Re: [wwwdocs] [patch] vectorizer webpage update

[Dorit Naishlos <DORIT at il dot ibm dot com>]

> This does not look like valid HTML, have you validated it?
yes I did

applied your comments and committed

thanks

dorit




                                                                                                                       
                      Gerald Pfeifer                                                                                   
                      <gerald@pfeifer.c        To:       Dorit Naishlos/Haifa/IBM@IBMIL                                
                      om>                      cc:       gcc-patches@gcc.gnu.org                                       
                                               Subject:  Re: [wwwdocs] [patch] vectorizer webpage update               
                      19/08/2004 12:21                                                                                 
                                                                                                                       




Hi Dorit,

On Wed, 18 Aug 2004, Dorit Naishlos wrote:
> update with current status of vectorizer on lno-branch and mainline.
> (attached both the diff and the new html doc)

would you mind attaching patches as text instead of
Application/OCTET-STREAM?

Much of the patch is actually mostly reformatting, which also made
reviewing a bit harder.  Feel free to commit such reformatings without
explicit approval in the future, as a separate patch.

The patch is fine if you address the following issues in the
version you are going to commit.

Is the term "powerpc platforms" (in this capitalization) really correct?

!     <p>Examples of loops that can currently be vectorized by
!     lno-branch (unless indicated otherwise). "feature" indicates
!     the vectorization capabilities demonstrated by the
!     example.</p><strong><a name="example1" id=
!     "example1">example1:</a></strong>

This does not look like valid HTML, have you validated it?  I suspect
that <strong> needs to be embedded in <p>.

Is there any specific reason you are using both "name=..." and "id=..."?
In my experience, just one should suffice, and it will make the source
easier to maintain and read.

I have to admit I find

   + </pre><a name="example10" id=
   + "example10"><strong>example10</strong>:</a>
   +     <pre>

somewhat hard to read.  Why do you indent <pre>, but not </pre>?  Would
you mind changing that (not to indent either one) and leave an empty line
before every <pre>?

!                 <li>The loop bound (number of iterations) in known
!                 and divides by the vectorization factor.</li>

Typo. "is" instead of "in".

!                 <li>Supported memory accesses are onedimensional

"one-dimensional"?

!                 arrays, whose alignment can be forced (not extern,
!                 and stack_boundary of target platform allows it),
                       ^^^^^^^^^^^^^^

"stack boundary"?

!                 are <strong>multi</strong>dimensional arrays, whose

"multi-dimensional"?

!                 alignment can be forced (not extern), and aligned

By extern do you refer to the C/C++ extern keyword?  If so, please use
<code>extern</code>.

!                 and/or/xor/1's-complement, according to available
!                 vector support in the target platform.</li>

"on the target platform"

!             <p>New test-cases:</p>

According to codingconventions.html, this should be "testcase".

Gerald
--
Gerald Pfeifer (Jerry)   gerald@pfeifer.com
http://www.pfeifer.com/gerald/