[llvm-commits] [PATCH] opt -Os/-Oz

[Patrik Hägglund H]

> (Any chance you're still interested in it Patrik?)

No. Our llvm target is still in its infancy, and optimization pragmas for specific functions or blocks of code in the fronted, is not on the agenda yet.

Therefore, I don't see the point of the optsize attribute. The only reason we use it in our frontend is that -Oz + optsize - for some strange reason I havn't investigated yet, but I should - gives smaller code than -Oz alone. When that issue is solved, we will not use optsize anymore.

/Patrik Hägglund

________________________________
From: Chandler Carruth [mailto:chandlerc at google.com]
Sent: den 15 maj 2012 23:44
To: Evan Cheng
Cc: Patrik Hägglund H; <llvm-commits at cs.uiuc.edu>
Subject: Re: [llvm-commits] [PATCH] opt -Os/-Oz

On Tue, May 15, 2012 at 12:42 PM, Evan Cheng <evan.cheng at apple.com<mailto:evan.cheng at apple.com>> wrote:
While adding opt options is useful, the thing we really need is to model -Oz as a new attribute. Right now we have OptimizeForSize, it really should be split into two: OptimizeForSizeAndSpeed and OptimizeForSize. The former is what OptimizeForSize currently is (for -Os), the later should be for -Oz.

Completely agree. The only reason for adding this is to enable testing of -Oz-behavior at all; I think fixing the attributes relating to this should be an orthogonal fix? (Any chance you're still interested in it Patrik?)
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.llvm.org/pipermail/llvm-commits/attachments/20120516/d094e6c7/attachment.html>