Commit ac3dbb4d authored by Sebastien Zwickert's avatar Sebastien Zwickert Committed by Luca Barbato

hwaccel: OS X Video Decoder Acceleration (VDA) support.

Signed-off-by: default avatarLuca Barbato <lu_zero@gentoo.org>
parent d8a09169
......@@ -101,6 +101,7 @@ easier to use. The changes are:
- Encrypted OMA files support
- Discworld II BMV decoding support
- VBLE Decoder
- OS X Video Decoder Acceleration (VDA) support
version 0.7:
......
......@@ -108,6 +108,7 @@ Configuration options:
--disable-mdct disable MDCT code
--disable-rdft disable RDFT code
--enable-vaapi enable VAAPI code
--enable-vda enable VDA code
--enable-vdpau enable VDPAU code
--disable-dxva2 disable DXVA2 code
--enable-runtime-cpudetect detect cpu capabilities at runtime (bigger binary)
......@@ -978,6 +979,7 @@ CONFIG_LIST="
swscale_alpha
thumb
vaapi
vda
vdpau
version3
x11grab
......@@ -1294,6 +1296,7 @@ h264_decoder_select="golomb h264dsp h264pred"
h264_dxva2_hwaccel_deps="dxva2api_h"
h264_dxva2_hwaccel_select="dxva2 h264_decoder"
h264_vaapi_hwaccel_select="vaapi h264_decoder"
h264_vda_hwaccel_select="vda h264_decoder"
h264_vdpau_decoder_select="vdpau h264_decoder"
imc_decoder_select="fft mdct sinewin"
jpegls_decoder_select="golomb"
......@@ -1390,6 +1393,7 @@ zmbv_decoder_select="zlib"
zmbv_encoder_select="zlib"
vaapi_deps="va_va_h"
vda_deps="VideoDecodeAcceleration_VDADecoder_h pthreads"
vdpau_deps="vdpau_vdpau_h vdpau_vdpau_x11_h"
# parsers
......@@ -2988,6 +2992,11 @@ check_func XOpenDisplay -lX11 &&
check_func XShmCreateImage -lX11 -lXext &&
check_func XFixesGetCursorImage -lX11 -lXext -lXfixes
# check for VDA header
if ! disabled vda && check_header VideoDecodeAcceleration/VDADecoder.h; then
enable vda && add_ldflags -framework CoreFoundation -framework VideoDecodeAcceleration -framework QuartzCore
fi
if ! disabled vdpau && enabled vdpau_vdpau_h; then
check_cpp_condition \
vdpau/vdpau.h "defined VDP_DECODER_PROFILE_MPEG4_PART2_ASP" ||
......
NAME = avcodec
FFLIBS = avutil
HEADERS = avcodec.h avfft.h dxva2.h opt.h vaapi.h vdpau.h version.h xvmc.h
HEADERS = avcodec.h avfft.h dxva2.h opt.h vaapi.h vda.h vdpau.h version.h xvmc.h
OBJS = allcodecs.o \
audioconvert.o \
......@@ -45,6 +45,7 @@ RDFT-OBJS-$(CONFIG_HARDCODED_TABLES) += sin_tables.o
OBJS-$(CONFIG_RDFT) += rdft.o $(RDFT-OBJS-yes)
OBJS-$(CONFIG_SINEWIN) += sinewin.o
OBJS-$(CONFIG_VAAPI) += vaapi.o
OBJS-$(CONFIG_VDA) += vda.o
OBJS-$(CONFIG_VDPAU) += vdpau.o
# decoders/encoders/hardware accelerators
......@@ -180,6 +181,7 @@ OBJS-$(CONFIG_H264_DECODER) += h264.o \
mpegvideo.o error_resilience.o
OBJS-$(CONFIG_H264_DXVA2_HWACCEL) += dxva2_h264.o
OBJS-$(CONFIG_H264_VAAPI_HWACCEL) += vaapi_h264.o
OBJS-$(CONFIG_H264_VDA_HWACCEL) += vda_h264.o
OBJS-$(CONFIG_HUFFYUV_DECODER) += huffyuv.o
OBJS-$(CONFIG_HUFFYUV_ENCODER) += huffyuv.o
OBJS-$(CONFIG_IDCIN_DECODER) += idcinvideo.o
......@@ -670,6 +672,7 @@ SKIPHEADERS-$(CONFIG_DXVA2) += dxva2.h dxva2_internal.h
SKIPHEADERS-$(CONFIG_LIBDIRAC) += libdirac.h
SKIPHEADERS-$(CONFIG_LIBSCHROEDINGER) += libschroedinger.h
SKIPHEADERS-$(CONFIG_VAAPI) += vaapi_internal.h
SKIPHEADERS-$(CONFIG_VDA) += vda.h vda_internal.h
SKIPHEADERS-$(CONFIG_VDPAU) += vdpau.h
SKIPHEADERS-$(CONFIG_XVMC) += xvmc.h
SKIPHEADERS-$(HAVE_W32THREADS) += w32pthreads.h
......
