From 2827295703c74e3c119df9a435aa856e268c2ea9 Mon Sep 17 00:00:00 2001 From: wm4 Date: Fri, 9 Aug 2013 14:01:30 +0200 Subject: video: add vaapi decode and output support This is based on the MPlayer VA API patches. To be exact it's based on a very stripped down version of commit f1ad459a263f8537f6c from git://gitorious.org/vaapi/mplayer.git. This doesn't contain useless things like benchmarking hacks and the demo code for GLX interop. Also, unlike in the original patch, decoding and video output are split into separate source files (the separation between decoding and display also makes pixel format hacks unnecessary). On the other hand, some features not present in the original patch were added, like screenshot support. VA API is rather bad for actual video output. Dealing with older libva versions or the completely broken vdpau backend doesn't help. OSD is low quality and should be rather slow. In some cases, only either OSD or subtitles can be shown at the same time (because OSD is drawn first, OSD is prefered). Also, libva can't decide whether it accepts straight or premultiplied alpha for OSD sub-pictures: the vdpau backend seems to assume premultiplied, while a native vaapi driver uses straight. So I picked straight alpha. It doesn't matter much, because the blending code for straight alpha I added to img_convert.c is probably buggy, and ASS subtitles might be blended incorrectly. Really good video output with VA API would probably use OpenGL and the GL interop features, but at this point you might just use vo_opengl. (Patches for making HW decoding with vo_opengl have a chance of being accepted.) Despite these issues, decoding seems to work ok. I still got tearing on the Intel system I tested (Intel(R) Core(TM) i3-2350M). It was also tested with the vdpau vaapi wrapper on a nvidia system; however this was rather broken. (Fortunately, there is no reason to use mpv's VAAPI support over native VDPAU.) --- video/decode/dec_video.h | 1 + video/decode/lavc.h | 1 + video/decode/vaapi.c | 414 ++++++++++++++++++ video/decode/vd_lavc.c | 4 + video/fmt-conversion.c | 5 + video/img_format.c | 3 + video/img_format.h | 14 +- video/out/vo.c | 4 + video/out/vo.h | 2 +- video/out/vo_vaapi.c | 1054 ++++++++++++++++++++++++++++++++++++++++++++++ video/vaapi.h | 73 ++++ 11 files changed, 1572 insertions(+), 3 deletions(-) create mode 100644 video/decode/vaapi.c create mode 100644 video/out/vo_vaapi.c create mode 100644 video/vaapi.h (limited to 'video') diff --git a/video/decode/dec_video.h b/video/decode/dec_video.h index 4ba052afd1..021abaaf22 100644 --- a/video/decode/dec_video.h +++ b/video/decode/dec_video.h @@ -51,6 +51,7 @@ extern int divx_quality; // The VO can set the context pointer for supported APIs. struct mp_hwdec_info { struct mp_vdpau_ctx *vdpau_ctx; + struct mp_vaapi_ctx *vaapi_ctx; }; #endif /* MPLAYER_DEC_VIDEO_H */ diff --git a/video/decode/lavc.h b/video/decode/lavc.h index 94973cb2d8..4252034b2c 100644 --- a/video/decode/lavc.h +++ b/video/decode/lavc.h @@ -17,6 +17,7 @@ enum hwdec_type { HWDEC_VDPAU = 1, HWDEC_VDA = 2, HWDEC_CRYSTALHD = 3, + HWDEC_VAAPI = 4, }; typedef struct lavc_ctx { diff --git a/video/decode/vaapi.c b/video/decode/vaapi.c new file mode 100644 index 0000000000..baa13588c9 --- /dev/null +++ b/video/decode/vaapi.c @@ -0,0 +1,414 @@ +/* + * This file is part of mpv. + * + * With some chunks from original MPlayer VAAPI patch: + * Copyright (C) 2008-2009 Splitted-Desktop Systems + * + * mpv is free software; you can redistribute it and/or modify + * it under the terms of the GNU General Public License as published by + * the Free Software Foundation; either version 2 of the License, or + * (at your option) any later version. + * + * mpv is distributed in the hope that it will be useful, + * but WITHOUT ANY WARRANTY; without even the implied warranty of + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the + * GNU General Public License for more details. + * + * You should have received a copy of the GNU General Public License along + * with mpv. If not, see . + */ + +#include +#include + +#include +#include +#include + +#include "lavc.h" +#include "mpvcore/mp_common.h" +#include "mpvcore/av_common.h" +#include "video/fmt-conversion.h" +#include "video/vaapi.h" +#include "video/decode/dec_video.h" + +/* + * The VAAPI decoder can work only with surfaces passed to the decoder at + * creation time. This means all surfaces have to be created in advance. + * So, additionally to the maximum number of reference frames, we need + * surfaces for: + * - 1 decode frame + * - decoding 1 frame ahead (done by generic playback code) + * - keeping the reference to the previous frame (done by vo_vaapi.c) + * Note that redundant additional surfaces also might allow for some + * buffering (i.e. not trying to reuse a surface while it's busy). + */ +#define ADDTIONAL_SURFACES 3 + +// Magic number taken from original MPlayer vaapi patch. +#define MAX_DECODER_SURFACES 21 + +#define MAX_SURFACES (MAX_DECODER_SURFACES + ADDTIONAL_SURFACES) + +struct priv { + struct mp_vaapi_ctx *ctx; + VADisplay display; + + // libavcodec shared struct + struct vaapi_context *va_context; + struct vaapi_context va_context_storage; + + int format, w, h; + VASurfaceID surfaces[MAX_SURFACES]; +}; + +struct profile_entry { + enum AVCodecID av_codec; + int ff_profile; + VAProfile va_profile; + int maxrefs; +}; + +#define PE(av_codec_id, ff_profile, va_dcoder_profile, maxrefs) \ + {AV_CODEC_ID_ ## av_codec_id, \ + FF_PROFILE_ ## ff_profile, \ + VAProfile ## va_dcoder_profile, \ + maxrefs} + +static const struct profile_entry profiles[] = { + PE(MPEG2VIDEO, MPEG2_SIMPLE, MPEG2Simple, 2), + PE(MPEG2VIDEO, UNKNOWN, MPEG2Main, 2), + PE(H264, H264_BASELINE, H264Baseline, 16), + PE(H264, H264_CONSTRAINED_BASELINE, H264ConstrainedBaseline, 16), + PE(H264, H264_MAIN, H264Main, 16), + PE(H264, UNKNOWN, H264High, 16), + PE(WMV3, VC1_SIMPLE, VC1Simple, 2), + PE(WMV3, VC1_MAIN, VC1Main, 2), + PE(WMV3, UNKNOWN, VC1Advanced, 2), + PE(VC1, VC1_SIMPLE, VC1Simple, 2), + PE(VC1, VC1_MAIN, VC1Main, 