#include <tbb/parallel_for.h>
#include <tbb/parallel_for_each.h>
#include <tbb/concurrent_queue.h>
#include <tbb/parallel_for.h>
#include <tbb/parallel_for_each.h>
#include <tbb/concurrent_queue.h>
std::array<const uint8_t*, 4> data;
core::image_transform transform;
item()
std::array<const uint8_t*, 4> data;
core::image_transform transform;
item()
// 100% accurate blending with correct rounding.
inline xmm::s8_x blend(xmm::s8_x d, xmm::s8_x s)
// 100% accurate blending with correct rounding.
inline xmm::s8_x blend(xmm::s8_x d, xmm::s8_x s)
// C(S, D) = S + D - (((T >> 8) + T) >> 8);
// T(S, D) = S * D[A] + 0x80
auto aaaa = s8_x::shuffle(d, s8_x(15, 15, 15, 15, 11, 11, 11, 11, 7, 7, 7, 7, 3, 3, 3, 3));
d = s8_x(u8_x::min(u8_x(d), u8_x(aaaa))); // Overflow guard. Some source files have color values which incorrectly exceed pre-multiplied alpha values, e.g. red(255) > alpha(254).
// C(S, D) = S + D - (((T >> 8) + T) >> 8);
// T(S, D) = S * D[A] + 0x80
auto aaaa = s8_x::shuffle(d, s8_x(15, 15, 15, 15, 11, 11, 11, 11, 7, 7, 7, 7, 3, 3, 3, 3));
d = s8_x(u8_x::min(u8_x(d), u8_x(aaaa))); // Overflow guard. Some source files have color values which incorrectly exceed pre-multiplied alpha values, e.g. red(255) > alpha(254).
- auto xaxa = s16_x(aaaa) >> 8;
-
- auto t1 = s16_x::multiply_low(s16_x(s) & 0x00FF, xaxa) + 0x80;
+ auto xaxa = s16_x(aaaa) >> 8;
+
+ auto t1 = s16_x::multiply_low(s16_x(s) & 0x00FF, xaxa) + 0x80;
auto argb = s8_x::blend(xyxy, yxyx, s8_x(-1, 0, -1, 0));
return s8_x(s) + (d - argb);
}
auto argb = s8_x::blend(xyxy, yxyx, s8_x(-1, 0, -1, 0));
return s8_x(s) + (d - argb);
}
template<typename temporal, typename alignment>
static void kernel(uint8_t* dest, const uint8_t* source, size_t count)
template<typename temporal, typename alignment>
static void kernel(uint8_t* dest, const uint8_t* source, size_t count)
{
auto s0 = s8_x::load<temporal_tag, alignment>(dest+n+0);
auto s1 = s8_x::load<temporal_tag, alignment>(dest+n+16);
auto d0 = s8_x::load<temporal_tag, alignment>(source+n+0);
auto d1 = s8_x::load<temporal_tag, alignment>(source+n+16);
{
auto s0 = s8_x::load<temporal_tag, alignment>(dest+n+0);
auto s1 = s8_x::load<temporal_tag, alignment>(dest+n+16);
auto d0 = s8_x::load<temporal_tag, alignment>(source+n+0);
auto d1 = s8_x::load<temporal_tag, alignment>(source+n+16);
auto argb0 = blend(d0, s0);
auto argb1 = blend(d1, s1);
s8_x::store<temporal, alignment>(argb0, dest+n+0 );
s8_x::store<temporal, alignment>(argb1, dest+n+16);
auto argb0 = blend(d0, s0);
auto argb1 = blend(d1, s1);
s8_x::store<temporal, alignment>(argb0, dest+n+0 );
s8_x::store<temporal, alignment>(argb1, dest+n+16);
}
template<typename temporal>
static void kernel(uint8_t* dest, const uint8_t* source, size_t count)
}
template<typename temporal>
static void kernel(uint8_t* dest, const uint8_t* source, size_t count)
kernel<temporal_tag, unaligned_tag>(dest, source, count);
else
kernel<temporal_tag, aligned_tag>(dest, source, count);
kernel<temporal_tag, unaligned_tag>(dest, source, count);
else
kernel<temporal_tag, aligned_tag>(dest, source, count);
{
tbb::concurrent_unordered_map<int64_t, tbb::concurrent_bounded_queue<std::shared_ptr<SwsContext>>> sws_devices_;
tbb::concurrent_bounded_queue<spl::shared_ptr<buffer>> temp_buffers_;
{
tbb::concurrent_unordered_map<int64_t, tbb::concurrent_bounded_queue<std::shared_ptr<SwsContext>>> sws_devices_;
tbb::concurrent_bounded_queue<spl::shared_ptr<buffer>> temp_buffers_;
- {
- convert(items, format_desc.width, format_desc.height);
-
+ {
+ if (format_desc != format_desc_)
+ {
+ format_desc_ = format_desc;
+ sws_devices_.clear();
+ }
+
+ convert(items, format_desc.width, format_desc.height);
+
// Remove first field stills.
boost::range::remove_erase_if(items, [&](const item& item)
{
return item.transform.is_still && item.transform.field_mode == format_desc.field_mode; // only us last field for stills.
