summaryrefslogtreecommitdiffstats
path: root/src/video_core/renderer_opengl
diff options
context:
space:
mode:
authorReinUsesLisp <reinuseslisp@airmail.cc>2020-05-11 21:35:04 +0200
committerReinUsesLisp <reinuseslisp@airmail.cc>2020-06-24 07:36:14 +0200
commit32485917ba7cb7b2f0cad766c0897365294650a7 (patch)
tree48805f8321d9352203664a2fb28e6e504684b11d /src/video_core/renderer_opengl
parentgl_device: Expose NV_vertex_buffer_unified_memory except on Turing (diff)
downloadyuzu-32485917ba7cb7b2f0cad766c0897365294650a7.tar
yuzu-32485917ba7cb7b2f0cad766c0897365294650a7.tar.gz
yuzu-32485917ba7cb7b2f0cad766c0897365294650a7.tar.bz2
yuzu-32485917ba7cb7b2f0cad766c0897365294650a7.tar.lz
yuzu-32485917ba7cb7b2f0cad766c0897365294650a7.tar.xz
yuzu-32485917ba7cb7b2f0cad766c0897365294650a7.tar.zst
yuzu-32485917ba7cb7b2f0cad766c0897365294650a7.zip
Diffstat (limited to 'src/video_core/renderer_opengl')
-rw-r--r--src/video_core/renderer_opengl/gl_buffer_cache.cpp24
-rw-r--r--src/video_core/renderer_opengl/gl_buffer_cache.h20
-rw-r--r--src/video_core/renderer_opengl/gl_rasterizer.cpp44
-rw-r--r--src/video_core/renderer_opengl/gl_stream_buffer.cpp11
-rw-r--r--src/video_core/renderer_opengl/gl_stream_buffer.h11
5 files changed, 70 insertions, 40 deletions
diff --git a/src/video_core/renderer_opengl/gl_buffer_cache.cpp b/src/video_core/renderer_opengl/gl_buffer_cache.cpp
index ad0577a4f..e09b47f57 100644
--- a/src/video_core/renderer_opengl/gl_buffer_cache.cpp
+++ b/src/video_core/renderer_opengl/gl_buffer_cache.cpp
@@ -22,21 +22,28 @@ using Maxwell = Tegra::Engines::Maxwell3D::Regs;
MICROPROFILE_DEFINE(OpenGL_Buffer_Download, "OpenGL", "Buffer Download", MP_RGB(192, 192, 128));
-Buffer::Buffer(VAddr cpu_addr, const std::size_t size) : VideoCommon::BufferBlock{cpu_addr, size} {
+Buffer::Buffer(const Device& device, VAddr cpu_addr, std::size_t size)
+ : VideoCommon::BufferBlock{cpu_addr, size} {
gl_buffer.Create();
glNamedBufferData(gl_buffer.handle, static_cast<GLsizeiptr>(size), nullptr, GL_DYNAMIC_DRAW);
+ if (device.HasVertexBufferUnifiedMemory()) {
+ glMakeNamedBufferResidentNV(gl_buffer.handle, GL_READ_WRITE);
+ glGetNamedBufferParameterui64vNV(gl_buffer.handle, GL_BUFFER_GPU_ADDRESS_NV, &gpu_address);
+ }
}
Buffer::~Buffer() = default;
OGLBufferCache::OGLBufferCache(RasterizerOpenGL& rasterizer, Core::System& system,
- const Device& device, std::size_t stream_size)
- : GenericBufferCache{rasterizer, system, std::make_unique<OGLStreamBuffer>(stream_size, true)} {
+ const Device& device_, std::size_t stream_size)
+ : GenericBufferCache{rasterizer, system,
+ std::make_unique<OGLStreamBuffer>(device_, stream_size, true)},
+ device{device_} {
if (!device.HasFastBufferSubData()) {
return;
}
- static constexpr auto size = static_cast<GLsizeiptr>(Maxwell::MaxConstBufferSize);
+ static constexpr GLsizeiptr size = static_cast<GLsizeiptr>(Maxwell::MaxConstBufferSize);
glCreateBuffers(static_cast<GLsizei>(std::size(cbufs)), std::data(cbufs));
for (const GLuint cbuf : cbufs) {
glNamedBufferData(cbuf, size, nullptr, GL_STREAM_DRAW);
@@ -48,11 +55,11 @@ OGLBufferCache::~OGLBufferCache() {
}
std::shared_ptr<Buffer> OGLBufferCache::CreateBlock(VAddr cpu_addr, std::size_t size) {
