summaryrefslogtreecommitdiffstats
path: root/src/video_core
diff options
context:
space:
mode:
Diffstat (limited to 'src/video_core')
-rw-r--r--src/video_core/CMakeLists.txt3
-rw-r--r--src/video_core/buffer_cache/buffer_block.h42
-rw-r--r--src/video_core/buffer_cache/buffer_cache.h145
-rw-r--r--src/video_core/buffer_cache/map_interval.h12
-rw-r--r--src/video_core/engines/maxwell_3d.h67
-rw-r--r--src/video_core/engines/shader_bytecode.h22
-rw-r--r--src/video_core/engines/shader_header.h55
-rw-r--r--src/video_core/gpu.cpp10
-rw-r--r--src/video_core/gpu.h24
-rw-r--r--src/video_core/gpu_asynch.cpp15
-rw-r--r--src/video_core/gpu_asynch.h15
-rw-r--r--src/video_core/gpu_synch.cpp23
-rw-r--r--src/video_core/gpu_synch.h16
-rw-r--r--src/video_core/gpu_thread.cpp21
-rw-r--r--src/video_core/gpu_thread.h25
-rw-r--r--src/video_core/memory_manager.cpp93
-rw-r--r--src/video_core/memory_manager.h5
-rw-r--r--src/video_core/query_cache.h37
-rw-r--r--src/video_core/rasterizer_cache.h44
-rw-r--r--src/video_core/rasterizer_interface.h6
-rw-r--r--src/video_core/renderer_base.h3
-rw-r--r--src/video_core/renderer_opengl/gl_buffer_cache.cpp8
-rw-r--r--src/video_core/renderer_opengl/gl_buffer_cache.h4
-rw-r--r--src/video_core/renderer_opengl/gl_device.cpp26
-rw-r--r--src/video_core/renderer_opengl/gl_device.h5
-rw-r--r--src/video_core/renderer_opengl/gl_rasterizer.cpp27
-rw-r--r--src/video_core/renderer_opengl/gl_rasterizer.h6
-rw-r--r--src/video_core/renderer_opengl/gl_shader_cache.cpp49
-rw-r--r--src/video_core/renderer_opengl/gl_shader_cache.h8
-rw-r--r--src/video_core/renderer_opengl/gl_shader_decompiler.cpp54
-rw-r--r--src/video_core/renderer_opengl/gl_texture_cache.cpp211
-rw-r--r--src/video_core/renderer_opengl/gl_texture_cache.h6
-rw-r--r--src/video_core/renderer_opengl/renderer_opengl.cpp65
-rw-r--r--src/video_core/renderer_opengl/renderer_opengl.h10
-rw-r--r--src/video_core/renderer_vulkan/declarations.h2
-rw-r--r--src/video_core/renderer_vulkan/renderer_vulkan.cpp267
-rw-r--r--src/video_core/renderer_vulkan/renderer_vulkan.h21
-rw-r--r--src/video_core/renderer_vulkan/vk_buffer_cache.cpp8
-rw-r--r--src/video_core/renderer_vulkan/vk_buffer_cache.h4
-rw-r--r--src/video_core/renderer_vulkan/vk_device.cpp135
-rw-r--r--src/video_core/renderer_vulkan/vk_device.h21
-rw-r--r--src/video_core/renderer_vulkan/vk_pipeline_cache.cpp38
-rw-r--r--src/video_core/renderer_vulkan/vk_pipeline_cache.h7
-rw-r--r--src/video_core/renderer_vulkan/vk_rasterizer.cpp15
-rw-r--r--src/video_core/renderer_vulkan/vk_rasterizer.h6
-rw-r--r--src/video_core/renderer_vulkan/vk_shader_decompiler.cpp48
-rw-r--r--src/video_core/renderer_vulkan/vk_texture_cache.cpp34
-rw-r--r--src/video_core/renderer_vulkan/wrapper.cpp750
-rw-r--r--src/video_core/renderer_vulkan/wrapper.h987
-rw-r--r--src/video_core/shader/decode/arithmetic_integer.cpp15
-rw-r--r--src/video_core/shader/decode/conversion.cpp15
-rw-r--r--src/video_core/shader/decode/image.cpp360
-rw-r--r--src/video_core/shader/decode/memory.cpp63
-rw-r--r--src/video_core/shader/decode/other.cpp48
-rw-r--r--src/video_core/shader/node.h16
-rw-r--r--src/video_core/shader/node_helper.cpp14
-rw-r--r--src/video_core/shader/shader_ir.cpp3
-rw-r--r--src/video_core/shader/shader_ir.h4
-rw-r--r--src/video_core/surface.h97
-rw-r--r--src/video_core/texture_cache/surface_base.cpp81
-rw-r--r--src/video_core/texture_cache/surface_base.h49
-rw-r--r--src/video_core/texture_cache/surface_params.cpp34
-rw-r--r--src/video_core/texture_cache/surface_params.h36
-rw-r--r--src/video_core/texture_cache/texture_cache.h127
-rw-r--r--src/video_core/textures/texture.cpp80
-rw-r--r--src/video_core/textures/texture.h46
-rw-r--r--src/video_core/video_core.cpp28
-rw-r--r--src/video_core/video_core.h11
68 files changed, 3538 insertions, 1094 deletions
diff --git a/src/video_core/CMakeLists.txt b/src/video_core/CMakeLists.txt
index 91df062d7..f7febd6a2 100644
--- a/src/video_core/CMakeLists.txt
+++ b/src/video_core/CMakeLists.txt
@@ -148,6 +148,7 @@ add_library(video_core STATIC
textures/convert.h
textures/decoders.cpp
textures/decoders.h
+ textures/texture.cpp
textures/texture.h
video_core.cpp
video_core.h
@@ -210,6 +211,8 @@ if (ENABLE_VULKAN)
renderer_vulkan/vk_texture_cache.h
renderer_vulkan/vk_update_descriptor.cpp
renderer_vulkan/vk_update_descriptor.h
+ renderer_vulkan/wrapper.cpp
+ renderer_vulkan/wrapper.h
)
target_include_directories(video_core PRIVATE sirit ../../externals/Vulkan-Headers/include)
diff --git a/src/video_core/buffer_cache/buffer_block.h b/src/video_core/buffer_cache/buffer_block.h
index 4b9193182..e35ee0b67 100644
--- a/src/video_core/buffer_cache/buffer_block.h
+++ b/src/video_core/buffer_cache/buffer_block.h
@@ -15,37 +15,29 @@ namespace VideoCommon {
class BufferBlock {
public:
- bool Overlaps(const CacheAddr start, const CacheAddr end) const {
- return (cache_addr < end) && (cache_addr_end > start);
+ bool Overlaps(const VAddr start, const VAddr end) const {
+ return (cpu_addr < end) && (cpu_addr_end > start);
}
- bool IsInside(const CacheAddr other_start, const CacheAddr other_end) const {
- return cache_addr <= other_start && other_end <= cache_addr_end;
+ bool IsInside(const VAddr other_start, const VAddr other_end) const {
+ return cpu_addr <= other_start && other_end <= cpu_addr_end;
}
- u8* GetWritableHostPtr() const {
- return FromCacheAddr(cache_addr);
+ std::size_t GetOffset(const VAddr in_addr) {
+ return static_cast<std::size_t>(in_addr - cpu_addr);
}
- u8* GetWritableHostPtr(std::size_t offset) const {
- return FromCacheAddr(cache_addr + offset);
+ VAddr GetCpuAddr() const {
+ return cpu_addr;
}
- std::size_t GetOffset(const CacheAddr in_addr) {
- return static_cast<std::size_t>(in_addr - cache_addr);
+ VAddr GetCpuAddrEnd() const {
+ return cpu_addr_end;
}
- CacheAddr GetCacheAddr() const {
- return cache_addr;
- }
-
- CacheAddr GetCacheAddrEnd() const {
- return cache_addr_end;
- }
-
- void SetCacheAddr(const CacheAddr new_addr) {
- cache_addr = new_addr;
- cache_addr_end = new_addr + size;
+ void SetCpuAddr(const VAddr new_addr) {
+ cpu_addr = new_addr;
+ cpu_addr_end = new_addr + size;
}
std::size_t GetSize() const {
@@ -61,14 +53,14 @@ public:
}
protected:
- explicit BufferBlock(CacheAddr cache_addr, const std::size_t size) : size{size} {
- SetCacheAddr(cache_addr);
+ explicit BufferBlock(VAddr cpu_addr, const std::size_t size) : size{size} {
+ SetCpuAddr(cpu_addr);
}
~BufferBlock() = default;
private:
- CacheAddr cache_addr{};
- CacheAddr cache_addr_end{};
+ VAddr cpu_addr{};
+ VAddr cpu_addr_end{};
std::size_t size{};
u64 epoch{};
};
diff --git a/src/video_core/buffer_cache/buffer_cache.h b/src/video_core/buffer_cache/buffer_cache.h
index 186aca61d..b57c0d4d4 100644
--- a/src/video_core/buffer_cache/buffer_cache.h
+++ b/src/video_core/buffer_cache/buffer_cache.h
@@ -19,6 +19,7 @@
#include "common/alignment.h"
#include "common/common_types.h"
#include "core/core.h"
+#include "core/memory.h"
#include "video_core/buffer_cache/buffer_block.h"
#include "video_core/buffer_cache/map_interval.h"
#include "video_core/memory_manager.h"
@@ -37,28 +38,45 @@ public:
bool is_written = false, bool use_fast_cbuf = false) {
std::lock_guard lock{mutex};
- auto& memory_manager = system.GPU().MemoryManager();
- const auto host_ptr = memory_manager.GetPointer(gpu_addr);
- if (!host_ptr) {
+ const std::optional<VAddr> cpu_addr_opt =
+ system.GPU().MemoryManager().GpuToCpuAddress(gpu_addr);
+
+ if (!cpu_addr_opt) {
return {GetEmptyBuffer(size), 0};
}
- const auto cache_addr = ToCacheAddr(host_ptr);
+
+ VAddr cpu_addr = *cpu_addr_opt;
// Cache management is a big overhead, so only cache entries with a given size.
// TODO: Figure out which size is the best for given games.
constexpr std::size_t max_stream_size = 0x800;
if (use_fast_cbuf || size < max_stream_size) {
- if (!is_written && !IsRegionWritten(cache_addr, cache_addr + size - 1)) {
+ if (!is_written && !IsRegionWritten(cpu_addr, cpu_addr + size - 1)) {
+ auto& memory_manager = system.GPU().MemoryManager();
if (use_fast_cbuf) {
- return ConstBufferUpload(host_ptr, size);
+ if (memory_manager.IsGranularRange(gpu_addr, size)) {
+ const auto host_ptr = memory_manager.GetPointer(gpu_addr);
+ return ConstBufferUpload(host_ptr, size);
+ } else {
+ staging_buffer.resize(size);
+ memory_manager.ReadBlockUnsafe(gpu_addr, staging_buffer.data(), size);
+ return ConstBufferUpload(staging_buffer.data(), size);
+ }
} else {
- return StreamBufferUpload(host_ptr, size, alignment);
+ if (memory_manager.IsGranularRange(gpu_addr, size)) {
+ const auto host_ptr = memory_manager.GetPointer(gpu_addr);
+ return StreamBufferUpload(host_ptr, size, alignment);
+ } else {
+ staging_buffer.resize(size);
+ memory_manager.ReadBlockUnsafe(gpu_addr, staging_buffer.data(), size);
+ return StreamBufferUpload(staging_buffer.data(), size, alignment);
+ }
}
}
}
- auto block = GetBlock(cache_addr, size);
- auto map = MapAddress(block, gpu_addr, cache_addr, size);
+ auto block = GetBlock(cpu_addr, size);
+ auto map = MapAddress(block, gpu_addr, cpu_addr, size);
if (is_written) {
map->MarkAsModified(true, GetModifiedTicks());
if (!map->IsWritten()) {
@@ -71,7 +89,7 @@ public:
}
}
- const u64 offset = static_cast<u64>(block->GetOffset(cache_addr));
+ const u64 offset = static_cast<u64>(block->GetOffset(cpu_addr));
return {ToHandle(block), offset};
}
@@ -112,7 +130,7 @@ public:
}
/// Write any cached resources overlapping the specified region back to memory
- void FlushRegion(CacheAddr addr, std::size_t size) {
+ void FlushRegion(VAddr addr, std::size_t size) {
std::lock_guard lock{mutex};
std::vector<MapInterval> objects = GetMapsInRange(addr, size);
@@ -127,7 +145,7 @@ public:
}
/// Mark the specified region as being invalidated
- void InvalidateRegion(CacheAddr addr, u64 size) {
+ void InvalidateRegion(VAddr addr, u64 size) {
std::lock_guard lock{mutex};
std::vector<MapInterval> objects = GetMapsInRange(addr, size);
@@ -152,7 +170,7 @@ protected:
virtual void WriteBarrier() = 0;
- virtual TBuffer CreateBlock(CacheAddr cache_addr, std::size_t size) = 0;
+ virtual TBuffer CreateBlock(VAddr cpu_addr, std::size_t size) = 0;
virtual void UploadBlockData(const TBuffer& buffer, std::size_t offset, std::size_t size,
const u8* data) = 0;
@@ -169,20 +187,17 @@ protected:
/// Register an object into the cache
void Register(const MapInterval& new_map, bool inherit_written = false) {
- const CacheAddr cache_ptr = new_map->GetStart();
- const std::optional<VAddr> cpu_addr =
- system.GPU().MemoryManager().GpuToCpuAddress(new_map->GetGpuAddress());
- if (!cache_ptr || !cpu_addr) {
+ const VAddr cpu_addr = new_map->GetStart();
+ if (!cpu_addr) {
LOG_CRITICAL(HW_GPU, "Failed to register buffer with unmapped gpu_address 0x{:016x}",
new_map->GetGpuAddress());
return;
}
const std::size_t size = new_map->GetEnd() - new_map->GetStart();
- new_map->SetCpuAddress(*cpu_addr);
new_map->MarkAsRegistered(true);
const IntervalType interval{new_map->GetStart(), new_map->GetEnd()};
mapped_addresses.insert({interval, new_map});
- rasterizer.UpdatePagesCachedCount(*cpu_addr, size, 1);
+ rasterizer.UpdatePagesCachedCount(cpu_addr, size, 1);
if (inherit_written) {
MarkRegionAsWritten(new_map->GetStart(), new_map->GetEnd() - 1);
new_map->MarkAsWritten(true);
@@ -192,7 +207,7 @@ protected:
/// Unregisters an object from the cache
void Unregister(MapInterval& map) {
const std::size_t size = map->GetEnd() - map->GetStart();
- rasterizer.UpdatePagesCachedCount(map->GetCpuAddress(), size, -1);
+ rasterizer.UpdatePagesCachedCount(map->GetStart(), size, -1);
map->MarkAsRegistered(false);
if (map->IsWritten()) {
UnmarkRegionAsWritten(map->GetStart(), map->GetEnd() - 1);
@@ -202,32 +217,39 @@ protected:
}
private:
- MapInterval CreateMap(const CacheAddr start, const CacheAddr end, const GPUVAddr gpu_addr) {
+ MapInterval CreateMap(const VAddr start, const VAddr end, const GPUVAddr gpu_addr) {
return std::make_shared<MapIntervalBase>(start, end, gpu_addr);
}
- MapInterval MapAddress(const TBuffer& block, const GPUVAddr gpu_addr,
- const CacheAddr cache_addr, const std::size_t size) {
+ MapInterval MapAddress(const TBuffer& block, const GPUVAddr gpu_addr, const VAddr cpu_addr,
+ const std::size_t size) {
- std::vector<MapInterval> overlaps = GetMapsInRange(cache_addr, size);
+ std::vector<MapInterval> overlaps = GetMapsInRange(cpu_addr, size);
if (overlaps.empty()) {
- const CacheAddr cache_addr_end = cache_addr + size;
- MapInterval new_map = CreateMap(cache_addr, cache_addr_end, gpu_addr);
- u8* host_ptr = FromCacheAddr(cache_addr);
- UploadBlockData(block, block->GetOffset(cache_addr), size, host_ptr);
+ auto& memory_manager = system.GPU().MemoryManager();
+ const VAddr cpu_addr_end = cpu_addr + size;
+ MapInterval new_map = CreateMap(cpu_addr, cpu_addr_end, gpu_addr);
+ if (memory_manager.IsGranularRange(gpu_addr, size)) {
+ u8* host_ptr = memory_manager.GetPointer(gpu_addr);
+ UploadBlockData(block, block->GetOffset(cpu_addr), size, host_ptr);
+ } else {
+ staging_buffer.resize(size);
+ memory_manager.ReadBlockUnsafe(gpu_addr, staging_buffer.data(), size);
+ UploadBlockData(block, block->GetOffset(cpu_addr), size, staging_buffer.data());
+ }
Register(new_map);
return new_map;
}
- const CacheAddr cache_addr_end = cache_addr + size;
+ const VAddr cpu_addr_end = cpu_addr + size;
if (overlaps.size() == 1) {
MapInterval& current_map = overlaps[0];
- if (current_map->IsInside(cache_addr, cache_addr_end)) {
+ if (current_map->IsInside(cpu_addr, cpu_addr_end)) {
return current_map;
}
}
- CacheAddr new_start = cache_addr;
- CacheAddr new_end = cache_addr_end;
+ VAddr new_start = cpu_addr;
+ VAddr new_end = cpu_addr_end;
bool write_inheritance = false;
bool modified_inheritance = false;
// Calculate new buffer parameters
@@ -237,7 +259,7 @@ private:
write_inheritance |= overlap->IsWritten();
modified_inheritance |= overlap->IsModified();
}
- GPUVAddr new_gpu_addr = gpu_addr + new_start - cache_addr;
+ GPUVAddr new_gpu_addr = gpu_addr + new_start - cpu_addr;
for (auto& overlap : overlaps) {
Unregister(overlap);
}
@@ -250,7 +272,7 @@ private:
return new_map;
}
- void UpdateBlock(const TBuffer& block, CacheAddr start, CacheAddr end,
+ void UpdateBlock(const TBuffer& block, VAddr start, VAddr end,
std::vector<MapInterval>& overlaps) {
const IntervalType base_interval{start, end};
IntervalSet interval_set{};
@@ -262,13 +284,15 @@ private:
for (auto& interval : interval_set) {
std::size_t size = interval.upper() - interval.lower();
if (size > 0) {
- u8* host_ptr = FromCacheAddr(interval.lower());
- UploadBlockData(block, block->GetOffset(interval.lower()), size, host_ptr);
+ staging_buffer.resize(size);
+ system.Memory().ReadBlockUnsafe(interval.lower(), staging_buffer.data(), size);
+ UploadBlockData(block, block->GetOffset(interval.lower()), size,
+ staging_buffer.data());
}
}
}
- std::vector<MapInterval> GetMapsInRange(CacheAddr addr, std::size_t size) {
+ std::vector<MapInterval> GetMapsInRange(VAddr addr, std::size_t size) {
if (size == 0) {
return {};
}
@@ -290,8 +314,9 @@ private:
void FlushMap(MapInterval map) {
std::size_t size = map->GetEnd() - map->GetStart();
TBuffer block = blocks[map->GetStart() >> block_page_bits];
- u8* host_ptr = FromCacheAddr(map->GetStart());
- DownloadBlockData(block, block->GetOffset(map->GetStart()), size, host_ptr);
+ staging_buffer.resize(size);
+ DownloadBlockData(block, block->GetOffset(map->GetStart()), size, staging_buffer.data());
+ system.Memory().WriteBlockUnsafe(map->GetStart(), staging_buffer.data(), size);
map->MarkAsModified(false, 0);
}
@@ -316,14 +341,14 @@ private:
TBuffer EnlargeBlock(TBuffer buffer) {
const std::size_t old_size = buffer->GetSize();
const std::size_t new_size = old_size + block_page_size;
- const CacheAddr cache_addr = buffer->GetCacheAddr();
- TBuffer new_buffer = CreateBlock(cache_addr, new_size);
+ const VAddr cpu_addr = buffer->GetCpuAddr();
+ TBuffer new_buffer = CreateBlock(cpu_addr, new_size);
CopyBlock(buffer, new_buffer, 0, 0, old_size);
buffer->SetEpoch(epoch);
pending_destruction.push_back(buffer);
- const CacheAddr cache_addr_end = cache_addr + new_size - 1;
- u64 page_start = cache_addr >> block_page_bits;
- const u64 page_end = cache_addr_end >> block_page_bits;
+ const VAddr cpu_addr_end = cpu_addr + new_size - 1;
+ u64 page_start = cpu_addr >> block_page_bits;
+ const u64 page_end = cpu_addr_end >> block_page_bits;
while (page_start <= page_end) {
blocks[page_start] = new_buffer;
++page_start;
@@ -334,9 +359,9 @@ private:
TBuffer MergeBlocks(TBuffer first, TBuffer second) {
const std::size_t size_1 = first->GetSize();
const std::size_t size_2 = second->GetSize();
- const CacheAddr first_addr = first->GetCacheAddr();
- const CacheAddr second_addr = second->GetCacheAddr();
- const CacheAddr new_addr = std::min(first_addr, second_addr);
+ const VAddr first_addr = first->GetCpuAddr();
+ const VAddr second_addr = second->GetCpuAddr();
+ const VAddr new_addr = std::min(first_addr, second_addr);
const std::size_t new_size = size_1 + size_2;
TBuffer new_buffer = CreateBlock(new_addr, new_size);
CopyBlock(first, new_buffer, 0, new_buffer->GetOffset(first_addr), size_1);
@@ -345,9 +370,9 @@ private:
second->SetEpoch(epoch);
pending_destruction.push_back(first);
pending_destruction.push_back(second);
- const CacheAddr cache_addr_end = new_addr + new_size - 1;
+ const VAddr cpu_addr_end = new_addr + new_size - 1;
u64 page_start = new_addr >> block_page_bits;
- const u64 page_end = cache_addr_end >> block_page_bits;
+ const u64 page_end = cpu_addr_end >> block_page_bits;
while (page_start <= page_end) {
blocks[page_start] = new_buffer;
++page_start;
@@ -355,18 +380,18 @@ private:
return new_buffer;
}
- TBuffer GetBlock(const CacheAddr cache_addr, const std::size_t size) {
+ TBuffer GetBlock(const VAddr cpu_addr, const std::size_t size) {
TBuffer found{};
- const CacheAddr cache_addr_end = cache_addr + size - 1;
- u64 page_start = cache_addr >> block_page_bits;
- const u64 page_end = cache_addr_end >> block_page_bits;
+ const VAddr cpu_addr_end = cpu_addr + size - 1;
+ u64 page_start = cpu_addr >> block_page_bits;
+ const u64 page_end = cpu_addr_end >> block_page_bits;
while (page_start <= page_end) {
auto it = blocks.find(page_start);
if (it == blocks.end()) {
if (found) {
found = EnlargeBlock(found);
} else {
- const CacheAddr start_addr = (page_start << block_page_bits);
+ const VAddr start_addr = (page_start << block_page_bits);
found = CreateBlock(start_addr, block_page_size);
blocks[page_start] = found;
}
@@ -386,7 +411,7 @@ private:
return found;
}
- void MarkRegionAsWritten(const CacheAddr start, const CacheAddr end) {
+ void MarkRegionAsWritten(const VAddr start, const VAddr end) {
u64 page_start = start >> write_page_bit;
const u64 page_end = end >> write_page_bit;
while (page_start <= page_end) {
@@ -400,7 +425,7 @@ private:
}
}
- void UnmarkRegionAsWritten(const CacheAddr start, const CacheAddr end) {
+ void UnmarkRegionAsWritten(const VAddr start, const VAddr end) {
u64 page_start = start >> write_page_bit;
const u64 page_end = end >> write_page_bit;
while (page_start <= page_end) {
@@ -416,7 +441,7 @@ private:
}
}
- bool IsRegionWritten(const CacheAddr start, const CacheAddr end) const {
+ bool IsRegionWritten(const VAddr start, const VAddr end) const {
u64 page_start = start >> write_page_bit;
const u64 page_end = end >> write_page_bit;
while (page_start <= page_end) {
@@ -440,8 +465,8 @@ private:
u64 buffer_offset = 0;
u64 buffer_offset_base = 0;
- using IntervalSet = boost::icl::interval_set<CacheAddr>;
- using IntervalCache = boost::icl::interval_map<CacheAddr, MapInterval>;
+ using IntervalSet = boost::icl::interval_set<VAddr>;
+ using IntervalCache = boost::icl::interval_map<VAddr, MapInterval>;
using IntervalType = typename IntervalCache::interval_type;
IntervalCache mapped_addresses;
@@ -456,6 +481,8 @@ private:
u64 epoch = 0;
u64 modified_ticks = 0;
+ std::vector<u8> staging_buffer;
+
std::recursive_mutex mutex;
};
diff --git a/src/video_core/buffer_cache/map_interval.h b/src/video_core/buffer_cache/map_interval.h
index 3a104d5cd..b0956029d 100644
--- a/src/video_core/buffer_cache/map_interval.h
+++ b/src/video_core/buffer_cache/map_interval.h
@@ -11,7 +11,7 @@ namespace VideoCommon {
class MapIntervalBase {
public:
- MapIntervalBase(const CacheAddr start, const CacheAddr end, const GPUVAddr gpu_addr)
+ MapIntervalBase(const VAddr start, const VAddr end, const GPUVAddr gpu_addr)
: start{start}, end{end}, gpu_addr{gpu_addr} {}
void SetCpuAddress(VAddr new_cpu_addr) {
@@ -26,7 +26,7 @@ public:
return gpu_addr;
}
- bool IsInside(const CacheAddr other_start, const CacheAddr other_end) const {
+ bool IsInside(const VAddr other_start, const VAddr other_end) const {
return (start <= other_start && other_end <= end);
}
@@ -46,11 +46,11 @@ public:
return is_registered;
}
- CacheAddr GetStart() const {
+ VAddr GetStart() const {
return start;
}
- CacheAddr GetEnd() const {
+ VAddr GetEnd() const {
return end;
}
@@ -76,8 +76,8 @@ public:
}
private:
- CacheAddr start;
- CacheAddr end;
+ VAddr start;
+ VAddr end;
GPUVAddr gpu_addr;
VAddr cpu_addr{};
bool is_written{};
diff --git a/src/video_core/engines/maxwell_3d.h b/src/video_core/engines/maxwell_3d.h
index d24c9f657..4637ddabd 100644
--- a/src/video_core/engines/maxwell_3d.h
+++ b/src/video_core/engines/maxwell_3d.h
@@ -312,6 +312,35 @@ public:
}
};
+ struct MsaaSampleLocation {
+ union {
+ BitField<0, 4, u32> x0;
+ BitField<4, 4, u32> y0;
+ BitField<8, 4, u32> x1;
+ BitField<12, 4, u32> y1;
+ BitField<16, 4, u32> x2;
+ BitField<20, 4, u32> y2;
+ BitField<24, 4, u32> x3;
+ BitField<28, 4, u32> y3;
+ };
+
+ constexpr std::pair<u32, u32> Location(int index) const {
+ switch (index) {
+ case 0:
+ return {x0, y0};
+ case 1:
+ return {x1, y1};
+ case 2:
+ return {x2, y2};
+ case 3:
+ return {x3, y3};
+ default:
+ UNREACHABLE();
+ return {0, 0};
+ }
+ }
+ };
+
enum class DepthMode : u32 {
MinusOneToOne = 0,
ZeroToOne = 1,
@@ -793,7 +822,13 @@ public:
u32 rt_separate_frag_data;
- INSERT_UNION_PADDING_WORDS(0xC);
+ INSERT_UNION_PADDING_WORDS(0x1);
+
+ u32 multisample_raster_enable;
+ u32 multisample_raster_samples;
+ std::array<u32, 4> multisample_sample_mask;
+
+ INSERT_UNION_PADDING_WORDS(0x5);
struct {
u32 address_high;
@@ -830,7 +865,16 @@ public:
std::array<VertexAttribute, NumVertexAttributes> vertex_attrib_format;
- INSERT_UNION_PADDING_WORDS(0xF);
+ std::array<MsaaSampleLocation, 4> multisample_sample_locations;
+
+ INSERT_UNION_PADDING_WORDS(0x2);
+
+ union {
+ BitField<0, 1, u32> enable;
+ BitField<4, 3, u32> target;
+ } multisample_coverage_to_color;
+
+ INSERT_UNION_PADDING_WORDS(0x8);
struct {
union {
@@ -943,7 +987,7 @@ public:
CounterReset counter_reset;
- INSERT_UNION_PADDING_WORDS(0x1);
+ u32 multisample_enable;
u32 zeta_enable;
@@ -1007,7 +1051,11 @@ public:
float polygon_offset_units;
- INSERT_UNION_PADDING_WORDS(0x11);
+ INSERT_UNION_PADDING_WORDS(0x4);
+
+ Tegra::Texture::MsaaMode multisample_mode;
+
+ INSERT_UNION_PADDING_WORDS(0xC);
union {
BitField<2, 1, u32> coord_origin;
@@ -1507,12 +1555,17 @@ ASSERT_REG_POSITION(stencil_back_func_ref, 0x3D5);
ASSERT_REG_POSITION(stencil_back_mask, 0x3D6);
ASSERT_REG_POSITION(stencil_back_func_mask, 0x3D7);
ASSERT_REG_POSITION(color_mask_common, 0x3E4);
-ASSERT_REG_POSITION(rt_separate_frag_data, 0x3EB);
ASSERT_REG_POSITION(depth_bounds, 0x3E7);
+ASSERT_REG_POSITION(rt_separate_frag_data, 0x3EB);
+ASSERT_REG_POSITION(multisample_raster_enable, 0x3ED);
+ASSERT_REG_POSITION(multisample_raster_samples, 0x3EE);
+ASSERT_REG_POSITION(multisample_sample_mask, 0x3EF);
ASSERT_REG_POSITION(zeta, 0x3F8);
ASSERT_REG_POSITION(clear_flags, 0x43E);
ASSERT_REG_POSITION(fill_rectangle, 0x44F);
ASSERT_REG_POSITION(vertex_attrib_format, 0x458);
+ASSERT_REG_POSITION(multisample_sample_locations, 0x478);
+ASSERT_REG_POSITION(multisample_coverage_to_color, 0x47E);
ASSERT_REG_POSITION(rt_control, 0x487);
ASSERT_REG_POSITION(zeta_width, 0x48a);
ASSERT_REG_POSITION(zeta_height, 0x48b);
@@ -1545,11 +1598,12 @@ ASSERT_REG_POSITION(samplecnt_enable, 0x545);
ASSERT_REG_POSITION(point_size, 0x546);
ASSERT_REG_POSITION(point_sprite_enable, 0x548);
ASSERT_REG_POSITION(counter_reset, 0x54C);
+ASSERT_REG_POSITION(multisample_enable, 0x54D);
ASSERT_REG_POSITION(zeta_enable, 0x54E);
ASSERT_REG_POSITION(multisample_control, 0x54F);
ASSERT_REG_POSITION(condition, 0x554);
ASSERT_REG_POSITION(tsc, 0x557);
-ASSERT_REG_POSITION(polygon_offset_factor, 0x55b);
+ASSERT_REG_POSITION(polygon_offset_factor, 0x55B);
ASSERT_REG_POSITION(tic, 0x55D);
ASSERT_REG_POSITION(stencil_two_side_enable, 0x565);
ASSERT_REG_POSITION(stencil_back_op_fail, 0x566);
@@ -1558,6 +1612,7 @@ ASSERT_REG_POSITION(stencil_back_op_zpass, 0x568);
ASSERT_REG_POSITION(stencil_back_func_func, 0x569);
ASSERT_REG_POSITION(framebuffer_srgb, 0x56E);
ASSERT_REG_POSITION(polygon_offset_units, 0x56F);
+ASSERT_REG_POSITION(multisample_mode, 0x574);
ASSERT_REG_POSITION(point_coord_replace, 0x581);
ASSERT_REG_POSITION(code_address, 0x582);
ASSERT_REG_POSITION(draw, 0x585);
diff --git a/src/video_core/engines/shader_bytecode.h b/src/video_core/engines/shader_bytecode.h
index 7400e1aa9..c66c66f6c 100644
--- a/src/video_core/engines/shader_bytecode.h
+++ b/src/video_core/engines/shader_bytecode.h
@@ -231,18 +231,6 @@ enum class AtomicOp : u64 {
Or = 6,
Xor = 7,
Exch = 8,
-};
-
-enum class GlobalAtomicOp : u64 {
- Add = 0,
- Min = 1,
- Max = 2,
- Inc = 3,
- Dec = 4,
- And = 5,
- Or = 6,
- Xor = 7,
- Exch = 8,
SafeAdd = 10,
};
@@ -1018,7 +1006,7 @@ union Instruction {
} stg;
union {
- BitField<52, 4, GlobalAtomicOp> operation;
+ BitField<52, 4, AtomicOp> operation;
BitField<49, 3, GlobalAtomicType> type;
BitField<28, 20, s64> offset;
} atom;
@@ -1777,6 +1765,7 @@ public:
BRK,
DEPBAR,
VOTE,
+ VOTE_VTG,
SHFL,
FSWZADD,
BFE_C,
@@ -1823,6 +1812,7 @@ public:
IPA,
OUT_R, // Emit vertex/primitive
ISBERD,
+ BAR,
MEMBAR,
VMAD,
VSETP,
@@ -1908,7 +1898,7 @@ public:
MOV_C,
MOV_R,
MOV_IMM,
- MOV_SYS,
+ S2R,
MOV32_IMM,
SHL_C,
SHL_R,
@@ -2092,6 +2082,7 @@ private:
INST("111000110000----", Id::EXIT, Type::Flow, "EXIT"),
INST("1111000011110---", Id::DEPBAR, Type::Synch, "DEPBAR"),
INST("0101000011011---", Id::VOTE, Type::Warp, "VOTE"),
+ INST("0101000011100---", Id::VOTE_VTG, Type::Warp, "VOTE_VTG"),
INST("1110111100010---", Id::SHFL, Type::Warp, "SHFL"),
INST("0101000011111---", Id::FSWZADD, Type::Warp, "FSWZADD"),
INST("1110111111011---", Id::LD_A, Type::Memory, "LD_A"),
@@ -2129,6 +2120,7 @@ private:
INST("11100000--------", Id::IPA, Type::Trivial, "IPA"),
INST("1111101111100---", Id::OUT_R, Type::Trivial, "OUT_R"),
INST("1110111111010---", Id::ISBERD, Type::Trivial, "ISBERD"),
+ INST("1111000010101---", Id::BAR, Type::Trivial, "BAR"),
INST("1110111110011---", Id::MEMBAR, Type::Trivial, "MEMBAR"),
INST("01011111--------", Id::VMAD, Type::Video, "VMAD"),
INST("0101000011110---", Id::VSETP, Type::Video, "VSETP"),
@@ -2201,7 +2193,7 @@ private:
INST("0100110010011---", Id::MOV_C, Type::Arithmetic, "MOV_C"),
INST("0101110010011---", Id::MOV_R, Type::Arithmetic, "MOV_R"),
INST("0011100-10011---", Id::MOV_IMM, Type::Arithmetic, "MOV_IMM"),
- INST("1111000011001---", Id::MOV_SYS, Type::Trivial, "MOV_SYS"),
+ INST("1111000011001---", Id::S2R, Type::Trivial, "S2R"),
INST("000000010000----", Id::MOV32_IMM, Type::ArithmeticImmediate, "MOV32_IMM"),
INST("0100110001100---", Id::FMNMX_C, Type::Arithmetic, "FMNMX_C"),
INST("0101110001100---", Id::FMNMX_R, Type::Arithmetic, "FMNMX_R"),
diff --git a/src/video_core/engines/shader_header.h b/src/video_core/engines/shader_header.h
index bc80661d8..72e2a33d5 100644
--- a/src/video_core/engines/shader_header.h
+++ b/src/video_core/engines/shader_header.h
@@ -4,6 +4,9 @@
#pragma once
+#include <array>
+#include <optional>
+
#include "common/bit_field.h"
#include "common/common_funcs.h"
#include "common/common_types.h"
@@ -16,7 +19,7 @@ enum class OutputTopology : u32 {
TriangleStrip = 7,
};
-enum class AttributeUse : u8 {
+enum class PixelImap : u8 {
Unused = 0,
Constant = 1,
Perspective = 2,
@@ -24,7 +27,7 @@ enum class AttributeUse : u8 {
};
// Documentation in:
-// http://download.nvidia.com/open-gpu-doc/Shader-Program-Header/1/Shader-Program-Header.html#ImapTexture
+// http://download.nvidia.com/open-gpu-doc/Shader-Program-Header/1/Shader-Program-Header.html
struct Header {
union {
BitField<0, 5, u32> sph_type;
@@ -59,8 +62,8 @@ struct Header {
union {
BitField<0, 12, u32> max_output_vertices;
BitField<12, 8, u32> store_req_start; // NOTE: not used by geometry shaders.
