void VDPAUDecoder::setupDecoder(AVCodecContext* pContext)
{
VdpStatus status;
// Create new decoder and mixer.
VdpDecoderProfile profile = 0;
switch (pContext->pix_fmt) {
case PIX_FMT_VDPAU_MPEG1:
profile = VDP_DECODER_PROFILE_MPEG1;
break;
case PIX_FMT_VDPAU_MPEG2:
profile = VDP_DECODER_PROFILE_MPEG2_MAIN;
break;
case PIX_FMT_VDPAU_H264:
profile = VDP_DECODER_PROFILE_H264_HIGH;
break;
case PIX_FMT_VDPAU_WMV3:
profile = VDP_DECODER_PROFILE_VC1_SIMPLE;
break;
case PIX_FMT_VDPAU_VC1:
profile = VDP_DECODER_PROFILE_VC1_SIMPLE;
break;
default:
AVG_ASSERT(false);
}
status = vdp_decoder_create(getVDPAUDevice(), profile, m_Size.x, m_Size.y, 16,
&m_VDPDecoder);
AVG_ASSERT(status == VDP_STATUS_OK);
m_PixFmt = pContext->pix_fmt;
VdpVideoMixerFeature features[] = {
VDP_VIDEO_MIXER_FEATURE_DEINTERLACE_TEMPORAL,
VDP_VIDEO_MIXER_FEATURE_DEINTERLACE_TEMPORAL_SPATIAL,
};
VdpVideoMixerParameter params[] = {
VDP_VIDEO_MIXER_PARAMETER_VIDEO_SURFACE_WIDTH,
VDP_VIDEO_MIXER_PARAMETER_VIDEO_SURFACE_HEIGHT,
VDP_VIDEO_MIXER_PARAMETER_CHROMA_TYPE,
VDP_VIDEO_MIXER_PARAMETER_LAYERS
};
VdpChromaType chroma = VDP_CHROMA_TYPE_420;
int numLayers = 0;
void const* paramValues [] = { &m_Size.x, &m_Size.y, &chroma, &numLayers };
status = vdp_video_mixer_create(getVDPAUDevice(), 2, features, 4, params,
paramValues, &m_VDPMixer);
AVG_ASSERT(status == VDP_STATUS_OK);
}
bool CVDPAU::ConfigVDPAU(AVCodecContext* avctx, int ref_frames)
{
FiniVDPAUOutput();
VdpStatus vdp_st;
VdpDecoderProfile vdp_decoder_profile;
vid_width = avctx->width;
vid_height = avctx->height;
surface_width = avctx->coded_width;
surface_height = avctx->coded_height;
past[1] = past[0] = current = future = NULL;
CLog::Log(LOGNOTICE, " (VDPAU) screenWidth:%i vidWidth:%i surfaceWidth:%i",OutWidth,vid_width,surface_width);
CLog::Log(LOGNOTICE, " (VDPAU) screenHeight:%i vidHeight:%i surfaceHeight:%i",OutHeight,vid_height,surface_height);
ReadFormatOf(avctx->pix_fmt, vdp_decoder_profile, vdp_chroma_type);
if(avctx->pix_fmt == PIX_FMT_VDPAU_H264)
{
max_references = ref_frames;
if (max_references > 16) max_references = 16;
if (max_references < 5) max_references = 5;
}
else
max_references = 2;
vdp_st = vdp_decoder_create(vdp_device,
vdp_decoder_profile,
surface_width,
surface_height,
max_references,
&decoder);
if (CheckStatus(vdp_st, __LINE__))
return false;
m_vdpauOutputMethod = OUTPUT_NONE;
vdpauConfigured = true;
return true;
}
static int Init(vlc_va_t *va, void **ctxp, vlc_fourcc_t *chromap,
int width, int height)
{
vlc_va_sys_t *sys = va->sys;
VdpStatus err;
width = (width + 1) & ~1;
height = (height + 3) & ~3;
sys->width = width;
sys->height = height;
unsigned surfaces = 2;
switch (sys->profile)
{
case VDP_DECODER_PROFILE_H264_BASELINE:
case VDP_DECODER_PROFILE_H264_MAIN:
case VDP_DECODER_PROFILE_H264_HIGH:
surfaces = 16;
break;
}
err = vdp_decoder_create(sys->vdp, sys->device, sys->profile, width,
