1
0
mirror of https://github.com/TomHarte/CLK.git synced 2024-07-09 21:29:53 +00:00

Merge pull request #82 from TomHarte/VideoRacing

Pulls the texture builder into flush/submit orthodoxy
This commit is contained in:
Thomas Harte 2016-12-06 19:13:05 -05:00 committed by GitHub
commit 0300ae4ec0
7 changed files with 179 additions and 177 deletions

View File

@ -108,8 +108,9 @@ Flywheel::SyncEvent CRT::get_next_horizontal_sync_event(bool hsync_is_requested,
#define source_amplitude() next_run[SourceVertexOffsetOfPhaseTimeAndAmplitude + 2]
#define source_phase_time() next_run[SourceVertexOffsetOfPhaseTimeAndAmplitude + 1]
void CRT::advance_cycles(unsigned int number_of_cycles, unsigned int source_divider, bool hsync_requested, bool vsync_requested, const bool vsync_charging, const Scan::Type type, uint16_t tex_x, uint16_t tex_y)
void CRT::advance_cycles(unsigned int number_of_cycles, bool hsync_requested, bool vsync_requested, const bool vsync_charging, const Scan::Type type)
{
std::unique_lock<std::mutex> output_lock = openGL_output_builder_.get_output_lock();
number_of_cycles *= time_multiplier_;
bool is_output_run = ((type == Scan::Type::Level) || (type == Scan::Type::Data));
@ -136,11 +137,8 @@ void CRT::advance_cycles(unsigned int number_of_cycles, unsigned int source_divi
if(next_run)
{
source_input_position_x1() = tex_x;
source_input_position_y() = tex_y;
// output_y and texture locations will be written later; we won't necessarily know what it is outside of the locked region
source_output_position_x1() = (uint16_t)horizontal_flywheel_->get_current_output_position();
// Don't write output_y now, write it later; we won't necessarily know what it is outside of the locked region
// source_output_position_y() = openGL_output_builder_->get_composite_output_y();
source_phase() = colour_burst_phase_;
source_amplitude() = colour_burst_amplitude_;
source_phase_time() = (uint8_t)colour_burst_time_; // assumption: burst was within the first 1/16 of the line
@ -162,10 +160,6 @@ void CRT::advance_cycles(unsigned int number_of_cycles, unsigned int source_divi
if(next_run)
{
// if this is a data run then advance the buffer pointer
if(type == Scan::Type::Data && source_divider) tex_x += next_run_length / (time_multiplier_ * source_divider);
source_input_position_x2() = tex_x;
source_output_position_x2() = (uint16_t)horizontal_flywheel_->get_current_output_position();
}
@ -190,16 +184,8 @@ void CRT::advance_cycles(unsigned int number_of_cycles, unsigned int source_divi
}
else
{
openGL_output_builder_.lock_output();
// Get and write all those previously unwritten output ys
uint16_t output_y = openGL_output_builder_.get_composite_output_y();
size_t size;
uint8_t *buffered_lines = openGL_output_builder_.array_builder.get_unflushed_input(size);
for(size_t position = 0; position < size; position += SourceVertexSize)
{
(*(uint16_t *)&buffered_lines[position + SourceVertexOffsetOfOutputStart + 2]) = output_y;
}
const uint16_t output_y = openGL_output_builder_.get_composite_output_y();
// Construct the output run
uint8_t *next_run = openGL_output_builder_.array_builder.get_output_storage(OutputVertexSize);
@ -210,9 +196,24 @@ void CRT::advance_cycles(unsigned int number_of_cycles, unsigned int source_divi
output_tex_y() = output_y;
output_x2() = (uint16_t)horizontal_flywheel_->get_current_output_position();