- BitField<24, 4, u32> reserved;
- BitField<12, 8, u32> store_req_end; // NOTE: not used by geometry shaders.
+ BitField<20, 4, u32> reserved;
+ BitField<24, 8, u32> store_req_end; // NOTE: not used by geometry shaders.
} common4{};
union {
@@ -93,17 +96,20 @@ struct Header {
struct {
INSERT_UNION_PADDING_BYTES(3); // ImapSystemValuesA
INSERT_UNION_PADDING_BYTES(1); // ImapSystemValuesB
+
union {
- BitField<0, 2, AttributeUse> x;
- BitField<2, 2, AttributeUse> y;
- BitField<4, 2, AttributeUse> w;
- BitField<6, 2, AttributeUse> z;
+ BitField<0, 2, PixelImap> x;
+ BitField<2, 2, PixelImap> y;
+ BitField<4, 2, PixelImap> z;
+ BitField<6, 2, PixelImap> w;
u8 raw;
} imap_generic_vector[32];
+
INSERT_UNION_PADDING_BYTES(2); // ImapColor
INSERT_UNION_PADDING_BYTES(2); // ImapSystemValuesC
INSERT_UNION_PADDING_BYTES(10); // ImapFixedFncTexture[10]
INSERT_UNION_PADDING_BYTES(2); // ImapReserved
+
struct {
u32 target;
union {
@@ -112,31 +118,30 @@ struct Header {
BitField<2, 30, u32> reserved;
};
} omap;
+
bool IsColorComponentOutputEnabled(u32 render_target, u32 component) const {
const u32 bit = render_target * 4 + component;
return omap.target & (1 << bit);
}
- AttributeUse GetAttributeIndexUse(u32 attribute, u32 index) const {
- return static_cast<AttributeUse>(
- (imap_generic_vector[attribute].raw >> (index * 2)) & 0x03);
- }
- AttributeUse GetAttributeUse(u32 attribute) const {
- AttributeUse result = AttributeUse::Unused;
- for (u32 i = 0; i < 4; i++) {
- const auto index = GetAttributeIndexUse(attribute, i);
- if (index == AttributeUse::Unused) {
- continue;
- }
- if (result == AttributeUse::Unused || result == index) {
- result = index;
+
+ PixelImap GetPixelImap(u32 attribute) const {
+ const auto get_index = [this, attribute](u32 index) {
+ return static_cast<PixelImap>(
+ (imap_generic_vector[attribute].raw >> (index * 2)) & 3);
+ };
+
+ std::optional<PixelImap> result;
+ for (u32 component = 0; component < 4; ++component) {
+ const PixelImap index = get_index(component);
+ if (index == PixelImap::Unused) {
continue;
}
- LOG_CRITICAL(HW_GPU, "Generic Attribute Conflict in Interpolation Mode");
- if (index == AttributeUse::Perspective) {
- result = index;
+ if (result && result != index) {
+ LOG_CRITICAL(HW_GPU, "Generic attribute conflict in interpolation mode");
}
+ result = index;
}
- return result;
+ return result.value_or(PixelImap::Unused);
}
} ps;
diff --git a/src/video_core/gpu.cpp b/src/video_core/gpu.cpp
index e8f763ce9..8acf2eda2 100644
--- a/src/video_core/gpu.cpp
+++ b/src/video_core/gpu.cpp
@@ -7,6 +7,7 @@
#include "core/core.h"
#include "core/core_timing.h"
#include "core/core_timing_util.h"
+#include "core/frontend/emu_window.h"
#include "core/memory.h"
#include "video_core/engines/fermi_2d.h"
#include "video_core/engines/kepler_compute.h"
@@ -16,14 +17,15 @@
#include "video_core/gpu.h"
#include "video_core/memory_manager.h"
#include "video_core/renderer_base.h"
+#include "video_core/video_core.h"
namespace Tegra {
MICROPROFILE_DEFINE(GPU_wait, "GPU", "Wait for the GPU", MP_RGB(128, 128, 192));
-GPU::GPU(Core::System& system, VideoCore::RendererBase& renderer, bool is_async)
- : system{system}, renderer{renderer}, is_async{is_async} {
- auto& rasterizer{renderer.Rasterizer()};
+GPU::GPU(Core::System& system, std::unique_ptr<VideoCore::RendererBase>&& renderer_, bool is_async)
+ : system{system}, renderer{std::move(renderer_)}, is_async{is_async} {
+ auto& rasterizer{renderer->Rasterizer()};
memory_manager = std::make_unique<Tegra::MemoryManager>(system, rasterizer);
dma_pusher = std::make_unique<Tegra::DmaPusher>(*this);
maxwell_3d = std::make_unique<Engines::Maxwell3D>(system, rasterizer, *memory_manager);
@@ -137,7 +139,7 @@ u64 GPU::GetTicks() const {
}
void GPU::FlushCommands() {
- renderer.Rasterizer().FlushCommands();
+ renderer->Rasterizer().FlushCommands();
}
// Note that, traditionally, methods are treated as 4-byte addressable locations, and hence
diff --git a/src/video_core/gpu.h b/src/video_core/gpu.h
index 64acb17df..1a2d747be 100644
--- a/src/video_core/gpu.h
+++ b/src/video_core/gpu.h
@@ -25,8 +25,11 @@ inline u8* FromCacheAddr(CacheAddr cache_addr) {
}
namespace Core {
-class System;
+namespace Frontend {
+class EmuWindow;
}
+class System;
+} // namespace Core
namespace VideoCore {
class RendererBase;
@@ -129,7 +132,8 @@ class MemoryManager;
class GPU {
public:
- explicit GPU(Core::System& system, VideoCore::RendererBase& renderer, bool is_async);
+ explicit GPU(Core::System& system, std::unique_ptr<VideoCore::RendererBase>&& renderer,
+ bool is_async);
virtual ~GPU();
@@ -174,6 +178,14 @@ public:
/// Returns a reference to the GPU DMA pusher.
Tegra::DmaPusher& DmaPusher();
+ VideoCore::RendererBase& Renderer() {
+ return *renderer;
+ }
+
+ const VideoCore::RendererBase& Renderer() const {
+ return *renderer;
+ }
+
// Waits for the GPU to finish working
virtual void WaitIdle() const = 0;
@@ -258,13 +270,13 @@ public:
virtual void SwapBuffers(const Tegra::FramebufferConfig* framebuffer) = 0;
/// Notify rasterizer that any caches of the specified region should be flushed to Switch memory
- virtual void FlushRegion(CacheAddr addr, u64 size) = 0;
+ virtual void FlushRegion(VAddr addr, u64 size) = 0;
/// Notify rasterizer that any caches of the specified region should be invalidated
- virtual void InvalidateRegion(CacheAddr addr, u64 size) = 0;
+ virtual void InvalidateRegion(VAddr addr, u64 size) = 0;
/// Notify rasterizer that any caches of the specified region should be flushed and invalidated
- virtual void FlushAndInvalidateRegion(CacheAddr addr, u64 size) = 0;
+ virtual void FlushAndInvalidateRegion(VAddr addr, u64 size) = 0;
protected:
virtual void TriggerCpuInterrupt(u32 syncpoint_id, u32 value) const = 0;
@@ -287,7 +299,7 @@ private:
protected:
std::unique_ptr<Tegra::DmaPusher> dma_pusher;
Core::System& system;
- VideoCore::RendererBase& renderer;
+ std::unique_ptr<VideoCore::RendererBase> renderer;
private:
std::unique_ptr<Tegra::MemoryManager> memory_manager;
diff --git a/src/video_core/gpu_asynch.cpp b/src/video_core/gpu_asynch.cpp
index 04222d060..cc434faf7 100644
--- a/src/video_core/gpu_asynch.cpp
+++ b/src/video_core/gpu_asynch.cpp
@@ -10,13 +10,16 @@
namespace VideoCommon {
-GPUAsynch::GPUAsynch(Core::System& system, VideoCore::RendererBase& renderer)
- : GPU(system, renderer, true), gpu_thread{system} {}
+GPUAsynch::GPUAsynch(Core::System& system, std::unique_ptr<VideoCore::RendererBase>&& renderer_,
+ std::unique_ptr<Core::Frontend::GraphicsContext>&& context)
+ : GPU(system, std::move(renderer_), true), gpu_thread{system}, gpu_context(std::move(context)),
+ cpu_context(renderer->GetRenderWindow().CreateSharedContext()) {}
GPUAsynch::~GPUAsynch() = default;
void GPUAsynch::Start() {
- gpu_thread.StartThread(renderer, *dma_pusher);
+ cpu_context->MakeCurrent();
+ gpu_thread.StartThread(*renderer, *gpu_context, *dma_pusher);
}
void GPUAsynch::PushGPUEntries(Tegra::CommandList&& entries) {
@@ -27,15 +30,15 @@ void GPUAsynch::SwapBuffers(const Tegra::FramebufferConfig* framebuffer) {
gpu_thread.SwapBuffers(framebuffer);
}
-void GPUAsynch::FlushRegion(CacheAddr addr, u64 size) {
+void GPUAsynch::FlushRegion(VAddr addr, u64 size) {
gpu_thread.FlushRegion(addr, size);
}
-void GPUAsynch::InvalidateRegion(CacheAddr addr, u64 size) {
+void GPUAsynch::InvalidateRegion(VAddr addr, u64 size) {
gpu_thread.InvalidateRegion(addr, size);
}
-void GPUAsynch::FlushAndInvalidateRegion(CacheAddr addr, u64 size) {
+void GPUAsynch::FlushAndInvalidateRegion(VAddr addr, u64 size) {
gpu_thread.FlushAndInvalidateRegion(addr, size);
}
diff --git a/src/video_core/gpu_asynch.h b/src/video_core/gpu_asynch.h
index 1241ade1d..03fd0eef0 100644
--- a/src/video_core/gpu_asynch.h
+++ b/src/video_core/gpu_asynch.h
@@ -7,6 +7,10 @@
#include "video_core/gpu.h"
#include "video_core/gpu_thread.h"
+namespace Core::Frontend {
+class GraphicsContext;
+}
+
namespace VideoCore {
class RendererBase;
} // namespace VideoCore
@@ -16,15 +20,16 @@ namespace VideoCommon {
/// Implementation of GPU interface that runs the GPU asynchronously
class GPUAsynch final : public Tegra::GPU {
public:
- explicit GPUAsynch(Core::System& system, VideoCore::RendererBase& renderer);
+ explicit GPUAsynch(Core::System& system, std::unique_ptr<VideoCore::RendererBase>&& renderer,
+ std::unique_ptr<Core::Frontend::GraphicsContext>&& context);
~GPUAsynch() override;
void Start() override;
void PushGPUEntries(Tegra::CommandList&& entries) override;
void SwapBuffers(const Tegra::FramebufferConfig* framebuffer) override;
- void FlushRegion(CacheAddr addr, u64 size) override;
- void InvalidateRegion(CacheAddr addr, u64 size) override;
- void FlushAndInvalidateRegion(CacheAddr addr, u64 size) override;
+ void FlushRegion(VAddr addr, u64 size) override;
+ void InvalidateRegion(VAddr addr, u64 size) override;
+ void FlushAndInvalidateRegion(VAddr addr, u64 size) override;
void WaitIdle() const override;
protected:
@@ -32,6 +37,8 @@ protected:
private:
GPUThread::ThreadManager gpu_thread;
+ std::unique_ptr<Core::Frontend::GraphicsContext> cpu_context;
+ std::unique_ptr<Core::Frontend::GraphicsContext> gpu_context;
};
} // namespace VideoCommon
diff --git a/src/video_core/gpu_synch.cpp b/src/video_core/gpu_synch.cpp
index d48221077..6f38a672a 100644
--- a/src/video_core/gpu_synch.cpp
+++ b/src/video_core/gpu_synch.cpp
@@ -7,12 +7,15 @@
namespace VideoCommon {
-GPUSynch::GPUSynch(Core::System& system, VideoCore::RendererBase& renderer)
- : GPU(system, renderer, false) {}
+GPUSynch::GPUSynch(Core::System& system, std::unique_ptr<VideoCore::RendererBase>&& renderer,
+ std::unique_ptr<Core::Frontend::GraphicsContext>&& context)
+ : GPU(system, std::move(renderer), false), context{std::move(context)} {}
GPUSynch::~GPUSynch() = default;
-void GPUSynch::Start() {}
+void GPUSynch::Start() {
+ context->MakeCurrent();
+}
void GPUSynch::PushGPUEntries(Tegra::CommandList&& entries) {
dma_pusher->Push(std::move(entries));
@@ -20,19 +23,19 @@ void GPUSynch::PushGPUEntries(Tegra::CommandList&& entries) {
}
void GPUSynch::SwapBuffers(const Tegra::FramebufferConfig* framebuffer) {
- renderer.SwapBuffers(framebuffer);
+ renderer->SwapBuffers(framebuffer);
}
-void GPUSynch::FlushRegion(CacheAddr addr, u64 size) {
- renderer.Rasterizer().FlushRegion(addr, size);
+void GPUSynch::FlushRegion(VAddr addr, u64 size) {
+ renderer->Rasterizer().FlushRegion(addr, size);
}
-void GPUSynch::InvalidateRegion(CacheAddr addr, u64 size) {
- renderer.Rasterizer().InvalidateRegion(addr, size);
+void GPUSynch::InvalidateRegion(VAddr addr, u64 size) {
+ renderer->Rasterizer().InvalidateRegion(addr, size);
}
-void GPUSynch::FlushAndInvalidateRegion(CacheAddr addr, u64 size) {
- renderer.Rasterizer().FlushAndInvalidateRegion(addr, size);
+void GPUSynch::FlushAndInvalidateRegion(VAddr addr, u64 size) {
+ renderer->Rasterizer().FlushAndInvalidateRegion(addr, size);
}
} // namespace VideoCommon
diff --git a/src/video_core/gpu_synch.h b/src/video_core/gpu_synch.h
index c71baee89..4a6e9a01d 100644
--- a/src/video_core/gpu_synch.h
+++ b/src/video_core/gpu_synch.h
@@ -6,6 +6,10 @@
#include "video_core/gpu.h"
+namespace Core::Frontend {
+class GraphicsContext;
+}
+
namespace VideoCore {
class RendererBase;
} // namespace VideoCore
@@ -15,20 +19,24 @@ namespace VideoCommon {
/// Implementation of GPU interface that runs the GPU synchronously
class GPUSynch final : public Tegra::GPU {
public:
- explicit GPUSynch(Core::System& system, VideoCore::RendererBase& renderer);
+ explicit GPUSynch(Core::System& system, std::unique_ptr<VideoCore::RendererBase>&& renderer,
+ std::unique_ptr<Core::Frontend::GraphicsContext>&& context);
~GPUSynch() override;
void Start() override;
void PushGPUEntries(Tegra::CommandList&& entries) override;
void SwapBuffers(const Tegra::FramebufferConfig* framebuffer) override;
- void FlushRegion(CacheAddr addr, u64 size) override;
- void InvalidateRegion(CacheAddr addr, u64 size) override;
- void FlushAndInvalidateRegion(CacheAddr addr, u64 size) override;
+ void FlushRegion(VAddr addr, u64 size) override;
+ void InvalidateRegion(VAddr addr, u64 size) override;
+ void FlushAndInvalidateRegion(VAddr addr, u64 size) override;
void WaitIdle() const override {}
protected:
void TriggerCpuInterrupt([[maybe_unused]] u32 syncpoint_id,
[[maybe_unused]] u32 value) const override {}
+
+private:
+ std::unique_ptr<Core::Frontend::GraphicsContext> context;
};
} // namespace VideoCommon
diff --git a/src/video_core/gpu_thread.cpp b/src/video_core/gpu_thread.cpp
index b1088af3d..10cda686b 100644
--- a/src/video_core/gpu_thread.cpp
+++ b/src/video_core/gpu_thread.cpp
@@ -5,7 +5,7 @@
#include "common/assert.h"
#include "common/microprofile.h"
#include "core/core.h"
-#include "core/frontend/scope_acquire_context.h"
+#include "core/frontend/emu_window.h"
#include "video_core/dma_pusher.h"
#include "video_core/gpu.h"
#include "video_core/gpu_thread.h"
@@ -14,8 +14,8 @@
namespace VideoCommon::GPUThread {
/// Runs the GPU thread
-static void RunThread(VideoCore::RendererBase& renderer, Tegra::DmaPusher& dma_pusher,
- SynchState& state) {
+static void RunThread(VideoCore::RendererBase& renderer, Core::Frontend::GraphicsContext& context,
+ Tegra::DmaPusher& dma_pusher, SynchState& state) {
MicroProfileOnThreadCreate("GpuThread");
// Wait for first GPU command before acquiring the window context
@@ -27,7 +27,7 @@ static void RunThread(VideoCore::RendererBase& renderer, Tegra::DmaPusher& dma_p
return;
}
- Core::Frontend::ScopeAcquireContext acquire_context{renderer.GetRenderWindow()};
+ auto current_context = context.Acquire();
CommandDataContainer next;
while (state.is_running) {
@@ -62,8 +62,11 @@ ThreadManager::~ThreadManager() {
thread.join();
}
-void ThreadManager::StartThread(VideoCore::RendererBase& renderer, Tegra::DmaPusher& dma_pusher) {
- thread = std::thread{RunThread, std::ref(renderer), std::ref(dma_pusher), std::ref(state)};
+void ThreadManager::StartThread(VideoCore::RendererBase& renderer,
+ Core::Frontend::GraphicsContext& context,
+ Tegra::DmaPusher& dma_pusher) {
+ thread = std::thread{RunThread, std::ref(renderer), std::ref(context), std::ref(dma_pusher),
+ std::ref(state)};
}
void ThreadManager::SubmitList(Tegra::CommandList&& entries) {
@@ -74,15 +77,15 @@ void ThreadManager::SwapBuffers(const Tegra::FramebufferConfig* framebuffer) {
PushCommand(SwapBuffersCommand(framebuffer ? std::make_optional(*framebuffer) : std::nullopt));
}
-void ThreadManager::FlushRegion(CacheAddr addr, u64 size) {
+void ThreadManager::FlushRegion(VAddr addr, u64 size) {
PushCommand(FlushRegionCommand(addr, size));
}
-void ThreadManager::InvalidateRegion(CacheAddr addr, u64 size) {
+void ThreadManager::InvalidateRegion(VAddr addr, u64 size) {
system.Renderer().Rasterizer().InvalidateRegion(addr, size);
}
-void ThreadManager::FlushAndInvalidateRegion(CacheAddr addr, u64 size) {
+void ThreadManager::FlushAndInvalidateRegion(VAddr addr, u64 size) {
// Skip flush on asynch mode, as FlushAndInvalidateRegion is not used for anything too important
InvalidateRegion(addr, size);
}
diff --git a/src/video_core/gpu_thread.h b/src/video_core/gpu_thread.h
index 882e2d9c7..cd74ad330 100644
--- a/src/video_core/gpu_thread.h
+++ b/src/video_core/gpu_thread.h
@@ -10,7 +10,6 @@
#include <optional>
#include <thread>
#include <variant>
-
#include "common/threadsafe_queue.h"
#include "video_core/gpu.h"
@@ -20,6 +19,9 @@ class DmaPusher;
} // namespace Tegra
namespace Core {
+namespace Frontend {
+class GraphicsContext;
+}
class System;
} // namespace Core
@@ -45,26 +47,26 @@ struct SwapBuffersCommand final {
/// Command to signal to the GPU thread to flush a region
struct FlushRegionCommand final {
- explicit constexpr FlushRegionCommand(CacheAddr addr, u64 size) : addr{addr}, size{size} {}
+ explicit constexpr FlushRegionCommand(VAddr addr, u64 size) : addr{addr}, size{size} {}
- CacheAddr addr;
+ VAddr addr;
u64 size;
};
/// Command to signal to the GPU thread to invalidate a region
struct InvalidateRegionCommand final {
- explicit constexpr InvalidateRegionCommand(CacheAddr addr, u64 size) : addr{addr}, size{size} {}
+ explicit constexpr InvalidateRegionCommand(VAddr addr, u64 size) : addr{addr}, size{size} {}
- CacheAddr addr;
+ VAddr addr;
u64 size;
};
/// Command to signal to the GPU thread to flush and invalidate a region
struct FlushAndInvalidateRegionCommand final {
- explicit constexpr FlushAndInvalidateRegionCommand(CacheAddr addr, u64 size)
+ explicit constexpr FlushAndInvalidateRegionCommand(VAddr addr, u64 size)
: addr{addr}, size{size} {}
- CacheAddr addr;
+ VAddr addr;
u64 size;
};
@@ -99,7 +101,8 @@ public:
~ThreadManager();
/// Creates and starts the GPU thread.
