forked from ShuriZma/suyu
gpu: Move GPUVAddr definition to common_types.
This commit is contained in:
parent
43b83d6b6a
commit
241563d15c
|
@ -40,10 +40,9 @@ using s64 = std::int64_t; ///< 64-bit signed int
|
||||||
using f32 = float; ///< 32-bit floating point
|
using f32 = float; ///< 32-bit floating point
|
||||||
using f64 = double; ///< 64-bit floating point
|
using f64 = double; ///< 64-bit floating point
|
||||||
|
|
||||||
// TODO: It would be nice to eventually replace these with strong types that prevent accidental
|
|
||||||
// conversion between each other.
|
|
||||||
using VAddr = u64; ///< Represents a pointer in the userspace virtual address space.
|
using VAddr = u64; ///< Represents a pointer in the userspace virtual address space.
|
||||||
using PAddr = u64; ///< Represents a pointer in the ARM11 physical address space.
|
using PAddr = u64; ///< Represents a pointer in the ARM11 physical address space.
|
||||||
|
using GPUVAddr = u64; ///< Represents a pointer in the GPU virtual address space.
|
||||||
|
|
||||||
using u128 = std::array<std::uint64_t, 2>;
|
using u128 = std::array<std::uint64_t, 2>;
|
||||||
static_assert(sizeof(u128) == 16, "u128 must be 128 bits wide");
|
static_assert(sizeof(u128) == 16, "u128 must be 128 bits wide");
|
||||||
|
|
|
@ -89,7 +89,7 @@ u32 nvhost_as_gpu::Remap(const std::vector<u8>& input, std::vector<u8>& output)
|
||||||
for (const auto& entry : entries) {
|
for (const auto& entry : entries) {
|
||||||
LOG_WARNING(Service_NVDRV, "remap entry, offset=0x{:X} handle=0x{:X} pages=0x{:X}",
|
LOG_WARNING(Service_NVDRV, "remap entry, offset=0x{:X} handle=0x{:X} pages=0x{:X}",
|
||||||
entry.offset, entry.nvmap_handle, entry.pages);
|
entry.offset, entry.nvmap_handle, entry.pages);
|
||||||
Tegra::GPUVAddr offset = static_cast<Tegra::GPUVAddr>(entry.offset) << 0x10;
|
GPUVAddr offset = static_cast<GPUVAddr>(entry.offset) << 0x10;
|
||||||
auto object = nvmap_dev->GetObject(entry.nvmap_handle);
|
auto object = nvmap_dev->GetObject(entry.nvmap_handle);
|
||||||
if (!object) {
|
if (!object) {
|
||||||
LOG_CRITICAL(Service_NVDRV, "nvmap {} is an invalid handle!", entry.nvmap_handle);
|
LOG_CRITICAL(Service_NVDRV, "nvmap {} is an invalid handle!", entry.nvmap_handle);
|
||||||
|
@ -102,7 +102,7 @@ u32 nvhost_as_gpu::Remap(const std::vector<u8>& input, std::vector<u8>& output)
|
||||||
u64 size = static_cast<u64>(entry.pages) << 0x10;
|
u64 size = static_cast<u64>(entry.pages) << 0x10;
|
||||||
ASSERT(size <= object->size);
|
ASSERT(size <= object->size);
|
||||||
|
|
||||||
Tegra::GPUVAddr returned = gpu.MemoryManager().MapBufferEx(object->addr, offset, size);
|
GPUVAddr returned = gpu.MemoryManager().MapBufferEx(object->addr, offset, size);
|
||||||
ASSERT(returned == offset);
|
ASSERT(returned == offset);
|
||||||
}
|
}
|
||||||
std::memcpy(output.data(), entries.data(), output.size());
|
std::memcpy(output.data(), entries.data(), output.size());
|
||||||
|
|
|
@ -13,9 +13,6 @@
|
||||||
|
|
||||||
namespace Tegra {
|
namespace Tegra {
|
||||||
|
|
||||||
/// Virtual addresses in the GPU's memory map are 64 bit.