- return std::make_shared<Buffer>(cpu_addr, size);
+ return std::make_shared<Buffer>(device, cpu_addr, size);
}
-GLuint OGLBufferCache::GetEmptyBuffer(std::size_t) {
- return 0;
+OGLBufferCache::BufferInfo OGLBufferCache::GetEmptyBuffer(std::size_t) {
+ return {0, 0, 0};
}
void OGLBufferCache::UploadBlockData(const Buffer& buffer, std::size_t offset, std::size_t size,
@@ -79,8 +86,9 @@ OGLBufferCache::BufferInfo OGLBufferCache::ConstBufferUpload(const void* raw_poi
std::size_t size) {
DEBUG_ASSERT(cbuf_cursor < std::size(cbufs));
const GLuint cbuf = cbufs[cbuf_cursor++];
+
glNamedBufferSubData(cbuf, 0, static_cast<GLsizeiptr>(size), raw_pointer);
- return {cbuf, 0};
+ return {cbuf, 0, 0};
}
} // namespace OpenGL
diff --git a/src/video_core/renderer_opengl/gl_buffer_cache.h b/src/video_core/renderer_opengl/gl_buffer_cache.h
index a49aaf9c4..6462cfae5 100644
--- a/src/video_core/renderer_opengl/gl_buffer_cache.h
+++ b/src/video_core/renderer_opengl/gl_buffer_cache.h
@@ -25,15 +25,20 @@ class RasterizerOpenGL;
class Buffer : public VideoCommon::BufferBlock {
public:
- explicit Buffer(VAddr cpu_addr, const std::size_t size);
+ explicit Buffer(const Device& device, VAddr cpu_addr, std::size_t size);
~Buffer();
- GLuint Handle() const {
+ GLuint Handle() const noexcept {
return gl_buffer.handle;
}
+ u64 Address() const noexcept {
+ return gpu_address;
+ }
+
private:
OGLBuffer gl_buffer;
+ u64 gpu_address = 0;
};
using GenericBufferCache = VideoCommon::BufferCache<Buffer, GLuint, OGLStreamBuffer>;
@@ -43,7 +48,7 @@ public:
const Device& device, std::size_t stream_size);
~OGLBufferCache();
- GLuint GetEmptyBuffer(std::size_t) override;
+ BufferInfo GetEmptyBuffer(std::size_t) override;
void Acquire() noexcept {
cbuf_cursor = 0;
@@ -64,10 +69,13 @@ protected:
BufferInfo ConstBufferUpload(const void* raw_pointer, std::size_t size) override;
private:
+ static constexpr std::size_t NUM_CBUFS = Tegra::Engines::Maxwell3D::Regs::MaxConstBuffers *
+ Tegra::Engines::Maxwell3D::Regs::MaxShaderProgram;
+
+ const Device& device;
+
std::size_t cbuf_cursor = 0;
- std::array<GLuint, Tegra::Engines::Maxwell3D::Regs::MaxConstBuffers *
- Tegra::Engines::Maxwell3D::Regs::MaxShaderProgram>
- cbufs;
+ std::array<GLuint, NUM_CBUFS> cbufs{};
};
} // namespace OpenGL
diff --git a/src/video_core/renderer_opengl/gl_rasterizer.cpp b/src/video_core/renderer_opengl/gl_rasterizer.cpp
index 2d6c11320..7cb378a71 100644
--- a/src/video_core/renderer_opengl/gl_rasterizer.cpp
+++ b/src/video_core/renderer_opengl/gl_rasterizer.cpp
@@ -253,8 +253,8 @@ void RasterizerOpenGL::SetupVertexBuffer() {
glBindVertexBuffer(static_cast<GLuint>(index), 0, 0, vertex_array.stride);
continue;
}
- const auto [vertex_buffer, vertex_buffer_offset] = buffer_cache.UploadMemory(start, size);
- glBindVertexBuffer(static_cast<GLuint>(index), vertex_buffer, vertex_buffer_offset,
+ const auto info = buffer_cache.UploadMemory(start, size);
+ glBindVertexBuffer(static_cast<GLuint>(index), info.handle, info.offset,
vertex_array.stride);
}
}
@@ -285,9 +285,9 @@ GLintptr RasterizerOpenGL::SetupIndexBuffer() {
MICROPROFILE_SCOPE(OpenGL_Index);
const auto& regs = system.GPU().Maxwell3D().regs;
const std::size_t size = CalculateIndexBufferSize();