- void StartThread(VideoCore::RendererBase& renderer, Tegra::DmaPusher& dma_pusher);
+ void StartThread(VideoCore::RendererBase& renderer, Core::Frontend::GraphicsContext& context,
+ Tegra::DmaPusher& dma_pusher);
/// Push GPU command entries to be processed
void SubmitList(Tegra::CommandList&& entries);
@@ -108,13 +111,13 @@ public:
void SwapBuffers(const Tegra::FramebufferConfig* framebuffer);
/// Notify rasterizer that any caches of the specified region should be flushed to Switch memory
- void FlushRegion(CacheAddr addr, u64 size);
+ void FlushRegion(VAddr addr, u64 size);
/// Notify rasterizer that any caches of the specified region should be invalidated
- void InvalidateRegion(CacheAddr addr, u64 size);
+ void InvalidateRegion(VAddr addr, u64 size);
/// Notify rasterizer that any caches of the specified region should be flushed and invalidated
- void FlushAndInvalidateRegion(CacheAddr addr, u64 size);
+ void FlushAndInvalidateRegion(VAddr addr, u64 size);
// Wait until the gpu thread is idle.
void WaitIdle() const;
diff --git a/src/video_core/memory_manager.cpp b/src/video_core/memory_manager.cpp
index f5d33f27a..a3389d0d2 100644
--- a/src/video_core/memory_manager.cpp
+++ b/src/video_core/memory_manager.cpp
@@ -81,12 +81,11 @@ GPUVAddr MemoryManager::UnmapBuffer(GPUVAddr gpu_addr, u64 size) {
ASSERT((gpu_addr & page_mask) == 0);
const u64 aligned_size{Common::AlignUp(size, page_size)};
- const CacheAddr cache_addr{ToCacheAddr(GetPointer(gpu_addr))};
const auto cpu_addr = GpuToCpuAddress(gpu_addr);
ASSERT(cpu_addr);
// Flush and invalidate through the GPU interface, to be asynchronous if possible.
- system.GPU().FlushAndInvalidateRegion(cache_addr, aligned_size);
+ system.GPU().FlushAndInvalidateRegion(*cpu_addr, aligned_size);
UnmapRange(gpu_addr, aligned_size);
ASSERT(system.CurrentProcess()
@@ -140,11 +139,11 @@ T MemoryManager::Read(GPUVAddr addr) const {
return {};
}
- const u8* page_pointer{page_table.pointers[addr >> page_bits]};
+ const u8* page_pointer{GetPointer(addr)};
if (page_pointer) {
// NOTE: Avoid adding any extra logic to this fast-path block
T value;
- std::memcpy(&value, &page_pointer[addr & page_mask], sizeof(T));
+ std::memcpy(&value, page_pointer, sizeof(T));
return value;
}
@@ -167,10 +166,10 @@ void MemoryManager::Write(GPUVAddr addr, T data) {
return;
}
- u8* page_pointer{page_table.pointers[addr >> page_bits]};
+ u8* page_pointer{GetPointer(addr)};
if (page_pointer) {
// NOTE: Avoid adding any extra logic to this fast-path block
- std::memcpy(&page_pointer[addr & page_mask], &data, sizeof(T));
+ std::memcpy(page_pointer, &data, sizeof(T));
return;
}
@@ -201,9 +200,12 @@ u8* MemoryManager::GetPointer(GPUVAddr addr) {
return {};
}
- u8* const page_pointer{page_table.pointers[addr >> page_bits]};
- if (page_pointer != nullptr) {
- return page_pointer + (addr & page_mask);
+ auto& memory = system.Memory();
+
+ const VAddr page_addr{page_table.backing_addr[addr >> page_bits]};
+
+ if (page_addr != 0) {
+ return memory.GetPointer(page_addr + (addr & page_mask));
}
LOG_ERROR(HW_GPU, "Unknown GetPointer @ 0x{:016X}", addr);
@@ -215,9 +217,12 @@ const u8* MemoryManager::GetPointer(GPUVAddr addr) const {
return {};
}
- const u8* const page_pointer{page_table.pointers[addr >> page_bits]};
- if (page_pointer != nullptr) {
- return page_pointer + (addr & page_mask);
+ const auto& memory = system.Memory();
+
+ const VAddr page_addr{page_table.backing_addr[addr >> page_bits]};
+
+ if (page_addr != 0) {
+ return memory.GetPointer(page_addr + (addr & page_mask));
}
LOG_ERROR(HW_GPU, "Unknown GetPointer @ 0x{:016X}", addr);
@@ -238,17 +243,19 @@ void MemoryManager::ReadBlock(GPUVAddr src_addr, void* dest_buffer, const std::s
std::size_t page_index{src_addr >> page_bits};
std::size_t page_offset{src_addr & page_mask};
+ auto& memory = system.Memory();
+
while (remaining_size > 0) {
const std::size_t copy_amount{
std::min(static_cast<std::size_t>(page_size) - page_offset, remaining_size)};
switch (page_table.attributes[page_index]) {
case Common::PageType::Memory: {
- const u8* src_ptr{page_table.pointers[page_index] + page_offset};
+ const VAddr src_addr{page_table.backing_addr[page_index] + page_offset};
// Flush must happen on the rasterizer interface, such that memory is always synchronous
// when it is read (even when in asynchronous GPU mode). Fixes Dead Cells title menu.
- rasterizer.FlushRegion(ToCacheAddr(src_ptr), copy_amount);
- std::memcpy(dest_buffer, src_ptr, copy_amount);
+ rasterizer.FlushRegion(src_addr, copy_amount);
+ memory.ReadBlockUnsafe(src_addr, dest_buffer, copy_amount);
break;
}
default:
@@ -268,13 +275,15 @@ void MemoryManager::ReadBlockUnsafe(GPUVAddr src_addr, void* dest_buffer,
std::size_t page_index{src_addr >> page_bits};
std::size_t page_offset{src_addr & page_mask};
+ auto& memory = system.Memory();
+
while (remaining_size > 0) {
const std::size_t copy_amount{
std::min(static_cast<std::size_t>(page_size) - page_offset, remaining_size)};
const u8* page_pointer = page_table.pointers[page_index];
if (page_pointer) {
- const u8* src_ptr{page_pointer + page_offset};
- std::memcpy(dest_buffer, src_ptr, copy_amount);
+ const VAddr src_addr{page_table.backing_addr[page_index] + page_offset};
+ memory.ReadBlockUnsafe(src_addr, dest_buffer, copy_amount);
} else {
std::memset(dest_buffer, 0, copy_amount);
}
@@ -290,17 +299,19 @@ void MemoryManager::WriteBlock(GPUVAddr dest_addr, const void* src_buffer, const
std::size_t page_index{dest_addr >> page_bits};
std::size_t page_offset{dest_addr & page_mask};
+ auto& memory = system.Memory();
+
while (remaining_size > 0) {
const std::size_t copy_amount{
std::min(static_cast<std::size_t>(page_size) - page_offset, remaining_size)};
switch (page_table.attributes[page_index]) {
case Common::PageType::Memory: {
- u8* dest_ptr{page_table.pointers[page_index] + page_offset};
+ const VAddr dest_addr{page_table.backing_addr[page_index] + page_offset};
// Invalidate must happen on the rasterizer interface, such that memory is always
// synchronous when it is written (even when in asynchronous GPU mode).
- rasterizer.InvalidateRegion(ToCacheAddr(dest_ptr), copy_amount);
- std::memcpy(dest_ptr, src_buffer, copy_amount);
+ rasterizer.InvalidateRegion(dest_addr, copy_amount);
+ memory.WriteBlockUnsafe(dest_addr, src_buffer, copy_amount);
break;
}
default:
@@ -320,13 +331,15 @@ void MemoryManager::WriteBlockUnsafe(GPUVAddr dest_addr, const void* src_buffer,
std::size_t page_index{dest_addr >> page_bits};
std::size_t page_offset{dest_addr & page_mask};
+ auto& memory = system.Memory();
+
while (remaining_size > 0) {
const std::size_t copy_amount{
std::min(static_cast<std::size_t>(page_size) - page_offset, remaining_size)};
u8* page_pointer = page_table.pointers[page_index];
if (page_pointer) {
- u8* dest_ptr{page_pointer + page_offset};
- std::memcpy(dest_ptr, src_buffer, copy_amount);
+ const VAddr dest_addr{page_table.backing_addr[page_index] + page_offset};
+ memory.WriteBlockUnsafe(dest_addr, src_buffer, copy_amount);
}
page_index++;
page_offset = 0;
@@ -336,33 +349,9 @@ void MemoryManager::WriteBlockUnsafe(GPUVAddr dest_addr, const void* src_buffer,
}
void MemoryManager::CopyBlock(GPUVAddr dest_addr, GPUVAddr src_addr, const std::size_t size) {
- std::size_t remaining_size{size};
- std::size_t page_index{src_addr >> page_bits};
- std::size_t page_offset{src_addr & page_mask};
-
- while (remaining_size > 0) {
- const std::size_t copy_amount{
- std::min(static_cast<std::size_t>(page_size) - page_offset, remaining_size)};
-
- switch (page_table.attributes[page_index]) {
- case Common::PageType::Memory: {
- // Flush must happen on the rasterizer interface, such that memory is always synchronous
- // when it is copied (even when in asynchronous GPU mode).
- const u8* src_ptr{page_table.pointers[page_index] + page_offset};
- rasterizer.FlushRegion(ToCacheAddr(src_ptr), copy_amount);
- WriteBlock(dest_addr, src_ptr, copy_amount);
- break;
- }
- default:
- UNREACHABLE();
- }
-
- page_index++;
- page_offset = 0;
- dest_addr += static_cast<VAddr>(copy_amount);
- src_addr += static_cast<VAddr>(copy_amount);
- remaining_size -= copy_amount;
- }
+ std::vector<u8> tmp_buffer(size);
+ ReadBlock(src_addr, tmp_buffer.data(), size);
+ WriteBlock(dest_addr, tmp_buffer.data(), size);
}
void MemoryManager::CopyBlockUnsafe(GPUVAddr dest_addr, GPUVAddr src_addr, const std::size_t size) {
@@ -371,6 +360,12 @@ void MemoryManager::CopyBlockUnsafe(GPUVAddr dest_addr, GPUVAddr src_addr, const
WriteBlockUnsafe(dest_addr, tmp_buffer.data(), size);
}
+bool MemoryManager::IsGranularRange(GPUVAddr gpu_addr, std::size_t size) {
+ const VAddr addr = page_table.backing_addr[gpu_addr >> page_bits];
+ const std::size_t page = (addr & Memory::PAGE_MASK) + size;
+ return page <= Memory::PAGE_SIZE;
+}
+
void MemoryManager::MapPages(GPUVAddr base, u64 size, u8* memory, Common::PageType type,
VAddr backing_addr) {
LOG_DEBUG(HW_GPU, "Mapping {} onto {:016X}-{:016X}", fmt::ptr(memory), base * page_size,
diff --git a/src/video_core/memory_manager.h b/src/video_core/memory_manager.h
index 073bdb491..0d9468535 100644
--- a/src/video_core/memory_manager.h
+++ b/src/video_core/memory_manager.h
@@ -97,6 +97,11 @@ public:
void WriteBlockUnsafe(GPUVAddr dest_addr, const void* src_buffer, std::size_t size);
void CopyBlockUnsafe(GPUVAddr dest_addr, GPUVAddr src_addr, std::size_t size);
+ /**
+ * IsGranularRange checks if a gpu region can be simply read with a pointer
+ */
+ bool IsGranularRange(GPUVAddr gpu_addr, std::size_t size);
+
private:
using VMAMap = std::map<GPUVAddr, VirtualMemoryArea>;
using VMAHandle = VMAMap::const_iterator;
diff --git a/src/video_core/query_cache.h b/src/video_core/query_cache.h
index e66054ed0..5ea2b01f2 100644
--- a/src/video_core/query_cache.h
+++ b/src/video_core/query_cache.h
@@ -98,12 +98,12 @@ public:
static_cast<QueryCache&>(*this),
VideoCore::QueryType::SamplesPassed}}} {}
- void InvalidateRegion(CacheAddr addr, std::size_t size) {
+ void InvalidateRegion(VAddr addr, std::size_t size) {
std::unique_lock lock{mutex};
FlushAndRemoveRegion(addr, size);
}
- void FlushRegion(CacheAddr addr, std::size_t size) {
+ void FlushRegion(VAddr addr, std::size_t size) {
std::unique_lock lock{mutex};
FlushAndRemoveRegion(addr, size);
}
@@ -117,14 +117,16 @@ public:
void Query(GPUVAddr gpu_addr, VideoCore::QueryType type, std::optional<u64> timestamp) {
std::unique_lock lock{mutex};
auto& memory_manager = system.GPU().MemoryManager();
- const auto host_ptr = memory_manager.GetPointer(gpu_addr);
+ const std::optional<VAddr> cpu_addr_opt = memory_manager.GpuToCpuAddress(gpu_addr);
+ ASSERT(cpu_addr_opt);
+ VAddr cpu_addr = *cpu_addr_opt;
- CachedQuery* query = TryGet(ToCacheAddr(host_ptr));
+ CachedQuery* query = TryGet(cpu_addr);
if (!query) {
- const auto cpu_addr = memory_manager.GpuToCpuAddress(gpu_addr);
- ASSERT_OR_EXECUTE(cpu_addr, return;);
+ ASSERT_OR_EXECUTE(cpu_addr_opt, return;);
+ const auto host_ptr = memory_manager.GetPointer(gpu_addr);
- query = Register(type, *cpu_addr, host_ptr, timestamp.has_value());
+ query = Register(type, cpu_addr, host_ptr, timestamp.has_value());
}
query->BindCounter(Stream(type).Current(), timestamp);
@@ -173,11 +175,11 @@ protected:
private:
/// Flushes a memory range to guest memory and removes it from the cache.
- void FlushAndRemoveRegion(CacheAddr addr, std::size_t size) {
+ void FlushAndRemoveRegion(VAddr addr, std::size_t size) {
const u64 addr_begin = static_cast<u64>(addr);
const u64 addr_end = addr_begin + static_cast<u64>(size);
const auto in_range = [addr_begin, addr_end](CachedQuery& query) {
- const u64 cache_begin = query.GetCacheAddr();
+ const u64 cache_begin = query.GetCpuAddr();
const u64 cache_end = cache_begin + query.SizeInBytes();
return cache_begin < addr_end && addr_begin < cache_end;
};
@@ -193,7 +195,7 @@ private:
if (!in_range(query)) {
continue;
}
- rasterizer.UpdatePagesCachedCount(query.CpuAddr(), query.SizeInBytes(), -1);
+ rasterizer.UpdatePagesCachedCount(query.GetCpuAddr(), query.SizeInBytes(), -1);
query.Flush();
}
contents.erase(std::remove_if(std::begin(contents), std::end(contents), in_range),
@@ -204,22 +206,21 @@ private:
/// Registers the passed parameters as cached and returns a pointer to the stored cached query.
CachedQuery* Register(VideoCore::QueryType type, VAddr cpu_addr, u8* host_ptr, bool timestamp) {
rasterizer.UpdatePagesCachedCount(cpu_addr, CachedQuery::SizeInBytes(timestamp), 1);
- const u64 page = static_cast<u64>(ToCacheAddr(host_ptr)) >> PAGE_SHIFT;
+ const u64 page = static_cast<u64>(cpu_addr) >> PAGE_SHIFT;
return &cached_queries[page].emplace_back(static_cast<QueryCache&>(*this), type, cpu_addr,
host_ptr);
}
/// Tries to a get a cached query. Returns nullptr on failure.
- CachedQuery* TryGet(CacheAddr addr) {
+ CachedQuery* TryGet(VAddr addr) {
const u64 page = static_cast<u64>(addr) >> PAGE_SHIFT;
const auto it = cached_queries.find(page);
if (it == std::end(cached_queries)) {
return nullptr;
}
auto& contents = it->second;
- const auto found =
- std::find_if(std::begin(contents), std::end(contents),
- [addr](auto& query) { return query.GetCacheAddr() == addr; });
+ const auto found = std::find_if(std::begin(contents), std::end(contents),
+ [addr](auto& query) { return query.GetCpuAddr() == addr; });
return found != std::end(contents) ? &*found : nullptr;
}
@@ -323,14 +324,10 @@ public:
timestamp = timestamp_;
}
- VAddr CpuAddr() const noexcept {
+ VAddr GetCpuAddr() const noexcept {
return cpu_addr;
}
- CacheAddr GetCacheAddr() const noexcept {
- return ToCacheAddr(host_ptr);
- }
-
u64 SizeInBytes() const noexcept {
return SizeInBytes(timestamp.has_value());
}
diff --git a/src/video_core/rasterizer_cache.h b/src/video_core/rasterizer_cache.h
index 6de1597a2..22987751e 100644
--- a/src/video_core/rasterizer_cache.h
+++ b/src/video_core/rasterizer_cache.h
@@ -18,22 +18,14 @@
class RasterizerCacheObject {
public:
- explicit RasterizerCacheObject(const u8* host_ptr)
- : host_ptr{host_ptr}, cache_addr{ToCacheAddr(host_ptr)} {}
+ explicit RasterizerCacheObject(const VAddr cpu_addr) : cpu_addr{cpu_addr} {}
virtual ~RasterizerCacheObject();
- CacheAddr GetCacheAddr() const {
- return cache_addr;
+ VAddr GetCpuAddr() const {
+ return cpu_addr;
}
- const u8* GetHostPtr() const {
- return host_ptr;
- }
-
- /// Gets the address of the shader in guest memory, required for cache management
- virtual VAddr GetCpuAddr() const = 0;
-
/// Gets the size of the shader in guest memory, required for cache management
virtual std::size_t GetSizeInBytes() const = 0;
@@ -68,8 +60,7 @@ private:
bool is_registered{}; ///< Whether the object is currently registered with the cache
bool is_dirty{}; ///< Whether the object is dirty (out of sync with guest memory)
u64 last_modified_ticks{}; ///< When the object was last modified, used for in-order flushing
- const u8* host_ptr{}; ///< Pointer to the memory backing this cached region
- CacheAddr cache_addr{}; ///< Cache address memory, unique from emulated virtual address space
+ VAddr cpu_addr{}; ///< Cpu address memory, unique from emulated virtual address space
};
template <class T>
@@ -80,7 +71,7 @@ public:
explicit RasterizerCache(VideoCore::RasterizerInterface& rasterizer) : rasterizer{rasterizer} {}
/// Write any cached resources overlapping the specified region back to memory
- void FlushRegion(CacheAddr addr, std::size_t size) {
+ void FlushRegion(VAddr addr, std::size_t size) {
std::lock_guard lock{mutex};
const auto& objects{GetSortedObjectsFromRegion(addr, size)};
@@ -90,7 +81,7 @@ public:
}
/// Mark the specified region as being invalidated
- void InvalidateRegion(CacheAddr addr, u64 size) {
+ void InvalidateRegion(VAddr addr, u64 size) {
std::lock_guard lock{mutex};
const auto& objects{GetSortedObjectsFromRegion(addr, size)};
@@ -114,27 +105,20 @@ public:
protected:
/// Tries to get an object from the cache with the specified cache address
- T TryGet(CacheAddr addr) const {
+ T TryGet(VAddr addr) const {
const auto iter = map_cache.find(addr);
if (iter != map_cache.end())
return iter->second;
return nullptr;
}
- T TryGet(const void* addr) const {
- const auto iter = map_cache.find(ToCacheAddr(addr));
- if (iter != map_cache.end())
- return iter->second;
- return nullptr;
- }
-
/// Register an object into the cache
virtual void Register(const T& object) {
std::lock_guard lock{mutex};
object->SetIsRegistered(true);
interval_cache.add({GetInterval(object), ObjectSet{object}});
- map_cache.insert({object->GetCacheAddr(), object});
+ map_cache.insert({object->GetCpuAddr(), object});
rasterizer.UpdatePagesCachedCount(object->GetCpuAddr(), object->GetSizeInBytes(), 1);
}
@@ -144,7 +128,7 @@ protected:
object->SetIsRegistered(false);
rasterizer.UpdatePagesCachedCount(object->GetCpuAddr(), object->GetSizeInBytes(), -1);
- const CacheAddr addr = object->GetCacheAddr();
+ const VAddr addr = object->GetCpuAddr();
interval_cache.subtract({GetInterval(object), ObjectSet{object}});
map_cache.erase(addr);
}
@@ -173,7 +157,7 @@ protected:
private:
/// Returns a list of cached objects from the specified memory region, ordered by access time
- std::vector<T> GetSortedObjectsFromRegion(CacheAddr addr, u64 size) {
+ std::vector<T> GetSortedObjectsFromRegion(VAddr addr, u64 size) {
if (size == 0) {
return {};
}
@@ -197,13 +181,13 @@ private:
}
using ObjectSet = std::set<T>;
- using ObjectCache = std::unordered_map<CacheAddr, T>;
- using IntervalCache = boost::icl::interval_map<CacheAddr, ObjectSet>;
+ using ObjectCache = std::unordered_map<VAddr, T>;
+ using IntervalCache = boost::icl::interval_map<VAddr, ObjectSet>;
using ObjectInterval = typename IntervalCache::interval_type;
static auto GetInterval(const T& object) {
- return ObjectInterval::right_open(object->GetCacheAddr(),
- object->GetCacheAddr() + object->GetSizeInBytes());
+ return ObjectInterval::right_open(object->GetCpuAddr(),
+ object->GetCpuAddr() + object->GetSizeInBytes());
}
ObjectCache map_cache;
diff --git a/src/video_core/rasterizer_interface.h b/src/video_core/rasterizer_interface.h
index 1a68e3caa..8ae5b9c4e 100644
--- a/src/video_core/rasterizer_interface.h
+++ b/src/video_core/rasterizer_interface.h
@@ -53,14 +53,14 @@ public:
virtual void FlushAll() = 0;
/// Notify rasterizer that any caches of the specified region should be flushed to Switch memory
- virtual void FlushRegion(CacheAddr addr, u64 size) = 0;
+ virtual void FlushRegion(VAddr addr, u64 size) = 0;
/// Notify rasterizer that any caches of the specified region should be invalidated
- virtual void InvalidateRegion(CacheAddr addr, u64 size) = 0;
+ virtual void InvalidateRegion(VAddr addr, u64 size) = 0;
/// Notify rasterizer that any caches of the specified region should be flushed to Switch memory
/// and invalidated
- virtual void FlushAndInvalidateRegion(CacheAddr addr, u64 size) = 0;
+ virtual void FlushAndInvalidateRegion(VAddr addr, u64 size) = 0;
/// Notify the rasterizer to send all written commands to the host GPU.