|
|
||||||
using GPUVAddr = u64;
|
|
||||||
|
|
||||||
class MemoryManager final {
|
class MemoryManager final {
|
||||||
public:
|
public:
|
||||||
MemoryManager();
|
MemoryManager();
|
||||||
|
|
|
@ -21,8 +21,8 @@ CachedBufferEntry::CachedBufferEntry(VAddr cpu_addr, std::size_t size, GLintptr
|
||||||
OGLBufferCache::OGLBufferCache(RasterizerOpenGL& rasterizer, std::size_t size)
|
OGLBufferCache::OGLBufferCache(RasterizerOpenGL& rasterizer, std::size_t size)
|
||||||
: RasterizerCache{rasterizer}, stream_buffer(size, true) {}
|
: RasterizerCache{rasterizer}, stream_buffer(size, true) {}
|
||||||
|
|
||||||
GLintptr OGLBufferCache::UploadMemory(Tegra::GPUVAddr gpu_addr, std::size_t size,
|
GLintptr OGLBufferCache::UploadMemory(GPUVAddr gpu_addr, std::size_t size, std::size_t alignment,
|
||||||
std::size_t alignment, bool cache) {
|
bool cache) {
|
||||||
auto& memory_manager = Core::System::GetInstance().GPU().MemoryManager();
|
auto& memory_manager = Core::System::GetInstance().GPU().MemoryManager();
|
||||||
|
|
||||||
// Cache management is a big overhead, so only cache entries with a given size.
|
// Cache management is a big overhead, so only cache entries with a given size.
|
||||||
|
|
|
@ -58,7 +58,7 @@ public:
|
||||||
|
|
||||||
/// Uploads data from a guest GPU address. Returns host's buffer offset where it's been
|
/// Uploads data from a guest GPU address. Returns host's buffer offset where it's been
|
||||||
/// allocated.
|
/// allocated.
|
||||||
GLintptr UploadMemory(Tegra::GPUVAddr gpu_addr, std::size_t size, std::size_t alignment = 4,
|
GLintptr UploadMemory(GPUVAddr gpu_addr, std::size_t size, std::size_t alignment = 4,
|
||||||
bool cache = true);
|
bool cache = true);
|
||||||
|
|
||||||
/// Uploads from a host memory. Returns host's buffer offset where it's been allocated.
|
/// Uploads from a host memory. Returns host's buffer offset where it's been allocated.
|
||||||
|
|
|
@ -46,7 +46,7 @@ GlobalRegion GlobalRegionCacheOpenGL::TryGetReservedGlobalRegion(CacheAddr addr,
|
||||||
return search->second;
|
return search->second;
|
||||||
}
|
}
|
||||||
|
|
||||||
GlobalRegion GlobalRegionCacheOpenGL::GetUncachedGlobalRegion(Tegra::GPUVAddr addr, u32 size,
|
GlobalRegion GlobalRegionCacheOpenGL::GetUncachedGlobalRegion(GPUVAddr addr, u32 size,
|
||||||
u8* host_ptr) {
|
u8* host_ptr) {
|
||||||
GlobalRegion region{TryGetReservedGlobalRegion(ToCacheAddr(host_ptr), size)};
|
GlobalRegion region{TryGetReservedGlobalRegion(ToCacheAddr(host_ptr), size)};
|
||||||
if (!region) {
|
if (!region) {
|
||||||
|
|
|
@ -66,7 +66,7 @@ public:
|
||||||
|
|
||||||
private:
|
private:
|
||||||
GlobalRegion TryGetReservedGlobalRegion(CacheAddr addr, u32 size) const;
|
GlobalRegion TryGetReservedGlobalRegion(CacheAddr addr, u32 size) const;
|
||||||
GlobalRegion GetUncachedGlobalRegion(Tegra::GPUVAddr addr, u32 size, u8* host_ptr);
|
GlobalRegion GetUncachedGlobalRegion(GPUVAddr addr, u32 size, u8* host_ptr);
|
||||||
void ReserveGlobalRegion(GlobalRegion region);