height, surfaces, &sys->context->decoder);
if (err != VDP_STATUS_OK)
{
msg_Err(va, "%s creation failure: %s", "decoder",
vdp_get_error_string(sys->vdp, err));
sys->context->decoder = VDP_INVALID_HANDLE;
return VLC_EGENERIC;
}
*ctxp = sys->context;
/* TODO: select better chromas when appropriate */
*chromap = VLC_CODEC_VDPAU_VIDEO_420;
return VLC_SUCCESS;
}
bool CVDPAU::ConfigVDPAU(AVCodecContext* avctx, int ref_frames)
{
FiniVDPAUOutput();
VdpStatus vdp_st;
VdpDecoderProfile vdp_decoder_profile;
vid_width = avctx->width;
vid_height = avctx->height;
past[1] = past[0] = current = future = NULL;
CLog::Log(LOGNOTICE, " (VDPAU) screenWidth:%i vidWidth:%i",OutWidth,vid_width);
CLog::Log(LOGNOTICE, " (VDPAU) screenHeight:%i vidHeight:%i",OutHeight,vid_height);
ReadFormatOf(avctx->pix_fmt, vdp_decoder_profile, vdp_chroma_type);
if(avctx->pix_fmt == PIX_FMT_VDPAU_H264)
{
max_references = ref_frames;
if (max_references > 16) max_references = 16;
if (max_references < 5) max_references = 5;
}
else
max_references = 2;
vdp_st = vdp_decoder_create(vdp_device,
vdp_decoder_profile,
vid_width,
vid_height,
max_references,
&decoder);
CHECK_VDPAU_RETURN(vdp_st, false);
vdp_st = vdp_presentation_queue_target_create_x11(vdp_device,
m_Pixmap, //x_window,
&vdp_flip_target);
CHECK_VDPAU_RETURN(vdp_st, false);
vdp_st = vdp_presentation_queue_create(vdp_device,
vdp_flip_target,
&vdp_flip_queue);
CHECK_VDPAU_RETURN(vdp_st, false);
totalAvailableOutputSurfaces = 0;
int tmpMaxOutputSurfaces = NUM_OUTPUT_SURFACES;
if (vid_width == FULLHD_WIDTH)
tmpMaxOutputSurfaces = NUM_OUTPUT_SURFACES_FOR_FULLHD;
// Creation of outputSurfaces
for (int i = 0; i < NUM_OUTPUT_SURFACES && i < tmpMaxOutputSurfaces; i++)
{
vdp_st = vdp_output_surface_create(vdp_device,
VDP_RGBA_FORMAT_B8G8R8A8,
OutWidth,
OutHeight,
&outputSurfaces[i]);
CHECK_VDPAU_RETURN(vdp_st, false);
totalAvailableOutputSurfaces++;
}
CLog::Log(LOGNOTICE, " (VDPAU) Total Output Surfaces Available: %i of a max (tmp: %i const: %i)",
totalAvailableOutputSurfaces,
tmpMaxOutputSurfaces,
NUM_OUTPUT_SURFACES);
surfaceNum = presentSurfaceNum = 0;
outputSurface = outputSurfaces[surfaceNum];
vdpauConfigured = true;
return true;
}
bool CVDPAU::Open(AVCodecContext* avctx, const enum PixelFormat, unsigned int surfaces)
{
if(avctx->width == 0
|| avctx->height == 0)
{
CLog::Log(LOGWARNING,"(VDPAU) no width/height available, can't init");
return false;
}
if (!dl_handle)
{
dl_handle = dlopen("libvdpau.so.1", RTLD_LAZY);
if (!dl_handle)
{
const char* error = dlerror();
if (!error)
error = "dlerror() returned NULL";
CLog::Log(LOGNOTICE,"(VDPAU) Unable to get handle to libvdpau: %s", error);
//g_application.m_guiDialogKaiToast.QueueNotification(CGUIDialogKaiToast::Error, "VDPAU", error, 10000);
return false;
}
}
InitVDPAUProcs();
if (vdp_device != VDP_INVALID_HANDLE)