virtual void FlushCommands() = 0;
diff --git a/src/video_core/renderer_base.h b/src/video_core/renderer_base.h
index 5ec99a126..1d85219b6 100644
--- a/src/video_core/renderer_base.h
+++ b/src/video_core/renderer_base.h
@@ -46,7 +46,8 @@ public:
/// Draws the latest frame to the window waiting timeout_ms for a frame to arrive (Renderer
/// specific implementation)
- virtual void TryPresent(int timeout_ms) = 0;
+ /// Returns true if a frame was drawn
+ virtual bool TryPresent(int timeout_ms) = 0;
// Getter/setter functions:
// ------------------------
diff --git a/src/video_core/renderer_opengl/gl_buffer_cache.cpp b/src/video_core/renderer_opengl/gl_buffer_cache.cpp
index 0375fca17..4eb37a96c 100644
--- a/src/video_core/renderer_opengl/gl_buffer_cache.cpp
+++ b/src/video_core/renderer_opengl/gl_buffer_cache.cpp
@@ -21,8 +21,8 @@ using Maxwell = Tegra::Engines::Maxwell3D::Regs;
MICROPROFILE_DEFINE(OpenGL_Buffer_Download, "OpenGL", "Buffer Download", MP_RGB(192, 192, 128));
-CachedBufferBlock::CachedBufferBlock(CacheAddr cache_addr, const std::size_t size)
- : VideoCommon::BufferBlock{cache_addr, size} {
+CachedBufferBlock::CachedBufferBlock(VAddr cpu_addr, const std::size_t size)
+ : VideoCommon::BufferBlock{cpu_addr, size} {
gl_buffer.Create();
glNamedBufferData(gl_buffer.handle, static_cast<GLsizeiptr>(size), nullptr, GL_DYNAMIC_DRAW);
}
@@ -47,8 +47,8 @@ OGLBufferCache::~OGLBufferCache() {
glDeleteBuffers(static_cast<GLsizei>(std::size(cbufs)), std::data(cbufs));
}
-Buffer OGLBufferCache::CreateBlock(CacheAddr cache_addr, std::size_t size) {
- return std::make_shared<CachedBufferBlock>(cache_addr, size);
+Buffer OGLBufferCache::CreateBlock(VAddr cpu_addr, std::size_t size) {
+ return std::make_shared<CachedBufferBlock>(cpu_addr, size);
}
void OGLBufferCache::WriteBarrier() {
diff --git a/src/video_core/renderer_opengl/gl_buffer_cache.h b/src/video_core/renderer_opengl/gl_buffer_cache.h
index 8c7145443..d94a11252 100644
--- a/src/video_core/renderer_opengl/gl_buffer_cache.h
+++ b/src/video_core/renderer_opengl/gl_buffer_cache.h
@@ -31,7 +31,7 @@ using GenericBufferCache = VideoCommon::BufferCache<Buffer, GLuint, OGLStreamBuf
class CachedBufferBlock : public VideoCommon::BufferBlock {
public:
- explicit CachedBufferBlock(CacheAddr cache_addr, const std::size_t size);
+ explicit CachedBufferBlock(VAddr cpu_addr, const std::size_t size);
~CachedBufferBlock();
const GLuint* GetHandle() const {
@@ -55,7 +55,7 @@ public:
}
protected:
- Buffer CreateBlock(CacheAddr cache_addr, std::size_t size) override;
+ Buffer CreateBlock(VAddr cpu_addr, std::size_t size) override;
void WriteBarrier() override;
diff --git a/src/video_core/renderer_opengl/gl_device.cpp b/src/video_core/renderer_opengl/gl_device.cpp
index 1a2e2a9f7..c286502ba 100644
--- a/src/video_core/renderer_opengl/gl_device.cpp
+++ b/src/video_core/renderer_opengl/gl_device.cpp
@@ -131,6 +131,31 @@ std::array<Device::BaseBindings, Tegra::Engines::MaxShaderTypes> BuildBaseBindin
return bindings;
}
+bool IsASTCSupported() {
+ static constexpr std::array formats = {
+ GL_COMPRESSED_RGBA_ASTC_4x4_KHR, GL_COMPRESSED_RGBA_ASTC_5x4_KHR,
+ GL_COMPRESSED_RGBA_ASTC_5x5_KHR, GL_COMPRESSED_RGBA_ASTC_6x5_KHR,
+ GL_COMPRESSED_RGBA_ASTC_6x6_KHR, GL_COMPRESSED_RGBA_ASTC_8x5_KHR,
+ GL_COMPRESSED_RGBA_ASTC_8x6_KHR, GL_COMPRESSED_RGBA_ASTC_8x8_KHR,
+ GL_COMPRESSED_RGBA_ASTC_10x5_KHR, GL_COMPRESSED_RGBA_ASTC_10x6_KHR,
+ GL_COMPRESSED_RGBA_ASTC_10x8_KHR, GL_COMPRESSED_RGBA_ASTC_10x10_KHR,
+ GL_COMPRESSED_RGBA_ASTC_12x10_KHR, GL_COMPRESSED_RGBA_ASTC_12x12_KHR,
+ GL_COMPRESSED_SRGB8_ALPHA8_ASTC_4x4_KHR, GL_COMPRESSED_SRGB8_ALPHA8_ASTC_5x4_KHR,
+ GL_COMPRESSED_SRGB8_ALPHA8_ASTC_5x5_KHR, GL_COMPRESSED_SRGB8_ALPHA8_ASTC_6x5_KHR,
+ GL_COMPRESSED_SRGB8_ALPHA8_ASTC_6x6_KHR, GL_COMPRESSED_SRGB8_ALPHA8_ASTC_8x5_KHR,
+ GL_COMPRESSED_SRGB8_ALPHA8_ASTC_8x6_KHR, GL_COMPRESSED_SRGB8_ALPHA8_ASTC_8x8_KHR,
+ GL_COMPRESSED_SRGB8_ALPHA8_ASTC_10x5_KHR, GL_COMPRESSED_SRGB8_ALPHA8_ASTC_10x6_KHR,
+ GL_COMPRESSED_SRGB8_ALPHA8_ASTC_10x8_KHR, GL_COMPRESSED_SRGB8_ALPHA8_ASTC_10x10_KHR,
+ GL_COMPRESSED_SRGB8_ALPHA8_ASTC_12x10_KHR, GL_COMPRESSED_SRGB8_ALPHA8_ASTC_12x12_KHR,
+ };
+ return std::find_if_not(formats.begin(), formats.end(), [](GLenum format) {
+ GLint supported;
+ glGetInternalformativ(GL_TEXTURE_2D, format, GL_INTERNALFORMAT_SUPPORTED, 1,
+ &supported);
+ return supported == GL_TRUE;
+ }) == formats.end();
+}
+
} // Anonymous namespace
Device::Device() : base_bindings{BuildBaseBindings()} {
@@ -152,6 +177,7 @@ Device::Device() : base_bindings{BuildBaseBindings()} {
has_shader_ballot = GLAD_GL_ARB_shader_ballot;
has_vertex_viewport_layer = GLAD_GL_ARB_shader_viewport_layer_array;
has_image_load_formatted = HasExtension(extensions, "GL_EXT_shader_image_load_formatted");
+ has_astc = IsASTCSupported();
has_variable_aoffi = TestVariableAoffi();
has_component_indexing_bug = is_amd;
has_precise_bug = TestPreciseBug();
diff --git a/src/video_core/renderer_opengl/gl_device.h b/src/video_core/renderer_opengl/gl_device.h
index d73b099d0..a55050cb5 100644
--- a/src/video_core/renderer_opengl/gl_device.h
+++ b/src/video_core/renderer_opengl/gl_device.h
@@ -64,6 +64,10 @@ public:
return has_image_load_formatted;
}
+ bool HasASTC() const {
+ return has_astc;
+ }
+
bool HasVariableAoffi() const {
return has_variable_aoffi;
}
@@ -97,6 +101,7 @@ private:
bool has_shader_ballot{};
bool has_vertex_viewport_layer{};
bool has_image_load_formatted{};
+ bool has_astc{};
bool has_variable_aoffi{};
bool has_component_indexing_bug{};
bool has_precise_bug{};
diff --git a/src/video_core/renderer_opengl/gl_rasterizer.cpp b/src/video_core/renderer_opengl/gl_rasterizer.cpp
index 826eee7df..368f399df 100644
--- a/src/video_core/renderer_opengl/gl_rasterizer.cpp
+++ b/src/video_core/renderer_opengl/gl_rasterizer.cpp
@@ -386,11 +386,14 @@ void RasterizerOpenGL::ConfigureClearFramebuffer(bool using_color_fb, bool using
texture_cache.GuardRenderTargets(true);
View color_surface;
if (using_color_fb) {
- color_surface = texture_cache.GetColorBufferSurface(regs.clear_buffers.RT, false);
+ const std::size_t index = regs.clear_buffers.RT;
+ color_surface = texture_cache.GetColorBufferSurface(index, true);
+ texture_cache.MarkColorBufferInUse(index);
}
View depth_surface;
if (using_depth_fb || using_stencil_fb) {
- depth_surface = texture_cache.GetDepthBufferSurface(false);
+ depth_surface = texture_cache.GetDepthBufferSurface(true);
+ texture_cache.MarkDepthBufferInUse();
}
texture_cache.GuardRenderTargets(false);
@@ -444,6 +447,7 @@ void RasterizerOpenGL::Clear() {
}
SyncRasterizeEnable();
+ SyncStencilTestState();
if (regs.clear_flags.scissor) {
SyncScissorTest();
@@ -652,9 +656,9 @@ void RasterizerOpenGL::Query(GPUVAddr gpu_addr, VideoCore::QueryType type,
void RasterizerOpenGL::FlushAll() {}
-void RasterizerOpenGL::FlushRegion(CacheAddr addr, u64 size) {
+void RasterizerOpenGL::FlushRegion(VAddr addr, u64 size) {
MICROPROFILE_SCOPE(OpenGL_CacheManagement);
- if (!addr || !size) {
+ if (addr == 0 || size == 0) {
return;
}
texture_cache.FlushRegion(addr, size);
@@ -662,9 +666,9 @@ void RasterizerOpenGL::FlushRegion(CacheAddr addr, u64 size) {
query_cache.FlushRegion(addr, size);
}
-void RasterizerOpenGL::InvalidateRegion(CacheAddr addr, u64 size) {
+void RasterizerOpenGL::InvalidateRegion(VAddr addr, u64 size) {
MICROPROFILE_SCOPE(OpenGL_CacheManagement);
- if (!addr || !size) {
+ if (addr == 0 || size == 0) {
return;
}
texture_cache.InvalidateRegion(addr, size);
@@ -673,7 +677,7 @@ void RasterizerOpenGL::InvalidateRegion(CacheAddr addr, u64 size) {
query_cache.InvalidateRegion(addr, size);
}
-void RasterizerOpenGL::FlushAndInvalidateRegion(CacheAddr addr, u64 size) {
+void RasterizerOpenGL::FlushAndInvalidateRegion(VAddr addr, u64 size) {
if (Settings::values.use_accurate_gpu_emulation) {
FlushRegion(addr, size);
}
@@ -712,8 +716,7 @@ bool RasterizerOpenGL::AccelerateDisplay(const Tegra::FramebufferConfig& config,
MICROPROFILE_SCOPE(OpenGL_CacheManagement);
- const auto surface{
- texture_cache.TryFindFramebufferSurface(system.Memory().GetPointer(framebuffer_addr))};
+ const auto surface{texture_cache.TryFindFramebufferSurface(framebuffer_addr)};
if (!surface) {
return {};
}
@@ -1052,12 +1055,8 @@ void RasterizerOpenGL::SyncStencilTestState() {
flags[Dirty::StencilTest] = false;
const auto& regs = gpu.regs;
- if (!regs.stencil_enable) {
- glDisable(GL_STENCIL_TEST);
- return;
- }
+ oglEnable(GL_STENCIL_TEST, regs.stencil_enable);
- glEnable(GL_STENCIL_TEST);
glStencilFuncSeparate(GL_FRONT, MaxwellToGL::ComparisonOp(regs.stencil_front_func_func),
regs.stencil_front_func_ref, regs.stencil_front_func_mask);
glStencilOpSeparate(GL_FRONT, MaxwellToGL::StencilOp(regs.stencil_front_op_fail),
diff --git a/src/video_core/renderer_opengl/gl_rasterizer.h b/src/video_core/renderer_opengl/gl_rasterizer.h
index 2d3be2437..212dad852 100644
--- a/src/video_core/renderer_opengl/gl_rasterizer.h
+++ b/src/video_core/renderer_opengl/gl_rasterizer.h
@@ -65,9 +65,9 @@ public:
void ResetCounter(VideoCore::QueryType type) override;
void Query(GPUVAddr gpu_addr, VideoCore::QueryType type, std::optional<u64> timestamp) override;
void FlushAll() override;
- void FlushRegion(CacheAddr addr, u64 size) override;
- void InvalidateRegion(CacheAddr addr, u64 size) override;
- void FlushAndInvalidateRegion(CacheAddr addr, u64 size) override;
+ void FlushRegion(VAddr addr, u64 size) override;
+ void InvalidateRegion(VAddr addr, u64 size) override;
+ void FlushAndInvalidateRegion(VAddr addr, u64 size) override;
void FlushCommands() override;
void TickFrame() override;
bool AccelerateSurfaceCopy(const Tegra::Engines::Fermi2D::Regs::Surface& src,
diff --git a/src/video_core/renderer_opengl/gl_shader_cache.cpp b/src/video_core/renderer_opengl/gl_shader_cache.cpp
index e3d31c3eb..6d2ff20f9 100644
--- a/src/video_core/renderer_opengl/gl_shader_cache.cpp
+++ b/src/video_core/renderer_opengl/gl_shader_cache.cpp
@@ -214,11 +214,11 @@ std::unordered_set<GLenum> GetSupportedFormats() {
} // Anonymous namespace
-CachedShader::CachedShader(const u8* host_ptr, VAddr cpu_addr, std::size_t size_in_bytes,
+CachedShader::CachedShader(VAddr cpu_addr, std::size_t size_in_bytes,
std::shared_ptr<VideoCommon::Shader::Registry> registry,
ShaderEntries entries, std::shared_ptr<OGLProgram> program)
- : RasterizerCacheObject{host_ptr}, registry{std::move(registry)}, entries{std::move(entries)},
- cpu_addr{cpu_addr}, size_in_bytes{size_in_bytes}, program{std::move(program)} {}
+ : RasterizerCacheObject{cpu_addr}, registry{std::move(registry)}, entries{std::move(entries)},
+ size_in_bytes{size_in_bytes}, program{std::move(program)} {}
CachedShader::~CachedShader() = default;
@@ -254,9 +254,8 @@ Shader CachedShader::CreateStageFromMemory(const ShaderParameters& params,
entry.bindless_samplers = registry->GetBindlessSamplers();
params.disk_cache.SaveEntry(std::move(entry));
- return std::shared_ptr<CachedShader>(new CachedShader(params.host_ptr, params.cpu_addr,
- size_in_bytes, std::move(registry),
- MakeEntries(ir), std::move(program)));
+ return std::shared_ptr<CachedShader>(new CachedShader(
+ params.cpu_addr, size_in_bytes, std::move(registry), MakeEntries(ir), std::move(program)));
}
Shader CachedShader::CreateKernelFromMemory(const ShaderParameters& params, ProgramCode code) {
@@ -279,17 +278,16 @@ Shader CachedShader::CreateKernelFromMemory(const ShaderParameters& params, Prog
entry.bindless_samplers = registry->GetBindlessSamplers();
params.disk_cache.SaveEntry(std::move(entry));
- return std::shared_ptr<CachedShader>(new CachedShader(params.host_ptr, params.cpu_addr,
- size_in_bytes, std::move(registry),
- MakeEntries(ir), std::move(program)));
+ return std::shared_ptr<CachedShader>(new CachedShader(
+ params.cpu_addr, size_in_bytes, std::move(registry), MakeEntries(ir), std::move(program)));
}
Shader CachedShader::CreateFromCache(const ShaderParameters& params,
const PrecompiledShader& precompiled_shader,
std::size_t size_in_bytes) {
- return std::shared_ptr<CachedShader>(new CachedShader(
- params.host_ptr, params.cpu_addr, size_in_bytes, precompiled_shader.registry,
- precompiled_shader.entries, precompiled_shader.program));
+ return std::shared_ptr<CachedShader>(
+ new CachedShader(params.cpu_addr, size_in_bytes, precompiled_shader.registry,
+ precompiled_shader.entries, precompiled_shader.program));
}
ShaderCacheOpenGL::ShaderCacheOpenGL(RasterizerOpenGL& rasterizer, Core::System& system,
@@ -327,8 +325,7 @@ void ShaderCacheOpenGL::LoadDiskCache(const std::atomic_bool& stop_loading,
const auto worker = [&](Core::Frontend::GraphicsContext* context, std::size_t begin,
std::size_t end) {
- context->MakeCurrent();
- SCOPE_EXIT({ return context->DoneCurrent(); });
+ const auto scope = context->Acquire();
for (std::size_t i = begin; i < end; ++i) {
if (stop_loading) {
@@ -450,12 +447,14 @@ Shader ShaderCacheOpenGL::GetStageProgram(Maxwell::ShaderProgram program) {
const GPUVAddr address{GetShaderAddress(system, program)};
// Look up shader in the cache based on address
- const auto host_ptr{memory_manager.GetPointer(address)};
- Shader shader{TryGet(host_ptr)};
+ const auto cpu_addr{memory_manager.GpuToCpuAddress(address)};
+ Shader shader{cpu_addr ? TryGet(*cpu_addr) : nullptr};
if (shader) {
return last_shaders[static_cast<std::size_t>(program)] = shader;
}
+ const auto host_ptr{memory_manager.GetPointer(address)};
+
// No shader found - create a new one
ProgramCode code{GetShaderCode(memory_manager, address, host_ptr)};
ProgramCode code_b;
@@ -466,9 +465,9 @@ Shader ShaderCacheOpenGL::GetStageProgram(Maxwell::ShaderProgram program) {
const auto unique_identifier = GetUniqueIdentifier(
GetShaderType(program), program == Maxwell::ShaderProgram::VertexA, code, code_b);
- const auto cpu_addr{*memory_manager.GpuToCpuAddress(address)};
- const ShaderParameters params{system, disk_cache, device,
- cpu_addr, host_ptr, unique_identifier};
+
+ const ShaderParameters params{system, disk_cache, device,
+ *cpu_addr, host_ptr, unique_identifier};
const auto found = runtime_cache.find(unique_identifier);
if (found == runtime_cache.end()) {
@@ -485,18 +484,20 @@ Shader ShaderCacheOpenGL::GetStageProgram(Maxwell::ShaderProgram program) {
Shader ShaderCacheOpenGL::GetComputeKernel(GPUVAddr code_addr) {
auto& memory_manager{system.GPU().MemoryManager()};
- const auto host_ptr{memory_manager.GetPointer(code_addr)};
- auto kernel = TryGet(host_ptr);
+ const auto cpu_addr{memory_manager.GpuToCpuAddress(code_addr)};
+
+ auto kernel = cpu_addr ? TryGet(*cpu_addr) : nullptr;
if (kernel) {
return kernel;
}
+ const auto host_ptr{memory_manager.GetPointer(code_addr)};
// No kernel found, create a new one
auto code{GetShaderCode(memory_manager, code_addr, host_ptr)};
const auto unique_identifier{GetUniqueIdentifier(ShaderType::Compute, false, code)};
- const auto cpu_addr{*memory_manager.GpuToCpuAddress(code_addr)};
- const ShaderParameters params{system, disk_cache, device,
- cpu_addr, host_ptr, unique_identifier};
+
+ const ShaderParameters params{system, disk_cache, device,
+ *cpu_addr, host_ptr, unique_identifier};
const auto found = runtime_cache.find(unique_identifier);
if (found == runtime_cache.end()) {
diff --git a/src/video_core/renderer_opengl/gl_shader_cache.h b/src/video_core/renderer_opengl/gl_shader_cache.h
index 4935019fc..c836df5bd 100644
--- a/src/video_core/renderer_opengl/gl_shader_cache.h
+++ b/src/video_core/renderer_opengl/gl_shader_cache.h
@@ -65,11 +65,6 @@ public:
/// Gets the GL program handle for the shader
GLuint GetHandle() const;
- /// Returns the guest CPU address of the shader
- VAddr GetCpuAddr() const override {
- return cpu_addr;
- }
-
/// Returns the size in bytes of the shader
std::size_t GetSizeInBytes() const override {
return size_in_bytes;
@@ -90,13 +85,12 @@ public:
std::size_t size_in_bytes);
private:
- explicit CachedShader(const u8* host_ptr, VAddr cpu_addr, std::size_t size_in_bytes,
+ explicit CachedShader(VAddr cpu_addr, std::size_t size_in_bytes,
std::shared_ptr<VideoCommon::Shader::Registry> registry,
ShaderEntries entries, std::shared_ptr<OGLProgram> program);
std::shared_ptr<VideoCommon::Shader::Registry> registry;
ShaderEntries entries;
- VAddr cpu_addr = 0;
std::size_t size_in_bytes = 0;
std::shared_ptr<OGLProgram> program;
};
diff --git a/src/video_core/renderer_opengl/gl_shader_decompiler.cpp b/src/video_core/renderer_opengl/gl_shader_decompiler.cpp
index 8aa4a7ac9..160ae4340 100644
--- a/src/video_core/renderer_opengl/gl_shader_decompiler.cpp
+++ b/src/video_core/renderer_opengl/gl_shader_decompiler.cpp
@@ -31,11 +31,11 @@ namespace {
using Tegra::Engines::ShaderType;
using Tegra::Shader::Attribute;
-using Tegra::Shader::AttributeUse;
using Tegra::Shader::Header;
using Tegra::Shader::IpaInterpMode;
using Tegra::Shader::IpaMode;
using Tegra::Shader::IpaSampleMode;
+using Tegra::Shader::PixelImap;
using Tegra::Shader::Register;
using VideoCommon::Shader::BuildTransformFeedback;
using VideoCommon::Shader::Registry;
@@ -702,20 +702,19 @@ private:
code.AddNewLine();
}
- std::string GetInputFlags(AttributeUse attribute) {
+ const char* GetInputFlags(PixelImap attribute) {
switch (attribute) {
- case AttributeUse::Perspective:
- // Default, Smooth
- return {};
- case AttributeUse::Constant:
- return "flat ";
- case AttributeUse::ScreenLinear:
- return "noperspective ";
- default:
- case AttributeUse::Unused:
- UNIMPLEMENTED_MSG("Unknown attribute usage index={}", static_cast<u32>(attribute));
- return {};
+ case PixelImap::Perspective:
+ return "smooth";
+ case PixelImap::Constant:
+ return "flat";
+ case PixelImap::ScreenLinear:
+ return "noperspective";
+ case PixelImap::Unused:
+ break;
}
+ UNIMPLEMENTED_MSG("Unknown attribute usage index={}", static_cast<int>(attribute));
+ return {};
}
void DeclareInputAttributes() {
@@ -749,8 +748,8 @@ private:
std::string suffix;
if (stage == ShaderType::Fragment) {
- const auto input_mode{header.ps.GetAttributeUse(location)};
- if (skip_unused && input_mode == AttributeUse::Unused) {
+ const auto input_mode{header.ps.GetPixelImap(location)};
+ if (input_mode == PixelImap::Unused) {
return;
}
suffix = GetInputFlags(input_mode);
@@ -927,7 +926,7 @@ private:
const u32 address{generic_base + index * generic_stride + element * element_stride};
const bool declared = stage != ShaderType::Fragment ||
- header.ps.GetAttributeUse(index) != AttributeUse::Unused;
+ header.ps.GetPixelImap(index) != PixelImap::Unused;
const std::string value =
declared ? ReadAttribute(attribute, element).AsFloat() : "0.0f";
code.AddLine("case 0x{:X}U: return {};", address, value);
@@ -1142,8 +1141,7 @@ private:
GetSwizzle(element)),
Type::Float};
case ShaderType::Fragment:
- return {element == 3 ? "1.0f" : ("gl_FragCoord"s + GetSwizzle(element)),
- Type::Float};
+ return {"gl_FragCoord"s + GetSwizzle(element), Type::Float};
default:
UNREACHABLE();
}
@@ -2114,6 +2112,10 @@ private:
template <const std::string_view& opname, Type type>
Expression Atomic(Operation operation) {
+ if ((opname == Func::Min || opname == Func::Max) && type == Type::Int) {
+ UNIMPLEMENTED_MSG("Unimplemented Min & Max for atomic operations");
+ return {};
+ }
return {fmt::format("atomic{}({}, {})", opname, Visit(operation[0]).GetCode(),
Visit(operation[1]).As(type)),
type};
@@ -2307,6 +2309,8 @@ private:
~Func() = delete;
static constexpr std::string_view Add = "Add";
+ static constexpr std::string_view Min = "Min";
+ static constexpr std::string_view Max = "Max";
static constexpr std::string_view And = "And";
static constexpr std::string_view Or = "Or";
static constexpr std::string_view Xor = "Xor";
@@ -2457,7 +2461,21 @@ private:
&GLSLDecompiler::AtomicImage<Func::Xor>,
&GLSLDecompiler::AtomicImage<Func::Exchange>,
+ &GLSLDecompiler::Atomic<Func::Exchange, Type::Uint>,
&GLSLDecompiler::Atomic<Func::Add, Type::Uint>,
+ &GLSLDecompiler::Atomic<Func::Min, Type::Uint>,
+ &GLSLDecompiler::Atomic<Func::Max, Type::Uint>,
+ &GLSLDecompiler::Atomic<Func::And, Type::Uint>,
+ &GLSLDecompiler::Atomic<Func::Or, Type::Uint>,
+ &GLSLDecompiler::Atomic<Func::Xor, Type::Uint>,
+
+ &GLSLDecompiler::Atomic<Func::Exchange, Type::Int>,
+ &GLSLDecompiler::Atomic<Func::Add, Type::Int>,
+ &GLSLDecompiler::Atomic<Func::Min, Type::Int>,
+ &GLSLDecompiler::Atomic<Func::Max, Type::Int>,
+ &GLSLDecompiler::Atomic<Func::And, Type::Int>,
+ &GLSLDecompiler::Atomic<Func::Or, Type::Int>,
+ &GLSLDecompiler::Atomic<Func::Xor, Type::Int>,
&GLSLDecompiler::Branch,
&GLSLDecompiler::BranchIndirect,
diff --git a/src/video_core/renderer_opengl/gl_texture_cache.cpp b/src/video_core/renderer_opengl/gl_texture_cache.cpp
index f424e3000..36590a6d0 100644
--- a/src/video_core/renderer_opengl/gl_texture_cache.cpp
+++ b/src/video_core/renderer_opengl/gl_texture_cache.cpp
@@ -24,7 +24,6 @@ using Tegra::Texture::SwizzleSource;
using VideoCore::MortonSwizzleMode;
using VideoCore::Surface::PixelFormat;
-using VideoCore::Surface::SurfaceCompression;
using VideoCore::Surface::SurfaceTarget;
using VideoCore::Surface::SurfaceType;
@@ -37,102 +36,100 @@ namespace {
struct FormatTuple {
GLint internal_format;
- GLenum format;
- GLenum type;
- bool compressed;
+ GLenum format = GL_NONE;
+ GLenum type = GL_NONE;
};
constexpr std::array<FormatTuple, VideoCore::Surface::MaxPixelFormat> tex_format_tuples = {{
- {GL_RGBA8, GL_RGBA, GL_UNSIGNED_INT_8_8_8_8_REV, false}, // ABGR8U
- {GL_RGBA8_SNORM, GL_RGBA, GL_BYTE, false}, // ABGR8S
- {GL_RGBA8UI, GL_RGBA_INTEGER, GL_UNSIGNED_BYTE, false}, // ABGR8UI
- {GL_RGB565, GL_RGB, GL_UNSIGNED_SHORT_5_6_5_REV, false}, // B5G6R5U
- {GL_RGB10_A2, GL_RGBA, GL_UNSIGNED_INT_2_10_10_10_REV, false}, // A2B10G10R10U
- {GL_RGB5_A1, GL_RGBA, GL_UNSIGNED_SHORT_1_5_5_5_REV, false}, // A1B5G5R5U
- {GL_R8, GL_RED, GL_UNSIGNED_BYTE, false}, // R8U
- {GL_R8UI, GL_RED_INTEGER, GL_UNSIGNED_BYTE, false}, // R8UI
- {GL_RGBA16F, GL_RGBA, GL_HALF_FLOAT, false}, // RGBA16F
- {GL_RGBA16, GL_RGBA, GL_UNSIGNED_SHORT, false}, // RGBA16U
- {GL_RGBA16_SNORM, GL_RGBA, GL_SHORT, false}, // RGBA16S
- {GL_RGBA16UI, GL_RGBA_INTEGER, GL_UNSIGNED_SHORT, false}, // RGBA16UI
- {GL_R11F_G11F_B10F, GL_RGB, GL_UNSIGNED_INT_10F_11F_11F_REV, false}, // R11FG11FB10F
- {GL_RGBA32UI, GL_RGBA_INTEGER, GL_UNSIGNED_INT, false}, // RGBA32UI
- {GL_COMPRESSED_RGBA_S3TC_DXT1_EXT, GL_RGBA, GL_UNSIGNED_INT_8_8_8_8, true}, // DXT1
- {GL_COMPRESSED_RGBA_S3TC_DXT3_EXT, GL_RGBA, GL_UNSIGNED_INT_8_8_8_8, true}, // DXT23
- {GL_COMPRESSED_RGBA_S3TC_DXT5_EXT, GL_RGBA, GL_UNSIGNED_INT_8_8_8_8, true}, // DXT45
- {GL_COMPRESSED_RED_RGTC1, GL_RED, GL_UNSIGNED_INT_8_8_8_8, true}, // DXN1
- {GL_COMPRESSED_RG_RGTC2, GL_RG, GL_UNSIGNED_INT_8_8_8_8, true}, // DXN2UNORM
- {GL_COMPRESSED_SIGNED_RG_RGTC2, GL_RG, GL_INT, true}, // DXN2SNORM
- {GL_COMPRESSED_RGBA_BPTC_UNORM, GL_RGBA, GL_UNSIGNED_INT_8_8_8_8, true}, // BC7U
- {GL_COMPRESSED_RGB_BPTC_UNSIGNED_FLOAT, GL_RGB, GL_UNSIGNED_INT_8_8_8_8, true}, // BC6H_UF16
- {GL_COMPRESSED_RGB_BPTC_SIGNED_FLOAT, GL_RGB, GL_UNSIGNED_INT_8_8_8_8, true}, // BC6H_SF16
- {GL_RGBA8, GL_RGBA, GL_UNSIGNED_BYTE, false}, // ASTC_2D_4X4
- {GL_RGBA8, GL_BGRA, GL_UNSIGNED_BYTE, false}, // BGRA8