......@@ -57,6 +57,7 @@ void avcodec_register_all(void)
REGISTER_HWACCEL (H263_VAAPI, h263_vaapi);
REGISTER_HWACCEL (H264_DXVA2, h264_dxva2);
REGISTER_HWACCEL (H264_VAAPI, h264_vaapi);
REGISTER_HWACCEL (H264_VDA, h264_vda);
REGISTER_HWACCEL (MPEG2_DXVA2, mpeg2_dxva2);
REGISTER_HWACCEL (MPEG2_VAAPI, mpeg2_vaapi);
REGISTER_HWACCEL (MPEG4_VAAPI, mpeg4_vaapi);
......
......@@ -56,6 +56,7 @@ static const uint8_t div6[QP_MAX_NUM+1]={
static const enum PixelFormat hwaccel_pixfmt_list_h264_jpeg_420[] = {
PIX_FMT_DXVA2_VLD,
PIX_FMT_VAAPI_VLD,
PIX_FMT_VDA_VLD,
PIX_FMT_YUVJ420P,
PIX_FMT_NONE
};
......
......@@ -118,6 +118,7 @@ const enum PixelFormat ff_pixfmt_list_420[] = {
const enum PixelFormat ff_hwaccel_pixfmt_list_420[] = {
PIX_FMT_DXVA2_VLD,
PIX_FMT_VAAPI_VLD,
PIX_FMT_VDA_VLD,
PIX_FMT_YUV420P,
PIX_FMT_NONE
};
......
/*
* VDA hardware acceleration
*
* copyright (c) 2011 Sebastien Zwickert
*
* This file is part of Libav.
*
* Libav is free software; you can redistribute it and/or
* modify it under the terms of the GNU Lesser General Public
* License as published by the Free Software Foundation; either
* version 2.1 of the License, or (at your option) any later version.
*
* Libav 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
* Lesser General Public License for more details.
*
* You should have received a copy of the GNU Lesser General Public
* License along with Libav; if not, write to the Free Software
* Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
*/
#include <pthread.h>
#include <CoreFoundation/CFDictionary.h>
#include <CoreFoundation/CFNumber.h>
#include <CoreFoundation/CFData.h>
#include <CoreFoundation/CFString.h>
#include "libavutil/avutil.h"
#include "vda_internal.h"
/* helper to create a dictionary according to the given pts */
static CFDictionaryRef vda_dictionary_with_pts(int64_t i_pts)
{
CFStringRef key = CFSTR("FF_VDA_DECODER_PTS_KEY");
CFNumberRef value = CFNumberCreate(kCFAllocatorDefault,
kCFNumberSInt64Type, &i_pts);
CFDictionaryRef user_info = CFDictionaryCreate(kCFAllocatorDefault,
(const void **)&key,
(const void **)&value,
1,
&kCFTypeDictionaryKeyCallBacks,
&kCFTypeDictionaryValueCallBacks);
CFRelease(value);
return user_info;
}
/* helper to retrieve the pts from the given dictionary */
static int64_t vda_pts_from_dictionary(CFDictionaryRef user_info)
{
CFNumberRef pts;
int64_t outValue = 0;
if (!user_info)
return 0;
pts = CFDictionaryGetValue(user_info, CFSTR("FF_VDA_DECODER_PTS_KEY"));
if (pts)
CFNumberGetValue(pts, kCFNumberSInt64Type, &outValue);
return outValue;
}
/* Remove and release all frames from the queue. */
static void vda_clear_queue(struct vda_context *vda_ctx)
{
vda_frame *top_frame;
pthread_mutex_lock(&vda_ctx->queue_mutex);
while (vda_ctx->queue) {
top_frame = vda_ctx->queue;
vda_ctx->queue = top_frame->next_frame;
ff_vda_release_vda_frame(top_frame);
}
pthread_mutex_unlock(&vda_ctx->queue_mutex);
}
/* Decoder callback that adds the VDA frame to the queue in display order. */
static void vda_decoder_callback(void *vda_hw_ctx,
CFDictionaryRef user_info,
OSStatus status,
uint32_t infoFlags,
CVImageBufferRef image_buffer)
{
struct vda_context *vda_ctx = vda_hw_ctx;
vda_frame *new_frame;
vda_frame *queue_walker;
if (!image_buffer)
return;
