runnable/arrayop.d execution test
FAIL: runnable/opover2.d execution test
FAIL: runnable/template1.d execution test
FAIL: runnable/test11.d execution test
FAIL: runnable/test12.d execution test
FAIL: runnable/testmath.d execution test
FAIL: runnable/testmmfile.d execution test

I have not actually looked into any of the failures yet. This is just 
for informational purposes to let people know how far along the ARM port is.

-- 
Alex Rønne Petersen
alex@lycus.org
http://lycus.org
July 25, 2012
Re: Comparison between GDC test suite results on x86 and ARM
On 25 July 2012 01:24, Alex Rønne Petersen <alex@lycus.org> wrote:
> Hi,
>
> I just ran the GDC test suite on x86 and ARM. I have attached the results to
> this message.
>
> Quick summary:
>
> x86:
>
> # of expected passes            1287
> # of unexpected failures        21
> # of unresolved testcases       17
>
> ARM:
>
> # of expected passes            1280
> # of unexpected failures        26
> # of unresolved testcases       18
>
> Not bad - the ARM port is progressing nicely.
>
> As for specific files...
>
> These tests fail on x86 but *not* on ARM:
>
> FAIL: fail_compilation/fail6451.d output-exists-not fail6451.s
> FAIL: runnable/test42.d (internal compiler error)
>
> These tests fail on ARM but *not* on x86:
>
> FAIL: runnable/arrayop.d execution test
> FAIL: runnable/opover2.d execution test
> FAIL: runnable/template1.d execution test
> FAIL: runnable/test11.d execution test
> FAIL: runnable/test12.d execution test
> FAIL: runnable/testmath.d execution test
> FAIL: runnable/testmmfile.d execution test
>
> I have not actually looked into any of the failures yet. This is just for
> informational purposes to let people know how far along the ARM port is.
>

In the <build>/gcc/testsuite/d directory, there is a .log file which
will contain the reasons for the test failure.

Most failures are simply because the bindings to the gcc testsuite
isn't up to scratch yet, whilst I'd be interested to look into that
ICE you're getting there. :-)

-- 
Iain Buclaw

*(p < e ? p++ : p) = (c & 0x0f) + '0';
July 25, 2012
Re: Comparison between GDC test suite results on x86 and ARM
Am Wed, 25 Jul 2012 02:24:58 +0200
schrieb Alex Rønne Petersen <alex@lycus.org>:

> Hi,
> 
> I just ran the GDC test suite on x86 and ARM. I have attached the 
> results to this message.
> 
> Quick summary:
> 
> x86:
> 
> # of expected passes		1287
> # of unexpected failures	21
> # of unresolved testcases	17
> 
> ARM:
> 
> # of expected passes		1280
> # of unexpected failures	26
> # of unresolved testcases	18
> 
> Not bad - the ARM port is progressing nicely.

That's indeed quite good. On what platform did you run those tests?
Soft/Hard float, thumb mode / normal mode, etc?

Did anyone already run the druntime/phobos unittests?
July 25, 2012
Re: Comparison between GDC test suite results on x86 and ARM
On 25-07-2012 19:47, Johannes Pfau wrote:
> Am Wed, 25 Jul 2012 02:24:58 +0200
> schrieb Alex Rønne Petersen <alex@lycus.org>:
>
>> Hi,
>>
>> I just ran the GDC test suite on x86 and ARM. I have attached the
>> results to this message.
>>
>> Quick summary:
>>
>> x86:
>>
>> # of expected passes		1287
>> # of unexpected failures	21
>> # of unresolved testcases	17
>>
>> ARM:
>>
>> # of expected passes		1280
>> # of unexpected failures	26
>> # of unresolved testcases	18
>>
>> Not bad - the ARM port is progressing nicely.
>
> That's indeed quite good. On what platform did you run those tests?
> Soft/Hard float, thumb mode / normal mode, etc?
>
> Did anyone already run the druntime/phobos unittests?
>

This was on an Efika MX with a Freescale i.MX515 (ARM Cortex A8) 
processor, running armel (normal mode) with soft float.

I might give it a shot in a Debian armhf chroot at some point.