|
void ReserveGlobalRegion(GlobalRegion region);
|
||||||
|
|
||||||
std::unordered_map<CacheAddr, GlobalRegion> reserve;
|
std::unordered_map<CacheAddr, GlobalRegion> reserve;
|
||||||
|
|
|
@ -40,8 +40,7 @@ GLintptr PrimitiveAssembler::MakeQuadArray(u32 first, u32 count) {
|
||||||
return index_offset;
|
return index_offset;
|
||||||
}
|
}
|
||||||
|
|
||||||
GLintptr PrimitiveAssembler::MakeQuadIndexed(Tegra::GPUVAddr gpu_addr, std::size_t index_size,
|
GLintptr PrimitiveAssembler::MakeQuadIndexed(GPUVAddr gpu_addr, std::size_t index_size, u32 count) {
|
||||||
u32 count) {
|
|
||||||
const std::size_t map_size{CalculateQuadSize(count)};
|
const std::size_t map_size{CalculateQuadSize(count)};
|
||||||
auto [dst_pointer, index_offset] = buffer_cache.ReserveMemory(map_size);
|
auto [dst_pointer, index_offset] = buffer_cache.ReserveMemory(map_size);
|
||||||
|
|
||||||
|
|
|
@ -24,7 +24,7 @@ public:
|
||||||
|
|
||||||
GLintptr MakeQuadArray(u32 first, u32 count);
|
GLintptr MakeQuadArray(u32 first, u32 count);
|
||||||
|
|
||||||
GLintptr MakeQuadIndexed(Tegra::GPUVAddr gpu_addr, std::size_t index_size, u32 count);
|
GLintptr MakeQuadIndexed(GPUVAddr gpu_addr, std::size_t index_size, u32 count);
|
||||||
|
|
||||||
private:
|
private:
|
||||||
OGLBufferCache& buffer_cache;
|
OGLBufferCache& buffer_cache;
|
||||||
|
|
|
@ -225,8 +225,8 @@ void RasterizerOpenGL::SetupVertexBuffer(GLuint vao) {
|
||||||
if (!vertex_array.IsEnabled())
|
if (!vertex_array.IsEnabled())
|
||||||
continue;
|
continue;
|
||||||
|
|
||||||
const Tegra::GPUVAddr start = vertex_array.StartAddress();
|
const GPUVAddr start = vertex_array.StartAddress();
|
||||||
const Tegra::GPUVAddr end = regs.vertex_array_limit[index].LimitAddress();
|
const GPUVAddr end = regs.vertex_array_limit[index].LimitAddress();
|
||||||
|
|
||||||
ASSERT(end > start);
|
ASSERT(end > start);
|
||||||
const u64 size = end - start + 1;
|
const u64 size = end - start + 1;
|
||||||
|
@ -421,8 +421,8 @@ std::size_t RasterizerOpenGL::CalculateVertexArraysSize() const {
|
||||||
if (!regs.vertex_array[index].IsEnabled())
|
if (!regs.vertex_array[index].IsEnabled())
|
||||||
continue;
|
continue;
|
||||||
|
|
||||||
const Tegra::GPUVAddr start = regs.vertex_array[index].StartAddress();
|
const GPUVAddr start = regs.vertex_array[index].StartAddress();
|
||||||
const Tegra::GPUVAddr end = regs.vertex_array_limit[index].LimitAddress();
|
const GPUVAddr end = regs.vertex_array_limit[index].LimitAddress();
|
||||||
|
|
||||||
ASSERT(end > start);
|
ASSERT(end > start);
|
||||||
size += end - start + 1;
|
size += end - start + 1;
|
||||||
|
|
|
@ -55,7 +55,7 @@ static void ApplyTextureDefaults(GLuint texture, u32 max_mip_level) {
|
||||||
}
|
}
|
||||||
}
|
}
|
||||||
|
|
||||||
void SurfaceParams::InitCacheParameters(Tegra::GPUVAddr gpu_addr_) {
|
void SurfaceParams::InitCacheParameters(GPUVAddr gpu_addr_) {
|
||||||
auto& memory_manager{Core::System::GetInstance().GPU().MemoryManager()};
|
auto& memory_manager{Core::System::GetInstance().GPU().MemoryManager()};
