[MPlayer-dev-eng] AMD64 assembler support

[Aurelien Jacobs]

On Sun, 31 Jul 2005 20:34:01 +0100
Johan Groth <johan.groth at cell-telecom.co.uk> wrote:

> Aurelien Jacobs wrote:
> > On Sat, 30 Jul 2005 17:52:18 +0100
> > Johan Groth <johan.groth at cell-telecom.co.uk> wrote:
> > 
> > 
> >>Rich Felker wrote:
> >>
> >>
> >>>Please use REG_a, REG_b, etc. names like elsewhere in MPlayer, for
> >>>consistency. Using EAX, EBX, etc. for the macros is confusing
> >>>because those are the names of the 32bit registers, and on amd64
> >>>they'll be defined to the 64bit regs..
> >>
> >>Ok, fixed in the attached patch.
> > 
> > 
> > I've reworked your patch (cleaned, fixed...).
> > It seems pretty good now, so I will probably commit.
> 
> I would like to see one small change before commit and that is in
> function:
> 
> static int a52_resample_2F_2R_LFE_to_6_MMX(float * _f, int16_t * s16){
> in resample_mmx.c
> 
> There is a line in that function that reads:
> 		"leal (%%"REG_si", %%"REG_si", 4), %%edi\n\t"
> 
> It must of course be:
> 	"lea (%%"REG_si", %%"REG_si", 2), %%"REG_di"	\n\t"
> 
> because rdi is used in a couple indirect address instructions further
> down.

Ok, I missed some of them (they was also missing in the original patch).
New patch fix this issue and all the previous ones.

Aurel
-------------- next part --------------
An embedded and charset-unspecified text was scrubbed...
Name: liba52_amd64.diff
URL: <http://lists.mplayerhq.hu/pipermail/mplayer-dev-eng/attachments/20050731/401e9cb9/attachment.asc>