}
openGL_output_builder_.array_builder.flush();
openGL_output_builder_.unlock_output();
openGL_output_builder_.array_builder.flush(
[output_y, this] (uint8_t *input_buffer, size_t input_size, uint8_t *output_buffer, size_t output_size)
{
openGL_output_builder_.texture_builder.flush(
[output_y, input_buffer] (const std::vector<TextureBuilder::WriteArea> &write_areas, size_t number_of_write_areas)
{
for(size_t run = 0; run < number_of_write_areas; run++)
{
*(uint16_t *)&input_buffer[run * SourceVertexSize + SourceVertexOffsetOfInputStart + 0] = write_areas[run].x;
*(uint16_t *)&input_buffer[run * SourceVertexSize + SourceVertexOffsetOfInputStart + 2] = write_areas[run].y;
*(uint16_t *)&input_buffer[run * SourceVertexSize + SourceVertexOffsetOfEnds + 0] = write_areas[run].x + write_areas[run].length;
}
});
for(size_t position = 0; position < input_size; position += SourceVertexSize)
{
(*(uint16_t *)&input_buffer[position + SourceVertexOffsetOfOutputStart + 2]) = output_y;
}
});
}
is_writing_composite_run_ ^= true;
}
@ -284,7 +285,7 @@ void CRT::output_scan(const Scan *const scan)
// TODO: inspect raw data for potential colour burst if required
sync_period_ = is_receiving_sync_ ? (sync_period_ + scan->number_of_cycles) : 0;
advance_cycles(scan->number_of_cycles, scan->source_divider, hsync_requested, vsync_requested, this_is_sync, scan->type, scan->tex_x, scan->tex_y);
advance_cycles(scan->number_of_cycles, hsync_requested, vsync_requested, this_is_sync, scan->type);
}
/*
@ -310,24 +311,11 @@ void CRT::output_blank(unsigned int number_of_cycles)
void CRT::output_level(unsigned int number_of_cycles)
{
if(!openGL_output_builder_.array_builder.is_full())
{
Scan scan {
.type = Scan::Type::Level,
.number_of_cycles = number_of_cycles,
.tex_x = openGL_output_builder_.texture_builder.get_last_write_x_position(),
.tex_y = openGL_output_builder_.texture_builder.get_last_write_y_position()
};
output_scan(&scan);
}
else
{
Scan scan {
.type = Scan::Type::Blank,
.number_of_cycles = number_of_cycles
};
output_scan(&scan);
}
Scan scan {
.type = Scan::Type::Level,
.number_of_cycles = number_of_cycles,
};
output_scan(&scan);
}
void CRT::output_colour_burst(unsigned int number_of_cycles, uint8_t phase, uint8_t amplitude)
@ -343,26 +331,12 @@ void CRT::output_colour_burst(unsigned int number_of_cycles, uint8_t phase, uint
void CRT::output_data(unsigned int number_of_cycles, unsigned int source_divider)
{
if(!openGL_output_builder_.array_builder.is_full())
{
openGL_output_builder_.texture_builder.reduce_previous_allocation_to(number_of_cycles / source_divider);
Scan scan {
.type = Scan::Type::Data,
.number_of_cycles = number_of_cycles,
.tex_x = openGL_output_builder_.texture_builder.get_last_write_x_position(),
.tex_y = openGL_output_builder_.texture_builder.get_last_write_y_position(),
.source_divider = source_divider
};
output_scan(&scan);
}
else
{
Scan scan {
.type = Scan::Type::Blank,
.number_of_cycles = number_of_cycles
};
output_scan(&scan);
}
openGL_output_builder_.texture_builder.reduce_previous_allocation_to(number_of_cycles / source_divider);
Scan scan {
.type = Scan::Type::Data,
.number_of_cycles = number_of_cycles,
};
output_scan(&scan);
}
Outputs::CRT::Rect CRT::get_rect_for_area(int first_line_after_sync, int number_of_lines, int first_cycle_after_sync, int number_of_cycles, float aspect_ratio)