2), + PE(VC1, UNKNOWN, VC1Advanced, 2), + // No idea whether these are correct + PE(MPEG4, MPEG4_SIMPLE, MPEG4Simple, 2), + PE(MPEG4, MPEG4_MAIN, MPEG4Main, 2), + PE(MPEG4, UNKNOWN, MPEG4AdvancedSimple, 2), +}; + +static const struct profile_entry *find_codec(enum AVCodecID id, int ff_profile) +{ + for (int n = 0; n < MP_ARRAY_SIZE(profiles); n++) { + if (profiles[n].av_codec == id && + (profiles[n].ff_profile == ff_profile || + profiles[n].ff_profile == FF_PROFILE_UNKNOWN)) + { + return &profiles[n]; + } + } + return NULL; +} + + +static const char *str_va_profile(VAProfile profile) +{ + switch (profile) { +#define PROFILE(profile) \ + case VAProfile##profile: return "VAProfile" #profile + PROFILE(MPEG2Simple); + PROFILE(MPEG2Main); + PROFILE(MPEG4Simple); + PROFILE(MPEG4AdvancedSimple); + PROFILE(MPEG4Main); + PROFILE(H264Baseline); + PROFILE(H264Main); + PROFILE(H264High); + PROFILE(VC1Simple); + PROFILE(VC1Main); + PROFILE(VC1Advanced); +#undef PROFILE + } + return ""; +} + +static int find_entrypoint(int format, VAEntrypoint *ep, int num_ep) +{ + int entrypoint = -1; + switch (format) { + case IMGFMT_VAAPI: entrypoint = VAEntrypointVLD; break; + case IMGFMT_VAAPI_MPEG2_IDCT: entrypoint = VAEntrypointIDCT; break; + case IMGFMT_VAAPI_MPEG2_MOCO: entrypoint = VAEntrypointMoComp; break; + } + for (int n = 0; n < num_ep; n++) { + if (ep[n] == entrypoint) + return entrypoint; + } + return -1; +} + +static int is_direct_mapping(VADisplay display) +{ + VADisplayAttribute attr; + VAStatus status; + +#if VA_CHECK_VERSION(0,34,0) + attr.type = VADisplayAttribRenderMode; + attr.flags = VA_DISPLAY_ATTRIB_GETTABLE; + + status = vaGetDisplayAttributes(display, &attr, 1); + if (status == VA_STATUS_SUCCESS) + return !(attr.value & (VA_RENDER_MODE_LOCAL_OVERLAY| + VA_RENDER_MODE_EXTERNAL_OVERLAY)); +#else + /* If the driver doesn't make a copy of the VA surface for + display, then we have to retain it until it's no longer the + visible surface. In other words, if the driver is using + DirectSurface mode, we don't want to decode the new surface + into the previous one that was used for display. */ + attr.type = VADisplayAttribDirectSurface; + attr.flags = VA_DISPLAY_ATTRIB_GETTABLE; + + status = vaGetDisplayAttributes(display, &attr, 1); + if (status == VA_STATUS_SUCCESS) + return !attr.value; +#endif + return 0; +} + +// Make vo_vaapi.c pool the required number of surfaces. +// This is very touchy: vo_vaapi.c must not free surfaces while we decode, +// and we must allocate only surfaces that were passed to the decoder on +// creation. +// We achieve this by deleting all previous surfaces, then allocate every +// surface needed. Then we free these surfaces, and rely on the fact that +// vo_vaapi.c keeps the released surfaces in the pool, and only allocates +// new surfaces out of that pool. +static int preallocate_surfaces(struct lavc_ctx *ctx, int va_rt_format, int num) +{ + struct priv *p = ctx->hwdec_priv; + int res = -1; + + struct mp_image *tmp_surfaces[MAX_SURFACES] = {0}; + + p->ctx->flush(p->ctx); // free previously allocated surfaces + + for (int n = 0; n < num; n++) { + tmp_surfaces[n] = p->ctx->get_surface(p->ctx, va_rt_format, p->format, + p->w, p->h); + if (!tmp_surfaces[n]) + goto done; + p->surfaces[n] = (uintptr_t)tmp_surfaces[n]->planes[3]; + } + res = 0; + +done: + for (int n = 0; n < num; n++) + talloc_free(tmp_surfaces[n]); + return res; +} + +static void destroy_decoder(struct lavc_ctx *ctx) +{ + struct priv *p = ctx->hwdec_priv; + + if (p->va_context->context_id != VA_INVALID_ID) { + vaDestroyContext(p->display, p->va_context->context_id); + p->va_context->context_id = VA_INVALID_ID; + } + + if (p->va_context->config_id != VA_INVALID_ID) { + vaDestroyConfig(p->display, p->va_context->config_id); + p->va_context->config_id = VA_INVALID_ID; + } + + for (int n = 0; n < MAX_SURFACES; n++) + p->surfaces[n] = VA_INVALID_ID; +} + +static int create_decoder(struct lavc_ctx *ctx) +{ + void *tmp = talloc_new(NULL); + + struct priv *p = ctx->hwdec_priv; + VAStatus status; + int res = -1; + + assert(IMGFMT_IS_VAAPI(p->format)); + + destroy_decoder(ctx); + + const struct profile_entry *pe = find_codec(ctx->avctx->codec_id, + ctx->avctx->profile); + if (!pe) { + mp_msg(MSGT_VO, MSGL_ERR, "[vaapi] Unknown codec!\n"); + goto error; + } + + int num_profiles = vaMaxNumProfiles(p->display); + VAProfile *va_profiles = talloc_zero_array(tmp, VAProfile, num_profiles); + status = vaQueryConfigProfiles(p->display, va_profiles, &num_profiles); + if (!check_va_status(status, "vaQueryConfigProfiles()")) + goto error; + mp_msg(MSGT_VO, MSGL_DBG2, "[vaapi] %d profiles available:\n", num_profiles); + for (int i = 0; i < num_profiles; i++) + mp_msg(MSGT_VO, MSGL_DBG2, " %s\n", str_va_profile(va_profiles[i])); + + bool profile_found = false; + for (int i = 0; i < num_profiles; i++) { + if (pe->va_profile == va_profiles[i]) { + profile_found = true; + break; + } + } + if (!profile_found) { + mp_msg(MSGT_VO, MSGL_ERR, "[vaapi] Profile '%s' not available.\n", + str_va_profile(pe->va_profile)); + goto error; + } + + int num_surfaces = pe->maxrefs; + if (!is_direct_mapping(p->display)) { + mp_msg(MSGT_VO, MSGL_V, "[vaapi] No direct mapping.\n"); + // Note: not sure why it has to be *=2 rather than +=1. + num_surfaces *= 2; + } + num_surfaces = MPMIN(num_surfaces, MAX_DECODER_SURFACES) + ADDTIONAL_SURFACES; + + if (num_surfaces > MAX_SURFACES) { + mp_msg(MSGT_VO, MSGL_ERR, "[vaapi] Internal error: too many surfaces.\n"); + goto error; + } + + if (preallocate_surfaces(ctx, VA_RT_FORMAT_YUV420, num_surfaces) < 0) { + mp_msg(MSGT_VO, MSGL_ERR, "[vaapi] Could not allocate surfaces.\n"); + goto error; + } + + int num_ep = vaMaxNumEntrypoints(p->display); + VAEntrypoint *ep = talloc_zero_array(tmp, VAEntrypoint, num_ep); + status = vaQueryConfigEntrypoints(p->display, pe->va_profile, ep, &num_ep); + if (!check_va_status(status, "vaQueryConfigEntrypoints()")) + goto error; + + VAEntrypoint entrypoint = find_entrypoint(p->format, ep, num_ep); + if (entrypoint < 0) { + mp_msg(MSGT_VO, MSGL_ERR, "[vaapi] Could not find VA entrypoint.