llvm dev - Apr 2013 - [LLVMdev] Is r174746 broken on ARM?

On 04/10/2013 10:22 PM, Hal Finkel wrote:
>> Okay, great! Can you please generate a test case? In case you don't
>> know, an easy way is this: place an assert that will crash the code
>> generation in the pre-dec case you've highlighted, then use
bugpoint
>> to produce a reduced test case for the crash.
>
> [list not cc'd]
>
> Were you able to make any progress on this?
Yes, a kind of - attached at http://llvm.org/bugs/show_bug.cgi?id=15581#c8.

Dmitry

----- Original Message -----
> From: "Dmitry Antipov" <antipov at dev.rtsoft.ru>
> To: "Hal Finkel" <hfinkel at anl.gov>
> Cc: llvmdev at cs.uiuc.edu
> Sent: Thursday, April 11, 2013 1:48:07 AM
> Subject: Re: [LLVMdev] Is r174746 broken on ARM?
> 
> On 04/10/2013 10:22 PM, Hal Finkel wrote:
> 
> >> Okay, great! Can you please generate a test case? In case you
> >> don't
> >> know, an easy way is this: place an assert that will crash the
> >> code
> >> generation in the pre-dec case you've highlighted, then use
> >> bugpoint
> >> to produce a reduced test case for the crash.
> >
> > [list not cc'd]
> >
> > Were you able to make any progress on this?
> 
> Yes, a kind of - attached at
> http://llvm.org/bugs/show_bug.cgi?id=15581#c8.
I see that you've posted a proposed fix including a test case, great! In the
future, you should post these to the llvm-commits list for review. In this case,
I have only one comment:

Is is safe to add something like this?
assert(AM == ISD::PRE_DEC && OV <= 0 && "Offset for
PRE_DEC must be negative");

Do you have commit access?

Thanks again,
Hal
> 
> Dmitry
> 
>