[x264-devel] Re: [PATCH] IDCT8 and SA8D routines in Altivec

[Loren Merritt]

On Mon, 27 Nov 2006, Guillaume Poirier wrote:

> Attached patch implements $SUBJ.
> It's basically the same code as the one I posted a few month ago,
> except that this code doesn't take advantage of some memory alignments
> that would make the code faster.
>
> These routines are already faster than existing C code that's why I'm
> offering them for inclusion, hoping that I have some time in the
> future to write code that takes advantage of aligned loads and stores.
> (I'm way to busy right now though)
>
> I've been given an svn write access by Antoine Cellerier and Eric
> Petit so I can commit it whenever you guys feel OK about this.

No complaints about the content, just cosmetics.

inconsistent indentation:

+    vec_s16_t b0v = vec_add(a0v, a2v); \
+    /* int    b2  =         a0 - a2; */\
+   vec_s16_t  b2v = vec_sub(a0v, a2v); \
+    /* int    b1  =         a1 + a3; */\
+    vec_s16_t b1v = vec_add(a1v, a3v); \

+    vec_s16_t diff0v, diff1v, diff2v, diff3v,
+        diff4v, diff5v, diff6v, diff7v;

+    vec_s16_t sa8d0v, sa8d1v, sa8d2v, sa8d3v,
+        sa8d4v, sa8d5v, sa8d6v, sa8d7v;

and some functions have cuddled braces while others have the open brace on 
its own line.

--Loren Merritt

-- 
This is the x264-devel mailing-list
To unsubscribe, go to: http://developers.videolan.org/lists.html