View File

@ -53,10 +53,6 @@ class CRT {
} type;
unsigned int number_of_cycles;
union {
struct {
unsigned int source_divider;
uint16_t tex_x, tex_y;
};
struct {
uint8_t phase, amplitude;
};
@ -69,7 +65,7 @@ class CRT {
bool is_writing_composite_run_;
// the outer entry point for dispatching output_sync, output_blank, output_level and output_data
void advance_cycles(unsigned int number_of_cycles, unsigned int source_divider, bool hsync_requested, bool vsync_requested, const bool vsync_charging, const Scan::Type type, uint16_t tex_x, uint16_t tex_y);
void advance_cycles(unsigned int number_of_cycles, bool hsync_requested, bool vsync_requested, const bool vsync_charging, const Scan::Type type);
// the inner entry point that determines whether and when the next sync event will occur within
// the current output window
@ -192,6 +188,7 @@ class CRT {
*/
inline uint8_t *allocate_write_area(size_t required_length)
{
std::unique_lock<std::mutex> output_lock = openGL_output_builder_.get_output_lock();
return openGL_output_builder_.texture_builder.allocate_write_area(required_length);
}

View File

@ -23,9 +23,7 @@ ArrayBuilder::ArrayBuilder(size_t input_size, size_t output_size, std::function<
bool ArrayBuilder::is_full()
{
bool was_full;
buffer_mutex_.lock();
was_full = is_full_;
buffer_mutex_.unlock();
return was_full;
}
@ -34,30 +32,23 @@ uint8_t *ArrayBuilder::get_input_storage(size_t size)
return get_storage(size, input_);
}
uint8_t *ArrayBuilder::get_unflushed_input(size_t &size)
{
return input_.get_unflushed(size);
}
uint8_t *ArrayBuilder::get_output_storage(size_t size)
{
return get_storage(size, output_);
}
uint8_t *ArrayBuilder::get_unflushed_output(size_t &size)
void ArrayBuilder::flush(const std::function<void(uint8_t *input, size_t input_size, uint8_t *output, size_t output_size)> &function)
{
return output_.get_unflushed(size);
}
void ArrayBuilder::flush()
{
buffer_mutex_.lock();
if(!is_full_)
{
size_t input_size, output_size;
uint8_t *input = input_.get_unflushed(input_size);
uint8_t *output = output_.get_unflushed(output_size);
function(input, input_size, output, output_size);
input_.flush();
output_.flush();
}
buffer_mutex_.unlock();
}
void ArrayBuilder::bind_input()
@ -74,7 +65,6 @@ ArrayBuilder::Submission ArrayBuilder::submit()
{
ArrayBuilder::Submission submission;
buffer_mutex_.lock();
submission.input_size = input_.submit(true);
submission.output_size = output_.submit(false);
if(is_full_)
@ -83,7 +73,6 @@ ArrayBuilder::Submission ArrayBuilder::submit()
input_.reset();
output_.reset();
}
buffer_mutex_.unlock();
return submission;
}
@ -110,10 +99,8 @@ ArrayBuilder::Buffer::~Buffer()
uint8_t *ArrayBuilder::get_storage(size_t size, Buffer &buffer)
{
buffer_mutex_.lock();
uint8_t *pointer = buffer.get_storage(size);
if(!pointer) is_full_ = true;
buffer_mutex_.unlock();
return pointer;
}