|
||||||
|
|
||||||
gpu_addr = gpu_addr_;
|
gpu_addr = gpu_addr_;
|
||||||
|
@ -222,7 +222,7 @@ std::size_t SurfaceParams::InnerMemorySize(bool force_gl, bool layer_only,
|
||||||
}
|
}
|
||||||
|
|
||||||
/*static*/ SurfaceParams SurfaceParams::CreateForDepthBuffer(
|
/*static*/ SurfaceParams SurfaceParams::CreateForDepthBuffer(
|
||||||
u32 zeta_width, u32 zeta_height, Tegra::GPUVAddr zeta_address, Tegra::DepthFormat format,
|
u32 zeta_width, u32 zeta_height, GPUVAddr zeta_address, Tegra::DepthFormat format,
|
||||||
u32 block_width, u32 block_height, u32 block_depth,
|
u32 block_width, u32 block_height, u32 block_depth,
|
||||||
Tegra::Engines::Maxwell3D::Regs::InvMemoryLayout type) {
|
Tegra::Engines::Maxwell3D::Regs::InvMemoryLayout type) {
|
||||||
SurfaceParams params{};
|
SurfaceParams params{};
|
||||||
|
@ -980,11 +980,11 @@ void RasterizerCacheOpenGL::FastLayeredCopySurface(const Surface& src_surface,
|
||||||
const auto& init_params{src_surface->GetSurfaceParams()};
|
const auto& init_params{src_surface->GetSurfaceParams()};
|
||||||
const auto& dst_params{dst_surface->GetSurfaceParams()};
|
const auto& dst_params{dst_surface->GetSurfaceParams()};
|
||||||
auto& memory_manager{Core::System::GetInstance().GPU().MemoryManager()};
|
auto& memory_manager{Core::System::GetInstance().GPU().MemoryManager()};
|
||||||
Tegra::GPUVAddr address{init_params.gpu_addr};
|
GPUVAddr address{init_params.gpu_addr};
|
||||||
const std::size_t layer_size{dst_params.LayerMemorySize()};
|
const std::size_t layer_size{dst_params.LayerMemorySize()};
|
||||||
for (u32 layer = 0; layer < dst_params.depth; layer++) {
|
for (u32 layer = 0; layer < dst_params.depth; layer++) {
|
||||||
for (u32 mipmap = 0; mipmap < dst_params.max_mip_level; mipmap++) {
|
for (u32 mipmap = 0; mipmap < dst_params.max_mip_level; mipmap++) {
|
||||||
const Tegra::GPUVAddr sub_address{address + dst_params.GetMipmapLevelOffset(mipmap)};
|
const GPUVAddr sub_address{address + dst_params.GetMipmapLevelOffset(mipmap)};
|
||||||
const Surface& copy{TryGet(memory_manager.GetPointer(sub_address))};
|
const Surface& copy{TryGet(memory_manager.GetPointer(sub_address))};
|
||||||
if (!copy) {
|
if (!copy) {
|
||||||
continue;
|
continue;
|
||||||
|
@ -1244,10 +1244,9 @@ static std::optional<u32> TryFindBestMipMap(std::size_t memory, const SurfacePar
|
||||||
return {};
|
return {};
|
||||||
}
|
}
|
||||||
|
|
||||||
static std::optional<u32> TryFindBestLayer(Tegra::GPUVAddr addr, const SurfaceParams params,
|
static std::optional<u32> TryFindBestLayer(GPUVAddr addr, const SurfaceParams params, u32 mipmap) {
|
||||||
u32 mipmap) {
|
|
||||||
const std::size_t size{params.LayerMemorySize()};
|
const std::size_t size{params.LayerMemorySize()};
|
||||||
Tegra::GPUVAddr start{params.gpu_addr + params.GetMipmapLevelOffset(mipmap)};
|
GPUVAddr start{params.gpu_addr + params.GetMipmapLevelOffset(mipmap)};
|
||||||
for (u32 i = 0; i < params.depth; i++) {
|
for (u32 i = 0; i < params.depth; i++) {