if (vda_ctx->cv_pix_fmt_type != CVPixelBufferGetPixelFormatType(image_buffer))
return;
if (!(new_frame = av_mallocz(sizeof(vda_frame))))
return;
new_frame->next_frame = NULL;
new_frame->cv_buffer = CVPixelBufferRetain(image_buffer);
new_frame->pts = vda_pts_from_dictionary(user_info);
pthread_mutex_lock(&vda_ctx->queue_mutex);
queue_walker = vda_ctx->queue;
if (!queue_walker || new_frame->pts < queue_walker->pts) {
/* we have an empty queue, or this frame earlier than the current queue head */
new_frame->next_frame = queue_walker;
vda_ctx->queue = new_frame;
} else {
/* walk the queue and insert this frame where it belongs in display order */
vda_frame *next_frame;
while (1) {
next_frame = queue_walker->next_frame;
if (!next_frame || new_frame->pts < next_frame->pts) {
new_frame->next_frame = next_frame;
queue_walker->next_frame = new_frame;
break;
}
queue_walker = next_frame;
}
}
pthread_mutex_unlock(&vda_ctx->queue_mutex);
}
int ff_vda_create_decoder(struct vda_context *vda_ctx,
uint8_t *extradata,
int extradata_size)
{
OSStatus status = kVDADecoderNoErr;
CFNumberRef height;
CFNumberRef width;
CFNumberRef format;
CFDataRef avc_data;
CFMutableDictionaryRef config_info;
CFMutableDictionaryRef buffer_attributes;
CFMutableDictionaryRef io_surface_properties;
CFNumberRef cv_pix_fmt;
pthread_mutex_init(&vda_ctx->queue_mutex, NULL);
config_info = CFDictionaryCreateMutable(kCFAllocatorDefault,
4,
&kCFTypeDictionaryKeyCallBacks,
&kCFTypeDictionaryValueCallBacks);
height = CFNumberCreate(kCFAllocatorDefault, kCFNumberSInt32Type, &vda_ctx->height);
width = CFNumberCreate(kCFAllocatorDefault, kCFNumberSInt32Type, &vda_ctx->width);
format = CFNumberCreate(kCFAllocatorDefault, kCFNumberSInt32Type, &vda_ctx->format);
avc_data = CFDataCreate(kCFAllocatorDefault, extradata, extradata_size);
CFDictionarySetValue(config_info, kVDADecoderConfiguration_Height, height);
CFDictionarySetValue(config_info, kVDADecoderConfiguration_Width, width);
CFDictionarySetValue(config_info, kVDADecoderConfiguration_SourceFormat, format);
CFDictionarySetValue(config_info, kVDADecoderConfiguration_avcCData, avc_data);
buffer_attributes = CFDictionaryCreateMutable(kCFAllocatorDefault,
2,
&kCFTypeDictionaryKeyCallBacks,
&kCFTypeDictionaryValueCallBacks);
io_surface_properties = CFDictionaryCreateMutable(kCFAllocatorDefault,
0,
&kCFTypeDictionaryKeyCallBacks,
&kCFTypeDictionaryValueCallBacks);
cv_pix_fmt = CFNumberCreate(kCFAllocatorDefault,
kCFNumberSInt32Type,
&vda_ctx->cv_pix_fmt_type);
CFDictionarySetValue(buffer_attributes,
kCVPixelBufferPixelFormatTypeKey,
cv_pix_fmt);
CFDictionarySetValue(buffer_attributes,
kCVPixelBufferIOSurfacePropertiesKey,
io_surface_properties);
status = VDADecoderCreate(config_info,
buffer_attributes,
vda_decoder_callback,
vda_ctx,
&vda_ctx->decoder);
CFRelease(height);
CFRelease(width);
CFRelease(format);
CFRelease(avc_data);
CFRelease(config_info);
CFRelease(io_surface_properties);
CFRelease(cv_pix_fmt);
CFRelease(buffer_attributes);
if (kVDADecoderNoErr != status)
return status;
return 0;
}
int ff_vda_destroy_decoder(struct vda_context *vda_ctx)
{
OSStatus status = kVDADecoderNoErr;
if (vda_ctx->decoder)
status = VDADecoderDestroy(vda_ctx->decoder);
vda_clear_queue(vda_ctx);
