1
0
mirror of https://github.com/TomHarte/CLK.git synced 2024-07-06 01:28:57 +00:00

As messy as it is, this use glMapBufferRange to avoid explicit buffer submits.

This commit is contained in:
Thomas Harte 2016-03-16 22:29:22 -04:00
parent ca35a7e222
commit 0d27d3bb7f
4 changed files with 82 additions and 49 deletions

View File

@ -82,14 +82,18 @@ void OpenGLOutputBuilder::draw_frame(unsigned int output_width, unsigned int out
glTexImage2D(GL_TEXTURE_2D, 0, (GLint)format, InputBufferBuilderWidth, InputBufferBuilderHeight, 0, format, GL_UNSIGNED_BYTE, _buffer_builder->buffers[buffer].data);
}
prepare_composite_input_shader();
prepare_rgb_output_shader();
glGenVertexArrays(1, &output_vertex_array);
glGenBuffers(1, &output_array_buffer);
output_vertices_per_slice = 0;
prepare_composite_input_shader();
prepare_rgb_output_shader();
glBindBuffer(GL_ARRAY_BUFFER, output_array_buffer);
glBufferData(GL_ARRAY_BUFFER, buffer_size, NULL, GL_STREAM_DRAW);
_output_buffer_data = (uint8_t *)glMapBufferRange(GL_ARRAY_BUFFER, 0, buffer_size, GL_MAP_WRITE_BIT | GL_MAP_UNSYNCHRONIZED_BIT);
_output_buffer_data_pointer = 0;
glBindVertexArray(output_vertex_array);
prepare_output_vertex_array();
@ -144,8 +148,8 @@ void OpenGLOutputBuilder::draw_frame(unsigned int output_width, unsigned int out
// check for anything to decode from composite
if(_composite_src_runs->number_of_vertices)
{
composite_input_shader_program->bind();
_composite_src_runs->reset();
// composite_input_shader_program->bind();
// _composite_src_runs->reset();
}
// _output_mutex->unlock();
@ -156,24 +160,24 @@ void OpenGLOutputBuilder::draw_frame(unsigned int output_width, unsigned int out
// glGetIntegerv(GL_VIEWPORT, results);
// ensure array buffer is up to date
glBindBuffer(GL_ARRAY_BUFFER, output_array_buffer);
size_t max_number_of_vertices = 0;
for(int c = 0; c < NumberOfFields; c++)
{
max_number_of_vertices = std::max(max_number_of_vertices, _run_builders[c]->number_of_vertices);
}
if(output_vertices_per_slice < max_number_of_vertices)
{
output_vertices_per_slice = max_number_of_vertices;
glBufferData(GL_ARRAY_BUFFER, (GLsizeiptr)(max_number_of_vertices * OutputVertexSize * OutputVertexSize), NULL, GL_STREAM_DRAW);
for(unsigned int c = 0; c < NumberOfFields; c++)
{
uint8_t *data = &_run_builders[c]->_runs[0];
glBufferSubData(GL_ARRAY_BUFFER, (GLsizeiptr)(c * output_vertices_per_slice * OutputVertexSize), (GLsizeiptr)(_run_builders[c]->number_of_vertices * OutputVertexSize), data);
_run_builders[c]->uploaded_vertices = _run_builders[c]->number_of_vertices;
}
}
// glBindBuffer(GL_ARRAY_BUFFER, output_array_buffer);
// size_t max_number_of_vertices = 0;
// for(int c = 0; c < NumberOfFields; c++)
// {
// max_number_of_vertices = std::max(max_number_of_vertices, _run_builders[c]->number_of_vertices);
// }
// if(output_vertices_per_slice < max_number_of_vertices)
// {
// output_vertices_per_slice = max_number_of_vertices;
