RPCS3/rpcs3
1
0
Fork 0
mirror of https://github.com/RPCS3/rpcs3.git synced 2024-11-22 10:42:36 +01:00
Code Issues Releases Wiki Activity

rsx: Remove ARGB->D24S8 conversion shader which has been deprecated for years since compute capabilities were added to the emulator

Browse Source
This commit is contained in:
kd-11 2020-06-14 21:39:18 +03:00 committed by kd-11
parent 717e851a99
commit 8d8fb4a2e4
6 changed files with 9 additions and 281 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

91
rpcs3/Emu/RSX/GL/GLDraw.cpp
View File

@ -600,26 +600,6 @@ void GLGSRender::end()
m_frame_stats.setup_time += m_profiler.duration();
gl::command_context cmd{ gl_state };
gl::render_target *ds = std::get<1>(m_rtts.m_bound_depth_stencil);
// Handle special memory barrier for ARGB8->D24S8 in an active DSV
if (ds && ds->old_contents.size() == 1 &&
ds->old_contents[0].source->get_internal_format() == gl::texture::internal_format::rgba8)
{
gl_state.enable(GL_FALSE, GL_SCISSOR_TEST);
// TODO: Stencil transfer
gl::g_hw_blitter->fast_clear_image(cmd, ds, 1.f, 0xFF);
ds->old_contents[0].init_transfer(ds);
m_depth_converter.run(ds->old_contents[0].src_rect(),
ds->old_contents[0].dst_rect(),
ds->old_contents[0].source, ds);
ds->on_write();
}
// Active texture environment is used to decode shaders
m_profiler.start();
load_texture_env();
@ -645,71 +625,16 @@ void GLGSRender::end()
m_gl_texture_cache.release_uncached_temporary_subresources();
m_frame_stats.textures_upload_time += m_profiler.duration();
// Optionally do memory synchronization if the texture stage has not yet triggered this
if (true)//g_cfg.video.strict_rendering_mode)
gl_state.enable(GL_FALSE, GL_SCISSOR_TEST);
gl::command_context cmd{ gl_state };
if (auto ds = std::get<1>(m_rtts.m_bound_depth_stencil)) ds->write_barrier(cmd);
for (auto &rtt : m_rtts.m_bound_render_targets)
{
gl_state.enable(GL_FALSE, GL_SCISSOR_TEST);
if (ds) ds->write_barrier(cmd);
for (auto &rtt : m_rtts.m_bound_render_targets)
if (auto surface = std::get<1>(rtt))
{
if (auto surface = std::get<1>(rtt))
{
surface->write_barrier(cmd);
}
}
}
else
{
rsx::simple_array<int> buffers_to_clear;
bool clear_all_color = true;
bool clear_depth = false;
for (int index = 0; index < 4; index++)
{
if (m_rtts.m_bound_render_targets[index].first)
{
if (!m_rtts.m_bound_render_targets[index].second->dirty())
clear_all_color = false;
else
buffers_to_clear.push_back(index);
}
}
if (ds && ds->dirty())
{
clear_depth = true;
}
if (clear_depth || !buffers_to_clear.empty())
{
gl_state.enable(GL_FALSE, GL_SCISSOR_TEST);
GLenum mask = 0;
if (clear_depth)
{
gl_state.depth_mask(GL_TRUE);
gl_state.clear_depth(1.f);
gl_state.clear_stencil(255);
mask |= GL_DEPTH_BUFFER_BIT | GL_STENCIL_BUFFER_BIT;
}
if (clear_all_color)
mask |= GL_COLOR_BUFFER_BIT;
glClear(mask);
if (!buffers_to_clear.empty() && !clear_all_color)
{
GLfloat colors[] = { 0.f, 0.f, 0.f, 0.f };
//It is impossible for the render target to be type A or B here (clear all would have been flagged)
for (auto &i : buffers_to_clear)
glClearBufferfv(GL_COLOR, i, colors);
}
if (clear_depth)
gl_state.depth_mask(rsx::method_registers.depth_write_enabled());
surface->write_barrier(cmd);
}
}

2
rpcs3/Emu/RSX/GL/GLGSRender.cpp
View File

@ -313,7 +313,6 @@ void GLGSRender::on_init_thread()
glEnable(GL_CLIP_DISTANCE0 + 4);
glEnable(GL_CLIP_DISTANCE0 + 5);
m_depth_converter.create();
m_ui_renderer.create();
m_video_output_pass.create();
@ -460,7 +459,6 @@ void GLGSRender::on_exit()
m_null_textures.clear();
m_text_printer.close();
m_gl_texture_cache.destroy();
m_depth_converter.destroy();
m_ui_renderer.destroy();
m_video_output_pass.destroy();

