[FFmpeg-devel] pre discussion around Blackfin dct_quantize_bfin routine

Reimar Doeffinger Reimar.Doeffinger
Tue Jun 12 12:50:28 CEST 2007

On Tue, Jun 12, 2007 at 06:34:22AM -0400, Marc Hoffman wrote:
> On Jun 12, 2007, at 6:30 AM, Reimar Doeffinger wrote:
> >On Tue, Jun 12, 2007 at 06:25:02AM -0400, Marc Hoffman wrote:
> >[...]
> >>Argh the assembler is broken each of those lines need to be
> >>concatenated.
> >
> >I expected that, that's why I did not put a \n\t at the end, like that
> >the code the assembler sees is unchanged...
> I removed those \n\t as you mentioned and it works just fine but the  
> code seems noisy to me when I read it.

You could still try to align the closing ", maybe that improves it.
It was just a suggestion anyway, you must be able to read the code first
of all.
I also assumed that they were separate instruction, if they aren't
putting them on the same line makes more sense. Also at least I am also
fine with lines of up to around 120 characters. What I disliked most
though was having the code after || hidden so far to the right. Am i
right that those are load instruction executed in parallel?
What would you think about a syntax more like this?
>        "     r0.h=(a1 =r1.h*r0.h), r0.l=(a0 =r1.l*r0.l) (is)      "
>        "||   r1=[%1++];                                       \n\t"
Maybe the assembler even allows newlines somwhere in that (though I
doubt it)...
If the way you write it now feels perfect to you, leave it as it is,
I am looking at it as someone who does not know the semantics of the
symbols, so I of course like for the structure to help me out ;-)

Reimar Doeffinger

More information about the ffmpeg-devel mailing list