diff options
author | Diego Biurrun <diego@biurrun.de> | 2009-02-15 17:42:08 +0000 |
---|---|---|
committer | Diego Biurrun <diego@biurrun.de> | 2009-02-15 17:42:08 +0000 |
commit | f602b044b55b986eba91aeeb598321b55c59715e (patch) | |
tree | 4b92281c7e73b3cdb05840974130933868898a08 | |
parent | 11f8aa042dc4708d3d80465fa44aefd116a9bf77 (diff) | |
download | ffmpeg-f602b044b55b986eba91aeeb598321b55c59715e.tar.gz |
Rename xvmc_pixfmt_render structure to xvmc_pix_fmt.
Originally committed as revision 17339 to svn://svn.ffmpeg.org/ffmpeg/trunk
-rw-r--r-- | libavcodec/mpegvideo_xvmc.c | 14 | ||||
-rw-r--r-- | libavcodec/xvmc.h | 2 |
2 files changed, 8 insertions, 8 deletions
diff --git a/libavcodec/mpegvideo_xvmc.c b/libavcodec/mpegvideo_xvmc.c index bd2318d6bc..1310c92bbf 100644 --- a/libavcodec/mpegvideo_xvmc.c +++ b/libavcodec/mpegvideo_xvmc.c @@ -37,7 +37,7 @@ */ void ff_xvmc_init_block(MpegEncContext *s) { - struct xvmc_pixfmt_render *render = (struct xvmc_pixfmt_render*)s->current_picture.data[2]; + struct xvmc_pix_fmt *render = (struct xvmc_pix_fmt*)s->current_picture.data[2]; assert(render && render->unique_id == AV_XVMC_RENDER_MAGIC); s->block = (DCTELEM *)(render->data_blocks + render->next_free_data_block_num * 64); @@ -64,7 +64,7 @@ void ff_xvmc_pack_pblocks(MpegEncContext *s, int cbp) */ int ff_xvmc_field_start(MpegEncContext*s, AVCodecContext *avctx) { - struct xvmc_pixfmt_render *last, *next, *render = (struct xvmc_pixfmt_render*)s->current_picture.data[2]; + struct xvmc_pix_fmt *last, *next, *render = (struct xvmc_pix_fmt*)s->current_picture.data[2]; const int mb_block_count = 4 + (1 << s->chroma_format); assert(avctx); @@ -97,7 +97,7 @@ int ff_xvmc_field_start(MpegEncContext*s, AVCodecContext *avctx) case FF_I_TYPE: return 0; // no prediction from other frames case FF_B_TYPE: - next = (struct xvmc_pixfmt_render*)s->next_picture.data[2]; + next = (struct xvmc_pix_fmt*)s->next_picture.data[2]; if (!next) return -1; if (next->unique_id != AV_XVMC_RENDER_MAGIC) @@ -105,7 +105,7 @@ int ff_xvmc_field_start(MpegEncContext*s, AVCodecContext *avctx) render->p_future_surface = next->p_surface; // no return here, going to set forward prediction case FF_P_TYPE: - last = (struct xvmc_pixfmt_render*)s->last_picture.data[2]; + last = (struct xvmc_pix_fmt*)s->last_picture.data[2]; if (!last) last = render; // predict second field from the first if (last->unique_id != AV_XVMC_RENDER_MAGIC) @@ -123,7 +123,7 @@ return -1; */ void ff_xvmc_field_end(MpegEncContext *s) { - struct xvmc_pixfmt_render *render = (struct xvmc_pixfmt_render*)s->current_picture.data[2]; + struct xvmc_pix_fmt *render = (struct xvmc_pix_fmt*)s->current_picture.data[2]; assert(render); if (render->filled_mv_blocks_num > 0) @@ -133,7 +133,7 @@ void ff_xvmc_field_end(MpegEncContext *s) void ff_xvmc_decode_mb(MpegEncContext *s) { XvMCMacroBlock *mv_block; - struct xvmc_pixfmt_render *render; + struct xvmc_pix_fmt *render; int i, cbp, blocks_per_mb; const int mb_xy = s->mb_y * s->mb_stride + s->mb_x; @@ -160,7 +160,7 @@ void ff_xvmc_decode_mb(MpegEncContext *s) s->current_picture.qscale_table[mb_xy] = s->qscale; // start of XVMC-specific code - render = (struct xvmc_pixfmt_render*)s->current_picture.data[2]; + render = (struct xvmc_pix_fmt*)s->current_picture.data[2]; assert(render); assert(render->unique_id == AV_XVMC_RENDER_MAGIC); assert(render->mv_blocks); diff --git a/libavcodec/xvmc.h b/libavcodec/xvmc.h index 2c189ff80f..95343e9ad9 100644 --- a/libavcodec/xvmc.h +++ b/libavcodec/xvmc.h @@ -33,7 +33,7 @@ #define AV_XVMC_RENDER_MAGIC 0x1DC711C0 /**< magic value to ensure that regular pixel routines haven't corrupted the struct */ // 1337 IDCT MCo -struct xvmc_pixfmt_render { +struct xvmc_pix_fmt { /** Set by calling application Once set these values are not supposed to be modified. */ |