|
||||||
if (start == addr) {
|
if (start == addr) {
|
||||||
return {i};
|
return {i};
|
||||||
|
|
|
@ -210,7 +210,7 @@ struct SurfaceParams {
|
||||||
|
|
||||||
/// Creates SurfaceParams for a depth buffer configuration
|
/// Creates SurfaceParams for a depth buffer configuration
|
||||||
static SurfaceParams CreateForDepthBuffer(
|
static SurfaceParams CreateForDepthBuffer(
|
||||||
u32 zeta_width, u32 zeta_height, Tegra::GPUVAddr zeta_address, Tegra::DepthFormat format,
|
u32 zeta_width, u32 zeta_height, GPUVAddr zeta_address, Tegra::DepthFormat format,
|
||||||
u32 block_width, u32 block_height, u32 block_depth,
|
u32 block_width, u32 block_height, u32 block_depth,
|
||||||
Tegra::Engines::Maxwell3D::Regs::InvMemoryLayout type);
|
Tegra::Engines::Maxwell3D::Regs::InvMemoryLayout type);
|
||||||
|
|
||||||
|
@ -232,7 +232,7 @@ struct SurfaceParams {
|
||||||
}
|
}
|
||||||
|
|
||||||
/// Initializes parameters for caching, should be called after everything has been initialized
|
/// Initializes parameters for caching, should be called after everything has been initialized
|
||||||
void InitCacheParameters(Tegra::GPUVAddr gpu_addr);
|
void InitCacheParameters(GPUVAddr gpu_addr);
|
||||||
|
|
||||||
std::string TargetName() const {
|
std::string TargetName() const {
|
||||||
switch (target) {
|
switch (target) {
|
||||||
|
@ -297,7 +297,7 @@ struct SurfaceParams {
|
||||||
bool srgb_conversion;
|
bool srgb_conversion;
|
||||||
// Parameters used for caching
|
// Parameters used for caching
|
||||||
u8* host_ptr;
|
u8* host_ptr;
|
||||||
Tegra::GPUVAddr gpu_addr;
|
GPUVAddr gpu_addr;
|
||||||
std::size_t size_in_bytes;
|
std::size_t size_in_bytes;
|
||||||
std::size_t size_in_bytes_gl;
|
std::size_t size_in_bytes_gl;
|
||||||
|
|
||||||
|
|
|
@ -32,7 +32,7 @@ struct UnspecializedShader {
|
||||||
namespace {
|
namespace {
|
||||||
|
|
||||||
/// Gets the address for the specified shader stage program
|
/// Gets the address for the specified shader stage program
|
||||||
Tegra::GPUVAddr GetShaderAddress(Maxwell::ShaderProgram program) {
|
GPUVAddr GetShaderAddress(Maxwell::ShaderProgram program) {
|
||||||
const auto& gpu{Core::System::GetInstance().GPU().Maxwell3D()};
|
const auto& gpu{Core::System::GetInstance().GPU().Maxwell3D()};
|
||||||
const auto& shader_config{gpu.regs.shader_config[static_cast<std::size_t>(program)]};
|
const auto& shader_config{gpu.regs.shader_config[static_cast<std::size_t>(program)]};
|
||||||
return gpu.regs.code_address.CodeAddress() + shader_config.offset;
|
return gpu.regs.code_address.CodeAddress() + shader_config.offset;
|
||||||
|
@ -486,7 +486,7 @@ Shader ShaderCacheOpenGL::GetStageProgram(Maxwell::ShaderProgram program) {
|
||||||
}
|
}
|
||||||
|
|
||||||
auto& memory_manager{Core::System::GetInstance().GPU().MemoryManager()};
|
auto& memory_manager{Core::System::GetInstance().GPU().MemoryManager()};
|
||||||
const Tegra::GPUVAddr program_addr{GetShaderAddress(program)};
|
const GPUVAddr program_addr{GetShaderAddress(program)};
