forked from suyu/suyu
gl_buffer_cache: Return used buffer from Upload function
This commit is contained in:
parent
a6d2f52fc3
commit
b54fb8fc4c
4 changed files with 35 additions and 36 deletions
|
@ -4,6 +4,7 @@
|
|||
|
||||
#include <cstring>
|
||||
#include <memory>
|
||||
#include <utility>
|
||||
|
||||
#include "common/alignment.h"
|
||||
#include "core/core.h"
|
||||
|
@ -21,9 +22,10 @@ CachedBufferEntry::CachedBufferEntry(VAddr cpu_addr, std::size_t size, GLintptr
|
|||
OGLBufferCache::OGLBufferCache(RasterizerOpenGL& rasterizer, std::size_t size)
|
||||
: RasterizerCache{rasterizer}, stream_buffer(size, true) {}
|
||||
|
||||
GLintptr OGLBufferCache::UploadMemory(GPUVAddr gpu_addr, std::size_t size, std::size_t alignment,
|
||||
bool cache) {
|
||||
std::pair<GLuint, GLintptr> OGLBufferCache::UploadMemory(GPUVAddr gpu_addr, std::size_t size,
|
||||
std::size_t alignment, bool cache) {
|
||||
std::lock_guard lock{mutex};
|
||||
|
||||
auto& memory_manager = Core::System::GetInstance().GPU().MemoryManager();
|
||||
|
||||
// Cache management is a big overhead, so only cache entries with a given size.
|
||||
|
@ -35,7 +37,7 @@ GLintptr OGLBufferCache::UploadMemory(GPUVAddr gpu_addr, std::size_t size, std::
|
|||
auto entry = TryGet(host_ptr);
|
||||
if (entry) {
|
||||
if (entry->GetSize() >= size && entry->GetAlignment() == alignment) {
|
||||
return entry->GetOffset();
|
||||
return {stream_buffer.GetHandle(), entry->GetOffset()};
|
||||
}
|
||||
Unregister(entry);
|
||||
}
|
||||
|
@ -45,7 +47,7 @@ GLintptr OGLBufferCache::UploadMemory(GPUVAddr gpu_addr, std::size_t size, std::
|
|||
const GLintptr uploaded_offset = buffer_offset;
|
||||
|
||||
if (!host_ptr) {
|
||||
return uploaded_offset;
|
||||
return {stream_buffer.GetHandle(), uploaded_offset};
|
||||
}
|
||||
|
||||
std::memcpy(buffer_ptr, host_ptr, size);
|
||||
|
@ -58,11 +60,12 @@ GLintptr OGLBufferCache::UploadMemory(GPUVAddr gpu_addr, std::size_t size, std::
|
|||
Register(entry);
|
||||
}
|
||||
|
||||
return uploaded_offset;
|
||||
return {stream_buffer.GetHandle(), uploaded_offset};
|
||||
}
|
||||
|
||||
GLintptr OGLBufferCache::UploadHostMemory(const void* raw_pointer, std::size_t size,
|
||||
std::size_t alignment) {
|
||||
std::pair<GLuint, GLintptr> OGLBufferCache::UploadHostMemory(const void* raw_pointer,
|
||||
std::size_t size,
|
||||
std::size_t alignment) {
|
||||
std::lock_guard lock{mutex};
|
||||
AlignBuffer(alignment);
|
||||
std::memcpy(buffer_ptr, raw_pointer, size);
|
||||
|
@ -70,7 +73,7 @@ GLintptr OGLBufferCache::UploadHostMemory(const void* raw_pointer, std::size_t s
|
|||
|
||||
buffer_ptr += size;
|
||||
buffer_offset += size;
|
||||
return uploaded_offset;
|
||||
return {stream_buffer.GetHandle(), uploaded_offset};
|
||||
}
|
||||
|
||||
bool OGLBufferCache::Map(std::size_t max_size) {
|
||||
|
@ -89,10 +92,6 @@ void OGLBufferCache::Unmap() {
|
|||
stream_buffer.Unmap(buffer_offset - buffer_offset_base);
|
||||
}
|
||||
|
||||
GLuint OGLBufferCache::GetHandle() const {
|
||||
return stream_buffer.GetHandle();
|
||||
}
|
||||
|
||||
void OGLBufferCache::AlignBuffer(std::size_t alignment) {
|
||||
// Align the offset, not the mapped pointer
|
||||
const GLintptr offset_aligned =
|
||||
|
|
|
@ -7,6 +7,7 @@
|
|||
#include <cstddef>
|
||||
#include <memory>
|
||||
#include <tuple>
|
||||
#include <utility>
|
||||
|
||||
#include "common/common_types.h"
|
||||
#include "video_core/rasterizer_cache.h"
|
||||
|
@ -53,19 +54,18 @@ class OGLBufferCache final : public RasterizerCache<std::shared_ptr<CachedBuffer
|
|||
public:
|
||||
explicit OGLBufferCache(RasterizerOpenGL& rasterizer, std::size_t size);
|
||||
|
||||
/// Uploads data from a guest GPU address. Returns host's buffer offset where it's been
|
||||
/// allocated.