1
rpcs3/Emu/RSX/GL/GLGSRender.h
View File

@ -114,7 +114,6 @@ private:
bool manually_flush_ring_buffers = false;
gl::text_writer m_text_printer;
gl::depth_convert_pass m_depth_converter;
gl::ui_overlay_renderer m_ui_renderer;
gl::video_out_calibration_pass m_video_output_pass;

78
rpcs3/Emu/RSX/GL/GLOverlays.h
View File

@ -270,84 +270,6 @@ namespace gl
}
};
struct depth_convert_pass : public overlay_pass
{
depth_convert_pass()
{
vs_src =
"#version 420\n\n"
"uniform vec2 tex_scale;\n"
"out vec2 tc0;\n"
"\n"
"void main()\n"
"{\n"
" vec2 positions[] = {vec2(-1., -1.), vec2(1., -1.), vec2(-1., 1.), vec2(1., 1.)};\n"
" vec2 coords[] = {vec2(0., 0.), vec2(1., 0.), vec2(0., 1.), vec2(1., 1.)};\n"
" gl_Position = vec4(positions[gl_VertexID % 4], 0., 1.);\n"
" tc0 = coords[gl_VertexID % 4] * tex_scale;\n"
"}\n";
fs_src =
"#version 420\n\n"
"in vec2 tc0;\n"
"layout(binding=31) uniform sampler2D fs0;\n"
"\n"
"void main()\n"
"{\n"
" vec4 rgba_in = texture(fs0, tc0);\n"
" gl_FragDepth = rgba_in.w * 0.99609 + rgba_in.x * 0.00389 + rgba_in.y * 0.00002;\n"
"}\n";
}
void run(const areai& src_area, const areai& dst_area, gl::texture* source, gl::texture* target)
{
const auto src_ratio_x = f32(src_area.x2) / source->width();
const auto src_ratio_y = f32(src_area.y2) / source->height();
program_handle.uniforms["tex_scale"] = color2f(src_ratio_x, src_ratio_y);
saved_sampler_state saved(31, m_sampler);
glBindTexture(GL_TEXTURE_2D, source->id());
overlay_pass::run(static_cast<areau>(dst_area), target->id(), true);
}
};
struct rgba8_unorm_rg16_sfloat_convert_pass : public overlay_pass
{
//Not really needed since directly copying data via ARB_copy_image works out fine
rgba8_unorm_rg16_sfloat_convert_pass()
{
vs_src =
"#version 420\n\n"
"\n"
"void main()\n"
"{\n"
" vec2 positions[] = {vec2(-1., -1.), vec2(1., -1.), vec2(-1., 1.), vec2(1., 1.)};\n"
" gl_Position = vec4(positions[gl_VertexID % 4], 0., 1.);\n"
"}\n";
fs_src =
"#version 420\n\n"
"layout(binding=31) uniform sampler2D fs0;\n"
"layout(location=0) out vec4 ocol;\n"
"\n"
"void main()\n"
"{\n"
" uint value = packUnorm4x8(texelFetch(fs0, ivec2(gl_FragCoord.xy), 0).zyxw);\n"
" ocol.xy = unpackHalf2x16(value);\n"
"}\n";
}
void run(const areau& viewport, GLuint target, GLuint source)
{
saved_sampler_state saved(31, m_sampler);
glBindTexture(GL_TEXTURE_2D, source);
overlay_pass::run(viewport, target, false);
}
};
struct ui_overlay_renderer : public overlay_pass
{
u32 num_elements = 0;