});
// Remove first field stills.
boost::range::remove_erase_if(items, [&](const item& item)
{
return item.transform.is_still && item.transform.field_mode == format_desc.field_mode; // only us last field for stills.
});
auto result = spl::make_shared<buffer>(format_desc.size, 0);
if(format_desc.field_mode != core::field_mode::progressive)
auto result = spl::make_shared<buffer>(format_desc.size, 0);
if(format_desc.field_mode != core::field_mode::progressive)
draw(items, result->data(), format_desc.width, format_desc.height, core::field_mode::upper);
draw(items, result->data(), format_desc.width, format_desc.height, core::field_mode::lower);
}
draw(items, result->data(), format_desc.width, format_desc.height, core::field_mode::upper);
draw(items, result->data(), format_desc.width, format_desc.height, core::field_mode::lower);
}
return make_ready_future(array<const std::uint8_t>(result->data(), format_desc.size, true, result));
}
private:
void draw(std::vector<item> items, uint8_t* dest, std::size_t width, std::size_t height, core::field_mode field_mode)
return make_ready_future(array<const std::uint8_t>(result->data(), format_desc.size, true, result));
}
private:
void draw(std::vector<item> items, uint8_t* dest, std::size_t width, std::size_t height, core::field_mode field_mode)
auto start = field_mode == core::field_mode::lower ? 1 : 0;
auto step = field_mode == core::field_mode::progressive ? 1 : 2;
auto start = field_mode == core::field_mode::lower ? 1 : 0;
auto step = field_mode == core::field_mode::progressive ? 1 : 2;
// TODO: Add support for fill translations.
// TODO: Add support for mask rect.
// TODO: Add support for opacity.
// TODO: Add support for fill translations.
// TODO: Add support for mask rect.
// TODO: Add support for opacity.
for(std::size_t n = 0; n < items.size()-1; ++n)
kernel<xmm::temporal_tag>(dest + y*width*4, items[n].data.at(0) + y*width*4, width*4);
for(std::size_t n = 0; n < items.size()-1; ++n)
kernel<xmm::temporal_tag>(dest + y*width*4, items[n].data.at(0) + y*width*4, width*4);
void convert(std::vector<item>& source_items, int width, int height)
{
std::set<std::array<const uint8_t*, 4>> buffers;
void convert(std::vector<item>& source_items, int width, int height)
{
std::set<std::array<const uint8_t*, 4>> buffers;
auto dest_items = source_items;
tbb::parallel_for_each(buffers.begin(), buffers.end(), [&](const std::array<const uint8_t*, 4>& data)
auto dest_items = source_items;
tbb::parallel_for_each(buffers.begin(), buffers.end(), [&](const std::array<const uint8_t*, 4>& data)
auto pix_desc = std::find_if(source_items.begin(), source_items.end(), [&](const item& item){return item.data == data;})->pix_desc;
auto pix_desc = std::find_if(source_items.begin(), source_items.end(), [&](const item& item){return item.data == data;})->pix_desc;
-
- int64_t key = ((static_cast<int64_t>(input_av_frame->width) << 32) & 0xFFFF00000000) |
- ((static_cast<int64_t>(input_av_frame->height) << 16) & 0xFFFF0000) |
+
+ int64_t key = ((static_cast<int64_t>(input_av_frame->width) << 32) & 0xFFFF00000000) |
+ ((static_cast<int64_t>(input_av_frame->height) << 16) & 0xFFFF0000) |
- sws_device.reset(sws_getContext(input_av_frame->width, input_av_frame->height, static_cast<PixelFormat>(input_av_frame->format), width, height, PIX_FMT_BGRA, SWS_BILINEAR, nullptr, nullptr, ¶m), sws_freeContext);
+ sws_device.reset(sws_getContext(input_av_frame->width, input_av_frame->height, static_cast<AVPixelFormat>(input_av_frame->format), width, height, AVPixelFormat::AV_PIX_FMT_BGRA, SWS_BILINEAR, nullptr, nullptr, ¶m), sws_freeContext);
- spl::shared_ptr<AVFrame> dest_av_frame(avcodec_alloc_frame(), av_free);
- avcodec_get_frame_defaults(dest_av_frame.get());
- avpicture_fill(reinterpret_cast<AVPicture*>(dest_av_frame.get()), dest_frame->data(), PIX_FMT_BGRA, width, height);
-