{
SpewHardwareAvailable();
VdpDecoderProfile profile = 0;
if(avctx->codec_id == CODEC_ID_H264)
profile = VDP_DECODER_PROFILE_H264_HIGH;
#ifdef VDP_DECODER_PROFILE_MPEG4_PART2_ASP
else if(avctx->codec_id == CODEC_ID_MPEG4)
profile = VDP_DECODER_PROFILE_MPEG4_PART2_ASP;
#endif
if(profile)
{
if (!CDVDCodecUtils::IsVP3CompatibleWidth(avctx->width))
CLog::Log(LOGWARNING,"(VDPAU) width %i might not be supported because of hardware bug", avctx->width);
/* attempt to create a decoder with this width/height, some sizes are not supported by hw */
VdpStatus vdp_st;
vdp_st = vdp_decoder_create(vdp_device, profile, avctx->width, avctx->height, 5, &decoder);
if(vdp_st != VDP_STATUS_OK)
{
CLog::Log(LOGERROR, " (VDPAU) Error: %s(%d) checking for decoder support\n", vdp_get_error_string(vdp_st), vdp_st);
FiniVDPAUProcs();
return false;
}
vdp_decoder_destroy(decoder);
CheckStatus(vdp_st, __LINE__);
}
InitCSCMatrix(avctx->height);
MakePixmap(avctx->width,avctx->height);
/* finally setup ffmpeg */
avctx->get_buffer = CVDPAU::FFGetBuffer;
avctx->release_buffer = CVDPAU::FFReleaseBuffer;
avctx->draw_horiz_band = CVDPAU::FFDrawSlice;
avctx->slice_flags=SLICE_FLAG_CODED_ORDER|SLICE_FLAG_ALLOW_FIELD;
return true;
}
return false;
}
static int vdpau_decoder_init( producer_avformat self )
{
mlt_log_debug( MLT_PRODUCER_SERVICE(self->parent), "vdpau_decoder_init\n" );
int success = 1;
self->video_codec->opaque = self;
self->video_codec->get_format = vdpau_get_format;
self->video_codec->get_buffer = vdpau_get_buffer;
self->video_codec->release_buffer = vdpau_release_buffer;
self->video_codec->draw_horiz_band = vdpau_draw_horiz;
self->video_codec->slice_flags = SLICE_FLAG_CODED_ORDER | SLICE_FLAG_ALLOW_FIELD;
self->video_codec->pix_fmt = PIX_FMT_VDPAU_H264;
VdpDecoderProfile profile = VDP_DECODER_PROFILE_H264_HIGH;
uint32_t max_references = self->video_codec->refs;
pthread_mutex_lock( &mlt_sdl_mutex );
VdpStatus status = vdp_decoder_create( self->vdpau->device,
profile, self->video_codec->width, self->video_codec->height, max_references, &self->vdpau->decoder );
pthread_mutex_unlock( &mlt_sdl_mutex );
if ( status == VDP_STATUS_OK )
{
int i, n = FFMIN( self->video_codec->refs + 2, MAX_VDPAU_SURFACES );
self->vdpau->deque = mlt_deque_init();
for ( i = 0; i < n; i++ )
{
if ( VDP_STATUS_OK == vdp_surface_create( self->vdpau->device, VDP_CHROMA_TYPE_420,
self->video_codec->width, self->video_codec->height, &self->vdpau->render_states[i].surface ) )
{
mlt_log_debug( MLT_PRODUCER_SERVICE(self->parent), "successfully created VDPAU surface %x\n",
self->vdpau->render_states[i].surface );
mlt_deque_push_back( self->vdpau->deque, &self->vdpau->render_states[i] );
}
else
{
mlt_log_info( MLT_PRODUCER_SERVICE(self->parent), "failed to create VDPAU surface %dx%d\n",
self->video_codec->width, self->video_codec->height );
