[llvm] r194527 - delinearization of arrays

[Eric Christopher]

On Wed, Nov 13, 2013 at 2:46 PM, Sebastian Pop <spop at codeaurora.org> wrote:
> Eric Christopher wrote:
>> On Wed, Nov 13, 2013 at 12:15 PM, Sebastian Pop <spop at codeaurora.org> wrote:
>> > Hi,
>> >
>> > Eric Christopher wrote:
>> >> To pile on... comments man, comments. Lots and lots of algorithm etc
>> >> comments.
>> >
>> > The SCEV->delinearize algorithm computes the multiples of SCEV coefficients:
>> > that is a pattern matching of sub expressions in the stride and base of a SCEV
>> > corresponding to the computation of a GCD (greatest common divisor) of base and
>> > stride.
>> >
>> > Could you please indicate in the patch the places where you would like to see
>> > more comments?
>> >
>>
>> Pretty much...
>>
>> >> Comments on whatever you're testing in the testcases since
>> >
>> > I think the testcases are pretty explicit about what they are testing: if the
>> > description I gave in the previous email is not enough, let me know: I can give
>> > more details.
>> >
>>
>> All of this. Anything you felt you needed to write in that email
>> should be in the code. Block comments in front of functions, algorithm
>> descriptions with links, etc.
>
> See r194612, and let me know if there are other comments you would like to see
> in the patch.
>
>>
>> >> right now if they start failing I'm just going to delete the pass :)
>> >
>> > Just let me know before you do that, and I'll fix the testcases ;-)
>> >
>>
>> A short summary in each testcase of what the various lines mean would be good.
>
> I have added a general comment about what the testsuite looks for and how to
> read the outputs the testsuite contains in the comments of SCEV->delinearize in
> ScalarEvolution.cpp.  Let me know if that is not enough.
>

Seems reasonable if a bit odd on location for some of the explanations. :)

-eric

> Thanks for your review!
>
> Sebastian
> --
> Qualcomm Innovation Center, Inc. is a member of Code Aurora Forum,
> hosted by The Linux Foundation