\n"); + goto error; + } + + VAConfigAttrib attrib = { + .type = VAConfigAttribRTFormat, + }; + status = vaGetConfigAttributes(p->display, pe->va_profile, entrypoint, + &attrib, 1); + if (!check_va_status(status, "vaGetConfigAttributes()")) + goto error; + if ((attrib.value & VA_RT_FORMAT_YUV420) == 0) { + mp_msg(MSGT_VO, MSGL_ERR, "[vaapi] Chroma format not supported.\n"); + goto error; + } + + status = vaCreateConfig(p->display, pe->va_profile, entrypoint, &attrib, 1, + &p->va_context->config_id); + if (!check_va_status(status, "vaCreateConfig()")) + goto error; + + status = vaCreateContext(p->display, p->va_context->config_id, + p->w, p->h, VA_PROGRESSIVE, + p->surfaces, num_surfaces, + &p->va_context->context_id); + if (!check_va_status(status, "vaCreateContext()")) + goto error; + + res = 0; +error: + talloc_free(tmp); + return res; +} + +static struct mp_image *allocate_image(struct lavc_ctx *ctx, AVFrame *frame) +{ + struct priv *p = ctx->hwdec_priv; + int format = pixfmt2imgfmt(frame->format); + + if (!IMGFMT_IS_VAAPI(format)) + return NULL; + + // frame->width/height lie. Using them breaks with non-mod 16 video. + int w = ctx->avctx->width; + int h = ctx->avctx->height; + + if (format != p->format || w != p->w || h != p->h || + p->va_context->context_id == VA_INVALID_ID) + { + p->format = format; + p->w = w; + p->h = h; + if (create_decoder(ctx) < 0) + return NULL; + } + + struct mp_image *img = p->ctx->get_surface(p->ctx, VA_RT_FORMAT_YUV420, + format, p->w, p->h); + if (img) { + for (int n = 0; n < MAX_SURFACES; n++) { + if (p->surfaces[n] == (uintptr_t)img->planes[3]) + return img; + } + talloc_free(img); + } + mp_msg(MSGT_VO, MSGL_ERR, "[vaapi] Insufficient number of surfaces.\n"); + return NULL; +} + +static void uninit(struct lavc_ctx *ctx) +{ + struct priv *p = ctx->hwdec_priv; + + if (!p) + return; + + destroy_decoder(ctx); + + talloc_free(p); + ctx->hwdec_priv = NULL; +} + +static int init(struct lavc_ctx *ctx) +{ + struct priv *p = talloc_ptrtype(NULL, p); + *p = (struct priv) { + .ctx = ctx->hwdec_info->vaapi_ctx, + .va_context = &p->va_context_storage, + }; + ctx->hwdec_priv = p; + + p->display = p->ctx->display; + + p->va_context->display = p->display; + p->va_context->config_id = VA_INVALID_ID; + p->va_context->context_id = VA_INVALID_ID; + + ctx->avctx->hwaccel_context = p->va_context; + + return 0; +} + + +static int probe(struct vd_lavc_hwdec *hwdec, struct mp_hwdec_info *info, + const char *decoder) +{ + if (!info || !info->vaapi_ctx) + return HWDEC_ERR_NO_CTX; + if (!find_codec(mp_codec_to_av_codec_id(decoder), FF_PROFILE_UNKNOWN)) + return HWDEC_ERR_NO_CODEC; + return 0; +} + +const struct vd_lavc_hwdec mp_vd_lavc_vaapi = { + .type = HWDEC_VAAPI, + .image_formats = (const int[]) {IMGFMT_VAAPI, IMGFMT_VAAPI_MPEG2_IDCT, + IMGFMT_VAAPI_MPEG2_MOCO, 0}, + .probe = probe, + .init = init, + .uninit = uninit, + .allocate_image = allocate_image, +}; diff --git a/video/decode/vd_lavc.c b/video/decode/vd_lavc.c index 2fc7a1ea4c..639e46ebcc 100644 --- a/video/decode/vd_lavc.c +++ b/video/decode/vd_lavc.c @@ -84,6 +84,7 @@ const m_option_t lavc_decode_opts_conf[] = { const struct vd_lavc_hwdec mp_vd_lavc_vdpau; const struct vd_lavc_hwdec mp_vd_lavc_vdpau_old; +const struct vd_lavc_hwdec mp_vd_lavc_vaapi; static const struct vd_lavc_hwdec mp_vd_lavc_crystalhd = { .type = HWDEC_CRYSTALHD, @@ -113,6 +114,9 @@ static const struct vd_lavc_hwdec *hwdec_list[] = { #endif // CONFIG_VDPAU &mp_vd_lavc_vda, &mp_vd_lavc_crystalhd, +#if CONFIG_VAAPI + &mp_vd_lavc_vaapi, +#endif NULL }; diff --git a/video/fmt-conversion.c b/video/fmt-conversion.c index 509ff31535..842d95ce89 100644 --- a/video/fmt-conversion.c +++ b/video/fmt-conversion.c @@ -183,6 +183,11 @@ static const struct { // map to an arbitrary but existing vdpau format {IMGFMT_VDPAU, PIX_FMT_VDPAU_H264}, #endif + + {IMGFMT_VAAPI, PIX_FMT_VAAPI_VLD}, + {IMGFMT_VAAPI_MPEG2_IDCT, PIX_FMT_VAAPI_IDCT}, + {IMGFMT_VAAPI_MPEG2_MOCO, PIX_FMT_VAAPI_MOCO}, + {0, PIX_FMT_NONE} }; diff --git a/video/img_format.c b/video/img_format.c index 9422fbdd27..56d5ddc412 100644 --- a/video/img_format.c +++ b/video/img_format.c @@ -119,6 +119,9 @@ struct mp_imgfmt_entry mp_imgfmt_list[] = { FMT("vdpau_vc1", IMGFMT_VDPAU_VC1) FMT("vdpau_mpeg4", IMGFMT_VDPAU_MPEG4) FMT("vdpau", IMGFMT_VDPAU) + FMT("vaapi", IMGFMT_VAAPI) + FMT("vaapi_mpeg2_idct", IMGFMT_VAAPI_MPEG2_IDCT) + FMT("vaapi_mpeg2_moco", IMGFMT_VAAPI_MPEG2_MOCO) {0} }; diff --git a/video/img_format.h b/video/img_format.h index 7261971e43..e83f5ec25b 100644 --- a/video/img_format.h +++ b/video/img_format.h @@ -87,7 +87,6 @@ enum mp_imgfmt { IMGFMT_START = 1000, // Planar YUV formats - IMGFMT_444P, // 1x1 IMGFMT_422P, // 2x1 IMGFMT_440P, // 1x2 @@ -253,6 +252,14 @@ enum mp_imgfmt { IMGFMT_VDPAU_FIRST = IMGFMT_VDPAU, IMGFMT_VDPAU_LAST = IMGFMT_VDPAU_MPEG4, + IMGFMT_VAAPI, + IMGFMT_VAAPI_MPEG2_IDCT, + IMGFMT_VAAPI_MPEG2_MOCO, + + IMGFMT_VAAPI_FIRST = IMGFMT_VAAPI, + IMGFMT_VAAPI_LAST = IMGFMT_VAAPI_MPEG2_MOCO, + + IMGFMT_END, // Redundant format aliases for native endian access @@ -328,7 +335,10 @@ static inline bool IMGFMT_IS_RGB(unsigned int fmt) #define IMGFMT_IS_VDPAU(fmt) \ (((fmt) >= IMGFMT_VDPAU_FIRST) && ((fmt) <= IMGFMT_VDPAU_LAST)) -#define IMGFMT_IS_HWACCEL(fmt) IMGFMT_IS_VDPAU(fmt) +#define IMGFMT_IS_VAAPI(fmt) \ + (((fmt) >= IMGFMT_VAAPI_FIRST) && ((fmt) <= IMGFMT_VAAPI_LAST)) + +#define IMGFMT_IS_HWACCEL(fmt) (IMGFMT_IS_VDPAU(fmt) || IMGFMT_IS_VAAPI(fmt)) struct mp_imgfmt_entry { diff --git a/video/out/vo.c b/video/out/vo.c index 5a1b6f6fcd..592dfe6243 100644 --- a/video/out/vo.c +++ b/video/out/vo.c @@ -61,6 +61,7 @@ extern struct vo_driver video_out_direct3d; extern struct vo_driver video_out_direct3d_shaders; extern struct vo_driver video_out_sdl; extern struct vo_driver video_out_corevideo; +extern struct vo_driver video_out_vaapi; const struct vo_driver *video_out_drivers[] = { @@ -86,6 +87,9 @@ const struct vo_driver *video_out_drivers[] = #ifdef CONFIG_GL &video_out_opengl_old, #endif +#ifdef CONFIG_VAAPI + &video_out_vaapi, +#endif #ifdef