while ( mlt_deque_count( self->vdpau->deque ) )
{
struct vdpau_render_state *render = mlt_deque_pop_front( self->vdpau->deque );
vdp_surface_destroy( render->surface );
}
mlt_deque_close( self->vdpau->deque );
success = 0;
break;
}
}
if ( self->vdpau )
self->vdpau->b_age = self->vdpau->ip_age[0] = self->vdpau->ip_age[1] = 256*256*256*64; // magic from Avidemux
}
else
{
success = 0;
self->vdpau->decoder = VDP_INVALID_HANDLE;
mlt_log_error( MLT_PRODUCER_SERVICE(self->parent), "VDPAU failed to initialize decoder (%s)\n",
vdp_get_error_string( status ) );
}
return success;
}
int main(int argc, char **argv) {
int width = 1280, height = 544;
Display *display = XOpenDisplay(NULL);
Window root = XDefaultRootWindow(display);
Window window = XCreateSimpleWindow(
display, root, 0, 0, 1280, 544, 0, 0, 0);
XSelectInput(display, window, ExposureMask | KeyPressMask);
XMapWindow(display, window);
XSync(display, 0);
VdpDevice dev;
mark("vdp_device_create_x11\n");
int ret = vdp_device_create_x11(display, 0, &dev, &vdp_get_proc_address);
assert(ret == VDP_STATUS_OK);
#define get(id, func) \
ret = vdp_get_proc_address(dev, id, (void **)&func); \
assert(ret == VDP_STATUS_OK);
get(VDP_FUNC_ID_DECODER_CREATE, vdp_decoder_create);
get(VDP_FUNC_ID_DECODER_DESTROY, vdp_decoder_destroy);
get(VDP_FUNC_ID_DECODER_RENDER, vdp_decoder_render);
get(VDP_FUNC_ID_VIDEO_MIXER_CREATE, vdp_video_mixer_create);
get(VDP_FUNC_ID_VIDEO_MIXER_DESTROY, vdp_video_mixer_destroy);
get(VDP_FUNC_ID_VIDEO_MIXER_RENDER, vdp_video_mixer_render);
get(VDP_FUNC_ID_VIDEO_SURFACE_CREATE, vdp_video_surface_create);
get(VDP_FUNC_ID_VIDEO_SURFACE_DESTROY, vdp_video_surface_destroy);
get(VDP_FUNC_ID_VIDEO_SURFACE_GET_BITS_Y_CB_CR, vdp_video_surface_get_bits_ycbcr);
get(VDP_FUNC_ID_OUTPUT_SURFACE_CREATE, vdp_output_surface_create);
get(VDP_FUNC_ID_OUTPUT_SURFACE_DESTROY, vdp_output_surface_destroy);
get(VDP_FUNC_ID_OUTPUT_SURFACE_GET_BITS_NATIVE, vdp_output_surface_get_bits_native);
get(VDP_FUNC_ID_PRESENTATION_QUEUE_CREATE, vdp_presentation_queue_create);
get(VDP_FUNC_ID_PRESENTATION_QUEUE_DESTROY, vdp_presentation_queue_destroy);
get(VDP_FUNC_ID_PRESENTATION_QUEUE_DISPLAY, vdp_presentation_queue_display);
get(VDP_FUNC_ID_PRESENTATION_QUEUE_TARGET_CREATE_X11, vdp_presentation_queue_target_create_x11);
get(VDP_FUNC_ID_PRESENTATION_QUEUE_BLOCK_UNTIL_SURFACE_IDLE, vdp_presentation_queue_block_until_surface_idle);
get(VDP_FUNC_ID_PRESENTATION_QUEUE_GET_TIME, vdp_presentation_queue_get_time);
#undef get
VdpDecoder dec;
VdpVideoSurface video[16];
VdpOutputSurface output;
VdpPresentationQueue queue;
VdpPresentationQueueTarget target;
VdpVideoMixer mixer;
VdpVideoMixerFeature mixer_features[] = {
};
VdpVideoMixerParameter mixer_params[] = {
VDP_VIDEO_MIXER_PARAMETER_VIDEO_SURFACE_WIDTH,
VDP_VIDEO_MIXER_PARAMETER_VIDEO_SURFACE_HEIGHT,
VDP_VIDEO_MIXER_PARAMETER_CHROMA_TYPE
};