View File

@ -9,9 +9,9 @@
#ifndef ArrayBuilder_hpp
#define ArrayBuilder_hpp
#include <vector>
#include <mutex>
#include <functional>
#include <memory>
#include <vector>
#include "OpenGL.hpp"
@ -41,26 +41,17 @@ class ArrayBuilder {
/// @returns a pointer to the allocated area if allocation was possible; @c nullptr otherwise.
uint8_t *get_input_storage(size_t size);
/// Gets the size of and a pointer to all data so far added to the input set but not yet flushed.
/// @returns a pointer from which it is safe to access @c size elements, which contains all regions returned via
/// @c get_input_storage in FIFO order.
uint8_t *get_unflushed_input(size_t &size);
/// Attempts to add @c size bytes to the output set.
/// @returns a pointer to the allocated area if allocation was possible; @c nullptr otherwise.
uint8_t *get_output_storage(size_t size);
/// Gets the size of and a pointer to all data so far added to the output set but not yet flushed.
/// @returns a pointer from which it is safe to access @c size elements, which contains all regions returned via
/// @c get_input_storage in FIFO order.
uint8_t *get_unflushed_output(size_t &size);
/// @returns @c true if either of the input or output storage areas is currently exhausted; @c false otherwise.
bool is_full();
/// If neither input nor output was exhausted since the last flush, atomically commits both input and output
/// up to the currently allocated size for use upon the next @c submit. Otherwise acts as a no-op.
void flush();
/// up to the currently allocated size for use upon the next @c submit, giving the supplied function a
/// chance to perform last-minute processing. Otherwise acts as a no-op.
void flush(const std::function<void(uint8_t *input, size_t input_size, uint8_t *output, size_t output_size)> &);
/// Binds the input array to GL_ARRAY_BUFFER.
void bind_input();
@ -101,7 +92,6 @@ class ArrayBuilder {
} output_, input_;
uint8_t *get_storage(size_t size, Buffer &buffer);
std::mutex buffer_mutex_;
bool is_full_;
};

View File

@ -90,6 +90,7 @@ class OpenGLOutputBuilder {
GLsync fence_;
public:
// These two are protected by output_mutex_.
TextureBuilder texture_builder;
ArrayBuilder array_builder;
@ -98,12 +99,11 @@ class OpenGLOutputBuilder {
inline void set_colour_format(ColourSpace colour_space, unsigned int colour_cycle_numerator, unsigned int colour_cycle_denominator)
{
output_mutex_.lock();
std::lock_guard<std::mutex> output_guard(output_mutex_);
colour_space_ = colour_space;
colour_cycle_numerator_ = colour_cycle_numerator;
colour_cycle_denominator_ = colour_cycle_denominator;
set_colour_space_uniforms();
output_mutex_.unlock();
}
inline void set_visible_area(Rect visible_area)
@ -111,14 +111,9 @@ class OpenGLOutputBuilder {
visible_area_ = visible_area;
}
inline void lock_output()
inline std::unique_lock<std::mutex> get_output_lock()
{
output_mutex_.lock();
}
inline void unlock_output()
{
output_mutex_.unlock();
return std::unique_lock<std::mutex>(output_mutex_);
}
inline OutputDevice get_output_device()