- sws_scale(sws_device.get(), input_av_frame->data, input_av_frame->linesize, 0, input_av_frame->height, dest_av_frame->data, dest_av_frame->linesize);
+ auto dest_av_frame = ffmpeg::create_frame();
+ avpicture_fill(reinterpret_cast<AVPicture*>(dest_av_frame.get()), dest_frame->data(), AVPixelFormat::AV_PIX_FMT_BGRA, width, height);
+
+ sws_scale(sws_device.get(), input_av_frame->data, input_av_frame->linesize, 0, input_av_frame->height, dest_av_frame->data, dest_av_frame->linesize);
dest_items[n].data[0] = dest_frame->data();
dest_items[n].pix_desc = core::pixel_format_desc(core::pixel_format::bgra);
dest_items[n].data[0] = dest_frame->data();
dest_items[n].pix_desc = core::pixel_format_desc(core::pixel_format::bgra);
image_renderer renderer_;
std::vector<core::image_transform> transform_stack_;
std::vector<item> items_; // layer/stream/items
public:
image_renderer renderer_;
std::vector<core::image_transform> transform_stack_;
std::vector<item> items_; // layer/stream/items
public:
void push(const core::frame_transform& transform)
{
transform_stack_.push_back(transform_stack_.back()*transform.image_transform);
}
void push(const core::frame_transform& transform)
{
transform_stack_.push_back(transform_stack_.back()*transform.image_transform);
}
if(frame.pixel_format_desc().format == core::pixel_format::invalid)
return;
if(frame.pixel_format_desc().planes.empty())
return;
if(frame.pixel_format_desc().format == core::pixel_format::invalid)
return;
if(frame.pixel_format_desc().planes.empty())
return;
item.pix_desc = frame.pixel_format_desc();
item.transform = transform_stack_.back();
for(int n = 0; n < item.pix_desc.planes.size(); ++n)
item.pix_desc = frame.pixel_format_desc();
item.transform = transform_stack_.back();
for(int n = 0; n < item.pix_desc.planes.size(); ++n)
std::future<array<const std::uint8_t>> render(const core::video_format_desc& format_desc)
{
return renderer_(std::move(items_), format_desc);
}
std::future<array<const std::uint8_t>> render(const core::video_format_desc& format_desc)
{
return renderer_(std::move(items_), format_desc);
}
{
auto buf = spl::make_shared<buffer>(plane.size);
buffers.push_back(array<std::uint8_t>(buf->data(), plane.size, true, buf));
}
{
auto buf = spl::make_shared<buffer>(plane.size);
buffers.push_back(array<std::uint8_t>(buf->data(), plane.size, true, buf));
}
- return core::mutable_frame(std::move(buffers), core::audio_buffer(), tag, desc);
+ return core::mutable_frame(std::move(buffers), core::mutable_audio_buffer(), tag, desc, channel_layout);
image_mixer::~image_mixer(){}
void image_mixer::push(const core::frame_transform& transform){impl_->push(transform);}
void image_mixer::visit(const core::const_frame& frame){impl_->visit(frame);}
void image_mixer::pop(){impl_->pop();}
image_mixer::~image_mixer(){}
void image_mixer::push(const core::frame_transform& transform){impl_->push(transform);}
void image_mixer::visit(const core::const_frame& frame){impl_->visit(frame);}
void image_mixer::pop(){impl_->pop();}
-std::future<array<const std::uint8_t>> image_mixer::operator()(const core::video_format_desc& format_desc){return impl_->render(format_desc);}
-void image_mixer::begin_layer(core::blend_mode blend_mode){impl_->begin_layer(blend_mode);}
-void image_mixer::end_layer(){impl_->end_layer();}
-core::mutable_frame image_mixer::create_frame(const void* tag, const core::pixel_format_desc& desc) {return impl_->create_frame(tag, desc);}
+int image_mixer::get_max_frame_size() { return std::numeric_limits<int>::max(); }
+std::future<array<const std::uint8_t>> image_mixer::operator()(const core::video_format_desc& format_desc, bool /* straighten_alpha */){return impl_->render(format_desc);}
+core::mutable_frame image_mixer::create_frame(const void* tag, const core::pixel_format_desc& desc, const core::audio_channel_layout& channel_layout) {return impl_->create_frame(tag, desc, channel_layout);}