search for: offsetidx

...t; The relevant code is a little farther down: > > APInt OV = > cast<ConstantSDNode>(Offset)->getAPIntValue(); > if (AM == ISD::PRE_DEC) > OV = -OV; > > ConstantSDNode *CN = > cast<ConstantSDNode>(OtherUses[i]->getOperand(OffsetIdx)); > APInt CNV = CN->getAPIntValue(); > if (OtherUses[i]->getOpcode() == ISD::SUB && OffsetIdx == 1) > CNV += OV; > else > CNV -= OV; > > perhaps something here is not quite right. I suspect that the first snippet (where OV is inverte...

...nly believe that there is a problem somewhere. The relevant code is a little farther down: APInt OV = cast<ConstantSDNode>(Offset)->getAPIntValue(); if (AM == ISD::PRE_DEC) OV = -OV; ConstantSDNode *CN = cast<ConstantSDNode>(OtherUses[i]->getOperand(OffsetIdx)); APInt CNV = CN->getAPIntValue(); if (OtherUses[i]->getOpcode() == ISD::SUB && OffsetIdx == 1) CNV += OV; else CNV -= OV; perhaps something here is not quite right. -Hal > > Thanks in advance, > Dmitry > > > > >

Hello Hal, I have a strong suspicion that your constant folding optimization introduced at r174746 is broken on ARM. There is a bug about it: http://llvm.org/bugs/show_bug.cgi?id=15581 There is no such issue with 3.2, and reverting r174746 on top of r178740 also fixes the problem. I'm trying to fix it myself, but still have no good ideas; so it would be great to have an advice from you.

I've looked at the ExceptionDemo and the SysV AMD64 ABI and I'm unclear on who, and when, the 'exception_cleanup' function is called. I have some working exceptions but my cleanup routine is never called. In the ExceptionDemo I turned on debug output and saw that the delete function is also not called. -- edA-qa mort-ora-y -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- --

Hello all, can someone do a quick review of the one-line proposal fix attached to http://llvm.org/bugs/show_bug.cgi?id=15581? I'm not sure that I can explain why it works, but it (hopefully) is, so I really need a feedback with this. Thanks, Dmitry

...d on its position in the instruction using it? I can vaguely see there might be some argument that we have to be careful about the 4 variants "B+O", "B-O", "O+B", "O-B" but the code a few lines later seems to be trying to handle that ("if (SUB && OffsetIdx == 1)"). That's where I'd be inclined to place my change if this bit of code really is wrong. At the very least the change is subtle enough that I'd want a clear explanation in a nearby comment, I think. Hopefully that explanation can make it clear the patch is correct too. Also,...