[llvm-commits] [PATCH] Multidimensional Array Index Delinearization Analysis

[Sebastian Pop]

Hal Finkel wrote:
> I think that there should be a change in the code. If we have an input sub-expression that looks like:
> 
>  zext (x * 4) from i32 to i64
> 
> but by the time that visitConstant is called on the (i32 4), I could have GCD
> = (i64 8), for example. Then then we'd try calling
> APIntOps::GreatestCommonDivisor (or some other APInt operation) on (i32 4) and
> (i64 8), which will assert. However we do it, both constants need to be of the
> same bit length for the operation to work.

Thanks for the explanation, the attached patch corrects this problem.

Sebastian
-- 
Qualcomm Innovation Center, Inc. is a member of Code Aurora Forum,
hosted by The Linux Foundation
-------------- next part --------------
A non-text attachment was scrubbed...
Name: 0001-delinearization-of-arrays.patch
Type: text/x-diff
Size: 75464 bytes
Desc: not available
URL: <http://lists.llvm.org/pipermail/llvm-commits/attachments/20131103/12f4f3dc/attachment.patch>