[llvm-commits] [llvm] r127539 - /llvm/trunk/lib/Transforms/InstCombine/InstCombineLoadStoreAlloca.cpp

Duncan Sands baldrick at free.fr
Sat Mar 12 06:32:57 PST 2011


Hi Jin Gu,

> I sended testcase on previous e-mail as following:

it should be added to the testsuite.

> and
>
>>    %A = alloca i32
>>    %Ptr = getelementptr A, 0
>>    %Ptr2 = bitcast i32 *%Ptr, i32*
>>    %LI = load %Ptr2
>>    do other stuff with Ptr2 that modifies the contents of %A
>>    %ST = store %LI, %Ptr<-- would be wrong to remove the store
>
> "%Ptr2 = bitcast i32 *%Ptr, i32*" instruction is removed on InstCombine Pass
> so equivalentAddressValues does not check this instrucion.

I don't think you should rely on this.  I think you should rework the check so
that it doesn't make any such "global" assumptions (i.e. assumptions about what
code far away from this point does), only local assumptions (i.e. that can be
verified by just reading the function itself).  An alternative is to add some
assertions that check the assumptions you are relying on.

Ciao, Duncan.

PS: Please don't forget to add some descriptive comments.



More information about the llvm-commits mailing list