48
rpcs3/Emu/RSX/VK/VKDraw.cpp
View File

@ -908,52 +908,6 @@ void VKGSRender::end()
m_profiler.start();
// Check for data casts
// NOTE: This is deprecated and will be removed soon. The memory barrier invoked before rendering does this better
auto ds = std::get<1>(m_rtts.m_bound_depth_stencil);
if (ds && ds->old_contents.size() == 1 &&
ds->old_contents[0].source->info.format == VK_FORMAT_B8G8R8A8_UNORM)
{
auto key = vk::get_renderpass_key({ static_cast<vk::image*>(ds) });
auto render_pass = vk::get_renderpass(*m_device, key);
verify("Usupported renderpass configuration" HERE), render_pass != VK_NULL_HANDLE;
VkClearDepthStencilValue clear = { 1.f, 0xFF };
VkImageSubresourceRange range = { VK_IMAGE_ASPECT_DEPTH_BIT | VK_IMAGE_ASPECT_STENCIL_BIT, 0, 1, 0, 1 };
// Initialize source data
auto srcobj = vk::as_rtt(ds->old_contents[0].source);
srcobj->read_barrier(*m_current_command_buffer);
auto src = srcobj->get_surface(rsx::surface_access::read);
switch (src->current_layout)
{
case VK_IMAGE_LAYOUT_GENERAL:
case VK_IMAGE_LAYOUT_SHADER_READ_ONLY_OPTIMAL:
break;
//case VK_IMAGE_LAYOUT_COLOR_ATTACHMENT_OPTIMAL:
default:
src->change_layout(*m_current_command_buffer, VK_IMAGE_LAYOUT_SHADER_READ_ONLY_OPTIMAL);
break;
}
// Clear explicitly before starting the inheritance transfer
const bool preinitialized = (ds->current_layout == VK_IMAGE_LAYOUT_GENERAL);
if (!preinitialized) ds->push_layout(*m_current_command_buffer, VK_IMAGE_LAYOUT_TRANSFER_DST_OPTIMAL);
vkCmdClearDepthStencilImage(*m_current_command_buffer, ds->value, VK_IMAGE_LAYOUT_TRANSFER_DST_OPTIMAL, &clear, 1, &range);
if (!preinitialized) ds->pop_layout(*m_current_command_buffer);
// TODO: Stencil transfer
ds->old_contents[0].init_transfer(ds);
vk::get_overlay_pass<vk::depth_convert_pass>()->run(*m_current_command_buffer,
ds->old_contents[0].src_rect(),
ds->old_contents[0].dst_rect(),
src->get_view(0xAAE4, rsx::default_remap_vector),
ds, render_pass);
ds->on_write();
}
load_texture_env();
m_frame_stats.textures_upload_time += m_profiler.duration();
@ -1019,7 +973,7 @@ void VKGSRender::end()
vk::get_appropriate_topology(rsx::method_registers.current_draw_clause.primitive, primitive_emulated);
// Apply write memory barriers
if (ds) ds->write_barrier(*m_current_command_buffer);
if (auto ds = std::get<1>(m_rtts.m_bound_depth_stencil)) ds->write_barrier(*m_current_command_buffer);
for (auto &rtt : m_rtts.m_bound_render_targets)
{

70
rpcs3/Emu/RSX/VK/VKOverlays.h
View File

@ -396,76 +396,6 @@ namespace vk
}
};
// @Deprecated!!
struct depth_convert_pass : public overlay_pass
{
f32 src_scale_x;
f32 src_scale_y;
depth_convert_pass()
{
vs_src =
"#version 450\n"
"#extension GL_ARB_separate_shader_objects : enable\n"
"layout(push_constant) uniform static_data{ vec2 scale; };\n"
"layout(location=0) out vec2 tc0;\n"
"\n"
"void main()\n"
"{\n"
" vec2 positions[] = {vec2(-1., -1.), vec2(1., -1.), vec2(-1., 1.), vec2(1., 1.)};\n"
" vec2 coords[] = {vec2(0., 0.), vec2(1., 0.), vec2(0., 1.), vec2(1., 1.)};\n"
" gl_Position = vec4(positions[gl_VertexIndex % 4], 0., 1.);\n"
" tc0 = coords[gl_VertexIndex % 4] * scale;\n"
"}\n";
fs_src =
"#version 420\n"
"#extension GL_ARB_separate_shader_objects : enable\n"
"#extension GL_ARB_shader_stencil_export : enable\n\n"
"layout(set=0, binding=1) uniform sampler2D fs0;\n"
"layout(location=0) in vec2 tc0;\n"
"\n"
"void main()\n"
"{\n"
" vec4 rgba_in = texture(fs0, tc0);\n"
" gl_FragDepth = rgba_in.w * 0.99609 + rgba_in.x * 0.00389 + rgba_in.y * 0.00002;\n"
"}\n";
renderpass_config.set_depth_mask(true);
renderpass_config.enable_depth_test(VK_COMPARE_OP_ALWAYS);
}
std::vector<VkPushConstantRange> get_push_constants() override
{
VkPushConstantRange constant;
constant.stageFlags = VK_SHADER_STAGE_VERTEX_BIT;
constant.offset = 0;
constant.size = 8;
return { constant };
}
void update_uniforms(vk::command_buffer& cmd, vk::glsl::program* /*program*/) override
{
float dst[2] = { src_scale_x, src_scale_y };
vkCmdPushConstants(cmd, m_pipeline_layout, VK_SHADER_STAGE_VERTEX_BIT, 0, 8, dst);
}
void run(vk::command_buffer& cmd, const areai& src_area, const areai& dst_area, vk::image_view* src, vk::image* dst, VkRenderPass render_pass)
{
auto real_src = src->image();
verify(HERE), real_src;
src_scale_x = static_cast<f32>(src_area.x2) / real_src->width();
src_scale_y = static_cast<f32>(src_area.y2) / real_src->height();
// Coverage sampling disabled, but actually report correct number of samples
renderpass_config.set_multisample_state(dst->samples(), 0xFFFF, false, false, false);
overlay_pass::run(cmd, static_cast<areau>(dst_area), dst, src, render_pass);
}
};
struct ui_overlay_renderer : public overlay_pass
{
f32 m_time = 0.f;