[XviD-devel] [CVS Commit] Fixed quant4_intra_xmm and quant_intra_3dne bug for DC<0.

[Marco Al]

From: "Edouard Gomez" <ed.gomez at free.fr>

>     That was always decreasing the ressult by 1 for negative DC values.
>     A not so bad solution is simply to use a cmov instruction and choose
>     the right value according to the DC value. As these function were
>     for PIII and Athlon, we are sure we can use the cmov instruction.

Wouldnt it be faster to reuse the result in edi? (0 if positive, -1 if
negative)

Marco

PS. quantize_3dne better be awfull fast to be that unreadable :)