Oh, I forgot about svq3. It totally breaks svq3 and I never ran FATE on it.
Let me fix that tomorrow. Jason On Sun, Jun 12, 2011 at 7:22 PM, Michael Niedermayer <[email protected]> wrote: > On Thu, Jun 09, 2011 at 06:55:15PM -0700, Jason Garrett-Glaser wrote: >> Updated with a fix to high bit depth wpred. Ronald has been sent FATE >> samples. > > I tried to apply this but it breaks fate (some SVQ3 sample) > I just wanted to confirm that svq3 fate is working on your side ? > > your asm is also buggy, you need: (or something equivalent) > > diff --git a/libavcodec/x86/h264_i386.h b/libavcodec/x86/h264_i386.h > index 096dfd6..2f55879 100644 > --- a/libavcodec/x86/h264_i386.h > +++ b/libavcodec/x86/h264_i386.h > @@ -36,7 +36,7 @@ > #if ARCH_X86 && HAVE_7REGS && HAVE_EBX_AVAILABLE && > !defined(BROKEN_RELOCATIONS) > static int decode_significance_x86(CABACContext *c, int max_coeff, > uint8_t *significant_coeff_ctx_base, > - int *index, int last_off){ > + int *index, x86_reg last_off){ > void *end= significant_coeff_ctx_base + max_coeff - 1; > int minusstart= -(int)significant_coeff_ctx_base; > int minusindex= 4-(int)index; > @@ -92,7 +92,7 @@ static int decode_significance_x86(CABACContext *c, int > max_coeff, > > static int decode_significance_8x8_x86(CABACContext *c, > uint8_t *significant_coeff_ctx_base, > - int *index, int last_off, const > uint8_t *sig_off){ > + int *index, x86_reg last_off, const > uint8_t *sig_off){ > int minusindex= 4-(int)index; > int coeff_count; > x86_reg last=0; > > [...] > > -- > Michael GnuPG fingerprint: 9FF2128B147EF6730BADF133611EC787040B0FAB > > it is not once nor twice but times without number that the same ideas make > their appearance in the world. -- Aristotle > > -----BEGIN PGP SIGNATURE----- > Version: GnuPG v1.4.10 (GNU/Linux) > > iEYEARECAAYFAk31dFgACgkQYR7HhwQLD6t0JgCdE2nRAVknr3zzfAjq4CEGBOLZ > rxoAn0WgEzPXchfZSdPkYwAXj3MKmzh4 > =uTOy > -----END PGP SIGNATURE----- > > _______________________________________________ libav-devel mailing list [email protected] https://lists.libav.org/mailman/listinfo/libav-devel