|
||||
GLintptr UploadMemory(GPUVAddr gpu_addr, std::size_t size, std::size_t alignment = 4,
|
||||
bool cache = true);
|
||||
/// Uploads data from a guest GPU address. Returns the OpenGL buffer where it's located and its
|
||||
/// offset.
|
||||
std::pair<GLuint, GLintptr> UploadMemory(GPUVAddr gpu_addr, std::size_t size,
|
||||
std::size_t alignment = 4, bool cache = true);
|
||||
|
||||
/// Uploads from a host memory. Returns host's buffer offset where it's been allocated.
|
||||
GLintptr UploadHostMemory(const void* raw_pointer, std::size_t size, std::size_t alignment = 4);
|
||||
/// Uploads from a host memory. Returns the OpenGL buffer where it's located and its offset.
|
||||
std::pair<GLuint, GLintptr> UploadHostMemory(const void* raw_pointer, std::size_t size,
|
||||
std::size_t alignment = 4);
|
||||
|
||||
bool Map(std::size_t max_size);
|
||||
void Unmap();
|
||||
|
||||
GLuint GetHandle() const;
|
||||
|
||||
protected:
|
||||
void AlignBuffer(std::size_t alignment);
|
||||
|
||||
|
|
|
@ -129,8 +129,6 @@ GLuint RasterizerOpenGL::SetupVertexFormat() {
|
|||
state.draw.vertex_array = vao;
|
||||
state.ApplyVertexArrayState();
|
||||
|
||||
glVertexArrayElementBuffer(vao, buffer_cache.GetHandle());
|
||||
|
||||
// Use the vertex array as-is, assumes that the data is formatted correctly for OpenGL.
|
||||
// Enables the first 16 vertex attributes always, as we don't know which ones are actually
|
||||
// used until shader time. Note, Tegra technically supports 32, but we're capping this to 16
|
||||
|
@ -197,10 +195,10 @@ void RasterizerOpenGL::SetupVertexBuffer(GLuint vao) {
|
|||
|
||||
ASSERT(end > start);
|
||||
const u64 size = end - start + 1;
|
||||
const GLintptr vertex_buffer_offset = buffer_cache.UploadMemory(start, size);
|
||||
const auto [vertex_buffer, vertex_buffer_offset] = buffer_cache.UploadMemory(start, size);
|
||||
|
||||
// Bind the vertex array to the buffer at the current offset.
