[llvm-commits] [llvm] r152847 - /llvm/trunk/lib/Target/ARM/ARMFastISel.cpp
Chad Rosier
mcrosier at apple.com
Thu Mar 15 16:12:19 PDT 2012
Eli's comments have been addressed in r152869. Thanks, Eli!
Chad
On Mar 15, 2012, at 3:37 PM, Chad Rosier <mcrosier at apple.com> wrote:
>
> On Mar 15, 2012, at 3:06 PM, Eli Friedman <eli.friedman at gmail.com> wrote:
>
>> On Thu, Mar 15, 2012 at 2:40 PM, Chad Rosier <mcrosier at apple.com> wrote:
>>> Author: mcrosier
>>> Date: Thu Mar 15 16:40:23 2012
>>> New Revision: 152847
>>>
>>> URL: http://llvm.org/viewvc/llvm-project?rev=152847&view=rev
>>> Log:
>>> [fast-isel] Don't try to encode LONG_MIN using cmn instructions.
>>> rdar://11038907
>>>
>>> Modified:
>>> llvm/trunk/lib/Target/ARM/ARMFastISel.cpp
>>>
>>> Modified: llvm/trunk/lib/Target/ARM/ARMFastISel.cpp
>>> URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Target/ARM/ARMFastISel.cpp?rev=152847&r1=152846&r2=152847&view=diff
>>> ==============================================================================
>>> --- llvm/trunk/lib/Target/ARM/ARMFastISel.cpp (original)
>>> +++ llvm/trunk/lib/Target/ARM/ARMFastISel.cpp Thu Mar 15 16:40:23 2012
>>> @@ -1384,12 +1384,16 @@
>>> SrcVT == MVT::i1) {
>>> const APInt &CIVal = ConstInt->getValue();
>>> Imm = (isZExt) ? (int)CIVal.getZExtValue() : (int)CIVal.getSExtValue();
>>> - if (Imm < 0) {
>>> - isNegativeImm = true;
>>> - Imm = -Imm;
>>> + // We can't encode LONG_MIN (i.e., 0x80000000) as an immediate because
>>> + // there is no way to represent 2147483648 as a signed 32-bit int.
>>> + if (Imm != (int)0x80000000) {
>>> + if (Imm < 0) {
>>> + isNegativeImm = true;
>>> + Imm = -Imm;
>>> + }
>>> + UseImm = isThumb2 ? (ARM_AM::getT2SOImmVal(Imm) != -1) :
>>> + (ARM_AM::getSOImmVal(Imm) != -1);
>>> }
>>> - UseImm = isThumb2 ? (ARM_AM::getT2SOImmVal(Imm) != -1) :
>>> - (ARM_AM::getSOImmVal(Imm) != -1);
>>
>> Why not just skip the negation rather than forcing it not to use an
>> immediate? ARM_AM::getT2SOImmVal(INT_MIN) will succeed, AFAIK.
>
> Ah, yes. That would be a better fix. I'll double check if it works.
>
>> Also, missing test case.
>
> Coming up!
>
> Thanks, Eli.
>
>> -Eli
>
> _______________________________________________
> 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