// glBufferData(GL_ARRAY_BUFFER, (GLsizeiptr)(max_number_of_vertices * OutputVertexSize * OutputVertexSize), NULL, GL_STREAM_DRAW);
//
// for(unsigned int c = 0; c < NumberOfFields; c++)
// {
// uint8_t *data = &_run_builders[c]->_runs[0];
// glBufferSubData(GL_ARRAY_BUFFER, (GLsizeiptr)(c * output_vertices_per_slice * OutputVertexSize), (GLsizeiptr)(_run_builders[c]->number_of_vertices * OutputVertexSize), data);
// _run_builders[c]->uploaded_vertices = _run_builders[c]->number_of_vertices;
// }
// }
// switch to the output shader
if(rgb_shader_program)
@ -202,17 +206,29 @@ void OpenGLOutputBuilder::draw_frame(unsigned int output_width, unsigned int out
{
glUniform1f(timestampBaseUniform, (GLfloat)total_age);
if(_run_builders[run]->uploaded_vertices != _run_builders[run]->number_of_vertices)
{
uint8_t *data = &_run_builders[run]->_runs[_run_builders[run]->uploaded_vertices * OutputVertexSize];
glBufferSubData(GL_ARRAY_BUFFER,
(GLsizeiptr)(((run * output_vertices_per_slice) + _run_builders[run]->uploaded_vertices) * OutputVertexSize),
(GLsizeiptr)((_run_builders[run]->number_of_vertices - _run_builders[run]->uploaded_vertices) * OutputVertexSize), data);
_run_builders[run]->uploaded_vertices = _run_builders[run]->number_of_vertices;
}
// if(_run_builders[run]->uploaded_vertices != _run_builders[run]->number_of_vertices)
// {
// uint8_t *data = &_run_builders[run]->_runs[_run_builders[run]->uploaded_vertices * OutputVertexSize];
// glBufferSubData(GL_ARRAY_BUFFER,
// (GLsizeiptr)(((run * output_vertices_per_slice) + _run_builders[run]->uploaded_vertices) * OutputVertexSize),
// (GLsizeiptr)((_run_builders[run]->number_of_vertices - _run_builders[run]->uploaded_vertices) * OutputVertexSize), data);
// _run_builders[run]->uploaded_vertices = _run_builders[run]->number_of_vertices;
// }
// draw this frame
glDrawArrays(GL_TRIANGLE_STRIP, (GLint)(run * output_vertices_per_slice), (GLsizei)_run_builders[run]->number_of_vertices);
// glDrawArrays(GL_TRIANGLE_STRIP, (GLint)(run * output_vertices_per_slice), (GLsizei)_run_builders[run]->number_of_vertices);
GLsizei count = (GLsizei)_run_builders[run]->number_of_vertices;
GLsizei max_count = (GLsizei)((buffer_size - _run_builders[run]->start) / InputVertexSize);
if(count < max_count)
{
glDrawArrays(GL_TRIANGLE_STRIP, (GLint)(_run_builders[run]->start / InputVertexSize), count);
}
else
{
glDrawArrays(GL_TRIANGLE_STRIP, (GLint)(_run_builders[run]->start / InputVertexSize), max_count);
glDrawArrays(GL_TRIANGLE_STRIP, 0, count - max_count);
}
}
// advance back in time
@ -549,11 +565,11 @@ void OpenGLOutputBuilder::set_output_device(OutputDevice output_device)
{
_output_device = output_device;
for(int builder = 0; builder < NumberOfFields; builder++)
{
_run_builders[builder]->reset();
}
_composite_src_runs->reset();
// for(int builder = 0; builder < NumberOfFields; builder++)
// {
// _run_builders[builder]->reset();
// }
// _composite_src_runs->reset();
_composite_src_output_y = 0;
}
}