|
||||
glVertexArrayVertexBuffer(vao, index, buffer_cache.GetHandle(), vertex_buffer_offset,
|
||||
glVertexArrayVertexBuffer(vao, index, vertex_buffer, vertex_buffer_offset,
|
||||
vertex_array.stride);
|
||||
|
||||
if (regs.instanced_arrays.IsInstancingEnabled(index) && vertex_array.divisor != 0) {
|
||||
|
@ -215,12 +213,16 @@ void RasterizerOpenGL::SetupVertexBuffer(GLuint vao) {
|
|||
gpu.dirty_flags.vertex_array.reset();
|
||||
}
|
||||
|
||||
GLintptr RasterizerOpenGL::SetupIndexBuffer() {
|
||||
GLintptr RasterizerOpenGL::SetupIndexBuffer(GLuint vao) {
|
||||
if (accelerate_draw != AccelDraw::Indexed) {
|
||||
return 0;
|
||||
}
|
||||
MICROPROFILE_SCOPE(OpenGL_Index);
|
||||
const auto& regs = system.GPU().Maxwell3D().regs;
|
||||
return buffer_cache.UploadMemory(regs.index_array.IndexStart(), CalculateIndexBufferSize());
|
||||
const std::size_t size = CalculateIndexBufferSize();
|
||||
const auto [buffer, offset] = buffer_cache.UploadMemory(regs.index_array.IndexStart(), size);
|
||||
glVertexArrayElementBuffer(vao, buffer);
|
||||
return offset;
|
||||
}
|
||||
|
||||
DrawParameters RasterizerOpenGL::SetupDraw(GLintptr index_buffer_offset) {
|
||||
|
@ -235,7 +237,6 @@ DrawParameters RasterizerOpenGL::SetupDraw(GLintptr index_buffer_offset) {
|
|||
params.primitive_mode = MaxwellToGL::PrimitiveTopology(regs.draw.topology);
|
||||
|
||||
if (is_indexed) {
|
||||
MICROPROFILE_SCOPE(OpenGL_Index);
|
||||
params.index_format = MaxwellToGL::IndexFormat(regs.index_array.format);
|
||||
params.count = regs.index_array.count;
|
||||
params.index_buffer_offset = index_buffer_offset;
|
||||
|
@ -278,12 +279,11 @@ void RasterizerOpenGL::SetupShaders(GLenum primitive_mode) {
|
|||
|
||||
GLShader::MaxwellUniformData ubo{};
|
||||
ubo.SetFromRegs(gpu, stage);
|
||||
const GLintptr offset =
|
||||
const auto [buffer, offset] =
|
||||
buffer_cache.UploadHostMemory(&ubo, sizeof(ubo), device.GetUniformBufferAlignment());
|
||||
|
||||
// Bind the emulation info buffer
|
||||
bind_ubo_pushbuffer.Push(buffer_cache.GetHandle(), offset,
|
||||
static_cast<GLsizeiptr>(sizeof(ubo)));
|
||||
bind_ubo_pushbuffer.Push(buffer, offset, static_cast<GLsizeiptr>(sizeof(ubo)));
|
||||
|
||||
Shader shader{shader_cache.GetStageProgram(program)};
|
||||
|
||||
|
@ -651,11 +651,11 @@ void RasterizerOpenGL::DrawArrays() {
|
|||
}
|
||||
|
||||
// Prepare vertex array format.
|
||||
const GLuint vertex_array = SetupVertexFormat();
|
||||
const GLuint vao = SetupVertexFormat();
|
||||
|
||||
// Upload vertex and index data.
|
||||
SetupVertexBuffer(vertex_array);
|
||||
const GLintptr index_buffer_offset = SetupIndexBuffer();
|
||||
SetupVertexBuffer(vao);
|
||||
const GLintptr index_buffer_offset = SetupIndexBuffer(vao);
|
||||
|
||||
// Setup draw parameters. It will automatically choose what glDraw* method to use.
|
||||
const DrawParameters params = SetupDraw(index_buffer_offset);
|
||||
|
@ -791,8 +791,8 @@ void RasterizerOpenGL::SetupConstBuffer(const Tegra::Engines::ConstBufferInfo& b
|
|||
ASSERT_MSG(size <= MaxConstbufferSize, "Constant buffer is too big");
|
||||
|
||||
const std::size_t alignment = device.GetUniformBufferAlignment();
|
||||
const GLintptr offset = buffer_cache.UploadMemory(buffer.address, size, alignment);
|
||||
bind_ubo_pushbuffer.Push(buffer_cache.GetHandle(), offset, size);
|
||||
const auto [cbuf, offset] = buffer_cache.UploadMemory(buffer.address, size, alignment);
|
||||
bind_ubo_pushbuffer.Push(cbuf, offset, size);
|
||||
}
|
||||
|
||||
void RasterizerOpenGL::SetupGlobalRegions(Tegra::Engines::Maxwell3D::Regs::ShaderStage stage,
|
||||
|
|
|
@ -220,7 +220,7 @@ private:
|
|||
|
||||
void SetupVertexBuffer(GLuint vao);
|
||||
|
||||
GLintptr SetupIndexBuffer();
|
||||
GLintptr SetupIndexBuffer(GLuint vao);
|
||||
|
||||
DrawParameters SetupDraw(GLintptr index_buffer_offset);
|
||||
|
||||
|
|
Loading…
Reference in a new issue