CONFIG_X11 &video_out_x11, #endif diff --git a/video/out/vo.h b/video/out/vo.h index 4052d11773..8bb1bd5617 100644 --- a/video/out/vo.h +++ b/video/out/vo.h @@ -50,7 +50,7 @@ enum mp_voctrl { VOCTRL_SET_EQUALIZER, // struct voctrl_set_equalizer_args* VOCTRL_GET_EQUALIZER, // struct voctrl_get_equalizer_args* - /* for vdpau hardware decoding */ + /* for hardware decoding */ VOCTRL_GET_HWDEC_INFO, // struct mp_hwdec_info* VOCTRL_NEWFRAME, diff --git a/video/out/vo_vaapi.c b/video/out/vo_vaapi.c new file mode 100644 index 0000000000..48eac61b9b --- /dev/null +++ b/video/out/vo_vaapi.c @@ -0,0 +1,1054 @@ +/* + * VA API output module + * + * Copyright (C) 2008-2009 Splitted-Desktop Systems + * + * This file is part of MPlayer. + * + * MPlayer is free software; you can redistribute it and/or modify + * it under the terms of the GNU General Public License as published by + * the Free Software Foundation; either version 2 of the License, or + * (at your option) any later version. + * + * MPlayer is distributed in the hope that it will be useful, + * but WITHOUT ANY WARRANTY; without even the implied warranty of + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the + * GNU General Public License for more details. + * + * You should have received a copy of the GNU General Public License along + * with MPlayer; if not, write to the Free Software Foundation, Inc., + * 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA. + */ + +#include +#include + +#include +#include +#include + +#include +#include + +#include "config.h" +#include "mpvcore/mp_msg.h" +#include "video/out/vo.h" +#include "video/memcpy_pic.h" +#include "sub/sub.h" +#include "sub/img_convert.h" +#include "x11_common.h" + +#include "video/vfcap.h" +#include "video/mp_image.h" +#include "video/vaapi.h" +#include "video/decode/dec_video.h" + +#define STR_FOURCC(fcc) \ + (const char[]){(fcc), (fcc) >> 8u, (fcc) >> 16u, (fcc) >> 24u, 0} + +struct vaapi_surface { + VASurfaceID id; // VA_INVALID_ID if unallocated + int w, h, va_format; // parameters of allocated image (0/0/-1 unallocated) + VAImage image; // used for sofwtare decoding case + bool is_bound; // image bound to the surface? + bool is_used; // referenced by a mp_image + bool is_dead; // used, but deallocate VA objects as soon as possible + int order; // for LRU allocation + + // convenience shortcut for mp_image deallocation callback + struct priv *p; +}; + +struct vaapi_osd_image { + int w, h; + VAImage image; + VASubpictureID subpic_id; + bool is_used; +}; + +struct vaapi_subpic { + VASubpictureID id; + int src_x, src_y, src_w, src_h; + int dst_x, dst_y, dst_w, dst_h; +}; + +struct vaapi_osd_part { + bool active; + int bitmap_pos_id; + struct vaapi_osd_image image; + struct vaapi_subpic subpic; + struct osd_conv_cache *conv_cache; +}; + +#define MAX_OUTPUT_SURFACES 2 + +struct priv { + struct mp_log *log; + struct vo *vo; + VADisplay display; + struct mp_vaapi_ctx mpvaapi; + + struct mp_image_params image_params; + struct mp_rect src_rect; + struct mp_rect dst_rect; + struct mp_osd_res screen_osd_res; + + struct mp_image *output_surfaces[MAX_OUTPUT_SURFACES]; + struct mp_image *swdec_surfaces[MAX_OUTPUT_SURFACES]; + + int output_surface; + int visible_surface; + int deint; + int deint_type; + int scaling; + int force_scaled_osd; + + VAImageFormat osd_format; // corresponds to OSD_VA_FORMAT + struct vaapi_osd_part osd_parts[MAX_OSD_PARTS]; + bool osd_screen; + + int num_video_surfaces; + struct vaapi_surface **video_surfaces; + int video_surface_lru_counter; + + VAImageFormat *va_image_formats; + int va_num_image_formats; + VAImageFormat *va_subpic_formats; + unsigned int *va_subpic_flags; + int va_num_subpic_formats; + VADisplayAttribute *va_display_attrs; + int va_num_display_attrs; +}; + +#define OSD_VA_FORMAT VA_FOURCC_BGRA + +static const bool osd_formats[SUBBITMAP_COUNT] = { + // Actually BGRA, but only on little endian. + // This will break on big endian, I think. + [SUBBITMAP_RGBA_STR] = true, +}; + +struct fmtentry { + uint32_t va; + int mp; +}; +static struct fmtentry va_to_imgfmt[] = { + {VA_FOURCC('Y','V','1','2'), IMGFMT_420P}, + {VA_FOURCC('I','4','2','0'), IMGFMT_420P}, + {VA_FOURCC('I','Y','U','V'), IMGFMT_420P}, + {VA_FOURCC('N','V','1','2'), IMGFMT_NV12}, + // Note: not sure about endian issues (the mp formats are byte-addressed) + {VA_FOURCC_RGBA, IMGFMT_RGBA}, + {VA_FOURCC_BGRA, IMGFMT_BGRA}, + // Untested. + //{VA_FOURCC_UYVY, IMGFMT_UYVY}, + //{VA_FOURCC_YUY2, IMGFMT_YUYV}, + {0} +}; + + +static int va_fourcc_to_imgfmt(uint32_t fourcc) +{ + for (int n = 0; va_to_imgfmt[n].mp; n++) { + if (va_to_imgfmt[n].va == fourcc) + return va_to_imgfmt[n].mp; + } + return 0; +} + +static VAImageFormat *VAImageFormat_from_imgfmt(struct priv *p, int format) +{ + for (int i = 0; i < p->va_num_image_formats; i++) { + if (va_fourcc_to_imgfmt(p->va_image_formats[i].fourcc) == format) + return &p->va_image_formats[i]; + } + return NULL; +} + +static struct vaapi_surface *to_vaapi_surface(struct priv *p, + struct mp_image *img) +{ + if (!img || !IMGFMT_IS_VAAPI(img->imgfmt)) + return NULL; + // Note: we _could_ use planes[1] or planes[2] to store a vaapi_surface + // pointer, but I just don't trust libavcodec enough. + VASurfaceID id = (uintptr_t)img->planes[3]; + for (int n = 0; n < p->num_video_surfaces; n++) { + struct vaapi_surface *s = p->video_surfaces[n]; + if (s->id == id) + return s; + } + return NULL; +} + +static struct vaapi_surface *alloc_vaapi_surface(struct priv *p, int w, int h, + int va_format) +{ + VAStatus status; + + VASurfaceID id = VA_INVALID_ID; + status = vaCreateSurfaces(p->display, w, h, va_format, 1, &id); + if (!check_va_status(status, "vaCreateSurfaces()")) + return NULL; + + struct vaapi_surface *surface = NULL; + for (int n = 0; n < p->num_video_surfaces; n++) { + struct vaapi_surface *s = p->video_surfaces[n]; + if (s->id == VA_INVALID_ID) { + surface = s; + break; + } + } + if (!surface) { + surface = talloc_ptrtype(NULL, surface); + MP_TARRAY_APPEND(p, p->video_surfaces, p->num_video_surfaces, surface); + } + + *surface = (struct vaapi_surface) { + .id = id, + .image = { .image_id = VA_INVALID_ID, .buf = VA_INVALID_ID }, + .w = w, + .h = h, + .va_format = va_format, + .p = p, + }; + return surface; +} + +static void destroy_vaapi_surface(struct priv *p, struct vaapi_surface *s) +{ + if (!s || s->id == VA_INVALID_ID) + return; + assert(!s->is_used); + + if (s->image.image_id != VA_INVALID_ID) + vaDestroyImage(p->display, s->image.image_id); + vaDestroySurfaces(p->display, &s->id, 1); + s->id = VA_INVALID_ID; + s->w = 0; + s->h = 0; + s->va_format = -1; +} + +static struct vaapi_surface *get_vaapi_surface(struct priv *p, int w, int h, + int va_format) +{ + struct vaapi_surface *best = NULL; + + for (int n = 0; n < p->num_video_surfaces; n++) { + struct vaapi_surface *s = p->video_surfaces[n]; + if (!s->is_used && s->w == w && s->h == h && s->va_format == va_format) { + if (!best || best->order > s->order) + best = s; + } + } + + if (!best) + best = alloc_vaapi_surface(p, w, h, va_format); + + if (best) { + best->is_used = true; + best->order = ++p->video_surface_lru_counter; + } + return best; +} + +static void release_video_surface(void *ptr) +{ + struct vaapi_surface *surface = ptr; + surface->is_used = false; + if (surface->is_dead) + destroy_vaapi_surface(surface->p, surface); +} + +static struct mp_image *get_surface(struct mp_vaapi_ctx *ctx, int va_rt_format, + int mp_format, int w, int h) +{ + assert(IMGFMT_IS_VAAPI(mp_format)); + + struct vo *vo = ctx->priv; + struct priv *p = vo->priv; + + struct mp_image img = {0}; + mp_image_setfmt(&img, mp_format); + mp_image_set_size(&img, w, h); + + struct vaapi_surface *surface = get_vaapi_surface(p, w, h, va_rt_format); + if (!surface) + return NULL; + + // libavcodec probably wants it at [0] and [3] + // [1] and [2] are possibly free for own use. + for (int n = 0; n < 4; n++) + img.planes[n] = (void *)(uintptr_t)surface->id; + + return mp_image_new_custom_ref(&img, surface, release_video_surface); +} + +// This should be called only by code that is going to preallocate surfaces +// (and by uninit). Otherwise, hw decoder init might get confused by +// accidentally releasing hw decoder preallocated surfaces. +static void flush_surfaces(struct mp_vaapi_ctx *ctx) +{ + struct vo *vo = ctx->priv; + struct priv *p = vo->priv; + + for (int n = 0; n < p->num_video_surfaces; n++) { + struct vaapi_surface *s = p->video_surfaces[n]; + if (s->is_used) { + s->is_dead = true; + } else { + destroy_vaapi_surface(p, s); + } + } +} + +static void flush_output_surfaces(struct priv *p) +{ + for (int n = 0; n < MAX_OUTPUT_SURFACES; n++) { + talloc_free(p->output_surfaces[n]); + p->output_surfaces[n] = NULL; + } + p->output_surface = 0; + p->visible_surface = 0; +} + +// See flush_surfaces() remarks - the same applies. +static void free_video_specific(struct priv *p) +{ + flush_output_surfaces(p); + + for (int n = 0; n < MAX_OUTPUT_SURFACES; n++) { + talloc_free(p->swdec_surfaces[n]); + p->swdec_surfaces[n] = NULL; + } + + flush_surfaces(&p->mpvaapi); +} + +static int alloc_swdec_surfaces(struct priv *p, int w, int h, int format) +{ + VAStatus status; + + free_video_specific(p); + + VAImageFormat *image_format = VAImageFormat_from_imgfmt(p, format); + if (!image_format) + return -1; + for (int i = 0; i < MAX_OUTPUT_SURFACES; i++) { + // WTF: no mapping from VAImageFormat -> VA_RT_FORMAT_ + struct mp_image *img = + get_surface(&p->mpvaapi, VA_RT_FORMAT_YUV420, IMGFMT_VAAPI, w, h); + struct vaapi_surface *s = to_vaapi_surface(p, img); + if (!s) + return -1; + + if (s->image.image_id != VA_INVALID_ID) { + vaDestroyImage(p->display, s->image.image_id); + s->image.image_id = VA_INVALID_ID; + } + + status = vaDeriveImage(p->display, s->id, &s->image); + if (status == VA_STATUS_SUCCESS) { + /* vaDeriveImage() is supported, check format */ + if (s->image.format.fourcc == image_format->fourcc && + s->image.width == w && s->image.height == h) + { + s->is_bound = true; + MP_VERBOSE(p, "Using vaDeriveImage()\n"); + } else { + vaDestroyImage(p->display, s->image.image_id); + s->image.image_id = VA_INVALID_ID; + status = VA_STATUS_ERROR_OPERATION_FAILED; + } + } + if (status != VA_STATUS_SUCCESS) { + status = vaCreateImage(p->display, image_format, w, h, &s->image); + if (!check_va_status(status, "vaCreateImage()")) { + talloc_free(img); + return -1; + } + } + p->swdec_surfaces[i] = img; + } + return 0; +} + +static void resize(struct priv *p) +{ + vo_get_src_dst_rects(p->vo, &p->src_rect, &p->dst_rect, &p->screen_osd_res); + + // It's not clear whether this is needed; maybe not. + //vo_x11_clearwindow(p->vo, p->vo->x11->window); + + p->vo->want_redraw = true; +} + +static int reconfig(struct vo *vo, struct mp_image_params *params, int flags) +{ + struct priv *p = vo->priv; + + vo_x11_config_vo_window(vo, NULL, vo->dx, vo->dy, vo->dwidth, vo->dheight, + flags, "vaapi"); + + if (!IMGFMT_IS_VAAPI(params->imgfmt)) { + if (alloc_swdec_surfaces(p, params->w, params->h, params->imgfmt) < 0) + return -1; + } + + p->image_params = *params; + resize(p); + return 0; +} + +static int query_format(struct vo *vo, uint32_t format) +{ + struct priv *p = vo->priv; + + if (IMGFMT_IS_VAAPI(format) || VAImageFormat_from_imgfmt(p, format)) + return VFCAP_CSP_SUPPORTED | VFCAP_CSP_SUPPORTED_BY_HW; + + return 0; +} + +static inline int get_field_flags(struct priv *p, int i, int flags) +{ + return (p->deint && (flags & MP_IMGFIELD_INTERLACED) ? + (((!!(flags & MP_IMGFIELD_TOP_FIRST)) ^ i) == 0 ? + VA_BOTTOM_FIELD : VA_TOP_FIELD) : VA_FRAME_PICTURE); +} + +static inline int get_colorspace_flags(struct priv *p) +{ +#if USE_VAAPI_COLORSPACE + switch (p->image_params.colorspace) { + case MP_CSP_BT_601: return VA_SRC_BT601; + case MP_CSP_BT_709: return VA_SRC_BT709; + case MP_CSP_SMPTE_240M: return VA_SRC_SMPTE_240; + } +#endif + return 0; +} + +static bool render_to_screen(struct priv *p, struct mp_image *mpi) +{ + bool res = true; + VAStatus status; + + struct vaapi_surface *surface = to_vaapi_surface(p, mpi); + if (!surface) + return false; + + for (int n = 0; n < MAX_OSD_PARTS; n++) { + struct vaapi_osd_part *part = &p->osd_parts[n]; + if (part->active) { + struct vaapi_subpic *sp = &part->subpic; + int flags = 0; + if (p->osd_screen) + flags |= VA_SUBPICTURE_DESTINATION_IS_SCREEN_COORD; + status = vaAssociateSubpicture2(p->display, + sp->id, &surface->id, 1, + sp->src_x, sp->src_y, + sp->src_w, sp->src_h, + sp->dst_x, sp->dst_y, + sp->dst_w, sp->dst_h, + flags); + check_va_status(status, "vaAssociateSubpicture()"); + } + } + + for (int i = 0; i <= !!