pthread_mutex_destroy(&vda_ctx->queue_mutex);
if (kVDADecoderNoErr != status)
return status;
return 0;
}
vda_frame *ff_vda_queue_pop(struct vda_context *vda_ctx)
{
vda_frame *top_frame;
if (!vda_ctx->queue)
return NULL;
pthread_mutex_lock(&vda_ctx->queue_mutex);
top_frame = vda_ctx->queue;
vda_ctx->queue = top_frame->next_frame;
pthread_mutex_unlock(&vda_ctx->queue_mutex);
return top_frame;
}
void ff_vda_release_vda_frame(vda_frame *frame)
{
if (frame) {
CVPixelBufferRelease(frame->cv_buffer);
av_freep(&frame);
}
}
int ff_vda_decoder_decode(struct vda_context *vda_ctx,
uint8_t *bitstream,
int bitstream_size,
int64_t frame_pts)
{
OSStatus status = kVDADecoderNoErr;
CFDictionaryRef user_info;
CFDataRef coded_frame;
coded_frame = CFDataCreate(kCFAllocatorDefault, bitstream, bitstream_size);
user_info = vda_dictionary_with_pts(frame_pts);
status = VDADecoderDecode(vda_ctx->decoder, 0, coded_frame, user_info);
CFRelease(user_info);
CFRelease(coded_frame);
if (kVDADecoderNoErr != status)
return status;
return 0;
}
/*
* VDA HW acceleration
*
* copyright (c) 2011 Sebastien Zwickert
*
* This file is part of Libav.
*
* Libav is free software; you can redistribute it and/or
* modify it under the terms of the GNU Lesser General Public
* License as published by the Free Software Foundation; either
* version 2.1 of the License, or (at your option) any later version.
*
* Libav 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
* Lesser General Public License for more details.
*
* You should have received a copy of the GNU Lesser General Public
* License along with Libav; if not, write to the Free Software
* Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
*/
#ifndef AVCODEC_VDA_H
#define AVCODEC_VDA_H
#include <pthread.h>
#include <stdint.h>
// emmintrin.h is unable to compile with -std=c99 -Werror=missing-prototypes
// http://openradar.appspot.com/8026390
#undef __GNUC_STDC_INLINE__
#define Picture QuickdrawPicture
#include <VideoDecodeAcceleration/VDADecoder.h>
#undef Picture
/**
* This structure is used to store a decoded frame information and data.
*/
typedef struct vda_frame {
/**
* The PTS of the frame.
*
* - encoding: unused
* - decoding: Set/Unset by libavcodec.
*/
int64_t pts;
/**
* The CoreVideo buffer that contains the decoded data.
*
* - encoding: unused
* - decoding: Set/Unset by libavcodec.
*/
CVPixelBufferRef cv_buffer;
/**
* A pointer to the next frame.
*
* - encoding: unused
* - decoding: Set/Unset by libavcodec.
*/
struct vda_frame *next_frame;
} vda_frame;
/**
* This structure is used to provide the necessary configurations and data
* to the VDA Libav HWAccel implementation.
*
* The application must make it available as AVCodecContext.hwaccel_context.
*/
struct vda_context {
/**
* VDA decoder object.
*
* - encoding: unused
* - decoding: Set/Unset by libavcodec.
*/
VDADecoder decoder;
/**
* VDA frames queue ordered by presentation timestamp.
*
* - encoding: unused
* - decoding: Set/Unset by libavcodec.
*/
vda_frame *queue;
/**
* Mutex for locking queue operations.
*
* - encoding: unused
* - decoding: Set/Unset by libavcodec.
*/
pthread_mutex_t queue_mutex;
/**
* The frame width.
*
* - encoding: unused
* - decoding: Set/Unset by user.
*/
int width;
/**
* The frame height.