int zero = 0;
const void *mixer_param_vals[] = {
&width,
&height,
&zero
};
mark("vdp_decoder_create\n");
ret = vdp_decoder_create(dev, VDP_DECODER_PROFILE_H264_MAIN, 1280, 544, 6, &dec);
assert(ret == VDP_STATUS_OK);
int i;
for (i = 0; i < 16; i++) {
mark("vdp_video_surface_create: %d\n", i);
ret = vdp_video_surface_create(dev, VDP_CHROMA_TYPE_420, 1280, 544, &video[i]);
assert(ret == VDP_STATUS_OK);
mark(" <-- %d\n", video[i]);
}
mark("vdp_output_surface_create\n");
ret = vdp_output_surface_create(dev, VDP_RGBA_FORMAT_B8G8R8A8, 1280, 544, &output);
assert(ret == VDP_STATUS_OK);
mark("vdp_presentation_queue_target_create_x11\n");
ret = vdp_presentation_queue_target_create_x11(dev, window, &target);
assert(ret == VDP_STATUS_OK);
mark("vdp_presentation_queue_create\n");
ret = vdp_presentation_queue_create(dev, target, &queue);
assert(ret == VDP_STATUS_OK);
mark("vdp_video_mixer_create\n");
ret = vdp_video_mixer_create(dev, sizeof(mixer_features)/sizeof(mixer_features[0]), mixer_features, sizeof(mixer_params)/sizeof(mixer_params[0]), mixer_params, mixer_param_vals, &mixer);
assert(ret == VDP_STATUS_OK);
assert(argc > 1);
int fd = open(argv[1], O_RDONLY);
struct stat statbuf;
assert(fstat(fd, &statbuf) == 0);
void *addr = mmap(NULL, statbuf.st_size, PROT_READ, MAP_SHARED, fd, 0);
void *orig_addr = addr;
mark("mmap file addr: 0x%p size: 0x%lx\n", addr, statbuf.st_size);
//printf("mmap'd file of size: %ld\n", statbuf.st_size);
VdpPictureInfoH264 info = {
.slice_count = 1,
.field_order_cnt = { 65536, 65536 },
.is_reference = 1,
.frame_num = -1,
.field_pic_flag = 0,
.bottom_field_flag = 0,
.num_ref_frames = 6,
.mb_adaptive_frame_field_flag = 0,
.constrained_intra_pred_flag = 0,
.weighted_pred_flag = 0,
.weighted_bipred_idc = 0,
.frame_mbs_only_flag = 1,
.transform_8x8_mode_flag = 0,
.chroma_qp_index_offset = 0,
.second_chroma_qp_index_offset = 0,
.pic_init_qp_minus26 = 0,
.num_ref_idx_l0_active_minus1 = 0,
.num_ref_idx_l1_active_minus1 = 0,
.log2_max_frame_num_minus4 = 5,
.pic_order_cnt_type = 0,
.log2_max_pic_order_cnt_lsb_minus4 = 6,
.delta_pic_order_always_zero_flag = 0,
.direct_8x8_inference_flag = 1,
.entropy_coding_mode_flag = 1,
.pic_order_present_flag = 0,
.deblocking_filter_control_present_flag = 1,
.redundant_pic_cnt_present_flag = 0,
};
int j;
for (j = 0; j < 6; ++j) {
int k;
for (k = 0; k < 16; ++k)
info.scaling_lists_4x4[j][k] = 16;
}
for (j = 0; j < 2; ++j) {
int k;
for (k = 0; k < 64; ++k)
info.scaling_lists_8x8[j][k] = 16;
}
for (j = 0; j < 16; ++j)
info.referenceFrames[j].surface = VDP_INVALID_HANDLE;
mark("vdp_presentation_queue_get_time\n");
VdpTime t;
ret = vdp_presentation_queue_get_time(queue, &t);
assert(ret == VDP_STATUS_OK);
fprintf(stderr, "Start time: %ld\n", t);
int vframe = 0;
while ((addr - orig_addr) < statbuf.st_size) {
int size = ntohl(*(int *)addr);
addr += 4;