(p->deint > 1); i++) { + unsigned int flags = (get_field_flags(p, i, mpi->fields) | + get_colorspace_flags(p) | + p->scaling); + status = vaPutSurface(p->display, + surface->id, + p->vo->x11->window, + p->src_rect.x0, + p->src_rect.y0, + p->src_rect.x1 - p->src_rect.x0, + p->src_rect.y1 - p->src_rect.y0, + p->dst_rect.x0, + p->dst_rect.y0, + p->dst_rect.x1 - p->dst_rect.x0, + p->dst_rect.y1 - p->dst_rect.y0, + NULL, 0, + flags); + if (!check_va_status(status, "vaPutSurface()")) + res = false; + } + + for (int n = 0; n < MAX_OSD_PARTS; n++) { + struct vaapi_osd_part *part = &p->osd_parts[n]; + if (part->active) { + struct vaapi_subpic *sp = &part->subpic; + status = vaDeassociateSubpicture(p->display, sp->id, + &surface->id, 1); + check_va_status(status, "vaDeassociateSubpicture()"); + } + } + + return res; +} + +static void flip_page(struct vo *vo) +{ + struct priv *p = vo->priv; + + p->visible_surface = p->output_surface; + render_to_screen(p, p->output_surfaces[p->output_surface]); + p->output_surface = (p->output_surface + 1) % MAX_OUTPUT_SURFACES; +} + +static int map_image(struct priv *p, VAImage *va_image, int mpfmt, + struct mp_image *dst) +{ + VAStatus status; + + if (mpfmt != va_fourcc_to_imgfmt(va_image->format.fourcc)) + return -1; + + void *image_data = NULL; + status = vaMapBuffer(p->display, va_image->buf, &image_data); + if (!check_va_status(status, "vaMapBuffer()")) + return -1; + + *dst = (struct mp_image) {0}; + mp_image_setfmt(dst, mpfmt); + mp_image_set_size(dst, va_image->width, va_image->height); + + for (int p = 0; p < va_image->num_planes; p++) { + dst->stride[p] = va_image->pitches[p]; + dst->planes[p] = (uint8_t *)image_data + va_image->offsets[p]; + } + + if (va_image->format.fourcc == VA_FOURCC('Y','V','1','2')) { + FFSWAP(unsigned int, dst->stride[1], dst->stride[2]); + FFSWAP(uint8_t *, dst->planes[1], dst->planes[2]); + } + + return 0; +} + +static int unmap_image(struct priv *p, VAImage *va_image) +{ + VAStatus status; + + status = vaUnmapBuffer(p->display, va_image->buf); + return check_va_status(status, "vaUnmapBuffer()") ? 0 : -1; +} + +static int upload_surface(struct priv *p, struct vaapi_surface *va_surface, + struct mp_image *mpi) +{ + VAStatus status; + + if (va_surface->image.image_id == VA_INVALID_ID) + return -1; + + struct mp_image img; + if (map_image(p, &va_surface->image, mpi->imgfmt, &img) < 0) + return -1; + mp_image_copy(&img, mpi); + unmap_image(p, &va_surface->image); + + if (!va_surface->is_bound) { + status = vaPutImage2(p->display, va_surface->id, + va_surface->image.image_id, + 0, 0, mpi->w, mpi->h, + 0, 0, mpi->w, mpi->h); + if (!check_va_status(status, "vaPutImage()")) + return -1; + } + + return 0; +} + +static int try_get_surface(struct priv *p, VAImageFormat *fmt, + struct vaapi_surface *va_surface, + VAImage *out_image) +{ + VAStatus status; + + status = vaSyncSurface(p->display, va_surface->id); + if (!check_va_status(status, "vaSyncSurface()")) + return -2; + + int w = va_surface->w; + int h = va_surface->h; + + status = vaCreateImage(p->display, fmt, w, h, out_image); + if (!check_va_status(status, "vaCreateImage()")) + return -2; + + status = vaGetImage(p->display, va_surface->id, 0, 0, w, h, + out_image->image_id); + if (status != VA_STATUS_SUCCESS) { + vaDestroyImage(p->display, out_image->image_id); + return -1; + } + + return 0; +} + +static struct mp_image *download_surface(struct priv *p, + struct vaapi_surface *va_surface) +{ + // We have no clue which format will work, so try them all. + // This code is just for screenshots, so it's ok not to cache the right + // format (to prevent unnecessary work), and we don't attempt to use + // vaDeriveImage() for direct access either. + for (int i = 0; i < p->va_num_image_formats; i++) { + VAImageFormat *fmt = &p->va_image_formats[i]; + int mpfmt = va_fourcc_to_imgfmt(fmt->fourcc); + if (!mpfmt) + continue; + VAImage image; + int r = try_get_surface(p, fmt, va_surface, &image); + if (r == -1) + continue; + if (r < 0) + return NULL; + + struct mp_image *res = NULL; + struct mp_image tmp; + if (map_image(p, &image, mpfmt, &tmp) >= 0) { + res = mp_image_alloc(mpfmt, tmp.w, tmp.h); + mp_image_copy(res, &tmp); + unmap_image(p, &image); + } + vaDestroyImage(p->display, image.image_id); + return res; + } + + MP_ERR(p, "failed to get surface data.\n"); + return NULL; +} + +static void draw_image(struct vo *vo, mp_image_t *mpi) +{ + struct priv *p = vo->priv; + + if (!IMGFMT_IS_VAAPI(mpi->imgfmt)) { + struct mp_image *surface = p->swdec_surfaces[p->output_surface]; + struct vaapi_surface *va_surface = to_vaapi_surface(p, surface); + if (!va_surface) + return; + if (upload_surface(p, va_surface, mpi) < 0) + return; + mp_image_copy_attributes(surface, mpi); + mpi = surface; + } + + mp_image_setrefp(&p->output_surfaces[p->output_surface], mpi); +} + +static struct mp_image *get_screenshot(struct priv *p) +{ + struct vaapi_surface *va_surface = + to_vaapi_surface(p, p->output_surfaces[p->visible_surface]); + if (!va_surface) + return NULL; + struct mp_image *img = download_surface(p, va_surface); + if (!img) + return NULL; + struct mp_image_params params = p->image_params; + params.imgfmt = img->imgfmt; + mp_image_set_params(img, ¶ms); + return img; +} + +static bool redraw_frame(struct priv *p) +{ + p->output_surface = p->visible_surface; + return render_to_screen(p, p->output_surfaces[p->output_surface]); +} + +static void free_subpicture(struct priv *p, struct vaapi_osd_image *img) +{ + if (img->image.image_id != VA_INVALID_ID) + vaDestroyImage(p->display, img->image.image_id); + if (img->subpic_id != VA_INVALID_ID) + vaDestroySubpicture(p->display, img->subpic_id); + img->image.image_id = VA_INVALID_ID; + img->subpic_id = VA_INVALID_ID; +} + +static int new_subpicture(struct priv *p, int w, int h, + struct vaapi_osd_image *out) +{ + VAStatus status; + + free_subpicture(p, out); + + struct vaapi_osd_image m = { + .image = {.image_id = VA_INVALID_ID, .buf = VA_INVALID_ID}, + .subpic_id = VA_INVALID_ID, + .w = w, + .h = h, + }; + + status = vaCreateImage(p->display, &p->osd_format, w, h, &m.image); + if (!check_va_status(status, "vaCreateImage()")) + goto error; + status = vaCreateSubpicture(p->display, m.image.image_id, &m.subpic_id); + if (!check_va_status(status, "vaCreateSubpicture()")) + goto error; + + *out = m; + return 0; + +error: + free_subpicture(p, &m); + MP_ERR(p, "failed to allocate OSD sub-picture of size %dx%d.