llvm dev - Jun 2012 - [LLVMdev] DFAPacketizer with StateTrans != 0 Assertion

Hi Ivan,

The assertion was happening because I wasn't checking after the first 
attempt failed. The first packet was failing and so it was ended, and 
then the packetizer attempted to add it to the next packet without 
checking for available resources. However this highlights probably the 
real problem - my packetizer is unable to find resources for the first 
instruction, or any of my instructions for that matter. Have you ever 
seen this problem before?

thanks
sam

On 12/06/12 15:48, Ivan Llopard [via LLVM] wrote:
> Hi Sam,
>
> On 12/06/2012 15:33, sam wrote:
> > Hi,
> >
> > I'm trying to get the DFAPacketizer to work for my target but with
any
> > instruction I get the
> > 'Assertion `CachedTable.count(StateTrans) != 0' failed'
error and it
> crashes
> > out before packeting a single instruction.
>
> Do you reserve some resource without verification?
>
> Note that reserveResources() should be called iff canReserveResources()
> is true (the call order is important here) even for the trivial case (0
> instructions in the packet). See VLIWPacketizerList as an example on how
> to use the API's.
>
> Ivan
>
> >
> > I have a *GenDFAPacketizer.inc file and my packetizer pass checks 
> that the
> > table is not empty before proceeding. I also have a schedule file 
> with my
> > functional units and itineraries listed. Can anyone think of what I 
> may have
> > missed to cause this error?
> >
> > Cheers,
> > sam
> >
> > --
> > View this message in context: 
>
http://llvm.1065342.n5.nabble.com/DFAPacketizer-with-StateTrans-0-Assertion-tp46105.html
> > Sent from the LLVM - Dev mailing list archive at Nabble.com.
> > _______________________________________________
> > LLVM Developers mailing list
> > [hidden email]
</user/SendEmail.jtp?type=node&node=46108&i=0>
> http://llvm.cs.uiuc.edu
> > http://lists.cs.uiuc.edu/mailman/listinfo/llvmdev
> _______________________________________________
> LLVM Developers mailing list
> [hidden email] </user/SendEmail.jtp?type=node&node=46108&i=1>
> http://llvm.cs.uiuc.edu
> http://lists.cs.uiuc.edu/mailman/listinfo/llvmdev
>
>
> ------------------------------------------------------------------------
> If you reply to this email, your message will be added to the 
> discussion below:
>
http://llvm.1065342.n5.nabble.com/DFAPacketizer-with-StateTrans-0-Assertion-tp46105p46108.html
>
> To unsubscribe from DFAPacketizer with StateTrans != 0 Assertion, 
> click here 
>
<http://llvm.1065342.n5.nabble.com/template/NamlServlet.jtp?macro=unsubscribe_by_code&node=46105&code=Uy5QYXJrZXIzQGxib3JvLmFjLnVrfDQ2MTA1fC0xNTU1NTYxMTE=>.
> NAML 
>
<http://llvm.1065342.n5.nabble.com/template/NamlServlet.jtp?macro=macro_viewer&id=instant_html%21nabble%3Aemail.naml&base=nabble.naml.namespaces.BasicNamespace-nabble.view.web.template.NabbleNamespace-nabble.view.web.template.NodeNamespace&breadcrumbs=notify_subscribers%21nabble%3Aemail.naml-instant_emails%21nabble%3Aemail.naml-send_instant_email%21nabble%3Aemail.naml>
>


--
View this message in context:
http://llvm.1065342.n5.nabble.com/DFAPacketizer-with-StateTrans-0-Assertion-tp46105p46112.html
Sent from the LLVM - Dev mailing list archive at Nabble.com.
-------------- next part --------------
An HTML attachment was scrubbed...
URL:
<http://lists.llvm.org/pipermail/llvm-dev/attachments/20120612/92cd1b46/attachment.html>

Sam,

 > However this highlights probably the real problem - my packetizer is
 > unable to find resources for the first instruction, or any of my 
instructions
 > for that matter. Have you ever seen this problem before?

I would recommend looking at <Target>DFAStateInputTable in 
<Target>GenDFAPacketizer.inc. That table contains <input,
transition>
pairs for each state. Each row in the table marks the beginning of a new 
state. In your case, I'd suggest examining the first row in the table 
(the DFA's initial state is state 0 which corresponds to the first row) 
to figure out what the generated DFA is for your target.

-Anshu

---
Qualcomm Innovation Center, Inc is a member of the Code Aurora Forum

Hi sam,

On 12/06/2012 17:30, sam wrote:
> Hi Ivan,
>
> The assertion was happening because I wasn't checking after the first 
> attempt failed. The first packet was failing and so it was ended, and 
> then the packetizer attempted to add it to the next packet without 
> checking for available resources. However this highlights probably the 
> real problem - my packetizer is unable to find resources for the first 
> instruction, or any of my instructions for that matter. Have you ever 
> seen this problem before?
No. AFAIK, the DFA expects the instruction to have well defined 
itinerary with its resource utilisation. Does your first instruction 
have an defined itinerary?

Ivan
>
> thanks
> sam
>
> On 12/06/12 15:48, Ivan Llopard [via LLVM] wrote:
>> Hi Sam,
>>
>> On 12/06/2012 15:33, sam wrote:
>> > Hi,
>> >
>> > I'm trying to get the DFAPacketizer to work for my target but
with any
>> > instruction I get the
>> > 'Assertion `CachedTable.count(StateTrans) != 0'
failed' error and
>> it crashes
>> > out before packeting a single instruction.
>>
>> Do you reserve some resource without verification?
>>
>> Note that reserveResources() should be called iff canReserveResources()
>> is true (the call order is important here) even for the trivial case (0
>> instructions in the packet). See VLIWPacketizerList as an example on
how
>> to use the API's.
>>
>> Ivan
>>
>> >
>> > I have a *GenDFAPacketizer.inc file and my packetizer pass checks 
>> that the
>> > table is not empty before proceeding. I also have a schedule file 
>> with my
>> > functional units and itineraries listed. Can anyone think of what
I
>> may have
>> > missed to cause this error?
>> >
>> > Cheers,
>> > sam
>> >
>> > --
>> > View this message in context: 
>>
http://llvm.1065342.n5.nabble.com/DFAPacketizer-with-StateTrans-0-Assertion-tp46105.html
>> > Sent from the LLVM - Dev mailing list archive at Nabble.com.
>> > _______________________________________________
>> > LLVM Developers mailing list
>> > [hidden email]
</user/SendEmail.jtp?type=node&node=46108&i=0>
>> http://llvm.cs.uiuc.edu
>> > http://lists.cs.uiuc.edu/mailman/listinfo/llvmdev
>> _______________________________________________
>> LLVM Developers mailing list
>> [hidden email]
</user/SendEmail.jtp?type=node&node=46108&i=1>
>> http://llvm.cs.uiuc.edu
>> http://lists.cs.uiuc.edu/mailman/listinfo/llvmdev
>>
>>
>>
------------------------------------------------------------------------
>> If you reply to this email, your message will be added to the 
>> discussion below:
>>
http://llvm.1065342.n5.nabble.com/DFAPacketizer-with-StateTrans-0-Assertion-tp46105p46108.html
>>
>> To unsubscribe from DFAPacketizer with StateTrans != 0 Assertion, 
>> click here.
>> NAML 
>>
<http://llvm.1065342.n5.nabble.com/template/NamlServlet.jtp?macro=macro_viewer&id=instant_html%21nabble%3Aemail.naml&base=nabble.naml.namespaces.BasicNamespace-nabble.view.web.template.NabbleNamespace-nabble.view.web.template.NodeNamespace&breadcrumbs=notify_subscribers%21nabble%3Aemail.naml-instant_emails%21nabble%3Aemail.naml-send_instant_email%21nabble%3Aemail.naml>
>>
>
>
> ------------------------------------------------------------------------
> View this message in context: Re: DFAPacketizer with StateTrans != 0 
> Assertion 
>
<http://llvm.1065342.n5.nabble.com/DFAPacketizer-with-StateTrans-0-Assertion-tp46105p46112.html>
> Sent from the LLVM - Dev mailing list archive 
> <http://llvm.1065342.n5.nabble.com/LLVM-Dev-f3.html> at Nabble.com.
>
>
> _______________________________________________
> LLVM Developers mailing list
> LLVMdev at cs.uiuc.edu         http://llvm.cs.uiuc.edu
> http://lists.cs.uiuc.edu/mailman/listinfo/llvmdev
-------------- next part --------------
An HTML attachment was scrubbed...
URL:
<http://lists.llvm.org/pipermail/llvm-dev/attachments/20120612/82ec1f19/attachment.html>

Thanks Anshu and Ivan,

It appears the problem was arising from the naming and usage of functional
units across different subtargets. I have renamed the units for each target
and it now works as expected.

--
View this message in context:
http://llvm.1065342.n5.nabble.com/DFAPacketizer-with-StateTrans-0-Assertion-tp46105p46144.html
Sent from the LLVM - Dev mailing list archive at Nabble.com.