[llvm] r179480 - Document the SLP infrastructure.

[Hal Finkel]

----- Original Message -----
> From: "Nadav Rotem" <nrotem at apple.com>
> To: "Hal Finkel" <hfinkel at anl.gov>
> Cc: llvm-commits at cs.uiuc.edu
> Sent: Sunday, April 14, 2013 10:21:53 PM
> Subject: Re: [llvm] r179480 - Document the SLP infrastructure.
> 
> 
> 
> 
> On Apr 14, 2013, at 1:11 AM, Hal Finkel < hfinkel at anl.gov > wrote:
> 
> 
> This is great! We should probably change the Clang flags around
> sooner rather than later to avoid confusing the users. Currently
> -fslp-vectorize == bb-vectorize, but should probably map to this
> pass by default instead. Maybe -fslp-vectorize-aggressive for
> bb-vectorize?
> 
> Thanks again,
> Hal
> 
> 
> 
> Hi Hal,
> 
> 
> Thank you. Do you mind if we keep one -fslp-vectorize flag in clang
> and add another llvm flag for selecting the algorithm ?


> Maybe
> something like '-mllvm -bb-vectorize' ?

;) We have that already, kind of; the issue is that the pass-manager flag and the pass name need to be different. I suppose with the loop vectorizer we have -vectorize-loops and -loop-vectorize, so we could do something similar for bb-vectorize. Then we can have -vectorize map to the combination of slp-vectorize and loop-vectorize. Regardless, we should definitely change the -mllvm -vectorize == bb-vectorize mapping, because that is also confusing at this point.

 -Hal

> 
> 
> Thanks,
> Nadav