[Mono-dev] On what platform|cpu-arch|mono-version does full AOT actually work?

Peter Moberg peter.moberg at gmail.com
Tue Feb 17 00:15:09 EST 2009


Looks like you are right regarding the fact that I will be running  
into a lot of problems...

Going to sound like a broken record...but I did get latest from SVN...  
compiled everything for the iPhone ARM...Successfully started up Mono  
on the iPhone device...started to compile mscorlib (profile 1 as well  
as profile 2) in AOT mode.....BAM!...same assert as mentioned below  
triggered.

Since I am able to run Mono on the iPhone and I am performing the AOT  
compilation on the iPhone with its ARM processor I am bit surprised  
that it still doesn't work... the only thing I can think of is that  
there are a few files that are generated before compilation... e.g.  
cpu-arm.h and a few others in the arch folder..and that they are not  
correct on my side... if anyone have those files and are able to run  
on the ARM processor I would love to take a look at them to compare....


On Feb 16, 2009, at 8:19 PM, Zoltan Varga wrote:

> Hi,
>
>  mono on iphone is not really a supported configuration right now, it
> works for some people,
> but expect to run into lots of problems. Cross compiling mono for arm
> is especially fraught
> with problems, better to compile on the device itself.
>
>                Zoltan
>
> On Tue, Feb 17, 2009 at 3:12 AM, Peter Moberg  
> <peter.moberg at gmail.com> wrote:
>> Will try to do that...
>>
>> Regarding the files in mono/arch/arm... Do I have to regenerate  
>> these files
>> or can I use the h-files that are present? Since I am compiling on  
>> a Mac OS
>> X I would need to fake being on an ARM if I need to run anything in  
>> the arch
>> folder...
>>
>>
>> On Feb 16, 2009, at 8:01 PM, Zoltan Varga wrote:
>>
>>> Try updating to mono HEAD. SVN versions sometimes have bugs like  
>>> that,
>>> but SVN HEAD
>>> works for me when using AOT on ARM.
>>>
>>>               Zoltan
>>>
>>> On Tue, Feb 17, 2009 at 2:45 AM, mobbe <peter.moberg at gmail.com>  
>>> wrote:
>>>>
>>>> Nope...same results with a mscorlib.dll from the 1.0 folder. It  
>>>> gets
>>>> stuck in
>>>> the mono_local_regalloc method on the assert  checking for ins- 
>>>> >sreg !=
>>>> -1... Exactly the same message as before. It is trying to compile  
>>>> the
>>>> Equals
>>>> method
>>>>
>>>> The inst op code is 633 and both sreg1 and sreg2 is -1.. .
>>>>
>>>> I am running the system with build 125709 and compiling the  
>>>> mscorlib.dll
>>>> from the 1.0 folder...
>>>>
>>>>
>>>>
>>>>
>>>> mobbe wrote:
>>>>>
>>>>> I am indeed trying to compile a 2.0 version of the mscorlib. I  
>>>>> will grab
>>>>> the 1.0 version instead and see if I have better luck with that  
>>>>> one...if
>>>>> I
>>>>> can get that to work then I will migrate to use the code from  
>>>>> SVN and
>>>>> try
>>>>> with the 2.0 profile.
>>>>>
>>>>> Thanks,
>>>>>
>>>>>
>>>>> Zoltan Varga wrote:
>>>>>>
>>>>>> Hi,
>>>>>>
>>>>>> You might be trying to use full-aot with the 2.0 profile.  
>>>>>> Currently,
>>>>>> only the 1.0 mscorlib can
>>>>>> be full-aot-ed, altough 2.0 should work when using mono SVN.
>>>>>>
>>>>>>            Zoltan
>>>>>>
>>>>>> On Mon, Feb 16, 2009 at 3:48 PM, mobbe <peter.moberg at gmail.com>  
>>>>>> wrote:
>>>>>>>
>>>>>>> I compiled the Mono framework (the unmanaged pieces) so that I  
>>>>>>> could
>>>>>>> run
>>>>>>> it
>>>>>>> on my iPhone. I was successful in doing that and thought that  
>>>>>>> I would
>>>>>>> try to
>>>>>>> do AOT on the iPhone...transfer the temporary asm output to my  
>>>>>>> Mac to
>>>>>>> assemble and link it and from there keep going... however when
>>>>>>> performing
>>>>>>> the AOT compilation of mscorlib.dll Mono hit another assert...
>>>>>>> something
>>>>>>> with inst->sreg1 being negative 1 (-1) (Don't remember right  
>>>>>>> now what
>>>>>>> funtion it stopped in..will have to supply that information  
>>>>>>> when I get
>>>>>>> to my
>>>>>>> Mac later tonight).
>>>>>>>
>>>>>>> If I commented out the assert it went on a bit further but  
>>>>>>> stopped on
>>>>>>> a
>>>>>>> another assert when trying to emit the op-code saying that it  
>>>>>>> was
>>>>>>> expecting
>>>>>>> 0 bytes for the op-code ld_loadmembase but got 4...
>>>>>>>
>>>>>>> Anyone out there have an ideas of where I should be looking  
>>>>>>> next in
>>>>>>> order to
>>>>>>> try to solve this? It seems like the instruction set it is  
>>>>>>> using for
>>>>>>> ARM
>>>>>>> doesn't match the source codes assumptions. I am sure that I  
>>>>>>> am doing
>>>>>>> something wrong but don't know where to begin to look for the  
>>>>>>> issue.
>>>>>>> Maybe
>>>>>>> someone out there has a cpu-arm.h file that they know works  
>>>>>>> that I can
>>>>>>> compare with mine? Are there any other generated files  
>>>>>>> regarding the
>>>>>>> instruction set that I should double check?
>>>>>>>
>>>>>>> Thanks for your help in advance.
>>>>>>>
>>>>>>>
>>>>>>>
>>>>>>> Hi,
>>>>>>>
>>>>>>> full-aot currently only works on amd64/arm.
>>>>>>>
>>>>>>>          Zoltan
>>>>>>>
>>>>>>>
>>>>>>> --
>>>>>>> View this message in context:
>>>>>>>
>>>>>>> http://www.nabble.com/On-what-platform%7Ccpu-arch%7Cmono-version-does-full-AOT-actually-work--tp21988583p22038601.html
>>>>>>> Sent from the Mono - Dev mailing list archive at Nabble.com.
>>>>>>>
>>>>>>> _______________________________________________
>>>>>>> Mono-devel-list mailing list
>>>>>>> Mono-devel-list at lists.ximian.com
>>>>>>> http://lists.ximian.com/mailman/listinfo/mono-devel-list
>>>>>>>
>>>>>> _______________________________________________
>>>>>> Mono-devel-list mailing list
>>>>>> Mono-devel-list at lists.ximian.com
>>>>>> http://lists.ximian.com/mailman/listinfo/mono-devel-list
>>>>>>
>>>>>>
>>>>>
>>>>>
>>>>
>>>> --
>>>> View this message in context:
>>>> http://www.nabble.com/On-what-platform%7Ccpu-arch%7Cmono-version-does-full-AOT-actually-work--tp21988583p22049003.html
>>>> Sent from the Mono - Dev mailing list archive at Nabble.com.
>>>>
>>>> _______________________________________________
>>>> Mono-devel-list mailing list
>>>> Mono-devel-list at lists.ximian.com
>>>> http://lists.ximian.com/mailman/listinfo/mono-devel-list
>>>>
>>
>>



More information about the Mono-devel-list mailing list