[llvm] r182585 - Setting the default value (fixes CRT assertions about uninitialized variable use when doing debug MSVC builds), and fixing coding style.
Bill Wendling
isanbard at gmail.com
Thu May 23 14:08:05 PDT 2013
Hi Tom,
While a performance hit isn’t ideal, I can’t accept this patch right now because it’s not a show-stopper. I’ll let you add it to the dot-release.
-bw
On May 23, 2013, at 2:02 PM, Tom Stellard <tom at stellard.net> wrote:
> Hi Bill,
>
> In the worst case the use of this variable uninitialized will cause a
> very slight performance decrease in programs, so I don't really
> consider it a critical fix from that perspective. However, in general
> using uninitialized values is bad, so I'm OK with merging this to 3.3, but
> I'll leave the final decision up to you.
>
> In any case, I would still like to maintain a stable 3.3 branch after
> the release and do a dot release at some point in the future, so if this
> patch doesn't go into 3.3, I'll pull it in for 3.3.1.
>
> -Tom
>
> On Thu, May 23, 2013 at 04:45:50PM -0400, Aaron Ballman wrote:
>> This wouldn't be a regression since the R600 was experimental before,
>> but it does fix a pretty serious bug with using an uninitialized
>> variable.
>>
>> ~Aaron
>>
>> On Thu, May 23, 2013 at 4:38 PM, Bill Wendling <isanbard at gmail.com> wrote:
>>> We’re not accepting any more patches unless they fix serious regressions.
>>>
>>> -bw
>>>
>>> On May 23, 2013, at 11:33 AM, Aaron Ballman <aaron at aaronballman.com> wrote:
>>>
>>>> Should this be a candidate for 3.3?
>>>>
>>>> ~Aaron
>>>>
>>>> On Thu, May 23, 2013 at 10:55 AM, Aaron Ballman <aaron at aaronballman.com> wrote:
>>>>> Author: aaronballman
>>>>> Date: Thu May 23 09:55:00 2013
>>>>> New Revision: 182585
>>>>>
>>>>> URL: http://llvm.org/viewvc/llvm-project?rev=182585&view=rev
>>>>> Log:
>>>>> Setting the default value (fixes CRT assertions about uninitialized variable use when doing debug MSVC builds), and fixing coding style.
>>>>>
>>>>> Modified:
>>>>> llvm/trunk/lib/Target/R600/R600ControlFlowFinalizer.cpp
>>>>>
>>>>> Modified: llvm/trunk/lib/Target/R600/R600ControlFlowFinalizer.cpp
>>>>> URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Target/R600/R600ControlFlowFinalizer.cpp?rev=182585&r1=182584&r2=182585&view=diff
>>>>> ==============================================================================
>>>>> --- llvm/trunk/lib/Target/R600/R600ControlFlowFinalizer.cpp (original)
>>>>> +++ llvm/trunk/lib/Target/R600/R600ControlFlowFinalizer.cpp Thu May 23 09:55:00 2013
>>>>> @@ -320,7 +320,7 @@ public:
>>>>> virtual bool runOnMachineFunction(MachineFunction &MF) {
>>>>> unsigned MaxStack = 0;
>>>>> unsigned CurrentStack = 0;
>>>>> - bool hasPush;
>>>>> + bool HasPush = false;
>>>>> for (MachineFunction::iterator MB = MF.begin(), ME = MF.end(); MB != ME;
>>>>> ++MB) {
>>>>> MachineBasicBlock &MBB = *MB;
>>>>> @@ -349,7 +349,7 @@ public:
>>>>> case AMDGPU::CF_ALU_PUSH_BEFORE:
>>>>> CurrentStack++;
>>>>> MaxStack = std::max(MaxStack, CurrentStack);
>>>>> - hasPush = true;
>>>>> + HasPush = true;
>>>>> case AMDGPU::CF_ALU:
>>>>> I = MI;
>>>>> AluClauses.push_back(MakeALUClause(MBB, I));
>>>>> @@ -470,7 +470,7 @@ public:
>>>>> break;
>>>>> }
>>>>> }
>>>>> - MFI->StackSize = getHWStackSize(MaxStack, hasPush);
>>>>> + MFI->StackSize = getHWStackSize(MaxStack, HasPush);
>>>>> }
>>>>>
>>>>> return false;
>>>>>
>>>>>
>>>>> _______________________________________________
>>>>> 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
More information about the llvm-commits
mailing list