- {GL_RGBA32F, GL_RGBA, GL_FLOAT, false}, // RGBA32F
- {GL_RG32F, GL_RG, GL_FLOAT, false}, // RG32F
- {GL_R32F, GL_RED, GL_FLOAT, false}, // R32F
- {GL_R16F, GL_RED, GL_HALF_FLOAT, false}, // R16F
- {GL_R16, GL_RED, GL_UNSIGNED_SHORT, false}, // R16U
- {GL_R16_SNORM, GL_RED, GL_SHORT, false}, // R16S
- {GL_R16UI, GL_RED_INTEGER, GL_UNSIGNED_SHORT, false}, // R16UI
- {GL_R16I, GL_RED_INTEGER, GL_SHORT, false}, // R16I
- {GL_RG16, GL_RG, GL_UNSIGNED_SHORT, false}, // RG16
- {GL_RG16F, GL_RG, GL_HALF_FLOAT, false}, // RG16F
- {GL_RG16UI, GL_RG_INTEGER, GL_UNSIGNED_SHORT, false}, // RG16UI
- {GL_RG16I, GL_RG_INTEGER, GL_SHORT, false}, // RG16I
- {GL_RG16_SNORM, GL_RG, GL_SHORT, false}, // RG16S
- {GL_RGB32F, GL_RGB, GL_FLOAT, false}, // RGB32F
- {GL_SRGB8_ALPHA8, GL_RGBA, GL_UNSIGNED_INT_8_8_8_8_REV, false}, // RGBA8_SRGB
- {GL_RG8, GL_RG, GL_UNSIGNED_BYTE, false}, // RG8U
- {GL_RG8_SNORM, GL_RG, GL_BYTE, false}, // RG8S
- {GL_RG32UI, GL_RG_INTEGER, GL_UNSIGNED_INT, false}, // RG32UI
- {GL_RGB16F, GL_RGBA, GL_HALF_FLOAT, false}, // RGBX16F
- {GL_R32UI, GL_RED_INTEGER, GL_UNSIGNED_INT, false}, // R32UI
- {GL_R32I, GL_RED_INTEGER, GL_INT, false}, // R32I
- {GL_RGBA8, GL_RGBA, GL_UNSIGNED_BYTE, false}, // ASTC_2D_8X8
- {GL_RGBA8, GL_RGBA, GL_UNSIGNED_BYTE, false}, // ASTC_2D_8X5
- {GL_RGBA8, GL_RGBA, GL_UNSIGNED_BYTE, false}, // ASTC_2D_5X4
- {GL_SRGB8_ALPHA8, GL_BGRA, GL_UNSIGNED_BYTE, false}, // BGRA8
+ {GL_RGBA8, GL_RGBA, GL_UNSIGNED_INT_8_8_8_8_REV}, // ABGR8U
+ {GL_RGBA8_SNORM, GL_RGBA, GL_BYTE}, // ABGR8S
+ {GL_RGBA8UI, GL_RGBA_INTEGER, GL_UNSIGNED_BYTE}, // ABGR8UI
+ {GL_RGB565, GL_RGB, GL_UNSIGNED_SHORT_5_6_5_REV}, // B5G6R5U
+ {GL_RGB10_A2, GL_RGBA, GL_UNSIGNED_INT_2_10_10_10_REV}, // A2B10G10R10U
+ {GL_RGB5_A1, GL_RGBA, GL_UNSIGNED_SHORT_1_5_5_5_REV}, // A1B5G5R5U
+ {GL_R8, GL_RED, GL_UNSIGNED_BYTE}, // R8U
+ {GL_R8UI, GL_RED_INTEGER, GL_UNSIGNED_BYTE}, // R8UI
+ {GL_RGBA16F, GL_RGBA, GL_HALF_FLOAT}, // RGBA16F
+ {GL_RGBA16, GL_RGBA, GL_UNSIGNED_SHORT}, // RGBA16U
+ {GL_RGBA16_SNORM, GL_RGBA, GL_SHORT}, // RGBA16S
+ {GL_RGBA16UI, GL_RGBA_INTEGER, GL_UNSIGNED_SHORT}, // RGBA16UI
+ {GL_R11F_G11F_B10F, GL_RGB, GL_UNSIGNED_INT_10F_11F_11F_REV}, // R11FG11FB10F
+ {GL_RGBA32UI, GL_RGBA_INTEGER, GL_UNSIGNED_INT}, // RGBA32UI
+ {GL_COMPRESSED_RGBA_S3TC_DXT1_EXT}, // DXT1
+ {GL_COMPRESSED_RGBA_S3TC_DXT3_EXT}, // DXT23
+ {GL_COMPRESSED_RGBA_S3TC_DXT5_EXT}, // DXT45
+ {GL_COMPRESSED_RED_RGTC1}, // DXN1
+ {GL_COMPRESSED_RG_RGTC2}, // DXN2UNORM
+ {GL_COMPRESSED_SIGNED_RG_RGTC2}, // DXN2SNORM
+ {GL_COMPRESSED_RGBA_BPTC_UNORM}, // BC7U
+ {GL_COMPRESSED_RGB_BPTC_UNSIGNED_FLOAT}, // BC6H_UF16
+ {GL_COMPRESSED_RGB_BPTC_SIGNED_FLOAT}, // BC6H_SF16
+ {GL_COMPRESSED_RGBA_ASTC_4x4_KHR}, // ASTC_2D_4X4
+ {GL_RGBA8, GL_BGRA, GL_UNSIGNED_BYTE}, // BGRA8
+ {GL_RGBA32F, GL_RGBA, GL_FLOAT}, // RGBA32F
+ {GL_RG32F, GL_RG, GL_FLOAT}, // RG32F
+ {GL_R32F, GL_RED, GL_FLOAT}, // R32F
+ {GL_R16F, GL_RED, GL_HALF_FLOAT}, // R16F
+ {GL_R16, GL_RED, GL_UNSIGNED_SHORT}, // R16U
+ {GL_R16_SNORM, GL_RED, GL_SHORT}, // R16S
+ {GL_R16UI, GL_RED_INTEGER, GL_UNSIGNED_SHORT}, // R16UI
+ {GL_R16I, GL_RED_INTEGER, GL_SHORT}, // R16I
+ {GL_RG16, GL_RG, GL_UNSIGNED_SHORT}, // RG16
+ {GL_RG16F, GL_RG, GL_HALF_FLOAT}, // RG16F
+ {GL_RG16UI, GL_RG_INTEGER, GL_UNSIGNED_SHORT}, // RG16UI
+ {GL_RG16I, GL_RG_INTEGER, GL_SHORT}, // RG16I
+ {GL_RG16_SNORM, GL_RG, GL_SHORT}, // RG16S
+ {GL_RGB32F, GL_RGB, GL_FLOAT}, // RGB32F
+ {GL_SRGB8_ALPHA8, GL_RGBA, GL_UNSIGNED_INT_8_8_8_8_REV}, // RGBA8_SRGB
+ {GL_RG8, GL_RG, GL_UNSIGNED_BYTE}, // RG8U
+ {GL_RG8_SNORM, GL_RG, GL_BYTE}, // RG8S
+ {GL_RG32UI, GL_RG_INTEGER, GL_UNSIGNED_INT}, // RG32UI
+ {GL_RGB16F, GL_RGBA, GL_HALF_FLOAT}, // RGBX16F
+ {GL_R32UI, GL_RED_INTEGER, GL_UNSIGNED_INT}, // R32UI
+ {GL_R32I, GL_RED_INTEGER, GL_INT}, // R32I
+ {GL_COMPRESSED_RGBA_ASTC_8x8_KHR}, // ASTC_2D_8X8
+ {GL_COMPRESSED_RGBA_ASTC_8x5_KHR}, // ASTC_2D_8X5
+ {GL_COMPRESSED_RGBA_ASTC_5x4_KHR}, // ASTC_2D_5X4
+ {GL_SRGB8_ALPHA8, GL_BGRA, GL_UNSIGNED_BYTE}, // BGRA8
// Compressed sRGB formats
- {GL_COMPRESSED_SRGB_ALPHA_S3TC_DXT1_EXT, GL_RGBA, GL_UNSIGNED_INT_8_8_8_8, true}, // DXT1_SRGB
- {GL_COMPRESSED_SRGB_ALPHA_S3TC_DXT3_EXT, GL_RGBA, GL_UNSIGNED_INT_8_8_8_8, true}, // DXT23_SRGB
- {GL_COMPRESSED_SRGB_ALPHA_S3TC_DXT5_EXT, GL_RGBA, GL_UNSIGNED_INT_8_8_8_8, true}, // DXT45_SRGB
- {GL_COMPRESSED_SRGB_ALPHA_BPTC_UNORM, GL_RGBA, GL_UNSIGNED_INT_8_8_8_8, true}, // BC7U_SRGB
- {GL_RGBA4, GL_RGBA, GL_UNSIGNED_SHORT_4_4_4_4_REV, false}, // R4G4B4A4U
- {GL_SRGB8_ALPHA8, GL_RGBA, GL_UNSIGNED_BYTE, false}, // ASTC_2D_4X4_SRGB
- {GL_SRGB8_ALPHA8, GL_RGBA, GL_UNSIGNED_BYTE, false}, // ASTC_2D_8X8_SRGB
- {GL_SRGB8_ALPHA8, GL_RGBA, GL_UNSIGNED_BYTE, false}, // ASTC_2D_8X5_SRGB
- {GL_SRGB8_ALPHA8, GL_RGBA, GL_UNSIGNED_BYTE, false}, // ASTC_2D_5X4_SRGB
- {GL_RGBA8, GL_RGBA, GL_UNSIGNED_BYTE, false}, // ASTC_2D_5X5
- {GL_SRGB8_ALPHA8, GL_RGBA, GL_UNSIGNED_BYTE, false}, // ASTC_2D_5X5_SRGB
- {GL_RGBA8, GL_RGBA, GL_UNSIGNED_BYTE, false}, // ASTC_2D_10X8
- {GL_SRGB8_ALPHA8, GL_RGBA, GL_UNSIGNED_BYTE, false}, // ASTC_2D_10X8_SRGB
- {GL_RGBA8, GL_RGBA, GL_UNSIGNED_BYTE, false}, // ASTC_2D_6X6
- {GL_SRGB8_ALPHA8, GL_RGBA, GL_UNSIGNED_BYTE, false}, // ASTC_2D_6X6_SRGB
- {GL_RGBA8, GL_RGBA, GL_UNSIGNED_BYTE, false}, // ASTC_2D_10X10
- {GL_SRGB8_ALPHA8, GL_RGBA, GL_UNSIGNED_BYTE, false}, // ASTC_2D_10X10_SRGB
- {GL_RGBA8, GL_RGBA, GL_UNSIGNED_BYTE, false}, // ASTC_2D_12X12
- {GL_SRGB8_ALPHA8, GL_RGBA, GL_UNSIGNED_BYTE, false}, // ASTC_2D_12X12_SRGB
- {GL_RGBA8, GL_RGBA, GL_UNSIGNED_BYTE, false}, // ASTC_2D_8X6
- {GL_SRGB8_ALPHA8, GL_RGBA, GL_UNSIGNED_BYTE, false}, // ASTC_2D_8X6_SRGB
- {GL_RGBA8, GL_RGBA, GL_UNSIGNED_BYTE, false}, // ASTC_2D_6X5
- {GL_SRGB8_ALPHA8, GL_RGBA, GL_UNSIGNED_BYTE, false}, // ASTC_2D_6X5_SRGB
- {GL_RGB9_E5, GL_RGB, GL_UNSIGNED_INT_5_9_9_9_REV, false}, // E5B9G9R9F
+ {GL_COMPRESSED_SRGB_ALPHA_S3TC_DXT1_EXT}, // DXT1_SRGB
+ {GL_COMPRESSED_SRGB_ALPHA_S3TC_DXT3_EXT}, // DXT23_SRGB
+ {GL_COMPRESSED_SRGB_ALPHA_S3TC_DXT5_EXT}, // DXT45_SRGB
+ {GL_COMPRESSED_SRGB_ALPHA_BPTC_UNORM}, // BC7U_SRGB
+ {GL_RGBA4, GL_RGBA, GL_UNSIGNED_SHORT_4_4_4_4_REV}, // R4G4B4A4U
+ {GL_COMPRESSED_SRGB8_ALPHA8_ASTC_4x4_KHR}, // ASTC_2D_4X4_SRGB
+ {GL_COMPRESSED_SRGB8_ALPHA8_ASTC_8x8_KHR}, // ASTC_2D_8X8_SRGB
+ {GL_COMPRESSED_SRGB8_ALPHA8_ASTC_8x5_KHR}, // ASTC_2D_8X5_SRGB
+ {GL_COMPRESSED_SRGB8_ALPHA8_ASTC_5x4_KHR}, // ASTC_2D_5X4_SRGB
+ {GL_COMPRESSED_RGBA_ASTC_5x5_KHR}, // ASTC_2D_5X5
+ {GL_COMPRESSED_SRGB8_ALPHA8_ASTC_5x5_KHR}, // ASTC_2D_5X5_SRGB
+ {GL_COMPRESSED_RGBA_ASTC_10x8_KHR}, // ASTC_2D_10X8
+ {GL_COMPRESSED_SRGB8_ALPHA8_ASTC_10x8_KHR}, // ASTC_2D_10X8_SRGB
+ {GL_COMPRESSED_RGBA_ASTC_6x6_KHR}, // ASTC_2D_6X6
+ {GL_COMPRESSED_SRGB8_ALPHA8_ASTC_6x6_KHR}, // ASTC_2D_6X6_SRGB
+ {GL_COMPRESSED_RGBA_ASTC_10x10_KHR}, // ASTC_2D_10X10
+ {GL_COMPRESSED_SRGB8_ALPHA8_ASTC_10x10_KHR}, // ASTC_2D_10X10_SRGB
+ {GL_COMPRESSED_RGBA_ASTC_12x12_KHR}, // ASTC_2D_12X12
+ {GL_COMPRESSED_SRGB8_ALPHA8_ASTC_12x12_KHR}, // ASTC_2D_12X12_SRGB
+ {GL_COMPRESSED_RGBA_ASTC_8x6_KHR}, // ASTC_2D_8X6
+ {GL_COMPRESSED_SRGB8_ALPHA8_ASTC_8x6_KHR}, // ASTC_2D_8X6_SRGB
+ {GL_COMPRESSED_RGBA_ASTC_6x5_KHR}, // ASTC_2D_6X5
+ {GL_COMPRESSED_SRGB8_ALPHA8_ASTC_6x5_KHR}, // ASTC_2D_6X5_SRGB
+ {GL_RGB9_E5, GL_RGB, GL_UNSIGNED_INT_5_9_9_9_REV}, // E5B9G9R9F
// Depth formats
- {GL_DEPTH_COMPONENT32F, GL_DEPTH_COMPONENT, GL_FLOAT, false}, // Z32F
- {GL_DEPTH_COMPONENT16, GL_DEPTH_COMPONENT, GL_UNSIGNED_SHORT, false}, // Z16
+ {GL_DEPTH_COMPONENT32F, GL_DEPTH_COMPONENT, GL_FLOAT}, // Z32F
+ {GL_DEPTH_COMPONENT16, GL_DEPTH_COMPONENT, GL_UNSIGNED_SHORT}, // Z16
// DepthStencil formats
- {GL_DEPTH24_STENCIL8, GL_DEPTH_STENCIL, GL_UNSIGNED_INT_24_8, false}, // Z24S8
- {GL_DEPTH24_STENCIL8, GL_DEPTH_STENCIL, GL_UNSIGNED_INT_24_8, false}, // S8Z24
- {GL_DEPTH32F_STENCIL8, GL_DEPTH_STENCIL, GL_FLOAT_32_UNSIGNED_INT_24_8_REV, false}, // Z32FS8
+ {GL_DEPTH24_STENCIL8, GL_DEPTH_STENCIL, GL_UNSIGNED_INT_24_8}, // Z24S8
+ {GL_DEPTH24_STENCIL8, GL_DEPTH_STENCIL, GL_UNSIGNED_INT_24_8}, // S8Z24
+ {GL_DEPTH32F_STENCIL8, GL_DEPTH_STENCIL, GL_FLOAT_32_UNSIGNED_INT_24_8_REV}, // Z32FS8
}};
const FormatTuple& GetFormatTuple(PixelFormat pixel_format) {
ASSERT(static_cast<std::size_t>(pixel_format) < tex_format_tuples.size());
- const auto& format{tex_format_tuples[static_cast<std::size_t>(pixel_format)]};
- return format;
+ return tex_format_tuples[static_cast<std::size_t>(pixel_format)];
}
GLenum GetTextureTarget(const SurfaceTarget& target) {
@@ -242,13 +239,20 @@ OGLTexture CreateTexture(const SurfaceParams& params, GLenum target, GLenum inte
} // Anonymous namespace
-CachedSurface::CachedSurface(const GPUVAddr gpu_addr, const SurfaceParams& params)
- : VideoCommon::SurfaceBase<View>(gpu_addr, params) {
- const auto& tuple{GetFormatTuple(params.pixel_format)};
- internal_format = tuple.internal_format;
- format = tuple.format;
- type = tuple.type;
- is_compressed = tuple.compressed;
+CachedSurface::CachedSurface(const GPUVAddr gpu_addr, const SurfaceParams& params,
+ bool is_astc_supported)
+ : VideoCommon::SurfaceBase<View>(gpu_addr, params, is_astc_supported) {
+ if (is_converted) {
+ internal_format = params.srgb_conversion ? GL_SRGB8_ALPHA8 : GL_RGBA8;
+ format = GL_RGBA;
+ type = GL_UNSIGNED_BYTE;
+ } else {
+ const auto& tuple{GetFormatTuple(params.pixel_format)};
+ internal_format = tuple.internal_format;
+ format = tuple.format;
+ type = tuple.type;
+ is_compressed = params.IsCompressed();
+ }
target = GetTextureTarget(params.target);
texture = CreateTexture(params, target, internal_format, texture_buffer);
DecorateSurfaceName();
@@ -264,7 +268,7 @@ void CachedSurface::DownloadTexture(std::vector<u8>& staging_buffer) {
if (params.IsBuffer()) {
glGetNamedBufferSubData(texture_buffer.handle, 0,
- static_cast<GLsizeiptr>(params.GetHostSizeInBytes()),
+ static_cast<GLsizeiptr>(params.GetHostSizeInBytes(false)),
staging_buffer.data());
return;
}
@@ -272,9 +276,10 @@ void CachedSurface::DownloadTexture(std::vector<u8>& staging_buffer) {
SCOPE_EXIT({ glPixelStorei(GL_PACK_ROW_LENGTH, 0); });
for (u32 level = 0; level < params.emulated_levels; ++level) {
- glPixelStorei(GL_PACK_ALIGNMENT, std::min(8U, params.GetRowAlignment(level)));
+ glPixelStorei(GL_PACK_ALIGNMENT, std::min(8U, params.GetRowAlignment(level, is_converted)));
glPixelStorei(GL_PACK_ROW_LENGTH, static_cast<GLint>(params.GetMipWidth(level)));
- const std::size_t mip_offset = params.GetHostMipmapLevelOffset(level);
+ const std::size_t mip_offset = params.GetHostMipmapLevelOffset(level, is_converted);
+
u8* const mip_data = staging_buffer.data() + mip_offset;
const GLsizei size = static_cast<GLsizei>(params.GetHostMipmapSize(level));
if (is_compressed) {
@@ -294,14 +299,10 @@ void CachedSurface::UploadTexture(const std::vector<u8>& staging_buffer) {
}
void CachedSurface::UploadTextureMipmap(u32 level, const std::vector<u8>& staging_buffer) {
- glPixelStorei(GL_UNPACK_ALIGNMENT, std::min(8U, params.GetRowAlignment(level)));
+ glPixelStorei(GL_UNPACK_ALIGNMENT, std::min(8U, params.GetRowAlignment(level, is_converted)));
glPixelStorei(GL_UNPACK_ROW_LENGTH, static_cast<GLint>(params.GetMipWidth(level)));
- auto compression_type = params.GetCompressionType();
-
- const std::size_t mip_offset = compression_type == SurfaceCompression::Converted
- ? params.GetConvertedMipmapOffset(level)
- : params.GetHostMipmapLevelOffset(level);
+ const std::size_t mip_offset = params.GetHostMipmapLevelOffset(level, is_converted);
const u8* buffer{staging_buffer.data() + mip_offset};
if (is_compressed) {
const auto image_size{static_cast<GLsizei>(params.GetHostMipmapSize(level))};
@@ -482,7 +483,7 @@ OGLTextureView CachedSurfaceView::CreateTextureView() const {
TextureCacheOpenGL::TextureCacheOpenGL(Core::System& system,
VideoCore::RasterizerInterface& rasterizer,
const Device& device, StateTracker& state_tracker)
- : TextureCacheBase{system, rasterizer}, state_tracker{state_tracker} {
+ : TextureCacheBase{system, rasterizer, device.HasASTC()}, state_tracker{state_tracker} {
src_framebuffer.Create();
dst_framebuffer.Create();
}
@@ -490,7 +491,7 @@ TextureCacheOpenGL::TextureCacheOpenGL(Core::System& system,
TextureCacheOpenGL::~TextureCacheOpenGL() = default;
Surface TextureCacheOpenGL::CreateSurface(GPUVAddr gpu_addr, const SurfaceParams& params) {
- return std::make_shared<CachedSurface>(gpu_addr, params);
+ return std::make_shared<CachedSurface>(gpu_addr, params, is_astc_supported);
}
void TextureCacheOpenGL::ImageCopy(Surface& src_surface, Surface& dst_surface,
@@ -596,7 +597,7 @@ void TextureCacheOpenGL::BufferCopy(Surface& src_surface, Surface& dst_surface)
glBindBuffer(GL_PIXEL_PACK_BUFFER, copy_pbo_handle);
- if (source_format.compressed) {
+ if (src_surface->IsCompressed()) {
glGetCompressedTextureImage(src_surface->GetTexture(), 0, static_cast<GLsizei>(source_size),
nullptr);
} else {
@@ -610,7 +611,7 @@ void TextureCacheOpenGL::BufferCopy(Surface& src_surface, Surface& dst_surface)
const GLsizei width = static_cast<GLsizei>(dst_params.width);
const GLsizei height = static_cast<GLsizei>(dst_params.height);
const GLsizei depth = static_cast<GLsizei>(dst_params.depth);
- if (dest_format.compressed) {
+ if (dst_surface->IsCompressed()) {
LOG_CRITICAL(HW_GPU, "Compressed buffer copy is unimplemented!");
UNREACHABLE();
} else {
diff --git a/src/video_core/renderer_opengl/gl_texture_cache.h b/src/video_core/renderer_opengl/gl_texture_cache.h
index 6658c6ffd..02d9981a1 100644
--- a/src/video_core/renderer_opengl/gl_texture_cache.h
+++ b/src/video_core/renderer_opengl/gl_texture_cache.h
@@ -37,7 +37,7 @@ class CachedSurface final : public VideoCommon::SurfaceBase<View> {
friend CachedSurfaceView;
public:
- explicit CachedSurface(GPUVAddr gpu_addr, const SurfaceParams& params);
+ explicit CachedSurface(GPUVAddr gpu_addr, const SurfaceParams& params, bool is_astc_supported);
~CachedSurface();
void UploadTexture(const std::vector<u8>& staging_buffer) override;
@@ -51,6 +51,10 @@ public:
return texture.handle;
}
+ bool IsCompressed() const {
+ return is_compressed;
+ }
+
protected:
void DecorateSurfaceName() override;
diff --git a/src/video_core/renderer_opengl/renderer_opengl.cpp b/src/video_core/renderer_opengl/renderer_opengl.cpp
index fca5e3ec0..f1a28cc21 100644
--- a/src/video_core/renderer_opengl/renderer_opengl.cpp
+++ b/src/video_core/renderer_opengl/renderer_opengl.cpp
@@ -30,8 +30,6 @@ namespace OpenGL {
namespace {
-// If the size of this is too small, it ends up creating a soft cap on FPS as the renderer will have
-// to wait on available presentation frames.
constexpr std::size_t SWAP_CHAIN_SIZE = 3;
struct Frame {
@@ -214,7 +212,7 @@ public:
std::deque<Frame*> present_queue;
Frame* previous_frame{};
- FrameMailbox() : has_debug_tool{HasDebugTool()} {
+ FrameMailbox() {
for (auto& frame : swap_chain) {
free_queue.push(&frame);
}
@@ -285,13 +283,9 @@ public:
std::unique_lock lock{swap_chain_lock};
present_queue.push_front(frame);
present_cv.notify_one();
-
- DebugNotifyNextFrame();
}
Frame* TryGetPresentFrame(int timeout_ms) {
- DebugWaitForNextFrame();
-
std::unique_lock lock{swap_chain_lock};
// wait for new entries in the present_queue
present_cv.wait_for(lock, std::chrono::milliseconds(timeout_ms),
@@ -317,38 +311,12 @@ public:
previous_frame = frame;
return frame;
}
-
-private:
- std::mutex debug_synch_mutex;
- std::condition_variable debug_synch_condition;
- std::atomic_int frame_for_debug{};
- const bool has_debug_tool; // When true, using a GPU debugger, so keep frames in lock-step
-
- /// Signal that a new frame is available (called from GPU thread)
- void DebugNotifyNextFrame() {
- if (!has_debug_tool) {
- return;
- }
- frame_for_debug++;
- std::lock_guard lock{debug_synch_mutex};
- debug_synch_condition.notify_one();
- }
-
- /// Wait for a new frame to be available (called from presentation thread)
- void DebugWaitForNextFrame() {
- if (!has_debug_tool) {
- return;
- }
- const int last_frame = frame_for_debug;
- std::unique_lock lock{debug_synch_mutex};
- debug_synch_condition.wait(lock,
- [this, last_frame] { return frame_for_debug > last_frame; });
- }
};
-RendererOpenGL::RendererOpenGL(Core::Frontend::EmuWindow& emu_window, Core::System& system)
+RendererOpenGL::RendererOpenGL(Core::Frontend::EmuWindow& emu_window, Core::System& system,
+ Core::Frontend::GraphicsContext& context)
: VideoCore::RendererBase{emu_window}, emu_window{emu_window}, system{system},
- frame_mailbox{std::make_unique<FrameMailbox>()} {}
+ frame_mailbox{}, context{context}, has_debug_tool{HasDebugTool()} {}
RendererOpenGL::~RendererOpenGL() = default;
@@ -356,8 +324,6 @@ MICROPROFILE_DEFINE(OpenGL_RenderFrame, "OpenGL", "Render Frame", MP_RGB(128, 12
MICROPROFILE_DEFINE(OpenGL_WaitPresent, "OpenGL", "Wait For Present", MP_RGB(128, 128, 128));
void RendererOpenGL::SwapBuffers(const Tegra::FramebufferConfig* framebuffer) {
- render_window.PollEvents();
-
if (!framebuffer) {
return;
}
@@ -413,6 +379,13 @@ void RendererOpenGL::SwapBuffers(const Tegra::FramebufferConfig* framebuffer) {
m_current_frame++;
rasterizer->TickFrame();
}
+
+ render_window.PollEvents();
+ if (has_debug_tool) {
+ glBindFramebuffer(GL_DRAW_FRAMEBUFFER, 0);
+ Present(0);
+ context.SwapBuffers();
+ }
}
void RendererOpenGL::PrepareRendertarget(const Tegra::FramebufferConfig* framebuffer) {
@@ -480,6 +453,8 @@ void RendererOpenGL::LoadColorToActiveGLTexture(u8 color_r, u8 color_g, u8 color
}
void RendererOpenGL::InitOpenGLObjects() {
+ frame_mailbox = std::make_unique<FrameMailbox>();
+
glClearColor(Settings::values.bg_red, Settings::values.bg_green, Settings::values.bg_blue,
0.0f);
@@ -692,12 +667,21 @@ void RendererOpenGL::DrawScreen(const Layout::FramebufferLayout& layout) {
glDrawArrays(GL_TRIANGLE_STRIP, 0, 4);
}
-void RendererOpenGL::TryPresent(int timeout_ms) {
+bool RendererOpenGL::TryPresent(int timeout_ms) {
+ if (has_debug_tool) {
+ LOG_DEBUG(Render_OpenGL,
+ "Skipping presentation because we are presenting on the main context");
+ return false;
+ }
+ return Present(timeout_ms);
+}
+
+bool RendererOpenGL::Present(int timeout_ms) {
const auto& layout = render_window.GetFramebufferLayout();
auto frame = frame_mailbox->TryGetPresentFrame(timeout_ms);
if (!frame) {
LOG_DEBUG(Render_OpenGL, "TryGetPresentFrame returned no frame to present");
- return;
+ return false;
}
// Clearing before a full overwrite of a fbo can signal to drivers that they can avoid a
@@ -725,6 +709,7 @@ void RendererOpenGL::TryPresent(int timeout_ms) {
glFlush();
glBindFramebuffer(GL_READ_FRAMEBUFFER, 0);
+ return true;
}
void RendererOpenGL::RenderScreenshot() {
diff --git a/src/video_core/renderer_opengl/renderer_opengl.h b/src/video_core/renderer_opengl/renderer_opengl.h
index 33073ce5b..50b647661 100644
--- a/src/video_core/renderer_opengl/renderer_opengl.h
+++ b/src/video_core/renderer_opengl/renderer_opengl.h
@@ -55,13 +55,14 @@ class FrameMailbox;
class RendererOpenGL final : public VideoCore::RendererBase {
public:
- explicit RendererOpenGL(Core::Frontend::EmuWindow& emu_window, Core::System& system);
+ explicit RendererOpenGL(Core::Frontend::EmuWindow& emu_window, Core::System& system,
+ Core::Frontend::GraphicsContext& context);
~RendererOpenGL() override;
bool Init() override;
void ShutDown() override;
void SwapBuffers(const Tegra::FramebufferConfig* framebuffer) override;
- void TryPresent(int timeout_ms) override;
+ bool TryPresent(int timeout_ms) override;
private:
/// Initializes the OpenGL state and creates persistent objects.
@@ -89,8 +90,11 @@ private:
void PrepareRendertarget(const Tegra::FramebufferConfig* framebuffer);
+ bool Present(int timeout_ms);
+
Core::Frontend::EmuWindow& emu_window;
Core::System& system;
+ Core::Frontend::GraphicsContext& context;
StateTracker state_tracker{system};
@@ -115,6 +119,8 @@ private:
/// Frame presentation mailbox
std::unique_ptr<FrameMailbox> frame_mailbox;
+
+ bool has_debug_tool = false;
};
} // namespace OpenGL
diff --git a/src/video_core/renderer_vulkan/declarations.h b/src/video_core/renderer_vulkan/declarations.h
index 323bf6b39..89a035ca4 100644
--- a/src/video_core/renderer_vulkan/declarations.h
+++ b/src/video_core/renderer_vulkan/declarations.h
@@ -39,6 +39,7 @@ using UniqueFence = UniqueHandle<vk::Fence>;
using UniqueFramebuffer = UniqueHandle<vk::Framebuffer>;
using UniqueImage = UniqueHandle<vk::Image>;
using UniqueImageView = UniqueHandle<vk::ImageView>;
+using UniqueInstance = UniqueHandle<vk::Instance>;
using UniqueIndirectCommandsLayoutNVX = UniqueHandle<vk::IndirectCommandsLayoutNVX>;
using UniqueObjectTableNVX = UniqueHandle<vk::ObjectTableNVX>;
using UniquePipeline = UniqueHandle<vk::Pipeline>;
@@ -50,6 +51,7 @@ using UniqueSampler = UniqueHandle<vk::Sampler>;
using UniqueSamplerYcbcrConversion = UniqueHandle<vk::SamplerYcbcrConversion>;
using UniqueSemaphore = UniqueHandle<vk::Semaphore>;
using UniqueShaderModule = UniqueHandle<vk::ShaderModule>;
+using UniqueSurfaceKHR = UniqueHandle<vk::SurfaceKHR>;
using UniqueSwapchainKHR = UniqueHandle<vk::SwapchainKHR>;
using UniqueValidationCacheEXT = UniqueHandle<vk::ValidationCacheEXT>;
using UniqueDebugReportCallbackEXT = UniqueHandle<vk::DebugReportCallbackEXT>;
diff --git a/src/video_core/renderer_vulkan/renderer_vulkan.cpp b/src/video_core/renderer_vulkan/renderer_vulkan.cpp
index 42bb01418..9cdb4b627 100644
--- a/src/video_core/renderer_vulkan/renderer_vulkan.cpp
+++ b/src/video_core/renderer_vulkan/renderer_vulkan.cpp
@@ -2,13 +2,18 @@
// Licensed under GPLv2 or any later version
// Refer to the license.txt file included.
+#include <algorithm>
+#include <array>
+#include <cstring>
#include <memory>
#include <optional>
+#include <string>
#include <vector>
#include <fmt/format.h>
#include "common/assert.h"
+#include "common/dynamic_library.h"
#include "common/logging/log.h"
#include "common/telemetry.h"
#include "core/core.h"
@@ -30,15 +35,30 @@
#include "video_core/renderer_vulkan/vk_state_tracker.h"
#include "video_core/renderer_vulkan/vk_swapchain.h"
+// Include these late to avoid changing Vulkan-Hpp's dynamic dispatcher size
+#ifdef _WIN32
+#include <windows.h>
+// ensure include order
+#include <vulkan/vulkan_win32.h>
+#endif
+
+#ifdef __linux__
+#include <X11/Xlib.h>
+#include <vulkan/vulkan_wayland.h>
+#include <vulkan/vulkan_xlib.h>
+#endif
+
namespace Vulkan {
namespace {
+using Core::Frontend::WindowSystemType;
+
VkBool32 DebugCallback(VkDebugUtilsMessageSeverityFlagBitsEXT severity_,
VkDebugUtilsMessageTypeFlagsEXT type,
const VkDebugUtilsMessengerCallbackDataEXT* data,
[[maybe_unused]] void* user_data) {
- const vk::DebugUtilsMessageSeverityFlagBitsEXT severity{severity_};
+ const auto severity{static_cast<vk::DebugUtilsMessageSeverityFlagBitsEXT>(severity_)};
const char* message{data->pMessage};
if (severity & vk::DebugUtilsMessageSeverityFlagBitsEXT::eError) {
@@ -53,6 +73,110 @@ VkBool32 DebugCallback(VkDebugUtilsMessageSeverityFlagBitsEXT severity_,
return VK_FALSE;
}
+Common::DynamicLibrary OpenVulkanLibrary() {
+ Common::DynamicLibrary library;
+#ifdef __APPLE__
+ // Check if a path to a specific Vulkan library has been specified.