View File

@ -39,8 +39,11 @@ static const GLenum formatForDepth(size_t depth)
TextureBuilder::TextureBuilder(size_t bytes_per_pixel, GLenum texture_unit) :
bytes_per_pixel_(bytes_per_pixel),
next_write_x_position_(0),
next_write_y_position_(0)
write_areas_start_x_(0),
write_areas_start_y_(0),
is_full_(false),
did_submit_(false),
number_of_write_areas_(0)
{
image_.resize(bytes_per_pixel * InputBufferBuilderWidth * InputBufferBuilderHeight);
glGenTextures(1, &texture_name_);
@ -59,81 +62,81 @@ TextureBuilder::~TextureBuilder()
glDeleteTextures(1, &texture_name_);
}
inline uint8_t *TextureBuilder::pointer_to_location(uint16_t x, uint16_t y)
{
return &image_[((y * InputBufferBuilderWidth) + x) * bytes_per_pixel_];
}
uint8_t *TextureBuilder::allocate_write_area(size_t required_length)
{
if(next_write_y_position_ != InputBufferBuilderHeight)
if(is_full_) return nullptr;
uint16_t starting_x, starting_y;
if(!number_of_write_areas_)
{
last_allocation_amount_ = required_length;
if(next_write_x_position_ + required_length + 2 > InputBufferBuilderWidth)
{
next_write_x_position_ = 0;
next_write_y_position_++;
if(next_write_y_position_ == InputBufferBuilderHeight)
return nullptr;
}
write_x_position_ = next_write_x_position_ + 1;
write_y_position_ = next_write_y_position_;
write_target_pointer_ = (write_y_position_ * InputBufferBuilderWidth) + write_x_position_;
next_write_x_position_ += required_length + 2;
starting_x = write_areas_start_x_;
starting_y = write_areas_start_y_;
}
else
{
starting_x = write_areas_[number_of_write_areas_ - 1].x + write_areas_[number_of_write_areas_ - 1].length + 1;
starting_y = write_areas_[number_of_write_areas_ - 1].y;
}
else return nullptr;
return &image_[write_target_pointer_ * bytes_per_pixel_];
WriteArea next_write_area;
if(starting_x + required_length + 2 > InputBufferBuilderWidth)
{
starting_x = 0;
starting_y++;
if(starting_y == InputBufferBuilderHeight)
{
is_full_ = true;
return nullptr;
}
}
next_write_area.x = starting_x + 1;
next_write_area.y = starting_y;
next_write_area.length = (uint16_t)required_length;
if(number_of_write_areas_ < write_areas_.size())
write_areas_[number_of_write_areas_] = next_write_area;
else
write_areas_.push_back(next_write_area);
number_of_write_areas_++;
return pointer_to_location(next_write_area.x, next_write_area.y);
}
bool TextureBuilder::is_full()
{
return (next_write_y_position_ == InputBufferBuilderHeight);
return is_full_;
}
void TextureBuilder::reduce_previous_allocation_to(size_t actual_length)
{
if(next_write_y_position_ == InputBufferBuilderHeight) return;
if(is_full_) return;
uint8_t *const image_pointer = image_.data();
// correct if the writing cursor was reset while a client was writing
if(next_write_x_position_ == 0 && next_write_y_position_ == 0)
{
memmove(&image_pointer[bytes_per_pixel_], &image_pointer[write_target_pointer_ * bytes_per_pixel_], actual_length * bytes_per_pixel_);
write_target_pointer_ = 1;
last_allocation_amount_ = actual_length;
next_write_x_position_ = (uint16_t)(actual_length + 2);
write_x_position_ = 1;
write_y_position_ = 0;
}
WriteArea &write_area = write_areas_[number_of_write_areas_-1];
write_area.length = (uint16_t)actual_length;
// book end the allocation with duplicates of the first and last pixel, to protect
// against rounding errors when this run is drawn
memcpy( &image_pointer[(write_target_pointer_ - 1) * bytes_per_pixel_],
&image_pointer[write_target_pointer_ * bytes_per_pixel_],
uint8_t *start_pointer = pointer_to_location(write_area.x, write_area.y);
memcpy( &start_pointer[-bytes_per_pixel_],
start_pointer,
bytes_per_pixel_);
memcpy( &image_pointer[(write_target_pointer_ + actual_length) * bytes_per_pixel_],
&image_pointer[(write_target_pointer_ + actual_length - 1) * bytes_per_pixel_],
memcpy( &start_pointer[actual_length * bytes_per_pixel_],
&start_pointer[(actual_length - 1) * bytes_per_pixel_],
bytes_per_pixel_);
// return any allocated length that wasn't actually used to the available pool
next_write_x_position_ -= (last_allocation_amount_ - actual_length);
}
uint16_t TextureBuilder::get_last_write_x_position()
{
return write_x_position_;
}
uint16_t TextureBuilder::get_last_write_y_position()
{
return write_y_position_;
}
void TextureBuilder::submit()
{
uint16_t height = write_y_position_ + (next_write_x_position_ ? 1 : 0);
next_write_x_position_ = next_write_y_position_ = 0;
uint16_t height = write_areas_start_y_ + (write_areas_start_x_ ? 1 : 0);
did_submit_ = true;
glTexSubImage2D( GL_TEXTURE_2D, 0,
0, 0,
@ -141,3 +144,54 @@ void TextureBuilder::submit()
formatForDepth(bytes_per_pixel_), GL_UNSIGNED_BYTE,
image_.data());
}
void TextureBuilder::flush(const std::function<void(const std::vector<WriteArea> &write_areas, size_t count)> &function)
{
bool was_full = is_full_;
if(did_submit_)
{
write_areas_start_y_ = write_areas_start_x_ = 0;
is_full_ = false;
}
if(number_of_write_areas_ && !was_full)
{
if(write_areas_[0].x != write_areas_start_x_+1 || write_areas_[0].y != write_areas_start_y_)
{
for(size_t area = 0; area < number_of_write_areas_; area++)
{
WriteArea &write_area = write_areas_[area];
if(write_areas_start_x_ + write_area.length + 2 > InputBufferBuilderWidth)
{
write_areas_start_x_ = 0;
write_areas_start_y_ ++;
if(write_areas_start_y_ == InputBufferBuilderHeight)
{
is_full_ = true;
break;
}
}
memmove(
pointer_to_location(write_areas_start_x_, write_areas_start_y_),
pointer_to_location(write_area.x - 1, write_area.y),
(write_area.length + 2) * bytes_per_pixel_);
write_area.x = write_areas_start_x_ + 1;
write_area.y = write_areas_start_y_;
}
}
if(!is_full_)
{
function(write_areas_, number_of_write_areas_);
write_areas_start_x_ = write_areas_[number_of_write_areas_-1].x + write_areas_[number_of_write_areas_-1].length + 1;
write_areas_start_y_ = write_areas_[number_of_write_areas_-1].y;
}
}
did_submit_ = false;
number_of_write_areas_ = 0;
}