|
||||||
|
|
||||||
// Look up shader in the cache based on address
|
// Look up shader in the cache based on address
|
||||||
const auto& host_ptr{memory_manager.GetPointer(program_addr)};
|
const auto& host_ptr{memory_manager.GetPointer(program_addr)};
|
||||||
|
|
|
@ -39,8 +39,7 @@ VKBufferCache::VKBufferCache(Tegra::MemoryManager& tegra_memory_manager,
|
||||||
|
|
||||||
VKBufferCache::~VKBufferCache() = default;
|
VKBufferCache::~VKBufferCache() = default;
|
||||||
|
|
||||||
u64 VKBufferCache::UploadMemory(Tegra::GPUVAddr gpu_addr, std::size_t size, u64 alignment,
|
u64 VKBufferCache::UploadMemory(GPUVAddr gpu_addr, std::size_t size, u64 alignment, bool cache) {
|
||||||
bool cache) {
|
|
||||||
const auto cpu_addr{tegra_memory_manager.GpuToCpuAddress(gpu_addr)};
|
const auto cpu_addr{tegra_memory_manager.GpuToCpuAddress(gpu_addr)};
|
||||||
ASSERT_MSG(cpu_addr, "Invalid GPU address");
|
ASSERT_MSG(cpu_addr, "Invalid GPU address");
|
||||||
|
|
||||||
|
|
|
@ -68,8 +68,7 @@ public:
|
||||||
|
|
||||||
/// Uploads data from a guest GPU address. Returns host's buffer offset where it's been
|
/// Uploads data from a guest GPU address. Returns host's buffer offset where it's been
|
||||||
/// allocated.
|
/// allocated.
|
||||||
u64 UploadMemory(Tegra::GPUVAddr gpu_addr, std::size_t size, u64 alignment = 4,
|
u64 UploadMemory(GPUVAddr gpu_addr, std::size_t size, u64 alignment = 4, bool cache = true);
|
||||||
bool cache = true);
|
|
||||||
|
|
||||||
/// Uploads from a host memory. Returns host's buffer offset where it's been allocated.
|
/// Uploads from a host memory. Returns host's buffer offset where it's been allocated.
|
||||||
u64 UploadHostMemory(const u8* raw_pointer, std::size_t size, u64 alignment = 4);
|
u64 UploadHostMemory(const u8* raw_pointer, std::size_t size, u64 alignment = 4);
|
||||||
|
|
|
@ -261,7 +261,7 @@ void GraphicsSurfaceWidget::OnSurfaceSourceChanged(int new_value) {
|
||||||
|
|
||||||
void GraphicsSurfaceWidget::OnSurfaceAddressChanged(qint64 new_value) {
|
void GraphicsSurfaceWidget::OnSurfaceAddressChanged(qint64 new_value) {
|
||||||
if (surface_address != new_value) {
|
if (surface_address != new_value) {
|
||||||
surface_address = static_cast<Tegra::GPUVAddr>(new_value);
|
surface_address = static_cast<GPUVAddr>(new_value);
|
||||||
|
|
||||||
surface_source_list->setCurrentIndex(static_cast<int>(Source::Custom));
|
surface_source_list->setCurrentIndex(static_cast<int>(Source::Custom));
|
||||||
emit Update();
|
emit Update();
|
||||||
|
|
|
@ -87,7 +87,7 @@ private:
|
||||||
QPushButton* save_surface;
|
QPushButton* save_surface;
|
||||||
|
|
||||||
Source surface_source;
|
Source surface_source;
|
||||||
Tegra::GPUVAddr surface_address;
|
GPUVAddr surface_address;
|
||||||
unsigned surface_width;
|
unsigned surface_width;
|
||||||
unsigned surface_height;
|
unsigned surface_height;
|
||||||
Tegra::Texture::TextureFormat surface_format;
|
Tegra::Texture::TextureFormat surface_format;
|
||||||
|
|
Loading…
Reference in New Issue