+ char* libvulkan_env = getenv("LIBVULKAN_PATH");
+ if (!libvulkan_env || !library.Open(libvulkan_env)) {
+ // Use the libvulkan.dylib from the application bundle.
+ std::string filename = File::GetBundleDirectory() + "/Contents/Frameworks/libvulkan.dylib";
+ library.Open(filename.c_str());
+ }
+#else
+ std::string filename = Common::DynamicLibrary::GetVersionedFilename("vulkan", 1);
+ if (!library.Open(filename.c_str())) {
+ // Android devices may not have libvulkan.so.1, only libvulkan.so.
+ filename = Common::DynamicLibrary::GetVersionedFilename("vulkan");
+ library.Open(filename.c_str());
+ }
+#endif
+ return library;
+}
+
+UniqueInstance CreateInstance(Common::DynamicLibrary& library, vk::DispatchLoaderDynamic& dld,
+ WindowSystemType window_type = WindowSystemType::Headless,
+ bool enable_layers = false) {
+ if (!library.IsOpen()) {
+ LOG_ERROR(Render_Vulkan, "Vulkan library not available");
+ return UniqueInstance{};
+ }
+ PFN_vkGetInstanceProcAddr vkGetInstanceProcAddr;
+ if (!library.GetSymbol("vkGetInstanceProcAddr", &vkGetInstanceProcAddr)) {
+ LOG_ERROR(Render_Vulkan, "vkGetInstanceProcAddr not present in Vulkan");
+ return UniqueInstance{};
+ }
+ dld.init(vkGetInstanceProcAddr);
+
+ std::vector<const char*> extensions;
+ extensions.reserve(4);
+ switch (window_type) {
+ case Core::Frontend::WindowSystemType::Headless:
+ break;
+#ifdef _WIN32
+ case Core::Frontend::WindowSystemType::Windows:
+ extensions.push_back(VK_KHR_WIN32_SURFACE_EXTENSION_NAME);
+ break;
+#endif
+#ifdef __linux__
+ case Core::Frontend::WindowSystemType::X11:
+ extensions.push_back(VK_KHR_XLIB_SURFACE_EXTENSION_NAME);
+ break;
+ case Core::Frontend::WindowSystemType::Wayland:
+ extensions.push_back(VK_KHR_WAYLAND_SURFACE_EXTENSION_NAME);
+ break;
+#endif
+ default:
+ LOG_ERROR(Render_Vulkan, "Presentation not supported on this platform");
+ break;
+ }
+ if (window_type != Core::Frontend::WindowSystemType::Headless) {
+ extensions.push_back(VK_KHR_SURFACE_EXTENSION_NAME);
+ }
+ if (enable_layers) {
+ extensions.push_back(VK_EXT_DEBUG_UTILS_EXTENSION_NAME);
+ }
+
+ u32 num_properties;
+ if (vk::enumerateInstanceExtensionProperties(nullptr, &num_properties, nullptr, dld) !=
+ vk::Result::eSuccess) {
+ LOG_ERROR(Render_Vulkan, "Failed to query number of extension properties");
+ return UniqueInstance{};
+ }
+ std::vector<vk::ExtensionProperties> properties(num_properties);
+ if (vk::enumerateInstanceExtensionProperties(nullptr, &num_properties, properties.data(),
+ dld) != vk::Result::eSuccess) {
+ LOG_ERROR(Render_Vulkan, "Failed to query extension properties");
+ return UniqueInstance{};
+ }
+
+ for (const char* extension : extensions) {
+ const auto it =
+ std::find_if(properties.begin(), properties.end(), [extension](const auto& prop) {
+ return !std::strcmp(extension, prop.extensionName);
+ });
+ if (it == properties.end()) {
+ LOG_ERROR(Render_Vulkan, "Required instance extension {} is not available", extension);
+ return UniqueInstance{};
+ }
+ }
+
+ const vk::ApplicationInfo application_info("yuzu Emulator", VK_MAKE_VERSION(0, 1, 0),
+ "yuzu Emulator", VK_MAKE_VERSION(0, 1, 0),
+ VK_API_VERSION_1_1);
+ const std::array layers = {"VK_LAYER_LUNARG_standard_validation"};
+ const vk::InstanceCreateInfo instance_ci(
+ {}, &application_info, enable_layers ? static_cast<u32>(layers.size()) : 0, layers.data(),
+ static_cast<u32>(extensions.size()), extensions.data());
+ vk::Instance unsafe_instance;
+ if (vk::createInstance(&instance_ci, nullptr, &unsafe_instance, dld) != vk::Result::eSuccess) {
+ LOG_ERROR(Render_Vulkan, "Failed to create Vulkan instance");
+ return UniqueInstance{};
+ }
+ dld.init(unsafe_instance);
+ return UniqueInstance(unsafe_instance, {nullptr, dld});
+}
+
std::string GetReadableVersion(u32 version) {
return fmt::format("{}.{}.{}", VK_VERSION_MAJOR(version), VK_VERSION_MINOR(version),
VK_VERSION_PATCH(version));
@@ -141,32 +265,18 @@ void RendererVulkan::SwapBuffers(const Tegra::FramebufferConfig* framebuffer) {
render_window.PollEvents();
}
-void RendererVulkan::TryPresent(int /*timeout_ms*/) {
+bool RendererVulkan::TryPresent(int /*timeout_ms*/) {
// TODO (bunnei): ImplementMe
+ return true;
}
bool RendererVulkan::Init() {
- PFN_vkGetInstanceProcAddr vkGetInstanceProcAddr{};
- render_window.RetrieveVulkanHandlers(&vkGetInstanceProcAddr, &instance, &surface);
- const vk::DispatchLoaderDynamic dldi(instance, vkGetInstanceProcAddr);
-
- std::optional<vk::DebugUtilsMessengerEXT> callback;
- if (Settings::values.renderer_debug && dldi.vkCreateDebugUtilsMessengerEXT) {
- callback = CreateDebugCallback(dldi);
- if (!callback) {
- return false;
- }
- }
-
- if (!PickDevices(dldi)) {
- if (callback) {
- instance.destroy(*callback, nullptr, dldi);
- }
+ library = OpenVulkanLibrary();
+ instance = CreateInstance(library, dld, render_window.GetWindowInfo().type,
+ Settings::values.renderer_debug);
+ if (!instance || !CreateDebugCallback() || !CreateSurface() || !PickDevices()) {
return false;
}
- debug_callback = UniqueDebugUtilsMessengerEXT(
- *callback, vk::ObjectDestroy<vk::Instance, vk::DispatchLoaderDynamic>(
- instance, nullptr, device->GetDispatchLoader()));
Report();
@@ -175,7 +285,7 @@ bool RendererVulkan::Init() {
resource_manager = std::make_unique<VKResourceManager>(*device);
const auto& framebuffer = render_window.GetFramebufferLayout();
- swapchain = std::make_unique<VKSwapchain>(surface, *device);
+ swapchain = std::make_unique<VKSwapchain>(*surface, *device);
swapchain->Create(framebuffer.width, framebuffer.height, false);
state_tracker = std::make_unique<StateTracker>(system);
@@ -212,8 +322,10 @@ void RendererVulkan::ShutDown() {
device.reset();
}
-std::optional<vk::DebugUtilsMessengerEXT> RendererVulkan::CreateDebugCallback(
- const vk::DispatchLoaderDynamic& dldi) {
+bool RendererVulkan::CreateDebugCallback() {
+ if (!Settings::values.renderer_debug) {
+ return true;
+ }
const vk::DebugUtilsMessengerCreateInfoEXT callback_ci(
{},
vk::DebugUtilsMessageSeverityFlagBitsEXT::eError |
@@ -224,32 +336,88 @@ std::optional<vk::DebugUtilsMessengerEXT> RendererVulkan::CreateDebugCallback(
vk::DebugUtilsMessageTypeFlagBitsEXT::eValidation |
vk::DebugUtilsMessageTypeFlagBitsEXT::ePerformance,
&DebugCallback, nullptr);
- vk::DebugUtilsMessengerEXT callback;
- if (instance.createDebugUtilsMessengerEXT(&callback_ci, nullptr, &callback, dldi) !=
+ vk::DebugUtilsMessengerEXT unsafe_callback;
+ if (instance->createDebugUtilsMessengerEXT(&callback_ci, nullptr, &unsafe_callback, dld) !=
vk::Result::eSuccess) {
LOG_ERROR(Render_Vulkan, "Failed to create debug callback");
- return {};
+ return false;
+ }
+ debug_callback = UniqueDebugUtilsMessengerEXT(unsafe_callback, {*instance, nullptr, dld});
+ return true;
+}
+
+bool RendererVulkan::CreateSurface() {
+ [[maybe_unused]] const auto& window_info = render_window.GetWindowInfo();
+ VkSurfaceKHR unsafe_surface = nullptr;
+
+#ifdef _WIN32
+ if (window_info.type == Core::Frontend::WindowSystemType::Windows) {
+ const HWND hWnd = static_cast<HWND>(window_info.render_surface);
+ const VkWin32SurfaceCreateInfoKHR win32_ci{VK_STRUCTURE_TYPE_WIN32_SURFACE_CREATE_INFO_KHR,
+ nullptr, 0, nullptr, hWnd};
+ const auto vkCreateWin32SurfaceKHR = reinterpret_cast<PFN_vkCreateWin32SurfaceKHR>(
+ dld.vkGetInstanceProcAddr(*instance, "vkCreateWin32SurfaceKHR"));
+ if (!vkCreateWin32SurfaceKHR || vkCreateWin32SurfaceKHR(instance.get(), &win32_ci, nullptr,
+ &unsafe_surface) != VK_SUCCESS) {
+ LOG_ERROR(Render_Vulkan, "Failed to initialize Win32 surface");
+ return false;
+ }
+ }
+#endif
+#ifdef __linux__
+ if (window_info.type == Core::Frontend::WindowSystemType::X11) {
+ const VkXlibSurfaceCreateInfoKHR xlib_ci{
+ VK_STRUCTURE_TYPE_XLIB_SURFACE_CREATE_INFO_KHR, nullptr, 0,
+ static_cast<Display*>(window_info.display_connection),
+ reinterpret_cast<Window>(window_info.render_surface)};
+ const auto vkCreateXlibSurfaceKHR = reinterpret_cast<PFN_vkCreateXlibSurfaceKHR>(
+ dld.vkGetInstanceProcAddr(*instance, "vkCreateXlibSurfaceKHR"));
+ if (!vkCreateXlibSurfaceKHR || vkCreateXlibSurfaceKHR(instance.get(), &xlib_ci, nullptr,
+ &unsafe_surface) != VK_SUCCESS) {
+ LOG_ERROR(Render_Vulkan, "Failed to initialize Xlib surface");
+ return false;
+ }
+ }
+ if (window_info.type == Core::Frontend::WindowSystemType::Wayland) {
+ const VkWaylandSurfaceCreateInfoKHR wayland_ci{
+ VK_STRUCTURE_TYPE_WAYLAND_SURFACE_CREATE_INFO_KHR, nullptr, 0,
+ static_cast<wl_display*>(window_info.display_connection),
+ static_cast<wl_surface*>(window_info.render_surface)};
+ const auto vkCreateWaylandSurfaceKHR = reinterpret_cast<PFN_vkCreateWaylandSurfaceKHR>(
+ dld.vkGetInstanceProcAddr(*instance, "vkCreateWaylandSurfaceKHR"));
+ if (!vkCreateWaylandSurfaceKHR ||
+ vkCreateWaylandSurfaceKHR(instance.get(), &wayland_ci, nullptr, &unsafe_surface) !=
+ VK_SUCCESS) {
+ LOG_ERROR(Render_Vulkan, "Failed to initialize Wayland surface");
+ return false;
+ }
+ }
+#endif
+ if (!unsafe_surface) {
+ LOG_ERROR(Render_Vulkan, "Presentation not supported on this platform");
+ return false;
}
- return callback;
+
+ surface = UniqueSurfaceKHR(unsafe_surface, {*instance, nullptr, dld});
+ return true;
}
-bool RendererVulkan::PickDevices(const vk::DispatchLoaderDynamic& dldi) {
- const auto devices = instance.enumeratePhysicalDevices(dldi);
+bool RendererVulkan::PickDevices() {
+ const auto devices = instance->enumeratePhysicalDevices(dld);
- // TODO(Rodrigo): Choose device from config file
const s32 device_index = Settings::values.vulkan_device;
if (device_index < 0 || device_index >= static_cast<s32>(devices.size())) {
LOG_ERROR(Render_Vulkan, "Invalid device index {}!", device_index);
return false;
}
- const vk::PhysicalDevice physical_device = devices[device_index];
+ const vk::PhysicalDevice physical_device = devices[static_cast<std::size_t>(device_index)];
- if (!VKDevice::IsSuitable(dldi, physical_device, surface)) {
+ if (!VKDevice::IsSuitable(physical_device, *surface, dld)) {
return false;
}
- device = std::make_unique<VKDevice>(dldi, physical_device, surface);
- return device->Create(dldi, instance);
+ device = std::make_unique<VKDevice>(dld, physical_device, *surface);
+ return device->Create(*instance);
}
void RendererVulkan::Report() const {
@@ -275,4 +443,33 @@ void RendererVulkan::Report() const {
telemetry_session.AddField(field, "GPU_Vulkan_Extensions", extensions);
}
+std::vector<std::string> RendererVulkan::EnumerateDevices() {
+ // Avoid putting DispatchLoaderDynamic, it's too large
+ auto dld_memory = std::make_unique<vk::DispatchLoaderDynamic>();
+ auto& dld = *dld_memory;
+
+ Common::DynamicLibrary library = OpenVulkanLibrary();
+ UniqueInstance instance = CreateInstance(library, dld);
+ if (!instance) {
+ return {};
+ }
+
+ u32 num_devices;
+ if (instance->enumeratePhysicalDevices(&num_devices, nullptr, dld) != vk::Result::eSuccess) {
+ return {};
+ }
+ std::vector<vk::PhysicalDevice> devices(num_devices);
+ if (instance->enumeratePhysicalDevices(&num_devices, devices.data(), dld) !=
+ vk::Result::eSuccess) {
+ return {};
+ }
+
+ std::vector<std::string> names;
+ names.reserve(num_devices);
+ for (auto& device : devices) {
+ names.push_back(device.getProperties(dld).deviceName);
+ }
+ return names;
+}
+
} // namespace Vulkan
diff --git a/src/video_core/renderer_vulkan/renderer_vulkan.h b/src/video_core/renderer_vulkan/renderer_vulkan.h
index 3da08d2e4..42e253de5 100644
--- a/src/video_core/renderer_vulkan/renderer_vulkan.h
+++ b/src/video_core/renderer_vulkan/renderer_vulkan.h
@@ -6,8 +6,11 @@
#include <memory>
#include <optional>
+#include <string>
#include <vector>
+#include "common/dynamic_library.h"
+
#include "video_core/renderer_base.h"
#include "video_core/renderer_vulkan/declarations.h"
@@ -42,20 +45,26 @@ public:
bool Init() override;
void ShutDown() override;
void SwapBuffers(const Tegra::FramebufferConfig* framebuffer) override;
- void TryPresent(int timeout_ms) override;
+ bool TryPresent(int timeout_ms) override;
+
+ static std::vector<std::string> EnumerateDevices();
private:
- std::optional<vk::DebugUtilsMessengerEXT> CreateDebugCallback(
- const vk::DispatchLoaderDynamic& dldi);
+ bool CreateDebugCallback();
- bool PickDevices(const vk::DispatchLoaderDynamic& dldi);
+ bool CreateSurface();
+
+ bool PickDevices();
void Report() const;
Core::System& system;
- vk::Instance instance;
- vk::SurfaceKHR surface;
+ Common::DynamicLibrary library;
+ vk::DispatchLoaderDynamic dld;
+
+ UniqueInstance instance;
+ UniqueSurfaceKHR surface;
VKScreenInfo screen_info;
diff --git a/src/video_core/renderer_vulkan/vk_buffer_cache.cpp b/src/video_core/renderer_vulkan/vk_buffer_cache.cpp
index 1ba544943..326d74f29 100644
--- a/src/video_core/renderer_vulkan/vk_buffer_cache.cpp
+++ b/src/video_core/renderer_vulkan/vk_buffer_cache.cpp
@@ -42,8 +42,8 @@ auto CreateStreamBuffer(const VKDevice& device, VKScheduler& scheduler) {
} // Anonymous namespace
CachedBufferBlock::CachedBufferBlock(const VKDevice& device, VKMemoryManager& memory_manager,
- CacheAddr cache_addr, std::size_t size)
- : VideoCommon::BufferBlock{cache_addr, size} {
+ VAddr cpu_addr, std::size_t size)
+ : VideoCommon::BufferBlock{cpu_addr, size} {
const vk::BufferCreateInfo buffer_ci({}, static_cast<vk::DeviceSize>(size),
BufferUsage | vk::BufferUsageFlagBits::eTransferSrc |
vk::BufferUsageFlagBits::eTransferDst,
@@ -68,8 +68,8 @@ VKBufferCache::VKBufferCache(VideoCore::RasterizerInterface& rasterizer, Core::S
VKBufferCache::~VKBufferCache() = default;
-Buffer VKBufferCache::CreateBlock(CacheAddr cache_addr, std::size_t size) {
- return std::make_shared<CachedBufferBlock>(device, memory_manager, cache_addr, size);
+Buffer VKBufferCache::CreateBlock(VAddr cpu_addr, std::size_t size) {
+ return std::make_shared<CachedBufferBlock>(device, memory_manager, cpu_addr, size);
}
const vk::Buffer* VKBufferCache::ToHandle(const Buffer& buffer) {
diff --git a/src/video_core/renderer_vulkan/vk_buffer_cache.h b/src/video_core/renderer_vulkan/vk_buffer_cache.h
index 3f38eed0c..508214618 100644
--- a/src/video_core/renderer_vulkan/vk_buffer_cache.h
+++ b/src/video_core/renderer_vulkan/vk_buffer_cache.h
@@ -30,7 +30,7 @@ class VKScheduler;
class CachedBufferBlock final : public VideoCommon::BufferBlock {
public:
explicit CachedBufferBlock(const VKDevice& device, VKMemoryManager& memory_manager,
- CacheAddr cache_addr, std::size_t size);
+ VAddr cpu_addr, std::size_t size);
~CachedBufferBlock();
const vk::Buffer* GetHandle() const {
@@ -55,7 +55,7 @@ public:
protected:
void WriteBarrier() override {}
- Buffer CreateBlock(CacheAddr cache_addr, std::size_t size) override;
+ Buffer CreateBlock(VAddr cpu_addr, std::size_t size) override;
const vk::Buffer* ToHandle(const Buffer& buffer) override;
diff --git a/src/video_core/renderer_vulkan/vk_device.cpp b/src/video_core/renderer_vulkan/vk_device.cpp
index 28d2fbc4f..6f4ae9132 100644
--- a/src/video_core/renderer_vulkan/vk_device.cpp
+++ b/src/video_core/renderer_vulkan/vk_device.cpp
@@ -10,6 +10,7 @@
#include <string_view>
#include <thread>
#include <vector>
+
#include "common/assert.h"
#include "core/settings.h"
#include "video_core/renderer_vulkan/declarations.h"
@@ -35,20 +36,20 @@ void SetNext(void**& next, T& data) {
}
template <typename T>
-T GetFeatures(vk::PhysicalDevice physical, const vk::DispatchLoaderDynamic& dldi) {
+T GetFeatures(vk::PhysicalDevice physical, const vk::DispatchLoaderDynamic& dld) {
vk::PhysicalDeviceFeatures2 features;
T extension_features;
features.pNext = &extension_features;
- physical.getFeatures2(&features, dldi);
+ physical.getFeatures2(&features, dld);
return extension_features;
}
template <typename T>
-T GetProperties(vk::PhysicalDevice physical, const vk::DispatchLoaderDynamic& dldi) {
+T GetProperties(vk::PhysicalDevice physical, const vk::DispatchLoaderDynamic& dld) {
vk::PhysicalDeviceProperties2 properties;
T extension_properties;
properties.pNext = &extension_properties;
- physical.getProperties2(&properties, dldi);
+ physical.getProperties2(&properties, dld);
return extension_properties;
}
@@ -78,19 +79,19 @@ vk::FormatFeatureFlags GetFormatFeatures(vk::FormatProperties properties, Format
} // Anonymous namespace
-VKDevice::VKDevice(const vk::DispatchLoaderDynamic& dldi, vk::PhysicalDevice physical,
+VKDevice::VKDevice(const vk::DispatchLoaderDynamic& dld, vk::PhysicalDevice physical,
vk::SurfaceKHR surface)
- : physical{physical}, properties{physical.getProperties(dldi)},
- format_properties{GetFormatProperties(dldi, physical)} {
- SetupFamilies(dldi, surface);
- SetupFeatures(dldi);
+ : dld{dld}, physical{physical}, properties{physical.getProperties(dld)},
+ format_properties{GetFormatProperties(dld, physical)} {
+ SetupFamilies(surface);
+ SetupFeatures();
}
VKDevice::~VKDevice() = default;
-bool VKDevice::Create(const vk::DispatchLoaderDynamic& dldi, vk::Instance instance) {
+bool VKDevice::Create(vk::Instance instance) {
const auto queue_cis = GetDeviceQueueCreateInfos();
- const std::vector extensions = LoadExtensions(dldi);
+ const std::vector extensions = LoadExtensions();
vk::PhysicalDeviceFeatures2 features2;
void** next = &features2.pNext;
@@ -165,15 +166,13 @@ bool VKDevice::Create(const vk::DispatchLoaderDynamic& dldi, vk::Instance instan
nullptr);
device_ci.pNext = &features2;
- vk::Device dummy_logical;
- if (physical.createDevice(&device_ci, nullptr, &dummy_logical, dldi) != vk::Result::eSuccess) {
+ vk::Device unsafe_logical;
+ if (physical.createDevice(&device_ci, nullptr, &unsafe_logical, dld) != vk::Result::eSuccess) {
LOG_CRITICAL(Render_Vulkan, "Logical device failed to be created!");
return false;
}
-
- dld.init(instance, dldi.vkGetInstanceProcAddr, dummy_logical, dldi.vkGetDeviceProcAddr);
- logical = UniqueDevice(
- dummy_logical, vk::ObjectDestroy<vk::NoParent, vk::DispatchLoaderDynamic>(nullptr, dld));
+ dld.init(instance, dld.vkGetInstanceProcAddr, unsafe_logical);
+ logical = UniqueDevice(unsafe_logical, {nullptr, dld});
CollectTelemetryParameters();
@@ -235,20 +234,23 @@ void VKDevice::ReportLoss() const {
// *(VKGraphicsPipeline*)data[0]
}
-bool VKDevice::IsOptimalAstcSupported(const vk::PhysicalDeviceFeatures& features,
- const vk::DispatchLoaderDynamic& dldi) const {
+bool VKDevice::IsOptimalAstcSupported(const vk::PhysicalDeviceFeatures& features) const {
// Disable for now to avoid converting ASTC twice.
- return false;
static constexpr std::array astc_formats = {
- vk::Format::eAstc4x4SrgbBlock, vk::Format::eAstc8x8SrgbBlock,
- vk::Format::eAstc8x5SrgbBlock, vk::Format::eAstc5x4SrgbBlock,
+ vk::Format::eAstc4x4UnormBlock, vk::Format::eAstc4x4SrgbBlock,
+ vk::Format::eAstc5x4UnormBlock, vk::Format::eAstc5x4SrgbBlock,
vk::Format::eAstc5x5UnormBlock, vk::Format::eAstc5x5SrgbBlock,
- vk::Format::eAstc10x8UnormBlock, vk::Format::eAstc10x8SrgbBlock,
+ vk::Format::eAstc6x5UnormBlock, vk::Format::eAstc6x5SrgbBlock,
vk::Format::eAstc6x6UnormBlock, vk::Format::eAstc6x6SrgbBlock,
- vk::Format::eAstc10x10UnormBlock, vk::Format::eAstc10x10SrgbBlock,
- vk::Format::eAstc12x12UnormBlock, vk::Format::eAstc12x12SrgbBlock,
+ vk::Format::eAstc8x5UnormBlock, vk::Format::eAstc8x5SrgbBlock,
vk::Format::eAstc8x6UnormBlock, vk::Format::eAstc8x6SrgbBlock,
- vk::Format::eAstc6x5UnormBlock, vk::Format::eAstc6x5SrgbBlock};
+ vk::Format::eAstc8x8UnormBlock, vk::Format::eAstc8x8SrgbBlock,
+ vk::Format::eAstc10x5UnormBlock, vk::Format::eAstc10x5SrgbBlock,
+ vk::Format::eAstc10x6UnormBlock, vk::Format::eAstc10x6SrgbBlock,
+ vk::Format::eAstc10x8UnormBlock, vk::Format::eAstc10x8SrgbBlock,
+ vk::Format::eAstc10x10UnormBlock, vk::Format::eAstc10x10SrgbBlock,
+ vk::Format::eAstc12x10UnormBlock, vk::Format::eAstc12x10SrgbBlock,
+ vk::Format::eAstc12x12UnormBlock, vk::Format::eAstc12x12SrgbBlock};
if (!features.textureCompressionASTC_LDR) {
return false;
}
@@ -257,7 +259,7 @@ bool VKDevice::IsOptimalAstcSupported(const vk::PhysicalDeviceFeatures& features
vk::FormatFeatureFlagBits::eBlitDst | vk::FormatFeatureFlagBits::eTransferSrc |
vk::FormatFeatureFlagBits::eTransferDst};
for (const auto format : astc_formats) {
- const auto format_properties{physical.getFormatProperties(format, dldi)};
+ const auto format_properties{physical.getFormatProperties(format, dld)};
if (!(format_properties.optimalTilingFeatures & format_feature_usage)) {
return false;
}
@@ -276,11 +278,9 @@ bool VKDevice::IsFormatSupported(vk::Format wanted_format, vk::FormatFeatureFlag
return (supported_usage & wanted_usage) == wanted_usage;
}
-bool VKDevice::IsSuitable(const vk::DispatchLoaderDynamic& dldi, vk::PhysicalDevice physical,
- vk::SurfaceKHR surface) {
- bool is_suitable = true;
-
- constexpr std::array required_extensions = {
+bool VKDevice::IsSuitable(vk::PhysicalDevice physical, vk::SurfaceKHR surface,
+ const vk::DispatchLoaderDynamic& dld) {
+ static constexpr std::array required_extensions = {
VK_KHR_SWAPCHAIN_EXTENSION_NAME,
VK_KHR_16BIT_STORAGE_EXTENSION_NAME,
VK_KHR_8BIT_STORAGE_EXTENSION_NAME,
@@ -290,9 +290,10 @@ bool VKDevice::IsSuitable(const vk::DispatchLoaderDynamic& dldi, vk::PhysicalDev
VK_EXT_SHADER_SUBGROUP_VOTE_EXTENSION_NAME,
VK_EXT_HOST_QUERY_RESET_EXTENSION_NAME,
};
+ bool is_suitable = true;
std::bitset<required_extensions.size()> available_extensions{};
- for (const auto& prop : physical.enumerateDeviceExtensionProperties(nullptr, dldi)) {
+ for (const auto& prop : physical.enumerateDeviceExtensionProperties(nullptr, dld)) {
for (std::size_t i = 0; i < required_extensions.size(); ++i) {
if (available_extensions[i]) {
continue;
@@ -312,7 +313,7 @@ bool VKDevice::IsSuitable(const vk::DispatchLoaderDynamic& dldi, vk::PhysicalDev
}
bool has_graphics{}, has_present{};
- const auto queue_family_properties = physical.getQueueFamilyProperties(dldi);
+ const auto queue_family_properties = physical.getQueueFamilyProperties(dld);
for (u32 i = 0; i < static_cast<u32>(queue_family_properties.size()); ++i) {
const auto& family = queue_family_properties[i];
if (family.queueCount == 0) {
@@ -320,7 +321,7 @@ bool VKDevice::IsSuitable(const vk::DispatchLoaderDynamic& dldi, vk::PhysicalDev
}
has_graphics |=
(family.queueFlags & vk::QueueFlagBits::eGraphics) != static_cast<vk::QueueFlagBits>(0);
- has_present |= physical.getSurfaceSupportKHR(i, surface, dldi) != 0;
+ has_present |= physical.getSurfaceSupportKHR(i, surface, dld) != 0;
}
if (!has_graphics || !has_present) {
LOG_ERROR(Render_Vulkan, "Device lacks a graphics and present queue");
@@ -328,7 +329,7 @@ bool VKDevice::IsSuitable(const vk::DispatchLoaderDynamic& dldi, vk::PhysicalDev
}
// TODO(Rodrigo): Check if the device matches all requeriments.
