On Tue, 20 Sep 2016, Diego Biurrun wrote:
This avoids SIMD-optimized functions having to sign-extend their
stride argument manually to be able to do pointer arithmetic.
Also rename all such parameters to "stride" for consistency.
---
libavcodec/arm/vc1dsp_init_neon.c | 2 +-
libavcodec/vc1.c | 11 ++++++-----
libavcodec/vc1_block.c | 21 +++++++++++++--------
libavcodec/vc1_loopfilter.c | 8 +++++---
libavcodec/vc1_pred.c | 9 ++++++---
libavcodec/vc1dsp.c | 26 +++++++++++++-------------
libavcodec/vc1dsp.h | 16 ++++++++--------
libavcodec/x86/vc1dsp.asm | 22 +++++++++++-----------
libavcodec/x86/vc1dsp_init.c | 16 ++++++++--------
libavcodec/x86/vc1dsp_mmx.c | 21 +++++++++++----------
10 files changed, 82 insertions(+), 70 deletions(-)
diff --git a/libavcodec/x86/vc1dsp.asm b/libavcodec/x86/vc1dsp.asm
index adf08d7..9136ad9 100644
--- a/libavcodec/x86/vc1dsp.asm
+++ b/libavcodec/x86/vc1dsp.asm
@@ -237,19 +237,19 @@ cglobal vc1_h_loop_filter_internal
VC1_H_LOOP_FILTER 4, r4
ret
-; void ff_vc1_v_loop_filter4_mmxext(uint8_t *src, int stride, int pq)
+; void ff_vc1_v_loop_filter4_mmxext(uint8_t *src, ptrdiff_t stride, int pq)
cglobal vc1_v_loop_filter4, 3,5,0
START_V_FILTER
call vc1_v_loop_filter_internal
RET
I don't see the corresponding asm simplification as the commit message
touts. I.e., this is probably a latent bug; fix that first with the proper
sign extensions before scrambling things by changing the signature.
// Martin
_______________________________________________
libav-devel mailing list
libav-devel@libav.org
https://lists.libav.org/mailman/listinfo/libav-devel