- const auto [buffer, offset] = buffer_cache.UploadMemory(regs.index_array.IndexStart(), size);
- glBindBuffer(GL_ELEMENT_ARRAY_BUFFER, buffer);
- return offset;
+ const auto info = buffer_cache.UploadMemory(regs.index_array.IndexStart(), size);
+ glBindBuffer(GL_ELEMENT_ARRAY_BUFFER, info.handle);
+ return info.offset;
}
void RasterizerOpenGL::SetupShaders(GLenum primitive_mode) {
@@ -643,9 +643,9 @@ void RasterizerOpenGL::Draw(bool is_indexed, bool is_instanced) {
if (!device.UseAssemblyShaders()) {
MaxwellUniformData ubo;
ubo.SetFromRegs(gpu);
- const auto [buffer, offset] =
+ const auto info =
buffer_cache.UploadHostMemory(&ubo, sizeof(ubo), device.GetUniformBufferAlignment());
- glBindBufferRange(GL_UNIFORM_BUFFER, EmulationUniformBlockBinding, buffer, offset,
+ glBindBufferRange(GL_UNIFORM_BUFFER, EmulationUniformBlockBinding, info.handle, info.offset,
static_cast<GLsizeiptr>(sizeof(ubo)));
}
@@ -956,8 +956,7 @@ void RasterizerOpenGL::SetupConstBuffer(GLenum stage, u32 binding,
if (device.UseAssemblyShaders()) {
glBindBufferRangeNV(stage, entry.GetIndex(), 0, 0, 0);
} else {
- glBindBufferRange(GL_UNIFORM_BUFFER, binding,
- buffer_cache.GetEmptyBuffer(sizeof(float)), 0, sizeof(float));
+ glBindBufferRange(GL_UNIFORM_BUFFER, binding, 0, 0, sizeof(float));
}
return;
}
@@ -970,24 +969,25 @@ void RasterizerOpenGL::SetupConstBuffer(GLenum stage, u32 binding,
const std::size_t alignment = use_unified ? 4 : device.GetUniformBufferAlignment();
const GPUVAddr gpu_addr = buffer.address;
- auto [cbuf, offset] = buffer_cache.UploadMemory(gpu_addr, size, alignment, false, fast_upload);
+ auto info = buffer_cache.UploadMemory(gpu_addr, size, alignment, false, fast_upload);
if (device.UseAssemblyShaders()) {
UNIMPLEMENTED_IF(use_unified);
- if (offset != 0) {
+ if (info.offset != 0) {
const GLuint staging_cbuf = staging_cbufs[current_cbuf++];
- glCopyNamedBufferSubData(cbuf, staging_cbuf, offset, 0, size);
- cbuf = staging_cbuf;
- offset = 0;
+ glCopyNamedBufferSubData(info.handle, staging_cbuf, info.offset, 0, size);
+ info.handle = staging_cbuf;
+ info.offset = 0;
}
- glBindBufferRangeNV(stage, binding, cbuf, offset, size);
+ glBindBufferRangeNV(stage, binding, info.handle, info.offset, size);
return;
}
if (use_unified) {
- glCopyNamedBufferSubData(cbuf, unified_uniform_buffer.handle, offset, unified_offset, size);
+ glCopyNamedBufferSubData(info.handle, unified_uniform_buffer.handle, info.offset,
+ unified_offset, size);
} else {
- glBindBufferRange(GL_UNIFORM_BUFFER, binding, cbuf, offset, size);
+ glBindBufferRange(GL_UNIFORM_BUFFER, binding, info.handle, info.offset, size);
}
}
@@ -1023,9 +1023,8 @@ void RasterizerOpenGL::SetupComputeGlobalMemory(Shader* kernel) {
void RasterizerOpenGL::SetupGlobalMemory(u32 binding, const GlobalMemoryEntry& entry,
GPUVAddr gpu_addr, std::size_t size) {
const auto alignment{device.GetShaderStorageBufferAlignment()};
- const auto [ssbo, buffer_offset] =
- buffer_cache.UploadMemory(gpu_addr, size, alignment, entry.is_written);
- glBindBufferRange(GL_SHADER_STORAGE_BUFFER, binding, ssbo, buffer_offset,
+ const auto info = buffer_cache.UploadMemory(gpu_addr, size, alignment, entry.is_written);