- const auto properties{physical.getProperties(dldi)};
+ const auto properties{physical.getProperties(dld)};
const auto& limits{properties.limits};
constexpr u32 required_ubo_size = 65536;
@@ -345,7 +346,7 @@ bool VKDevice::IsSuitable(const vk::DispatchLoaderDynamic& dldi, vk::PhysicalDev
is_suitable = false;
}
- const auto features{physical.getFeatures(dldi)};
+ const auto features{physical.getFeatures(dld)};
const std::array feature_report = {
std::make_pair(features.vertexPipelineStoresAndAtomics, "vertexPipelineStoresAndAtomics"),
std::make_pair(features.independentBlend, "independentBlend"),
@@ -377,7 +378,7 @@ bool VKDevice::IsSuitable(const vk::DispatchLoaderDynamic& dldi, vk::PhysicalDev
return is_suitable;
}
-std::vector<const char*> VKDevice::LoadExtensions(const vk::DispatchLoaderDynamic& dldi) {
+std::vector<const char*> VKDevice::LoadExtensions() {
std::vector<const char*> extensions;
const auto Test = [&](const vk::ExtensionProperties& extension,
std::optional<std::reference_wrapper<bool>> status, const char* name,
@@ -408,7 +409,7 @@ std::vector<const char*> VKDevice::LoadExtensions(const vk::DispatchLoaderDynami
bool has_khr_shader_float16_int8{};
bool has_ext_subgroup_size_control{};
bool has_ext_transform_feedback{};
- for (const auto& extension : physical.enumerateDeviceExtensionProperties(nullptr, dldi)) {
+ for (const auto& extension : physical.enumerateDeviceExtensionProperties(nullptr, dld)) {
Test(extension, khr_uniform_buffer_standard_layout,
VK_KHR_UNIFORM_BUFFER_STANDARD_LAYOUT_EXTENSION_NAME, true);
Test(extension, has_khr_shader_float16_int8, VK_KHR_SHADER_FLOAT16_INT8_EXTENSION_NAME,
@@ -430,15 +431,15 @@ std::vector<const char*> VKDevice::LoadExtensions(const vk::DispatchLoaderDynami
if (has_khr_shader_float16_int8) {
is_float16_supported =
- GetFeatures<vk::PhysicalDeviceFloat16Int8FeaturesKHR>(physical, dldi).shaderFloat16;
+ GetFeatures<vk::PhysicalDeviceFloat16Int8FeaturesKHR>(physical, dld).shaderFloat16;
extensions.push_back(VK_KHR_SHADER_FLOAT16_INT8_EXTENSION_NAME);
}
if (has_ext_subgroup_size_control) {
const auto features =
- GetFeatures<vk::PhysicalDeviceSubgroupSizeControlFeaturesEXT>(physical, dldi);
+ GetFeatures<vk::PhysicalDeviceSubgroupSizeControlFeaturesEXT>(physical, dld);
const auto properties =
- GetProperties<vk::PhysicalDeviceSubgroupSizeControlPropertiesEXT>(physical, dldi);
+ GetProperties<vk::PhysicalDeviceSubgroupSizeControlPropertiesEXT>(physical, dld);
is_warp_potentially_bigger = properties.maxSubgroupSize > GuestWarpSize;
@@ -453,9 +454,9 @@ std::vector<const char*> VKDevice::LoadExtensions(const vk::DispatchLoaderDynami
if (has_ext_transform_feedback) {
const auto features =
- GetFeatures<vk::PhysicalDeviceTransformFeedbackFeaturesEXT>(physical, dldi);
+ GetFeatures<vk::PhysicalDeviceTransformFeedbackFeaturesEXT>(physical, dld);
const auto properties =
- GetProperties<vk::PhysicalDeviceTransformFeedbackPropertiesEXT>(physical, dldi);
+ GetProperties<vk::PhysicalDeviceTransformFeedbackPropertiesEXT>(physical, dld);
if (features.transformFeedback && features.geometryStreams &&
properties.maxTransformFeedbackStreams >= 4 && properties.maxTransformFeedbackBuffers &&
@@ -468,10 +469,10 @@ std::vector<const char*> VKDevice::LoadExtensions(const vk::DispatchLoaderDynami
return extensions;
}
-void VKDevice::SetupFamilies(const vk::DispatchLoaderDynamic& dldi, vk::SurfaceKHR surface) {
+void VKDevice::SetupFamilies(vk::SurfaceKHR surface) {
std::optional<u32> graphics_family_, present_family_;
- const auto queue_family_properties = physical.getQueueFamilyProperties(dldi);
+ const auto queue_family_properties = physical.getQueueFamilyProperties(dld);
for (u32 i = 0; i < static_cast<u32>(queue_family_properties.size()); ++i) {
if (graphics_family_ && present_family_)
break;
@@ -480,10 +481,12 @@ void VKDevice::SetupFamilies(const vk::DispatchLoaderDynamic& dldi, vk::SurfaceK
if (queue_family.queueCount == 0)
continue;
- if (queue_family.queueFlags & vk::QueueFlagBits::eGraphics)
+ if (queue_family.queueFlags & vk::QueueFlagBits::eGraphics) {
graphics_family_ = i;
- if (physical.getSurfaceSupportKHR(i, surface, dldi))
+ }
+ if (physical.getSurfaceSupportKHR(i, surface, dld)) {
present_family_ = i;
+ }
}
ASSERT(graphics_family_ && present_family_);
@@ -491,10 +494,10 @@ void VKDevice::SetupFamilies(const vk::DispatchLoaderDynamic& dldi, vk::SurfaceK
present_family = *present_family_;
}
-void VKDevice::SetupFeatures(const vk::DispatchLoaderDynamic& dldi) {
- const auto supported_features{physical.getFeatures(dldi)};
+void VKDevice::SetupFeatures() {
+ const auto supported_features{physical.getFeatures(dld)};
is_formatless_image_load_supported = supported_features.shaderStorageImageReadWithoutFormat;
- is_optimal_astc_supported = IsOptimalAstcSupported(supported_features, dldi);
+ is_optimal_astc_supported = IsOptimalAstcSupported(supported_features);
}
void VKDevice::CollectTelemetryParameters() {
@@ -522,7 +525,7 @@ std::vector<vk::DeviceQueueCreateInfo> VKDevice::GetDeviceQueueCreateInfos() con
}
std::unordered_map<vk::Format, vk::FormatProperties> VKDevice::GetFormatProperties(
- const vk::DispatchLoaderDynamic& dldi, vk::PhysicalDevice physical) {
+ const vk::DispatchLoaderDynamic& dld, vk::PhysicalDevice physical) {
static constexpr std::array formats{vk::Format::eA8B8G8R8UnormPack32,
vk::Format::eA8B8G8R8UintPack32,
vk::Format::eA8B8G8R8SnormPack32,
@@ -572,28 +575,38 @@ std::unordered_map<vk::Format, vk::FormatProperties> VKDevice::GetFormatProperti
vk::Format::eBc2SrgbBlock,
vk::Format::eBc3SrgbBlock,
vk::Format::eBc7SrgbBlock,
+ vk::Format::eAstc4x4UnormBlock,
vk::Format::eAstc4x4SrgbBlock,
- vk::Format::eAstc8x8SrgbBlock,
- vk::Format::eAstc8x5SrgbBlock,
+ vk::Format::eAstc5x4UnormBlock,
vk::Format::eAstc5x4SrgbBlock,
vk::Format::eAstc5x5UnormBlock,
vk::Format::eAstc5x5SrgbBlock,
- vk::Format::eAstc10x8UnormBlock,
- vk::Format::eAstc10x8SrgbBlock,
+ vk::Format::eAstc6x5UnormBlock,
+ vk::Format::eAstc6x5SrgbBlock,
vk::Format::eAstc6x6UnormBlock,
vk::Format::eAstc6x6SrgbBlock,
+ vk::Format::eAstc8x5UnormBlock,
+ vk::Format::eAstc8x5SrgbBlock,
+ vk::Format::eAstc8x6UnormBlock,
+ vk::Format::eAstc8x6SrgbBlock,
+ vk::Format::eAstc8x8UnormBlock,
+ vk::Format::eAstc8x8SrgbBlock,
+ vk::Format::eAstc10x5UnormBlock,
+ vk::Format::eAstc10x5SrgbBlock,
+ vk::Format::eAstc10x6UnormBlock,
+ vk::Format::eAstc10x6SrgbBlock,
+ vk::Format::eAstc10x8UnormBlock,
+ vk::Format::eAstc10x8SrgbBlock,
vk::Format::eAstc10x10UnormBlock,
vk::Format::eAstc10x10SrgbBlock,
+ vk::Format::eAstc12x10UnormBlock,
+ vk::Format::eAstc12x10SrgbBlock,
vk::Format::eAstc12x12UnormBlock,
vk::Format::eAstc12x12SrgbBlock,
- vk::Format::eAstc8x6UnormBlock,
- vk::Format::eAstc8x6SrgbBlock,
- vk::Format::eAstc6x5UnormBlock,
- vk::Format::eAstc6x5SrgbBlock,
vk::Format::eE5B9G9R9UfloatPack32};
std::unordered_map<vk::Format, vk::FormatProperties> format_properties;
for (const auto format : formats) {
- format_properties.emplace(format, physical.getFormatProperties(format, dldi));
+ format_properties.emplace(format, physical.getFormatProperties(format, dld));
}
return format_properties;
}
diff --git a/src/video_core/renderer_vulkan/vk_device.h b/src/video_core/renderer_vulkan/vk_device.h
index 6e656517f..d9d809852 100644
--- a/src/video_core/renderer_vulkan/vk_device.h
+++ b/src/video_core/renderer_vulkan/vk_device.h
@@ -22,12 +22,12 @@ const u32 GuestWarpSize = 32;
/// Handles data specific to a physical device.
class VKDevice final {
public:
- explicit VKDevice(const vk::DispatchLoaderDynamic& dldi, vk::PhysicalDevice physical,
+ explicit VKDevice(const vk::DispatchLoaderDynamic& dld, vk::PhysicalDevice physical,
vk::SurfaceKHR surface);
~VKDevice();
/// Initializes the device. Returns true on success.
- bool Create(const vk::DispatchLoaderDynamic& dldi, vk::Instance instance);
+ bool Create(vk::Instance instance);
/**
* Returns a format supported by the device for the passed requeriments.
@@ -188,18 +188,18 @@ public:
}
/// Checks if the physical device is suitable.
- static bool IsSuitable(const vk::DispatchLoaderDynamic& dldi, vk::PhysicalDevice physical,
- vk::SurfaceKHR surface);
+ static bool IsSuitable(vk::PhysicalDevice physical, vk::SurfaceKHR surface,
+ const vk::DispatchLoaderDynamic& dld);
private:
/// Loads extensions into a vector and stores available ones in this object.
- std::vector<const char*> LoadExtensions(const vk::DispatchLoaderDynamic& dldi);
+ std::vector<const char*> LoadExtensions();
/// Sets up queue families.
- void SetupFamilies(const vk::DispatchLoaderDynamic& dldi, vk::SurfaceKHR surface);
+ void SetupFamilies(vk::SurfaceKHR surface);
/// Sets up device features.
- void SetupFeatures(const vk::DispatchLoaderDynamic& dldi);
+ void SetupFeatures();
/// Collects telemetry information from the device.
void CollectTelemetryParameters();
@@ -208,8 +208,7 @@ private:
std::vector<vk::DeviceQueueCreateInfo> GetDeviceQueueCreateInfos() const;
/// Returns true if ASTC textures are natively supported.
- bool IsOptimalAstcSupported(const vk::PhysicalDeviceFeatures& features,
- const vk::DispatchLoaderDynamic& dldi) const;
+ bool IsOptimalAstcSupported(const vk::PhysicalDeviceFeatures& features) const;
/// Returns true if a format is supported.
bool IsFormatSupported(vk::Format wanted_format, vk::FormatFeatureFlags wanted_usage,
@@ -217,10 +216,10 @@ private:
/// Returns the device properties for Vulkan formats.
static std::unordered_map<vk::Format, vk::FormatProperties> GetFormatProperties(
- const vk::DispatchLoaderDynamic& dldi, vk::PhysicalDevice physical);
+ const vk::DispatchLoaderDynamic& dld, vk::PhysicalDevice physical);
- const vk::PhysicalDevice physical; ///< Physical device.
vk::DispatchLoaderDynamic dld; ///< Device function pointers.
+ vk::PhysicalDevice physical; ///< Physical device.
vk::PhysicalDeviceProperties properties; ///< Device properties.
UniqueDevice logical; ///< Logical device.
vk::Queue graphics_queue; ///< Main graphics queue.
diff --git a/src/video_core/renderer_vulkan/vk_pipeline_cache.cpp b/src/video_core/renderer_vulkan/vk_pipeline_cache.cpp
index 557b9d662..c2a426aeb 100644
--- a/src/video_core/renderer_vulkan/vk_pipeline_cache.cpp
+++ b/src/video_core/renderer_vulkan/vk_pipeline_cache.cpp
@@ -158,11 +158,11 @@ u32 FillDescriptorLayout(const ShaderEntries& entries,
} // Anonymous namespace
CachedShader::CachedShader(Core::System& system, Tegra::Engines::ShaderType stage,
- GPUVAddr gpu_addr, VAddr cpu_addr, u8* host_ptr,
- ProgramCode program_code, u32 main_offset)
- : RasterizerCacheObject{host_ptr}, gpu_addr{gpu_addr}, cpu_addr{cpu_addr},
- program_code{std::move(program_code)}, registry{stage, GetEngine(system, stage)},
- shader_ir{this->program_code, main_offset, compiler_settings, registry},
+ GPUVAddr gpu_addr, VAddr cpu_addr, ProgramCode program_code,
+ u32 main_offset)
+ : RasterizerCacheObject{cpu_addr}, gpu_addr{gpu_addr}, program_code{std::move(program_code)},
+ registry{stage, GetEngine(system, stage)}, shader_ir{this->program_code, main_offset,
+ compiler_settings, registry},
entries{GenerateShaderEntries(shader_ir)} {}
CachedShader::~CachedShader() = default;
@@ -201,19 +201,19 @@ std::array<Shader, Maxwell::MaxShaderProgram> VKPipelineCache::GetShaders() {
auto& memory_manager{system.GPU().MemoryManager()};
const GPUVAddr program_addr{GetShaderAddress(system, program)};
- const auto host_ptr{memory_manager.GetPointer(program_addr)};
- auto shader = TryGet(host_ptr);
+ const std::optional cpu_addr = memory_manager.GpuToCpuAddress(program_addr);
+ ASSERT(cpu_addr);
+ auto shader = cpu_addr ? TryGet(*cpu_addr) : nullptr;
if (!shader) {
+ const auto host_ptr{memory_manager.GetPointer(program_addr)};
+
// No shader found - create a new one
constexpr u32 stage_offset = 10;
const auto stage = static_cast<Tegra::Engines::ShaderType>(index == 0 ? 0 : index - 1);
auto code = GetShaderCode(memory_manager, program_addr, host_ptr, false);
- const std::optional cpu_addr = memory_manager.GpuToCpuAddress(program_addr);
- ASSERT(cpu_addr);
-
shader = std::make_shared<CachedShader>(system, stage, program_addr, *cpu_addr,
- host_ptr, std::move(code), stage_offset);
+ std::move(code), stage_offset);
Register(shader);
}
shaders[index] = std::move(shader);
@@ -253,18 +253,19 @@ VKComputePipeline& VKPipelineCache::GetComputePipeline(const ComputePipelineCach
auto& memory_manager = system.GPU().MemoryManager();
const auto program_addr = key.shader;
- const auto host_ptr = memory_manager.GetPointer(program_addr);
- auto shader = TryGet(host_ptr);
+ const auto cpu_addr = memory_manager.GpuToCpuAddress(program_addr);
+ ASSERT(cpu_addr);
+
+ auto shader = cpu_addr ? TryGet(*cpu_addr) : nullptr;
if (!shader) {
// No shader found - create a new one
- const auto cpu_addr = memory_manager.GpuToCpuAddress(program_addr);
- ASSERT(cpu_addr);
+ const auto host_ptr = memory_manager.GetPointer(program_addr);
auto code = GetShaderCode(memory_manager, program_addr, host_ptr, true);
constexpr u32 kernel_main_offset = 0;
shader = std::make_shared<CachedShader>(system, Tegra::Engines::ShaderType::Compute,
- program_addr, *cpu_addr, host_ptr, std::move(code),
+ program_addr, *cpu_addr, std::move(code),
kernel_main_offset);
Register(shader);
}
@@ -345,8 +346,9 @@ VKPipelineCache::DecompileShaders(const GraphicsPipelineCacheKey& key) {
}
const GPUVAddr gpu_addr = GetShaderAddress(system, program_enum);
- const auto host_ptr = memory_manager.GetPointer(gpu_addr);
- const auto shader = TryGet(host_ptr);
+ const auto cpu_addr = memory_manager.GpuToCpuAddress(gpu_addr);
+ ASSERT(cpu_addr);
+ const auto shader = TryGet(*cpu_addr);
ASSERT(shader);
const std::size_t stage = index == 0 ? 0 : index - 1; // Stage indices are 0 - 5
diff --git a/src/video_core/renderer_vulkan/vk_pipeline_cache.h b/src/video_core/renderer_vulkan/vk_pipeline_cache.h
index c4c112290..27c01732f 100644
--- a/src/video_core/renderer_vulkan/vk_pipeline_cache.h
+++ b/src/video_core/renderer_vulkan/vk_pipeline_cache.h
@@ -113,17 +113,13 @@ namespace Vulkan {
class CachedShader final : public RasterizerCacheObject {
public:
explicit CachedShader(Core::System& system, Tegra::Engines::ShaderType stage, GPUVAddr gpu_addr,
- VAddr cpu_addr, u8* host_ptr, ProgramCode program_code, u32 main_offset);
+ VAddr cpu_addr, ProgramCode program_code, u32 main_offset);
~CachedShader();
GPUVAddr GetGpuAddr() const {
return gpu_addr;
}
- VAddr GetCpuAddr() const override {
- return cpu_addr;
- }
-
std::size_t GetSizeInBytes() const override {
return program_code.size() * sizeof(u64);
}
@@ -149,7 +145,6 @@ private:
Tegra::Engines::ShaderType stage);
GPUVAddr gpu_addr{};
- VAddr cpu_addr{};
ProgramCode program_code;
VideoCommon::Shader::Registry registry;
VideoCommon::Shader::ShaderIR shader_ir;
diff --git a/src/video_core/renderer_vulkan/vk_rasterizer.cpp b/src/video_core/renderer_vulkan/vk_rasterizer.cpp
index 58c69b786..0a2ea4fd4 100644
--- a/src/video_core/renderer_vulkan/vk_rasterizer.cpp
+++ b/src/video_core/renderer_vulkan/vk_rasterizer.cpp
@@ -495,20 +495,26 @@ void RasterizerVulkan::Query(GPUVAddr gpu_addr, VideoCore::QueryType type,
void RasterizerVulkan::FlushAll() {}
-void RasterizerVulkan::FlushRegion(CacheAddr addr, u64 size) {
+void RasterizerVulkan::FlushRegion(VAddr addr, u64 size) {
+ if (addr == 0 || size == 0) {
+ return;
+ }
texture_cache.FlushRegion(addr, size);
buffer_cache.FlushRegion(addr, size);
query_cache.FlushRegion(addr, size);
}
-void RasterizerVulkan::InvalidateRegion(CacheAddr addr, u64 size) {
+void RasterizerVulkan::InvalidateRegion(VAddr addr, u64 size) {
+ if (addr == 0 || size == 0) {
+ return;
+ }
texture_cache.InvalidateRegion(addr, size);
pipeline_cache.InvalidateRegion(addr, size);
buffer_cache.InvalidateRegion(addr, size);
query_cache.InvalidateRegion(addr, size);
}
-void RasterizerVulkan::FlushAndInvalidateRegion(CacheAddr addr, u64 size) {
+void RasterizerVulkan::FlushAndInvalidateRegion(VAddr addr, u64 size) {
FlushRegion(addr, size);
InvalidateRegion(addr, size);
}
@@ -540,8 +546,7 @@ bool RasterizerVulkan::AccelerateDisplay(const Tegra::FramebufferConfig& config,
return false;
}
- const u8* host_ptr{system.Memory().GetPointer(framebuffer_addr)};
- const auto surface{texture_cache.TryFindFramebufferSurface(host_ptr)};
+ const auto surface{texture_cache.TryFindFramebufferSurface(framebuffer_addr)};
if (!surface) {
return false;
}
diff --git a/src/video_core/renderer_vulkan/vk_rasterizer.h b/src/video_core/renderer_vulkan/vk_rasterizer.h
index 3185868e9..f642dde76 100644
--- a/src/video_core/renderer_vulkan/vk_rasterizer.h
+++ b/src/video_core/renderer_vulkan/vk_rasterizer.h
@@ -118,9 +118,9 @@ public:
void ResetCounter(VideoCore::QueryType type) override;
void Query(GPUVAddr gpu_addr, VideoCore::QueryType type, std::optional<u64> timestamp) override;
void FlushAll() override;
- void FlushRegion(CacheAddr addr, u64 size) override;
- void InvalidateRegion(CacheAddr addr, u64 size) override;
- void FlushAndInvalidateRegion(CacheAddr addr, u64 size) override;
+ void FlushRegion(VAddr addr, u64 size) override;
+ void InvalidateRegion(VAddr addr, u64 size) override;
+ void FlushAndInvalidateRegion(VAddr addr, u64 size) override;
void FlushCommands() override;
void TickFrame() override;
bool AccelerateSurfaceCopy(const Tegra::Engines::Fermi2D::Regs::Surface& src,
diff --git a/src/video_core/renderer_vulkan/vk_shader_decompiler.cpp b/src/video_core/renderer_vulkan/vk_shader_decompiler.cpp
index 51ecb5567..b9f9e2714 100644
--- a/src/video_core/renderer_vulkan/vk_shader_decompiler.cpp
+++ b/src/video_core/renderer_vulkan/vk_shader_decompiler.cpp
@@ -35,7 +35,7 @@ namespace {
using Sirit::Id;
using Tegra::Engines::ShaderType;
using Tegra::Shader::Attribute;
-using Tegra::Shader::AttributeUse;
+using Tegra::Shader::PixelImap;
using Tegra::Shader::Register;
using namespace VideoCommon::Shader;
@@ -752,16 +752,16 @@ private:
if (stage != ShaderType::Fragment) {
continue;
}
- switch (header.ps.GetAttributeUse(location)) {
- case AttributeUse::Constant:
+ switch (header.ps.GetPixelImap(location)) {
+ case PixelImap::Constant:
Decorate(id, spv::Decoration::Flat);
break;
- case AttributeUse::ScreenLinear:
- Decorate(id, spv::Decoration::NoPerspective);
- break;
- case AttributeUse::Perspective:
+ case PixelImap::Perspective:
// Default
break;
+ case PixelImap::ScreenLinear:
+ Decorate(id, spv::Decoration::NoPerspective);
+ break;
default:
UNREACHABLE_MSG("Unused attribute being fetched");
}
@@ -1145,9 +1145,6 @@ private:
switch (attribute) {
case Attribute::Index::Position: {
if (stage == ShaderType::Fragment) {
- if (element == 3) {
- return {Constant(t_float, 1.0f), Type::Float};
- }
return {OpLoad(t_float, AccessElement(t_in_float, frag_coord, element)),
Type::Float};
}
@@ -1941,7 +1938,11 @@ private:
return {};
}
- Expression AtomicAdd(Operation operation) {
+ template <Id (Module::*func)(Id, Id, Id, Id, Id), Type result_type,
+ Type value_type = result_type>
+ Expression Atomic(Operation operation) {
+ const Id type_def = GetTypeDefinition(result_type);
+
Id pointer;
if (const auto smem = std::get_if<SmemNode>(&*operation[0])) {
pointer = GetSharedMemoryPointer(*smem);
@@ -1949,14 +1950,15 @@ private:
pointer = GetGlobalMemoryPointer(*gmem);
} else {
UNREACHABLE();
- return {Constant(t_uint, 0), Type::Uint};
+ return {Constant(type_def, 0), result_type};
}
+ const Id value = As(Visit(operation[1]), value_type);
+
const Id scope = Constant(t_uint, static_cast<u32>(spv::Scope::Device));
- const Id semantics = Constant(t_uint, 0U);
+ const Id semantics = Constant(type_def, 0);
- const Id value = AsUint(Visit(operation[1]));
- return {OpAtomicIAdd(t_uint, pointer, scope, semantics, value), Type::Uint};
+ return {(this->*func)(type_def, pointer, scope, semantics, value), result_type};
}
Expression Branch(Operation operation) {
@@ -2545,7 +2547,21 @@ private:
&SPIRVDecompiler::AtomicImageXor,
&SPIRVDecompiler::AtomicImageExchange,
- &SPIRVDecompiler::AtomicAdd,
+ &SPIRVDecompiler::Atomic<&Module::OpAtomicExchange, Type::Uint>,
+ &SPIRVDecompiler::Atomic<&Module::OpAtomicIAdd, Type::Uint>,
+ &SPIRVDecompiler::Atomic<&Module::OpAtomicUMin, Type::Uint>,
+ &SPIRVDecompiler::Atomic<&Module::OpAtomicUMax, Type::Uint>,
+ &SPIRVDecompiler::Atomic<&Module::OpAtomicAnd, Type::Uint>,
+ &SPIRVDecompiler::Atomic<&Module::OpAtomicOr, Type::Uint>,
+ &SPIRVDecompiler::Atomic<&Module::OpAtomicXor, Type::Uint>,
+
+ &SPIRVDecompiler::Atomic<&Module::OpAtomicExchange, Type::Int>,
+ &SPIRVDecompiler::Atomic<&Module::OpAtomicIAdd, Type::Int>,
+ &SPIRVDecompiler::Atomic<&Module::OpAtomicSMin, Type::Int>,
+ &SPIRVDecompiler::Atomic<&Module::OpAtomicSMax, Type::Int>,
+ &SPIRVDecompiler::Atomic<&Module::OpAtomicAnd, Type::Int>,
+ &SPIRVDecompiler::Atomic<&Module::OpAtomicOr, Type::Int>,
+ &SPIRVDecompiler::Atomic<&Module::OpAtomicXor, Type::Int>,
&SPIRVDecompiler::Branch,
&SPIRVDecompiler::BranchIndirect,
diff --git a/src/video_core/renderer_vulkan/vk_texture_cache.cpp b/src/video_core/renderer_vulkan/vk_texture_cache.cpp
index 26175921b..5b9b39670 100644
--- a/src/video_core/renderer_vulkan/vk_texture_cache.cpp
+++ b/src/video_core/renderer_vulkan/vk_texture_cache.cpp
@@ -35,7 +35,6 @@ using VideoCore::MortonSwizzleMode;
using Tegra::Texture::SwizzleSource;
using VideoCore::Surface::PixelFormat;
-using VideoCore::Surface::SurfaceCompression;
using VideoCore::Surface::SurfaceTarget;
namespace {
@@ -96,9 +95,10 @@ vk::ImageViewType GetImageViewType(SurfaceTarget target) {
return {};
}
-UniqueBuffer CreateBuffer(const VKDevice& device, const SurfaceParams& params) {
+UniqueBuffer CreateBuffer(const VKDevice& device, const SurfaceParams& params,
+ std::size_t host_memory_size) {
// TODO(Rodrigo): Move texture buffer creation to the buffer cache
- const vk::BufferCreateInfo buffer_ci({}, params.GetHostSizeInBytes(),
+ const vk::BufferCreateInfo buffer_ci({}, host_memory_size,
vk::BufferUsageFlagBits::eUniformTexelBuffer |
vk::BufferUsageFlagBits::eTransferSrc |
vk::BufferUsageFlagBits::eTransferDst,
@@ -110,12 +110,13 @@ UniqueBuffer CreateBuffer(const VKDevice& device, const SurfaceParams& params) {
vk::BufferViewCreateInfo GenerateBufferViewCreateInfo(const VKDevice& device,
const SurfaceParams& params,
- vk::Buffer buffer) {
+ vk::Buffer buffer,
+ std::size_t host_memory_size) {
ASSERT(params.IsBuffer());
const auto format =
MaxwellToVK::SurfaceFormat(device, FormatType::Buffer, params.pixel_format).format;
- return vk::BufferViewCreateInfo({}, buffer, format, 0, params.GetHostSizeInBytes());
+ return vk::BufferViewCreateInfo({}, buffer, format, 0, host_memory_size);
}
vk::ImageCreateInfo GenerateImageCreateInfo(const VKDevice& device, const SurfaceParams& params) {
@@ -169,14 +170,15 @@ CachedSurface::CachedSurface(Core::System& system, const VKDevice& device,
VKResourceManager& resource_manager, VKMemoryManager& memory_manager,
VKScheduler& scheduler, VKStagingBufferPool& staging_pool,
GPUVAddr gpu_addr, const SurfaceParams& params)
- : SurfaceBase<View>{gpu_addr, params}, system{system}, device{device},
- resource_manager{resource_manager}, memory_manager{memory_manager}, scheduler{scheduler},
- staging_pool{staging_pool} {
+ : SurfaceBase<View>{gpu_addr, params, device.IsOptimalAstcSupported()}, system{system},
+ device{device}, resource_manager{resource_manager},
+ memory_manager{memory_manager}, scheduler{scheduler}, staging_pool{staging_pool} {
if (params.IsBuffer()) {
- buffer = CreateBuffer(device, params);
+ buffer = CreateBuffer(device, params, host_memory_size);
commit = memory_manager.Commit(*buffer, false);
- const auto buffer_view_ci = GenerateBufferViewCreateInfo(device, params, *buffer);
+ const auto buffer_view_ci =
+ GenerateBufferViewCreateInfo(device, params, *buffer, host_memory_size);
format = buffer_view_ci.format;
const auto dev = device.GetLogical();
@@ -255,7 +257,7 @@ void CachedSurface::UploadBuffer(const std::vector<u8>& staging_buffer) {
std::memcpy(src_buffer.commit->Map(host_memory_size), staging_buffer.data(), host_memory_size);
scheduler.Record([src_buffer = *src_buffer.handle, dst_buffer = *buffer,
- size = params.GetHostSizeInBytes()](auto cmdbuf, auto& dld) {
+ size = host_memory_size](auto cmdbuf, auto& dld) {
const vk::BufferCopy copy(0, 0, size);
cmdbuf.copyBuffer(src_buffer, dst_buffer, {copy}, dld);
@@ -299,10 +301,7 @@ void CachedSurface::UploadImage(const std::vector<u8>& staging_buffer) {
vk::BufferImageCopy CachedSurface::GetBufferImageCopy(u32 level) const {
const u32 vk_depth = params.target == SurfaceTarget::Texture3D ? params.GetMipDepth(level) : 1;
- const auto compression_type = params.GetCompressionType();
- const std::size_t mip_offset = compression_type == SurfaceCompression::Converted
- ? params.GetConvertedMipmapOffset(level)
- : params.GetHostMipmapLevelOffset(level);
+ const std::size_t mip_offset = params.GetHostMipmapLevelOffset(level, is_converted);
return vk::BufferImageCopy(
mip_offset, 0, 0,
@@ -390,8 +389,9 @@ VKTextureCache::VKTextureCache(Core::System& system, VideoCore::RasterizerInterf
const VKDevice& device, VKResourceManager& resource_manager,
VKMemoryManager& memory_manager, VKScheduler& scheduler,
VKStagingBufferPool& staging_pool)
- : TextureCache(system, rasterizer), device{device}, resource_manager{resource_manager},
- memory_manager{memory_manager}, scheduler{scheduler}, staging_pool{staging_pool} {}
+ : TextureCache(system, rasterizer, device.IsOptimalAstcSupported()), device{device},
+ resource_manager{resource_manager}, memory_manager{memory_manager}, scheduler{scheduler},
+ staging_pool{staging_pool} {}
VKTextureCache::~VKTextureCache() = default;
diff --git a/src/video_core/renderer_vulkan/wrapper.cpp b/src/video_core/renderer_vulkan/wrapper.cpp
new file mode 100644
index 000000000..9b94dfff1
--- /dev/null
+++ b/src/video_core/renderer_vulkan/wrapper.cpp
@@ -0,0 +1,750 @@
+// Copyright 2020 yuzu Emulator Project
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#include <exception>
+#include <memory>
+#include <optional>
+#include <utility>
+#include <vector>
+
+#include "common/common_types.h"
+
+#include "video_core/renderer_vulkan/wrapper.h"
+
+namespace Vulkan::vk {
+
+namespace {
+
+template <typename T>
+bool Proc(T& result, const InstanceDispatch& dld, const char* proc_name,
+ VkInstance instance = nullptr) noexcept {
+ result = reinterpret_cast<T>(dld.vkGetInstanceProcAddr(instance, proc_name));
+ return result != nullptr;
+}
+
+template <typename T>
+void Proc(T& result, const DeviceDispatch& dld, const char* proc_name, VkDevice device) noexcept {
+ result = reinterpret_cast<T>(dld.vkGetDeviceProcAddr(device, proc_name));
+}
+
+void Load(VkDevice device, DeviceDispatch& dld) noexcept {
+#define X(name) Proc(dld.name, dld, #name, device)
+ X(vkAcquireNextImageKHR);
+ X(vkAllocateCommandBuffers);
+ X(vkAllocateDescriptorSets);
+ X(vkAllocateMemory);
+ X(vkBeginCommandBuffer);
+ X(vkBindBufferMemory);
+ X(vkBindImageMemory);
+ X(vkCmdBeginQuery);
+ X(vkCmdBeginRenderPass);
+ X(vkCmdBeginTransformFeedbackEXT);
+ X(vkCmdBindDescriptorSets);
+ X(vkCmdBindIndexBuffer);
+ X(vkCmdBindPipeline);
+ X(vkCmdBindTransformFeedbackBuffersEXT);
+ X(vkCmdBindVertexBuffers);
+ X(vkCmdBlitImage);
+ X(vkCmdClearAttachments);
+ X(vkCmdCopyBuffer);
+ X(vkCmdCopyBufferToImage);
+ X(vkCmdCopyImage);
+ X(vkCmdCopyImageToBuffer);
+ X(vkCmdDispatch);
+ X(vkCmdDraw);
+ X(vkCmdDrawIndexed);
+ X(vkCmdEndQuery);
+ X(vkCmdEndRenderPass);
+ X(vkCmdEndTransformFeedbackEXT);
+ X(vkCmdFillBuffer);
+ X(vkCmdPipelineBarrier);
+ X(vkCmdPushConstants);
+ X(vkCmdSetBlendConstants);
+ X(vkCmdSetCheckpointNV);
+ X(vkCmdSetDepthBias);
+ X(vkCmdSetDepthBounds);
+ X(vkCmdSetScissor);
+ X(vkCmdSetStencilCompareMask);
+ X(vkCmdSetStencilReference);
+ X(vkCmdSetStencilWriteMask);
+ X(vkCmdSetViewport);
+ X(vkCreateBuffer);
+ X(vkCreateBufferView);
+ X(vkCreateCommandPool);
+ X(vkCreateComputePipelines);
+ X(vkCreateDescriptorPool);
+ X(vkCreateDescriptorSetLayout);
+ X(vkCreateDescriptorUpdateTemplateKHR);
+ X(vkCreateFence);
+ X(vkCreateFramebuffer);
+ X(vkCreateGraphicsPipelines);
+ X(vkCreateImage);
+ X(vkCreateImageView);
+ X(vkCreatePipelineLayout);
+ X(vkCreateQueryPool);
+ X(vkCreateRenderPass);
+ X(vkCreateSampler);
+ X(vkCreateSemaphore);
+ X(vkCreateShaderModule);
+ X(vkCreateSwapchainKHR);
+ X(vkDestroyBuffer);
+ X(vkDestroyBufferView);
+ X(vkDestroyCommandPool);
+ X(vkDestroyDescriptorPool);
+ X(vkDestroyDescriptorSetLayout);
+ X(vkDestroyDescriptorUpdateTemplateKHR);
+ X(vkDestroyFence);
+ X(vkDestroyFramebuffer);
+ X(vkDestroyImage);
+ X(vkDestroyImageView);
+ X(vkDestroyPipeline);
+ X(vkDestroyPipelineLayout);
+ X(vkDestroyQueryPool);
+ X(vkDestroyRenderPass);
+ X(vkDestroySampler);
+ X(vkDestroySemaphore);
+ X(vkDestroyShaderModule);
+ X(vkDestroySwapchainKHR);
+ X(vkDeviceWaitIdle);
+ X(vkEndCommandBuffer);
+ X(vkFreeCommandBuffers);
+ X(vkFreeDescriptorSets);
+ X(vkFreeMemory);
+ X(vkGetBufferMemoryRequirements);
+ X(vkGetDeviceQueue);
+ X(vkGetFenceStatus);
+ X(vkGetImageMemoryRequirements);
+ X(vkGetQueryPoolResults);
+ X(vkGetQueueCheckpointDataNV);
+ X(vkMapMemory);
+ X(vkQueueSubmit);
+ X(vkResetFences);
+ X(vkResetQueryPoolEXT);
+ X(vkUnmapMemory);
+ X(vkUpdateDescriptorSetWithTemplateKHR);
+ X(vkUpdateDescriptorSets);
+ X(vkWaitForFences);
+#undef X
+}
+
+} // Anonymous namespace
+
+bool Load(InstanceDispatch& dld) noexcept {
+#define X(name) Proc(dld.name, dld, #name)
+ return X(vkCreateInstance) && X(vkEnumerateInstanceExtensionProperties);
+#undef X
+}
+
+bool Load(VkInstance instance, InstanceDispatch& dld) noexcept {
+#define X(name) Proc(dld.name, dld, #name, instance)
+ // These functions may fail to load depending on the enabled extensions.