*
* - encoding: unused
* - decoding: Set/Unset by user.
*/
int height;
/**
* The frame format.
*
* - encoding: unused
* - decoding: Set/Unset by user.
*/
int format;
/**
* The pixel format for output image buffers.
*
* - encoding: unused
* - decoding: Set/Unset by user.
*/
OSType cv_pix_fmt_type;
};
/** Create the video decoder. */
int ff_vda_create_decoder(struct vda_context *vda_ctx,
uint8_t *extradata,
int extradata_size);
/** Destroy the video decoder. */
int ff_vda_destroy_decoder(struct vda_context *vda_ctx);
/** Return the top frame of the queue. */
vda_frame *ff_vda_queue_pop(struct vda_context *vda_ctx);
/** Release the given frame. */
void ff_vda_release_vda_frame(vda_frame *frame);
#endif /* AVCODEC_VDA_H */
/*
* VDA H.264 hardware acceleration
*
* copyright (c) 2011 Sebastien Zwickert
*
* This file is part of Libav.
*
* Libav is free software; you can redistribute it and/or
* modify it under the terms of the GNU Lesser General Public
* License as published by the Free Software Foundation; either
* version 2.1 of the License, or (at your option) any later version.
*
* Libav 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
* Lesser General Public License for more details.
*
* You should have received a copy of the GNU Lesser General Public
* License along with Libav; if not, write to the Free Software
* Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
*/
#include "h264.h"
#include "h264data.h"
#include "vda_internal.h"
/* This structure is used to store the bitstream of the current frame. */
struct vda_picture_context {
uint8_t *bitstream;
int bitstream_size;
};
static int start_frame(AVCodecContext *avctx,
av_unused const uint8_t *buffer,
av_unused uint32_t size)
{
const H264Context *h = avctx->priv_data;
struct vda_context *vda_ctx = avctx->hwaccel_context;
struct vda_picture_context *pic_ctx = h->s.current_picture_ptr->f.hwaccel_picture_private;
if (!vda_ctx->decoder)
return -1;
pic_ctx->bitstream = NULL;
pic_ctx->bitstream_size = 0;
return 0;
}
static int decode_slice(AVCodecContext *avctx,
const uint8_t *buffer,
uint32_t size)
{
H264Context *h = avctx->priv_data;
struct vda_context *vda_ctx = avctx->hwaccel_context;
struct vda_picture_context *pic_ctx = h->s.current_picture_ptr->f.hwaccel_picture_private;
void *tmp;
if (!vda_ctx->decoder)
return -1;
tmp = av_realloc(pic_ctx->bitstream, pic_ctx->bitstream_size+size+4);
if (!tmp)
return AVERROR(ENOMEM);
pic_ctx->bitstream = tmp;
AV_WB32(pic_ctx->bitstream + pic_ctx->bitstream_size, size);
memcpy(pic_ctx->bitstream + pic_ctx->bitstream_size + 4, buffer, size);
pic_ctx->bitstream_size += size + 4;
return 0;
}
static int end_frame(AVCodecContext *avctx)
{
H264Context *h = avctx->priv_data;
struct vda_context *vda_ctx = avctx->hwaccel_context;
struct vda_picture_context *pic_ctx = h->s.current_picture_ptr->f.hwaccel_picture_private;
AVFrame *frame = &h->s.current_picture_ptr->f;
int status;
if (!vda_ctx->decoder || !pic_ctx->bitstream)
return -1;
status = ff_vda_decoder_decode(vda_ctx, pic_ctx->bitstream,
pic_ctx->bitstream_size,
frame->reordered_opaque);
if (status)
av_log(avctx, AV_LOG_ERROR, "Failed to decode frame (%d)\n", status);
av_freep(&pic_ctx->bitstream);
return status;
}
AVHWAccel ff_h264_vda_hwaccel = {
.name = "h264_vda",
.type = AVMEDIA_TYPE_VIDEO,
.id = CODEC_ID_H264,
.pix_fmt = PIX_FMT_VDA_VLD,
.capabilities = 0,
.start_frame = start_frame,
.decode_slice = decode_slice,
.end_frame = end_frame,
.priv_data_size = sizeof(struct vda_picture_context),
};
/*
* VDA hardware acceleration
*
* copyright (c) 2011 Sebastien Zwickert
*
* This file is part of Libav.
*
* Libav is free software; you can redistribute it and/or