\n", w, h); + return -1; +} + +static void draw_osd_cb(void *pctx, struct sub_bitmaps *imgs) +{ + struct priv *p = pctx; + + struct vaapi_osd_part *part = &p->osd_parts[imgs->render_index]; + if (imgs->bitmap_pos_id != part->bitmap_pos_id) { + part->bitmap_pos_id = imgs->bitmap_pos_id; + + osd_scale_rgba(part->conv_cache, imgs); + + struct mp_rect bb; + if (!mp_sub_bitmaps_bb(imgs, &bb)) + goto error; + + // Prevent filtering artifacts on borders + int pad = 2; + + int w = bb.x1 - bb.x0; + int h = bb.y1 - bb.y0; + if (part->image.w < w + pad || part->image.h < h + pad) { + int sw = MP_ALIGN_UP(w + pad, 64); + int sh = MP_ALIGN_UP(h + pad, 64); + if (new_subpicture(p, sw, sh, &part->image) < 0) + goto error; + } + + struct vaapi_osd_image *img = &part->image; + struct mp_image vaimg; + if (map_image(p, &img->image, IMGFMT_BGRA, &vaimg) < 0) + goto error; + + // Clear borders and regions uncovered by sub-bitmaps + mp_image_clear(&vaimg, 0, 0, w + pad, h + pad); + + for (int n = 0; n < imgs->num_parts; n++) { + struct sub_bitmap *sub = &imgs->parts[n]; + + // Note: nothing guarantees that the sub-bitmaps don't overlap. + // But in all currently existing cases, they don't. + // We simply hope that this won't change, and nobody will + // ever notice our little shortcut here. + + size_t dst = (sub->y - bb.y0) * vaimg.stride[0] + + (sub->x - bb.x0) * 4; + + memcpy_pic(vaimg.planes[0] + dst, sub->bitmap, sub->w * 4, sub->h, + vaimg.stride[0], sub->stride); + } + + if (unmap_image(p, &img->image) < 0) + goto error; + + part->subpic = (struct vaapi_subpic) { + .id = img->subpic_id, + .src_x = 0, .src_y = 0, + .src_w = w, .src_h = h, + .dst_x = bb.x0, .dst_y = bb.y0, + .dst_w = w, .dst_h = h, + }; + } + + part->active = true; + +error: + ; +} + +static void draw_osd(struct vo *vo, struct osd_state *osd) +{ + struct priv *p = vo->priv; + + if (!p->osd_format.fourcc) + return; + + struct mp_osd_res vid_res = { + .w = p->image_params.w, + .h = p->image_params.h, + .display_par = 1.0 / vo->aspdat.par, + .video_par = vo->aspdat.par, + }; + + struct mp_osd_res *res; + if (p->osd_screen) { + res = &p->screen_osd_res; + } else { + res = &vid_res; + } + + for (int n = 0; n < MAX_OSD_PARTS; n++) + p->osd_parts[n].active = false; + osd_draw(osd, *res, osd->vo_pts, 0, osd_formats, draw_osd_cb, p); +} + +static int get_displayattribtype(const char *name) +{ + if (!strcasecmp(name, "brightness")) + return VADisplayAttribBrightness; + else if (!strcasecmp(name, "contrast")) + return VADisplayAttribContrast; + else if (!strcasecmp(name, "saturation")) + return VADisplayAttribSaturation; + else if (!strcasecmp(name, "hue")) + return VADisplayAttribHue; + return -1; +} + +static VADisplayAttribute *get_display_attribute(struct priv *p, + const char *name) +{ + int type = get_displayattribtype(name); + for (int n = 0; n < p->va_num_display_attrs; n++) { + VADisplayAttribute *attr = &p->va_display_attrs[n]; + if (attr->type == type) + return attr; + } + return NULL; +} + +static int get_equalizer(struct priv *p, const char *name, int *value) +{ + VADisplayAttribute * const attr = get_display_attribute(p, name); + + if (!attr || !(attr->flags & VA_DISPLAY_ATTRIB_GETTABLE)) + return VO_NOTIMPL; + + /* normalize to -100 .. 100 range */ + int r = attr->max_value - attr->min_value; + if (r == 0) + return VO_NOTIMPL; + *value = ((attr->value - attr->min_value) * 200) / r - 100; + return VO_TRUE; +} + +static int set_equalizer(struct priv *p, const char *name, int value) +{ + VADisplayAttribute * const attr = get_display_attribute(p, name); + VAStatus status; + + if (!attr || !(attr->flags & VA_DISPLAY_ATTRIB_SETTABLE)) + return VO_NOTIMPL; + + /* normalize to attribute value range */ + int r = attr->max_value - attr->min_value; + if (r == 0) + return VO_NOTIMPL; + attr->value = ((value + 100) * r) / 200 + attr->min_value; + + status = vaSetDisplayAttributes(p->display, attr, 1); + if (!check_va_status(status, "vaSetDisplayAttributes()")) + return VO_FALSE; + return VO_TRUE; +} + +static int control(struct vo *vo, uint32_t request, void *data) +{ + struct priv *p = vo->priv; + + switch (request) { + case VOCTRL_GET_DEINTERLACE: + *(int*)data = !!p->deint; + return VO_TRUE; + case VOCTRL_SET_DEINTERLACE: + p->deint = *(int*)data ? p->deint_type : 0; + return VO_TRUE; + case VOCTRL_GET_HWDEC_INFO: { + struct mp_hwdec_info *arg = data; + arg->vaapi_ctx = &p->mpvaapi; + return true; + } + case VOCTRL_SET_EQUALIZER: { + struct voctrl_set_equalizer_args *eq = data; + return set_equalizer(p, eq->name, eq->value); + } + case VOCTRL_GET_EQUALIZER: { + struct voctrl_get_equalizer_args *eq = data; + return get_equalizer(p, eq->name, eq->valueptr); + } + case VOCTRL_REDRAW_FRAME: + return redraw_frame(p); + case VOCTRL_SCREENSHOT: { + struct voctrl_screenshot_args *args = data; + args->out_image = get_screenshot(p); + return true; + } + case VOCTRL_GET_PANSCAN: + return VO_TRUE; + case VOCTRL_SET_PANSCAN: + resize(p); + return VO_TRUE; + } + + int events = 0; + int r = vo_x11_control(vo, &events, request, data); + if (events & VO_EVENT_RESIZE) + resize(p); + if (events & VO_EVENT_EXPOSE) + vo->want_redraw = true; + return r; +} + +static void uninit(struct vo *vo) +{ + struct priv *p = vo->priv; + + free_video_specific(p); + + for (int n = 0; n < p->num_video_surfaces; n++) { + struct vaapi_surface *surface = p->video_surfaces[n]; + // Nothing is allowed to reference HW surfaces past VO lifetime. + assert(!surface->is_used); + talloc_free(surface); + } + p->num_video_surfaces = 0; + + for (int n = 0; n < MAX_OSD_PARTS; n++) { + struct vaapi_osd_part *part = &p->osd_parts[n]; + free_subpicture(p, &part->image); + } + + if (p->display) { + vaTerminate(p->display); + p->display = NULL; + } + + vo_x11_uninit(vo); +} + +static int preinit(struct vo *vo) +{ + struct priv *p = vo->priv; + p->vo = vo; + p->log = vo->log; + + VAStatus status; + + if (!vo_x11_init(vo)) + return -1; + + p->display = vaGetDisplay(vo->x11->display); + if (!p->display) + return -1; + + int major_version, minor_version; + status = vaInitialize(p->display, &major_version, &minor_version); + if (!check_va_status(status, "vaInitialize()")) + return -1; + MP_VERBOSE(vo, "VA API version %d.%d\n", major_version, minor_version); + + p->mpvaapi.display = p->display; + p->mpvaapi.priv = vo; + p->mpvaapi.flush = flush_surfaces; + p->mpvaapi.get_surface = get_surface; + + int max_image_formats = vaMaxNumImageFormats(p->display); + p->va_image_formats = talloc_array(vo, VAImageFormat, max_image_formats); + status = vaQueryImageFormats(p->display, p->va_image_formats, + &p->va_num_image_formats); + if (!check_va_status(status, "vaQueryImageFormats()")) + return -1; + MP_VERBOSE(vo, "%d image formats available:\n", p->va_num_image_formats); + for (int i = 0; i < p->va_num_image_formats; i++) + MP_VERBOSE(vo, " %s\n", STR_FOURCC(p->va_image_formats[i].fourcc)); + + int max_subpic_formats = vaMaxNumSubpictureFormats(p->display); + p->va_subpic_formats = talloc_array(vo, VAImageFormat, max_subpic_formats); + p->va_subpic_flags = talloc_array(vo, unsigned int, max_subpic_formats); + status = vaQuerySubpictureFormats(p->display, + p->va_subpic_formats, + p->va_subpic_flags, + &p->va_num_subpic_formats); + if (!check_va_status(status, "vaQuerySubpictureFormats()")) + p->va_num_subpic_formats = 0; + MP_VERBOSE(vo, "%d subpicture formats available:\n", + p->va_num_subpic_formats); + + for (int i = 0; i < p->va_num_subpic_formats; i++) { + MP_VERBOSE(vo, " %s, flags 0x%x\n", + STR_FOURCC(p->va_subpic_formats[i].fourcc), + p->va_subpic_flags[i]); + if (p->va_subpic_formats[i].fourcc == OSD_VA_FORMAT) { + p->osd_format = p->va_subpic_formats[i]; + if (!p->force_scaled_osd) { + p->osd_screen = + p->va_subpic_flags[i] & VA_SUBPICTURE_DESTINATION_IS_SCREEN_COORD; + } + } + } + + if (!p->osd_format.fourcc) + MP_ERR(vo, "OSD format not supported. Disabling OSD.\n"); + + for (int n = 0; n < MAX_OSD_PARTS; n++) { + struct vaapi_osd_part *part = &p->osd_parts[n]; + part->image.image.image_id = VA_INVALID_ID; + part->image.subpic_id = VA_INVALID_ID; + part->conv_cache = talloc_steal(vo, osd_conv_cache_new()); + } + + int max_display_attrs = vaMaxNumDisplayAttributes(p->display); + p->va_display_attrs = talloc_array(vo, VADisplayAttribute, max_display_attrs); + if (p->va_display_attrs) { + status = vaQueryDisplayAttributes(p->display, p->va_display_attrs, + &p->va_num_display_attrs); + if (!check_va_status(status, "vaQueryDisplayAttributes()")) + p->va_num_display_attrs = 0; + } + return 0; +} + +#define OPT_BASE_STRUCT struct priv + +const struct vo_driver video_out_vaapi = { + .info = &(const vo_info_t) { + "VA API with X11", + "vaapi", + "Gwenole Beauchesne and others", + "" + }, + .preinit = preinit, + .query_format = query_format, + .reconfig = reconfig, + .control = control, + .draw_image = draw_image, + .draw_osd = draw_osd, + .flip_page = flip_page, + .uninit = uninit, + .priv_size = sizeof(struct priv), + .priv_defaults = &(const struct priv) { + .scaling = VA_FILTER_SCALING_DEFAULT, + .deint_type = 2, + .deint = 0, + }, + .options = (const struct m_option[]) { +#if USE_VAAPI_SCALING + OPT_CHOICE("scaling", scaling, 0, + ({"default", VA_FILTER_SCALING_DEFAULT}, + {"fast", VA_FILTER_SCALING_FAST}, + {"hq", VA_FILTER_SCALING_HQ}, + {"nla", VA_FILTER_SCALING_NL_ANAMORPHIC})), +#endif + OPT_CHOICE("deint", deint_type, 0, + ({"no", 0}, + {"first-field", 1}, + {"bob", 2})), + OPT_FLAG("scaled-osd", force_scaled_osd, 0), + {0} + }, +}; diff --git a/video/vaapi.h b/video/vaapi.h new file mode 100644 index 0000000000..2fc149d524 --- /dev/null +++ b/video/vaapi.h @@ -0,0 +1,73 @@ +#ifndef MPV_VDPAU_H +#define MPV_VDPAU_H + +#include +#include + +#include + +/* Compatibility glue with VA-API >= 0.31 */ +#if defined VA_CHECK_VERSION +#if VA_CHECK_VERSION(0,31,0) +#define vaPutImage2 vaPutImage +#define vaAssociateSubpicture2 vaAssociateSubpicture +#endif +#endif + +/* Compatibility glue with VA-API >= 0.34 */ +#if VA_CHECK_VERSION(0,34,0) +#include +#endif + +/* Compatibility glue with upstream libva */ +#ifndef VA_SDS_VERSION +#define VA_SDS_VERSION 0 +#endif + +/* Compatibility glue with VA-API >= 0.30 */ +#ifndef VA_INVALID_ID +#define VA_INVALID_ID 0xffffffff +#endif +#ifndef VA_FOURCC +#define VA_FOURCC(ch0, ch1, ch2, ch3) \ + ((uint32_t)(uint8_t)(ch0) | \ + ((uint32_t)(uint8_t)(ch1) << 8) | \ + ((uint32_t)(uint8_t)(ch2) << 16) | \ + ((uint32_t)(uint8_t)(ch3) << 24 )) +#endif +#if defined VA_SRC_BT601 && defined VA_SRC_BT709 +# define USE_VAAPI_COLORSPACE 1 +#else +# define USE_VAAPI_COLORSPACE 0 +#endif + +/* Compatibility glue with VA-API >= 0.31.1 */ +#ifndef VA_SRC_SMPTE_240 +#define VA_SRC_SMPTE_240 0x00000040 +#endif +#if defined VA_FILTER_SCALING_MASK +# define USE_VAAPI_SCALING 1 +#else +# define USE_VAAPI_SCALING 0 +#endif + +#include "mpvcore/mp_msg.h" + +static inline bool check_va_status(VAStatus status, const char *msg) +{ + if (status != VA_STATUS_SUCCESS) { + mp_msg(MSGT_VO, MSGL_ERR, "[vaapi] %s: %s\n", msg, vaErrorStr(status)); + return false; + } + return true; +} + +struct mp_vaapi_ctx { + VADisplay display; + struct mp_image *(*get_surface)(struct mp_vaapi_ctx *ctx, int va_format, + int mp_format, int w, int h); + void (*flush)(struct mp_vaapi_ctx *ctx); + void *priv; // for VO +}; + +#endif -- cgit v1.2.3