int nal_type = (*(char *)addr) & 0x1F;
int nal_ref_idc = (*(char *)addr) >> 5;
if (nal_type != 1 && nal_type != 5) {
//fprintf(stderr, "Skipping NAL type %d, size: %d\n", nal_type, size);
addr += size;
continue;
}
//fprintf(stderr, "Processing NAL type %d, ref_idc: %d, size: %d\n", nal_type, nal_ref_idc, size);
int bit_offset = 8;
ue(addr, &bit_offset);
int slice_type = ue(addr, &bit_offset);
mark("nal_type: %d, ref_idc: %d, size: %d, slice_type: %d\n", nal_type, nal_ref_idc, size, slice_type);
//fprintf(stderr, "Slice type: %d\n", slice_type);
ue(addr, &bit_offset);
info.frame_num = read_bits(addr, &bit_offset, info.log2_max_frame_num_minus4 + 4);
if (nal_type == 5) {
ue(addr, &bit_offset);
info.frame_num = 0;
for (j = 0; j < 16; ++j)
info.referenceFrames[j].surface = VDP_INVALID_HANDLE;
}
uint32_t poc_lsb = read_bits(addr, &bit_offset, info.log2_max_pic_order_cnt_lsb_minus4 + 4);
info.field_order_cnt[0] = (1 << 16) + poc_lsb;
info.field_order_cnt[1] = (1 << 16) + poc_lsb;
info.is_reference = nal_ref_idc != 0;
VdpBitstreamBuffer buffer[2];
static const char header[3] = {0, 0, 1};
buffer[0].struct_version = VDP_BITSTREAM_BUFFER_VERSION;
buffer[0].bitstream = header;
buffer[0].bitstream_bytes = sizeof(header);
buffer[1].struct_version = VDP_BITSTREAM_BUFFER_VERSION;
buffer[1].bitstream = addr;
buffer[1].bitstream_bytes = size;
mark("vdp_decoder_render: %d\n", video[vframe]);
ret = vdp_decoder_render(dec, video[vframe], (void*)&info, 2, buffer);
assert(ret == VDP_STATUS_OK);
mark("vdp_video_mixer_render\n");
ret = vdp_video_mixer_render(
mixer,
VDP_INVALID_HANDLE, NULL,
VDP_VIDEO_MIXER_PICTURE_STRUCTURE_FRAME,
0, NULL,
video[vframe],
0, NULL,
NULL,
output,
NULL,
NULL,
0, NULL);
assert(ret == VDP_STATUS_OK);
t += 1000000000ULL;
mark("vdp_presentation_queue_display\n");
ret = vdp_presentation_queue_display(queue, output, 1280, 544, t);
assert(ret == VDP_STATUS_OK);
addr += size;
/*
uint32_t pitches[2] = {1280, 640 * 2};
uint8_t *data[2];
for (i = 0; i < 2; i++) {
data[i] = malloc(1280 * 544 / (i ? 2 : 1));
assert(data[i]);
}
ret = vdp_video_surface_get_bits_ycbcr(video[vframe], VDP_YCBCR_FORMAT_NV12, (void **)data, pitches);
assert(ret == VDP_STATUS_OK);
write(1, data[0], 1280 * 544);
for (i = 0; i < 1280 * 544 / 2; i+=2)
write(1, data[1] + i, 1);
for (i = 0; i < 1280 * 544 / 2; i+=2)
write(1, data[1] + i + 1, 1);
*/
if (info.is_reference) {
for (j = 5; j > 0; --j)
memcpy(&info.referenceFrames[j], &info.referenceFrames[j-1], sizeof(info.referenceFrames[0]));
info.referenceFrames[0].surface = video[vframe];
memcpy(info.referenceFrames[0].field_order_cnt, info.field_order_cnt, 2 * sizeof(uint32_t));
info.referenceFrames[0].frame_idx = info.frame_num;
info.referenceFrames[0].top_is_reference = 1;
info.referenceFrames[0].bottom_is_reference = 1;
}
vframe = (vframe + 1) % 16;
//if (vframe > 10) break;
}
return 0;
}
