Oopsy daisy! Sorry, and thanks for the revert. I will look into this.
From: Hans Wennborg
Sent: 18 September 2019 14:41
To: Sjoerd Meijer
Cc: cfe-commits
Subject: Re: r372082 - [Clang] Pragma vectorize_width() implies
vectorize(enable)
I've reverted th
I've reverted this in r372225 as it broke the Chromium build. It seems
it doesn't actually manage to enable vectorization of a simple loop at
-Oz, and at -Os it asserts. See details in my commit message.
On Tue, Sep 17, 2019 at 10:41 AM Sjoerd Meijer via cfe-commits
wrote:
>
> Author: sjoerdmeije
Author: hans
Date: Wed Sep 18 06:41:51 2019
New Revision: 372225
URL: http://llvm.org/viewvc/llvm-project?rev=372225&view=rev
Log:
Revert r372082 "[Clang] Pragma vectorize_width() implies vectorize(enable)"
This broke the Chromium build. Consider the following code:
float Sca
Author: sjoerdmeijer
Date: Tue Sep 17 01:43:11 2019
New Revision: 372082
URL: http://llvm.org/viewvc/llvm-project?rev=372082&view=rev
Log:
[Clang] Pragma vectorize_width() implies vectorize(enable)
Specifying the vectorization width was supposed to implicitly enable
vectorization, except that it