+ glBindBufferRange(GL_SHADER_STORAGE_BUFFER, binding, info.handle, info.offset,
static_cast<GLsizeiptr>(size));
}
@@ -1712,8 +1711,9 @@ void RasterizerOpenGL::EndTransformFeedback() {
const GLuint handle = transform_feedback_buffers[index].handle;
const GPUVAddr gpu_addr = binding.Address();
const std::size_t size = binding.buffer_size;
- const auto [dest_buffer, offset] = buffer_cache.UploadMemory(gpu_addr, size, 4, true);
- glCopyNamedBufferSubData(handle, dest_buffer, 0, offset, static_cast<GLsizeiptr>(size));
+ const auto info = buffer_cache.UploadMemory(gpu_addr, size, 4, true);
+ glCopyNamedBufferSubData(handle, info.handle, 0, info.offset,
+ static_cast<GLsizeiptr>(size));
}
}
diff --git a/src/video_core/renderer_opengl/gl_stream_buffer.cpp b/src/video_core/renderer_opengl/gl_stream_buffer.cpp
index aeafcfbfe..164df4feb 100644
--- a/src/video_core/renderer_opengl/gl_stream_buffer.cpp
+++ b/src/video_core/renderer_opengl/gl_stream_buffer.cpp
@@ -2,12 +2,13 @@
// Licensed under GPLv2 or any later version
// Refer to the license.txt file included.
-#include <deque>
+#include <tuple>
#include <vector>
#include "common/alignment.h"
#include "common/assert.h"
#include "common/microprofile.h"
+#include "video_core/renderer_opengl/gl_device.h"
#include "video_core/renderer_opengl/gl_stream_buffer.h"
MICROPROFILE_DEFINE(OpenGL_StreamBuffer, "OpenGL", "Stream Buffer Orphaning",
@@ -15,7 +16,8 @@ MICROPROFILE_DEFINE(OpenGL_StreamBuffer, "OpenGL", "Stream Buffer Orphaning",
namespace OpenGL {
-OGLStreamBuffer::OGLStreamBuffer(GLsizeiptr size, bool vertex_data_usage) : buffer_size(size) {
+OGLStreamBuffer::OGLStreamBuffer(const Device& device, GLsizeiptr size, bool vertex_data_usage)
+ : buffer_size(size) {
gl_buffer.Create();
GLsizeiptr allocate_size = size;
@@ -32,6 +34,11 @@ OGLStreamBuffer::OGLStreamBuffer(GLsizeiptr size, bool vertex_data_usage) : buff
glNamedBufferStorage(gl_buffer.handle, allocate_size, nullptr, flags);
mapped_ptr = static_cast<u8*>(
glMapNamedBufferRange(gl_buffer.handle, 0, buffer_size, flags | GL_MAP_FLUSH_EXPLICIT_BIT));
+
+ if (device.HasVertexBufferUnifiedMemory()) {
+ glMakeNamedBufferResidentNV(gl_buffer.handle, GL_READ_ONLY);
+ glGetNamedBufferParameterui64vNV(gl_buffer.handle, GL_BUFFER_GPU_ADDRESS_NV, &gpu_address);
+ }
}
OGLStreamBuffer::~OGLStreamBuffer() {
diff --git a/src/video_core/renderer_opengl/gl_stream_buffer.h b/src/video_core/renderer_opengl/gl_stream_buffer.h
index 826c2e361..e67a82980 100644
--- a/src/video_core/renderer_opengl/gl_stream_buffer.h
+++ b/src/video_core/renderer_opengl/gl_stream_buffer.h
@@ -11,9 +11,11 @@
namespace OpenGL {
+class Device;
+
class OGLStreamBuffer : private NonCopyable {
public:
- explicit OGLStreamBuffer(GLsizeiptr size, bool vertex_data_usage);
+ explicit OGLStreamBuffer(const Device& device, GLsizeiptr size, bool vertex_data_usage);
~OGLStreamBuffer();
/*
@@ -32,13 +34,18 @@ public:
return gl_buffer.handle;
}
- GLsizeiptr Size() const {
+ u64 Address() const {
+ return gpu_address;
+ }
+
+ GLsizeiptr Size() const noexcept {
return buffer_size;
}
private:
OGLBuffer gl_buffer;
+ GLuint64EXT gpu_address = 0;
GLintptr buffer_pos = 0;
GLsizeiptr buffer_size = 0;
GLintptr mapped_offset = 0;