View File

@ -10,6 +10,7 @@
#define Outputs_CRT_Internals_TextureBuilder_hpp
#include <cstdint>
#include <functional>
#include <memory>
#include <vector>
@ -40,12 +41,6 @@ class TextureBuilder {
/// and indicates that its actual final size was @c actual_length.
void reduce_previous_allocation_to(size_t actual_length);
/// @returns the start column for the most recent allocated write area.
uint16_t get_last_write_x_position();
/// @returns the row of the most recent allocated write area.
uint16_t get_last_write_y_position();
/// @returns @c true if all future calls to @c allocate_write_area will fail on account of the input texture
/// being full; @c false if calls may succeed.
bool is_full();
@ -53,23 +48,33 @@ class TextureBuilder {
/// Updates the currently-bound texture with all new data provided since the last @c submit.
void submit();
struct WriteArea {
uint16_t x, y, length;
};
/// Finalises all write areas allocated since the last call to @c flush. Only finalised areas will be
/// submitted upon the next @c submit. The supplied function will be called with a list of write areas
/// allocated, indicating their final resting locations and their lengths.
void flush(const std::function<void(const std::vector<WriteArea> &write_areas, size_t count)> &);
private:
// where pixel data will be put to the next time a write is requested
uint16_t next_write_x_position_, next_write_y_position_;
// the most recent position returned for pixel data writing
uint16_t write_x_position_, write_y_position_;
// details of the most recent allocation
size_t write_target_pointer_;
size_t last_allocation_amount_;
// the buffer size
size_t bytes_per_pixel_;
// the buffer
std::vector<uint8_t> image_;
GLuint texture_name_;
// the current list of write areas
std::vector<WriteArea> write_areas_;
size_t number_of_write_areas_;
bool is_full_;
bool did_submit_;
inline uint8_t *pointer_to_location(uint16_t x, uint16_t y);
// Usually: the start position for the current batch of write areas.
// Caveat: reset to the origin upon a submit. So used in comparison by flush to
// determine whether the current batch of write areas needs to be relocated.
uint16_t write_areas_start_x_, write_areas_start_y_;
};
}