-- 
Alex Rønne Petersen
alex@lycus.org
http://lycus.org
July 25, 2012
Re: Comparison between GDC test suite results on x86 and ARM
On 25-07-2012 09:11, Iain Buclaw wrote:
> On 25 July 2012 01:24, Alex Rønne Petersen <alex@lycus.org> wrote:
>> Hi,
>>
>> I just ran the GDC test suite on x86 and ARM. I have attached the results to
>> this message.
>>
>> Quick summary:
>>
>> x86:
>>
>> # of expected passes            1287
>> # of unexpected failures        21
>> # of unresolved testcases       17
>>
>> ARM:
>>
>> # of expected passes            1280
>> # of unexpected failures        26
>> # of unresolved testcases       18
>>
>> Not bad - the ARM port is progressing nicely.
>>
>> As for specific files...
>>
>> These tests fail on x86 but *not* on ARM:
>>
>> FAIL: fail_compilation/fail6451.d output-exists-not fail6451.s
>> FAIL: runnable/test42.d (internal compiler error)
>>
>> These tests fail on ARM but *not* on x86:
>>
>> FAIL: runnable/arrayop.d execution test
>> FAIL: runnable/opover2.d execution test
>> FAIL: runnable/template1.d execution test
>> FAIL: runnable/test11.d execution test
>> FAIL: runnable/test12.d execution test
>> FAIL: runnable/testmath.d execution test
>> FAIL: runnable/testmmfile.d execution test
>>
>> I have not actually looked into any of the failures yet. This is just for
>> informational purposes to let people know how far along the ARM port is.
>>
>
> In the <build>/gcc/testsuite/d directory, there is a .log file which
> will contain the reasons for the test failure.
>
> Most failures are simply because the bindings to the gcc testsuite
> isn't up to scratch yet, whilst I'd be interested to look into that
> ICE you're getting there. :-)
>

Appears to be this, if I'm reading this log correctly:

cc1d: internal compiler error: in hwi2toli, at d/d-codegen.cc:1107
Please submit a full bug report,
with preprocessed source if appropriate.
See <http://gcc.gnu.org/bugs.html> for instructions.
compiler exited with status 1
output is:
success
Not ported to x86-64 compatible varargs, yet.
myInt int
myBool bool
cc1d: internal compiler error: in hwi2toli, at d/d-codegen.cc:1107
Please submit a full bug report,
with preprocessed source if appropriate.
See <http://gcc.gnu.org/bugs.html> for instructions.

FAIL: runnable/test42.d (internal compiler error)

-- 
Alex Rønne Petersen
alex@lycus.org
http://lycus.org
July 26, 2012
Re: Comparison between GDC test suite results on x86 and ARM
On 25 July 2012 19:25, Alex Rønne Petersen <alex@lycus.org> wrote:
> On 25-07-2012 09:11, Iain Buclaw wrote:
>>
>> On 25 July 2012 01:24, Alex Rønne Petersen <alex@lycus.org> wrote:
>>>
>>> Hi,
>>>
>>> I just ran the GDC test suite on x86 and ARM. I have attached the results
>>> to
>>> this message.
>>>
>>> Quick summary:
>>>
>>> x86:
>>>
>>> # of expected passes            1287
>>> # of unexpected failures        21
>>> # of unresolved testcases       17
>>>
>>> ARM:
>>>
>>> # of expected passes            1280
>>> # of unexpected failures        26
>>> # of unresolved testcases       18
>>>
>>> Not bad - the ARM port is progressing nicely.
>>>
>>> As for specific files...
>>>
>>> These tests fail on x86 but *not* on ARM:
>>>
>>> FAIL: fail_compilation/fail6451.d output-exists-not fail6451.s
>>> FAIL: runnable/test42.d (internal compiler error)
>>>
>>> These tests fail on ARM but *not* on x86:
>>>
>>> FAIL: runnable/arrayop.d execution test
>>> FAIL: runnable/opover2.d execution test
>>> FAIL: runnable/template1.d execution test
>>> FAIL: runnable/test11.d execution test
>>> FAIL: runnable/test12.d execution test
>>> FAIL: runnable/testmath.d execution test
>>> FAIL: runnable/testmmfile.d execution test
>>>
>>> I have not actually looked into any of the failures yet. This is just for
>>> informational purposes to let people know how far along the ARM port is.
>>>
>>
>> In the <build>/gcc/testsuite/d directory, there is a .log file which
>> will contain the reasons for the test failure.
>>
>> Most failures are simply because the bindings to the gcc testsuite
>> isn't up to scratch yet, whilst I'd be interested to look into that
>> ICE you're getting there. :-)
>>
>
> Appears to be this, if I'm reading this log correctly:
>
> cc1d: internal compiler error: in hwi2toli, at d/d-codegen.cc:1107
> Please submit a full bug report,
> with preprocessed source if appropriate.
> See <http://gcc.gnu.org/bugs.html> for instructions.
> compiler exited with status 1
> output is:
> success
> Not ported to x86-64 compatible varargs, yet.
> myInt int
> myBool bool
> cc1d: internal compiler error: in hwi2toli, at d/d-codegen.cc:1107
> Please submit a full bug report,
> with preprocessed source if appropriate.
> See <http://gcc.gnu.org/bugs.html> for instructions.
>
>
> FAIL: runnable/test42.d (internal compiler error)
>

Yep, spotted it, and have fixed. :-)


-- 
Iain Buclaw

*(p < e ? p++ : p) = (c & 0x0f) + '0';
Top | Discussion index | About this forum | D home