[FFmpeg-devel] [PATCH] move h264 loopfilter strength code to yasm

[Reimar Döffinger]

On Fri, Sep 24, 2010 at 04:34:36PM -0400, Ronald S. Bultje wrote:
> Hi,
> 
> On Fri, Sep 24, 2010 at 3:30 PM, Alexander Strange
> <astrange at ithinksw.com> wrote:
> > "g" permits registers, so it could generate something like this:
> > ? por %rax(%rcx), %mm1
> 
> Is that even legal?
> 
> > You have to use "m" or MANGLE() in this situation.
> 
> So what would that work like?
> 
>                 "por "MANGLE(d_idx)"(%0), %%mm1 \n" // nnz[b] || nnz[bn]
> 
> gives
> 
> /var/folders/Rz/RzQTCSLsFPWQeOEO5EXsJE+++TI/-Tmp-//ccNBPRga.s:341:junk
> `(%rip)' after expression
> /var/folders/Rz/RzQTCSLsFPWQeOEO5EXsJE+++TI/-Tmp-//ccNBPRga.s:341:32-bit
> absolute addressing is not supported for x86-64
> /var/folders/Rz/RzQTCSLsFPWQeOEO5EXsJE+++TI/-Tmp-//ccNBPRga.s:341:cannot
> do signed 4 byte relocation

Have you tried looking at the generated assembler (e.g. compiling
with -S instead of -c)?
I guess the argument is of type int, so it will try to use
eax instead of rax.