View File

@ -47,6 +47,9 @@ const int InputBufferBuilderHeight = 1024;
const int IntermediateBufferWidth = 2048;
const int IntermediateBufferHeight = 2048;
const GLsizeiptr buffer_size = (GLsizeiptr)(312 * 6 * 6 * OutputVertexSize);
// Runs are divided discretely by vertical syncs in order to put a usable bounds on the uniform used to track
// run age; that therefore creates a discrete number of fields that are stored. This number should be the
// number of historic fields that are required fully to
@ -136,23 +139,30 @@ class OpenGLOutputBuilder {
inline uint8_t *get_next_input_run()
{
_output_mutex->lock();
return (_output_device == Monitor) ? _run_builders[_run_write_pointer]->get_next_run(6) : _composite_src_runs->get_next_run(2);
if (_output_buffer_data_pointer + 6 * InputVertexSize > buffer_size) _output_buffer_data_pointer = 0;
uint8_t *pointer = &_output_buffer_data[_output_buffer_data_pointer];
_output_buffer_data_pointer += 6 * InputVertexSize;
return pointer;
// _output_mutex->lock();
// return (_output_device == Monitor) ? _run_builders[_run_write_pointer]->get_next_run(6) : _composite_src_runs->get_next_run(2);
}
inline void complete_input_run()
{
_output_mutex->unlock();
_run_builders[_run_write_pointer]->number_of_vertices += 6;
// _output_mutex->unlock();
}
inline uint8_t *get_next_output_run()
{
_output_mutex->lock();
return (_output_device == Monitor) ? _run_builders[_run_write_pointer]->get_next_run(6) : _composite_src_runs->get_next_run(2);
// _output_mutex->lock();
// return (_output_device == Monitor) ? _run_builders[_run_write_pointer]->get_next_run(6) : _composite_src_runs->get_next_run(2);
return nullptr;
}
inline void complete_output_run()
{
// _output_mutex->unlock();
}
inline OutputDevice get_output_device()
@ -183,7 +193,9 @@ class OpenGLOutputBuilder {
inline void increment_field()
{
_run_write_pointer = (_run_write_pointer + 1)%NumberOfFields;
_run_builders[_run_write_pointer]->reset();
_run_builders[_run_write_pointer]->start = _output_buffer_data_pointer;
_run_builders[_run_write_pointer]->duration = 0;
_run_builders[_run_write_pointer]->number_of_vertices = 0;
}
inline void allocate_write_area(size_t required_length)
@ -228,6 +240,9 @@ class OpenGLOutputBuilder {
// TODO: update related uniforms
}
uint8_t *_output_buffer_data;
size_t _output_buffer_data_pointer;
};
}

View File

@ -11,7 +11,7 @@
using namespace Outputs::CRT;
void CRTRunBuilder::reset()
/*void CRTRunBuilder::reset()
{
number_of_vertices = 0;
uploaded_vertices = 0;
@ -31,3 +31,4 @@ uint8_t *CRTRunBuilder::get_next_run(size_t number_of_vertices_in_run)
return next_run;
}
*/

View File

@ -15,24 +15,25 @@ namespace Outputs {
namespace CRT {
struct CRTRunBuilder {
CRTRunBuilder(size_t vertex_size) : _vertex_size(vertex_size) { reset(); }
CRTRunBuilder(size_t vertex_size) : _vertex_size(vertex_size), duration(0), start(0), number_of_vertices(0) {} // reset();
// Resets the run builder.
void reset();
// void reset();
// Getter for new storage plus backing storage; in RGB mode input runs will map directly
// from the input buffer to the screen. In composite mode input runs will map from the
// input buffer to the processing buffer, and output runs will map from the processing
// buffer to the screen.
uint8_t *get_next_run(size_t number_of_vertices);
std::vector<uint8_t> _runs;
// uint8_t *get_next_run(size_t number_of_vertices);
// std::vector<uint8_t> _runs;
// Container for total length in cycles of all contained runs.
uint32_t duration;
size_t start;
// Storage for the length of run data uploaded so far; reset to zero by reset but otherwise
// entrusted to the CRT to update.
size_t uploaded_vertices;
// size_t uploaded_vertices;
size_t number_of_vertices;
private: