[LLVMdev] SelectionDAG viewers, filter-view-dags question

[Ryan Taylor]

Ah, I missed that subltety. Thanks.
On Aug 1, 2015 2:43 PM, "Mehdi Amini" <mehdi.amini at apple.com> wrote:

> The diff is not only the && and || but also the leading !:
>
>
> diff --git a/lib/CodeGen/SelectionDAG/SelectionDAGISel.cpp
> b/lib/CodeGen/SelectionDAG/SelectionDAGISel.cpp
> index 58f029fbe9fc..7ee06fc153b2 100644
> --- a/lib/CodeGen/SelectionDAG/SelectionDAGISel.cpp
> +++ b/lib/CodeGen/SelectionDAG/SelectionDAGISel.cpp
> @@ -659,7 +659,7 @@ void SelectionDAGISel::CodeGenAndEmitDAG() {
>    (void)BlockNumber;
>    bool MatchFilterBB = false; (void)MatchFilterBB;
>  #ifndef NDEBUG
> -  MatchFilterBB = (!FilterDAGBasicBlockName.empty() &&
> +  MatchFilterBB = (FilterDAGBasicBlockName.empty() ||
>                     FilterDAGBasicBlockName ==
>                         FuncInfo->MBB->getBasicBlock()->getName().str());
>  #endif
>
>
> —
> Mehdi
>
>
>
> On Aug 1, 2015, at 11:07 AM, Ryan Taylor <ryta1203 at gmail.com> wrote:
>
> Simply replacing the && with || did not fix the issue. This issue still
> exists after making those changes. There is maybe some other code that
> needs to be changed also?
>
> Thanks.
>
> On Mon, Jul 27, 2015 at 12:45 PM, Ryan Taylor <ryta1203 at gmail.com> wrote:
>
>> Daniel,
>>
>>   Ok, thanks. Simple fix. We'll just make correction in local copy for
>> now, one less thing to port later :)
>>
>> Thanks.
>>
>> On Mon, Jul 27, 2015 at 12:29 PM, Daniel Sanders <
>> Daniel.Sanders at imgtec.com> wrote:
>>
>>> I've just looked at my checkout of 3.6 and it looks like the fix wasn't
>>> merged. I don't have the revision number for the fix to hand but in
>>> lib/CodeGen/SelectionDAG/SelectionDAGISel.cpp, this:
>>>
>>>   MatchFilterBB = (!FilterDAGBasicBlockName.empty() &&
>>>
>>>                    FilterDAGBasicBlockName ==
>>>
>>>                        FuncInfo->MBB->getBasicBlock()->getName().str());
>>>
>>> Should be:
>>>
>>>   MatchFilterBB = (FilterDAGBasicBlockName.empty() ||
>>>
>>>                    FilterDAGBasicBlockName ==
>>>
>>>                        FuncInfo->MBB->getBasicBlock()->getName().str());
>>>
>>>
>>>
>>> The trunk has the correct code so the option should be ok for LLVM 3.7.
>>>
>>>
>>>
>>> *From:* Ryan Taylor [mailto:ryta1203 at gmail.com]
>>> *Sent:* 27 July 2015 17:20
>>> *To:* Daniel Sanders
>>> *Cc:* llvmdev at cs.uiuc.edu
>>> *Subject:* Re: [LLVMdev] SelectionDAG viewers, filter-view-dags question
>>>
>>>
>>>
>>> Daniel,
>>>
>>>
>>>
>>>   We are using 3.6. Someone also pointed out that it was mandatory in
>>> 3.6.2 but I have not verified that.
>>>
>>>
>>>
>>>
>>>
>>> On Mon, Jul 27, 2015 at 12:10 PM, Daniel Sanders <
>>> Daniel.Sanders at imgtec.com> wrote:
>>>
>>> It's not supposed to be. There was a short period where it was
>>> unintentionally mandatory but this was fixed after I pointed it out during
>>> post-commit review.
>>>
>>>
>>>
>>> Which version/revision are you using?
>>>
>>>
>>>
>>> *From:* llvmdev-bounces at cs.uiuc.edu [mailto:llvmdev-bounces at cs.uiuc.edu]
>>> *On Behalf Of *Ryan Taylor
>>> *Sent:* 27 July 2015 16:27
>>> *To:* llvmdev at cs.uiuc.edu
>>> *Subject:* [LLVMdev] SelectionDAG viewers, filter-view-dags question
>>>
>>>
>>>
>>> Is this option currently mandatory? If so, why? If not, I'm not sure
>>> what's been added that I need to do differently.
>>>
>>>
>>>
>>> -view-isel-dags opened just fine in dotty in 3.4 and now this does
>>> nothing without the filter-view-dags 'option' and now has a different
>>> priority program list or something.
>>>
>>>
>>>
>>> I'm just curious why this option should be mandatory?
>>>
>>>
>>>
>>> Thanks.
>>>
>>>
>>>
>>
>>
> _______________________________________________
> 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/20150801/58f48345/attachment.html>