[llvm-commits] [llvm] r56384 - in /llvm/trunk/lib/CodeGen: RegAllocLinearScan.cpp SimpleRegisterCoalescing.cpp

[Evan Cheng]

On Sep 23, 2008, at 4:31 PM, Dale Johannesen wrote:

>
> On Sep 23, 2008, at 4:27 PMPDT, Evan Cheng wrote:
>
>>
>> On Sep 23, 2008, at 3:17 PM, Dale Johannesen wrote:
>>>>
>>>
>>> This did basically Just Work (I haven't run the gcc testsuite yet,
>>> but
>>> it doesn't break anything in
>>> the llvm testsuite and fixes all the examples in our bug reports.)
>>> Unfortunately, this approach
>>> only works in the linear scan RA, while we use a different RA with -
>>> O0.... I guess I need to fix
>>> that before turning it on.
>>
>> Yay. Good thing I was wrong. :-) Would you like to add the support
>> that allows us to model it as operand constraint so we can test it on
>> more things?
>
> Are you talking about the "minor tblgen and TargetInstrDesc changes"
> described below?   Not sure I understand how that would work, and I
> don't really want to wait for it before checking in.

You don't have to wait for it. I'll take care of it some time this week.

Evan

>
>
>> Evan
>>
>>>
>>>
>>>> Owen, can your liveintervals renumbering work be impacted?
>>>> i.e. Does it assume def must be one cycle before use?
>>>>
>>>>>
>>>>>
>>>>>>> Can earlyclobber operands be used to model the armv4 "mul"
>>>>>>> register
>>>>>>> weirdness where the input and output regs can't be the same?
>>>>>>
>>>>>> We could, but the correct way to model it is to push the use down
>>>>>> one
>>>>>> cycle, not def up one cycle. The problem with early clobber is  
>>>>>> the
>>>>>> def
>>>>>> cannot overlap any of the uses (both Rm and Rs). In the ARM mul
>>>>>> case,
>>>>>> the def cannot overlap Rm only.
>>>>>
>>>>> Sure, I agree with that.  This is also true of most inline asms
>>>>> with
>>>>> multiple instructions (like the example I gave).  Using
>>>>> earlyclobber
>>>>> should be a conservative but correct solution (and much better  
>>>>> than
>>>>> using the 'multiply fixer' pass that used to exist).
>>>>>
>>>>> This would also make it easier to test that earlyclobber is  
>>>>> working
>>>>> right (multiplies are more frequent than & in asms), which is  
>>>>> why I
>>>>> brought it up. :)
>>>>
>>>> That will require some minor tblgen and TargetInstrDesc changes. If
>>>> someone want to do that, sure.
>>>>
>>>> Evan
>>>>
>>>>>
>>>>>
>>>>> -Chris
>>>>>
>>>>> _______________________________________________
>>>>> llvm-commits mailing list
>>>>> llvm-commits at cs.uiuc.edu
>>>>> http://lists.cs.uiuc.edu/mailman/listinfo/llvm-commits
>>>>
>>>> _______________________________________________
>>>> llvm-commits mailing list
>>>> llvm-commits at cs.uiuc.edu
>>>> http://lists.cs.uiuc.edu/mailman/listinfo/llvm-commits
>>>
>>> _______________________________________________
>>> llvm-commits mailing list
>>> llvm-commits at cs.uiuc.edu
>>> http://lists.cs.uiuc.edu/mailman/listinfo/llvm-commits
>>
>> _______________________________________________
>> llvm-commits mailing list
>> llvm-commits at cs.uiuc.edu
>> http://lists.cs.uiuc.edu/mailman/listinfo/llvm-commits
>
> _______________________________________________
> llvm-commits mailing list
> llvm-commits at cs.uiuc.edu
> http://lists.cs.uiuc.edu/mailman/listinfo/llvm-commits