[llvm-commits] [PATCH] Test cases for constant gep's with various address spaces.

[Chandler Carruth]

On Thu, Oct 25, 2012 at 2:07 PM, Micah Villmow <villmow at gmail.com> wrote:
>
>   The various files are just the same test but with various address spaces and pointer sizes being tested. This was done to make sure it is handling all the common cases correctly. Anything relevant to -as0 also applies to -as[1-5].

Then let's get the first one in, and you can send the rest with a
nice, smaller diff relative to the first one.

Growing the set of code in a review mid-review is really hard on the reviewers.

>
>
> ================
> Comment at: test/Transforms/InstCombine/opt-constant-fold-gep-0.ll:76-78
> @@ +75,5 @@
> +
> +; OPT: define i8 addrspace(1)* @goo8() nounwind {
> +; OPT:   ret i8 addrspace(1)* null
> +; OPT: }
> +; OPT: define i1 addrspace(2)* @goo1() nounwind {
> ----------------
> Chandler Carruth wrote:
>> Rather than testing the exact output of the function, you should sink this to live next to the function, and do a more weak assertion:
>>
>> define i8 addrspace(1)* @goo8() {
>> ; CHECK: @goo8
>> ; CHECK: ret i8 addrspace(1)* null
>>   ret ....
>> }
> This is legacy because of the test I inherited this from, I'll clean it up(Same with most of the other comments).
>
>
> http://llvm-reviews.chandlerc.com/D79
> _______________________________________________
> llvm-commits mailing list
> llvm-commits at cs.uiuc.edu
> http://lists.cs.uiuc.edu/mailman/listinfo/llvm-commits