+ // Don't return a failure on these.
+ X(vkCreateDebugUtilsMessengerEXT);
+ X(vkDestroyDebugUtilsMessengerEXT);
+ X(vkDestroySurfaceKHR);
+ X(vkGetPhysicalDeviceFeatures2KHR);
+ X(vkGetPhysicalDeviceProperties2KHR);
+ X(vkGetPhysicalDeviceSurfaceCapabilitiesKHR);
+ X(vkGetPhysicalDeviceSurfaceFormatsKHR);
+ X(vkGetPhysicalDeviceSurfacePresentModesKHR);
+ X(vkGetPhysicalDeviceSurfaceSupportKHR);
+ X(vkGetSwapchainImagesKHR);
+ X(vkQueuePresentKHR);
+
+ return X(vkCreateDevice) && X(vkDestroyDevice) && X(vkDestroyDevice) &&
+ X(vkEnumerateDeviceExtensionProperties) && X(vkEnumeratePhysicalDevices) &&
+ X(vkGetDeviceProcAddr) && X(vkGetPhysicalDeviceFormatProperties) &&
+ X(vkGetPhysicalDeviceMemoryProperties) && X(vkGetPhysicalDeviceProperties) &&
+ X(vkGetPhysicalDeviceQueueFamilyProperties);
+#undef X
+}
+
+const char* Exception::what() const noexcept {
+ return ToString(result);
+}
+
+const char* ToString(VkResult result) noexcept {
+ switch (result) {
+ case VkResult::VK_SUCCESS:
+ return "VK_SUCCESS";
+ case VkResult::VK_NOT_READY:
+ return "VK_NOT_READY";
+ case VkResult::VK_TIMEOUT:
+ return "VK_TIMEOUT";
+ case VkResult::VK_EVENT_SET:
+ return "VK_EVENT_SET";
+ case VkResult::VK_EVENT_RESET:
+ return "VK_EVENT_RESET";
+ case VkResult::VK_INCOMPLETE:
+ return "VK_INCOMPLETE";
+ case VkResult::VK_ERROR_OUT_OF_HOST_MEMORY:
+ return "VK_ERROR_OUT_OF_HOST_MEMORY";
+ case VkResult::VK_ERROR_OUT_OF_DEVICE_MEMORY:
+ return "VK_ERROR_OUT_OF_DEVICE_MEMORY";
+ case VkResult::VK_ERROR_INITIALIZATION_FAILED:
+ return "VK_ERROR_INITIALIZATION_FAILED";
+ case VkResult::VK_ERROR_DEVICE_LOST:
+ return "VK_ERROR_DEVICE_LOST";
+ case VkResult::VK_ERROR_MEMORY_MAP_FAILED:
+ return "VK_ERROR_MEMORY_MAP_FAILED";
+ case VkResult::VK_ERROR_LAYER_NOT_PRESENT:
+ return "VK_ERROR_LAYER_NOT_PRESENT";
+ case VkResult::VK_ERROR_EXTENSION_NOT_PRESENT:
+ return "VK_ERROR_EXTENSION_NOT_PRESENT";
+ case VkResult::VK_ERROR_FEATURE_NOT_PRESENT:
+ return "VK_ERROR_FEATURE_NOT_PRESENT";
+ case VkResult::VK_ERROR_INCOMPATIBLE_DRIVER:
+ return "VK_ERROR_INCOMPATIBLE_DRIVER";
+ case VkResult::VK_ERROR_TOO_MANY_OBJECTS:
+ return "VK_ERROR_TOO_MANY_OBJECTS";
+ case VkResult::VK_ERROR_FORMAT_NOT_SUPPORTED:
+ return "VK_ERROR_FORMAT_NOT_SUPPORTED";
+ case VkResult::VK_ERROR_FRAGMENTED_POOL:
+ return "VK_ERROR_FRAGMENTED_POOL";
+ case VkResult::VK_ERROR_OUT_OF_POOL_MEMORY:
+ return "VK_ERROR_OUT_OF_POOL_MEMORY";
+ case VkResult::VK_ERROR_INVALID_EXTERNAL_HANDLE:
+ return "VK_ERROR_INVALID_EXTERNAL_HANDLE";
+ case VkResult::VK_ERROR_SURFACE_LOST_KHR:
+ return "VK_ERROR_SURFACE_LOST_KHR";
+ case VkResult::VK_ERROR_NATIVE_WINDOW_IN_USE_KHR:
+ return "VK_ERROR_NATIVE_WINDOW_IN_USE_KHR";
+ case VkResult::VK_SUBOPTIMAL_KHR:
+ return "VK_SUBOPTIMAL_KHR";
+ case VkResult::VK_ERROR_OUT_OF_DATE_KHR:
+ return "VK_ERROR_OUT_OF_DATE_KHR";
+ case VkResult::VK_ERROR_INCOMPATIBLE_DISPLAY_KHR:
+ return "VK_ERROR_INCOMPATIBLE_DISPLAY_KHR";
+ case VkResult::VK_ERROR_VALIDATION_FAILED_EXT:
+ return "VK_ERROR_VALIDATION_FAILED_EXT";
+ case VkResult::VK_ERROR_INVALID_SHADER_NV:
+ return "VK_ERROR_INVALID_SHADER_NV";
+ case VkResult::VK_ERROR_INVALID_DRM_FORMAT_MODIFIER_PLANE_LAYOUT_EXT:
+ return "VK_ERROR_INVALID_DRM_FORMAT_MODIFIER_PLANE_LAYOUT_EXT";
+ case VkResult::VK_ERROR_FRAGMENTATION_EXT:
+ return "VK_ERROR_FRAGMENTATION_EXT";
+ case VkResult::VK_ERROR_NOT_PERMITTED_EXT:
+ return "VK_ERROR_NOT_PERMITTED_EXT";
+ case VkResult::VK_ERROR_INVALID_DEVICE_ADDRESS_EXT:
+ return "VK_ERROR_INVALID_DEVICE_ADDRESS_EXT";
+ case VkResult::VK_ERROR_FULL_SCREEN_EXCLUSIVE_MODE_LOST_EXT:
+ return "VK_ERROR_FULL_SCREEN_EXCLUSIVE_MODE_LOST_EXT";
+ }
+ return "Unknown";
+}
+
+void Destroy(VkInstance instance, const InstanceDispatch& dld) noexcept {
+ dld.vkDestroyInstance(instance, nullptr);
+}
+
+void Destroy(VkDevice device, const InstanceDispatch& dld) noexcept {
+ dld.vkDestroyDevice(device, nullptr);
+}
+
+void Destroy(VkDevice device, VkBuffer handle, const DeviceDispatch& dld) noexcept {
+ dld.vkDestroyBuffer(device, handle, nullptr);
+}
+
+void Destroy(VkDevice device, VkBufferView handle, const DeviceDispatch& dld) noexcept {
+ dld.vkDestroyBufferView(device, handle, nullptr);
+}
+
+void Destroy(VkDevice device, VkCommandPool handle, const DeviceDispatch& dld) noexcept {
+ dld.vkDestroyCommandPool(device, handle, nullptr);
+}
+
+void Destroy(VkDevice device, VkDescriptorPool handle, const DeviceDispatch& dld) noexcept {
+ dld.vkDestroyDescriptorPool(device, handle, nullptr);
+}
+
+void Destroy(VkDevice device, VkDescriptorSetLayout handle, const DeviceDispatch& dld) noexcept {
+ dld.vkDestroyDescriptorSetLayout(device, handle, nullptr);
+}
+
+void Destroy(VkDevice device, VkDescriptorUpdateTemplateKHR handle,
+ const DeviceDispatch& dld) noexcept {
+ dld.vkDestroyDescriptorUpdateTemplateKHR(device, handle, nullptr);
+}
+
+void Destroy(VkDevice device, VkDeviceMemory handle, const DeviceDispatch& dld) noexcept {
+ dld.vkFreeMemory(device, handle, nullptr);
+}
+
+void Destroy(VkDevice device, VkFence handle, const DeviceDispatch& dld) noexcept {
+ dld.vkDestroyFence(device, handle, nullptr);
+}
+
+void Destroy(VkDevice device, VkFramebuffer handle, const DeviceDispatch& dld) noexcept {
+ dld.vkDestroyFramebuffer(device, handle, nullptr);
+}
+
+void Destroy(VkDevice device, VkImage handle, const DeviceDispatch& dld) noexcept {
+ dld.vkDestroyImage(device, handle, nullptr);
+}
+
+void Destroy(VkDevice device, VkImageView handle, const DeviceDispatch& dld) noexcept {
+ dld.vkDestroyImageView(device, handle, nullptr);
+}
+
+void Destroy(VkDevice device, VkPipeline handle, const DeviceDispatch& dld) noexcept {
+ dld.vkDestroyPipeline(device, handle, nullptr);
+}
+
+void Destroy(VkDevice device, VkPipelineLayout handle, const DeviceDispatch& dld) noexcept {
+ dld.vkDestroyPipelineLayout(device, handle, nullptr);
+}
+
+void Destroy(VkDevice device, VkQueryPool handle, const DeviceDispatch& dld) noexcept {
+ dld.vkDestroyQueryPool(device, handle, nullptr);
+}
+
+void Destroy(VkDevice device, VkRenderPass handle, const DeviceDispatch& dld) noexcept {
+ dld.vkDestroyRenderPass(device, handle, nullptr);
+}
+
+void Destroy(VkDevice device, VkSampler handle, const DeviceDispatch& dld) noexcept {
+ dld.vkDestroySampler(device, handle, nullptr);
+}
+
+void Destroy(VkDevice device, VkSwapchainKHR handle, const DeviceDispatch& dld) noexcept {
+ dld.vkDestroySwapchainKHR(device, handle, nullptr);
+}
+
+void Destroy(VkDevice device, VkSemaphore handle, const DeviceDispatch& dld) noexcept {
+ dld.vkDestroySemaphore(device, handle, nullptr);
+}
+
+void Destroy(VkDevice device, VkShaderModule handle, const DeviceDispatch& dld) noexcept {
+ dld.vkDestroyShaderModule(device, handle, nullptr);
+}
+
+void Destroy(VkInstance instance, VkDebugUtilsMessengerEXT handle,
+ const InstanceDispatch& dld) noexcept {
+ dld.vkDestroyDebugUtilsMessengerEXT(instance, handle, nullptr);
+}
+
+void Destroy(VkInstance instance, VkSurfaceKHR handle, const InstanceDispatch& dld) noexcept {
+ dld.vkDestroySurfaceKHR(instance, handle, nullptr);
+}
+
+VkResult Free(VkDevice device, VkDescriptorPool handle, Span<VkDescriptorSet> sets,
+ const DeviceDispatch& dld) noexcept {
+ return dld.vkFreeDescriptorSets(device, handle, sets.size(), sets.data());
+}
+
+VkResult Free(VkDevice device, VkCommandPool handle, Span<VkCommandBuffer> buffers,
+ const DeviceDispatch& dld) noexcept {
+ dld.vkFreeCommandBuffers(device, handle, buffers.size(), buffers.data());
+ return VK_SUCCESS;
+}
+
+Instance Instance::Create(Span<const char*> layers, Span<const char*> extensions,
+ InstanceDispatch& dld) noexcept {
+ VkApplicationInfo application_info;
+ application_info.sType = VK_STRUCTURE_TYPE_APPLICATION_INFO;
+ application_info.pNext = nullptr;
+ application_info.pApplicationName = "yuzu Emulator";
+ application_info.applicationVersion = VK_MAKE_VERSION(0, 1, 0);
+ application_info.pEngineName = "yuzu Emulator";
+ application_info.engineVersion = VK_MAKE_VERSION(0, 1, 0);
+ application_info.apiVersion = VK_API_VERSION_1_1;
+
+ VkInstanceCreateInfo ci;
+ ci.sType = VK_STRUCTURE_TYPE_INSTANCE_CREATE_INFO;
+ ci.pNext = nullptr;
+ ci.flags = 0;
+ ci.pApplicationInfo = &application_info;
+ ci.enabledLayerCount = layers.size();
+ ci.ppEnabledLayerNames = layers.data();
+ ci.enabledExtensionCount = extensions.size();
+ ci.ppEnabledExtensionNames = extensions.data();
+
+ VkInstance instance;
+ if (dld.vkCreateInstance(&ci, nullptr, &instance) != VK_SUCCESS) {
+ // Failed to create the instance.
+ return {};
+ }
+ if (!Proc(dld.vkDestroyInstance, dld, "vkDestroyInstance", instance)) {
+ // We successfully created an instance but the destroy function couldn't be loaded.
+ // This is a good moment to panic.
+ return {};
+ }
+
+ return Instance(instance, dld);
+}
+
+std::optional<std::vector<VkPhysicalDevice>> Instance::EnumeratePhysicalDevices() {
+ u32 num;
+ if (dld->vkEnumeratePhysicalDevices(handle, &num, nullptr) != VK_SUCCESS) {
+ return std::nullopt;
+ }
+ std::vector<VkPhysicalDevice> physical_devices(num);
+ if (dld->vkEnumeratePhysicalDevices(handle, &num, physical_devices.data()) != VK_SUCCESS) {
+ return std::nullopt;
+ }
+ return physical_devices;
+}
+
+DebugCallback Instance::TryCreateDebugCallback(
+ PFN_vkDebugUtilsMessengerCallbackEXT callback) noexcept {
+ VkDebugUtilsMessengerCreateInfoEXT ci;
+ ci.sType = VK_STRUCTURE_TYPE_DEBUG_UTILS_MESSENGER_CREATE_INFO_EXT;
+ ci.pNext = nullptr;
+ ci.flags = 0;
+ ci.messageSeverity = VK_DEBUG_UTILS_MESSAGE_SEVERITY_ERROR_BIT_EXT |
+ VK_DEBUG_UTILS_MESSAGE_SEVERITY_WARNING_BIT_EXT |
+ VK_DEBUG_UTILS_MESSAGE_SEVERITY_INFO_BIT_EXT |
+ VK_DEBUG_UTILS_MESSAGE_SEVERITY_VERBOSE_BIT_EXT;
+ ci.messageType = VK_DEBUG_UTILS_MESSAGE_TYPE_GENERAL_BIT_EXT |
+ VK_DEBUG_UTILS_MESSAGE_TYPE_VALIDATION_BIT_EXT |
+ VK_DEBUG_UTILS_MESSAGE_TYPE_PERFORMANCE_BIT_EXT;
+ ci.pfnUserCallback = callback;
+ ci.pUserData = nullptr;
+
+ VkDebugUtilsMessengerEXT messenger;
+ if (dld->vkCreateDebugUtilsMessengerEXT(handle, &ci, nullptr, &messenger) != VK_SUCCESS) {
+ return {};
+ }
+ return DebugCallback(messenger, handle, *dld);
+}
+
+std::vector<VkCheckpointDataNV> Queue::GetCheckpointDataNV(const DeviceDispatch& dld) const {
+ if (!dld.vkGetQueueCheckpointDataNV) {
+ return {};
+ }
+ u32 num;
+ dld.vkGetQueueCheckpointDataNV(queue, &num, nullptr);
+ std::vector<VkCheckpointDataNV> checkpoints(num);
+ dld.vkGetQueueCheckpointDataNV(queue, &num, checkpoints.data());
+ return checkpoints;
+}
+
+void Buffer::BindMemory(VkDeviceMemory memory, VkDeviceSize offset) const {
+ Check(dld->vkBindBufferMemory(owner, handle, memory, offset));
+}
+
+void Image::BindMemory(VkDeviceMemory memory, VkDeviceSize offset) const {
+ Check(dld->vkBindImageMemory(owner, handle, memory, offset));
+}
+
+DescriptorSets DescriptorPool::Allocate(const VkDescriptorSetAllocateInfo& ai) const {
+ const std::size_t num = ai.descriptorSetCount;
+ std::unique_ptr sets = std::make_unique<VkDescriptorSet[]>(num);
+ switch (const VkResult result = dld->vkAllocateDescriptorSets(owner, &ai, sets.get())) {
+ case VK_SUCCESS:
+ return DescriptorSets(std::move(sets), num, owner, handle, *dld);
+ case VK_ERROR_OUT_OF_POOL_MEMORY:
+ return {};
+ default:
+ throw Exception(result);
+ }
+}
+
+CommandBuffers CommandPool::Allocate(std::size_t num_buffers, VkCommandBufferLevel level) const {
+ VkCommandBufferAllocateInfo ai;
+ ai.sType = VK_STRUCTURE_TYPE_COMMAND_BUFFER_ALLOCATE_INFO;
+ ai.pNext = nullptr;
+ ai.commandPool = handle;
+ ai.level = level;
+ ai.commandBufferCount = static_cast<u32>(num_buffers);
+
+ std::unique_ptr buffers = std::make_unique<VkCommandBuffer[]>(num_buffers);
+ switch (const VkResult result = dld->vkAllocateCommandBuffers(owner, &ai, buffers.get())) {
+ case VK_SUCCESS:
+ return CommandBuffers(std::move(buffers), num_buffers, owner, handle, *dld);
+ case VK_ERROR_OUT_OF_POOL_MEMORY:
+ return {};
+ default:
+ throw Exception(result);
+ }
+}
+
+std::vector<VkImage> SwapchainKHR::GetImages() const {
+ u32 num;
+ Check(dld->vkGetSwapchainImagesKHR(owner, handle, &num, nullptr));
+ std::vector<VkImage> images(num);
+ Check(dld->vkGetSwapchainImagesKHR(owner, handle, &num, images.data()));
+ return images;
+}
+
+Device Device::Create(VkPhysicalDevice physical_device, Span<VkDeviceQueueCreateInfo> queues_ci,
+ Span<const char*> enabled_extensions,
+ const VkPhysicalDeviceFeatures2& enabled_features,
+ DeviceDispatch& dld) noexcept {
+ VkDeviceCreateInfo ci;
+ ci.sType = VK_STRUCTURE_TYPE_DEVICE_CREATE_INFO;
+ ci.pNext = &enabled_features;
+ ci.flags = 0;
+ ci.queueCreateInfoCount = queues_ci.size();
+ ci.pQueueCreateInfos = queues_ci.data();
+ ci.enabledLayerCount = 0;
+ ci.ppEnabledLayerNames = nullptr;
+ ci.enabledExtensionCount = enabled_extensions.size();
+ ci.ppEnabledExtensionNames = enabled_extensions.data();
+ ci.pEnabledFeatures = nullptr;
+
+ VkDevice device;
+ if (dld.vkCreateDevice(physical_device, &ci, nullptr, &device) != VK_SUCCESS) {
+ return {};
+ }
+ Load(device, dld);
+ return Device(device, dld);
+}
+
+Queue Device::GetQueue(u32 family_index) const noexcept {
+ VkQueue queue;
+ dld->vkGetDeviceQueue(handle, family_index, 0, &queue);
+ return Queue(queue, *dld);
+}
+
+Buffer Device::CreateBuffer(const VkBufferCreateInfo& ci) const {
+ VkBuffer object;
+ Check(dld->vkCreateBuffer(handle, &ci, nullptr, &object));
+ return Buffer(object, handle, *dld);
+}
+
+BufferView Device::CreateBufferView(const VkBufferViewCreateInfo& ci) const {
+ VkBufferView object;
+ Check(dld->vkCreateBufferView(handle, &ci, nullptr, &object));
+ return BufferView(object, handle, *dld);
+}
+
+Image Device::CreateImage(const VkImageCreateInfo& ci) const {
+ VkImage object;
+ Check(dld->vkCreateImage(handle, &ci, nullptr, &object));
+ return Image(object, handle, *dld);
+}
+
+ImageView Device::CreateImageView(const VkImageViewCreateInfo& ci) const {
+ VkImageView object;
+ Check(dld->vkCreateImageView(handle, &ci, nullptr, &object));
+ return ImageView(object, handle, *dld);
+}
+
+Semaphore Device::CreateSemaphore() const {
+ VkSemaphoreCreateInfo ci;
+ ci.sType = VK_STRUCTURE_TYPE_SEMAPHORE_CREATE_INFO;
+ ci.pNext = nullptr;
+ ci.flags = 0;
+
+ VkSemaphore object;
+ Check(dld->vkCreateSemaphore(handle, &ci, nullptr, &object));
+ return Semaphore(object, handle, *dld);
+}
+
+Fence Device::CreateFence(const VkFenceCreateInfo& ci) const {
+ VkFence object;
+ Check(dld->vkCreateFence(handle, &ci, nullptr, &object));
+ return Fence(object, handle, *dld);
+}
+
+DescriptorPool Device::CreateDescriptorPool(const VkDescriptorPoolCreateInfo& ci) const {
+ VkDescriptorPool object;
+ Check(dld->vkCreateDescriptorPool(handle, &ci, nullptr, &object));
+ return DescriptorPool(object, handle, *dld);
+}
+
+RenderPass Device::CreateRenderPass(const VkRenderPassCreateInfo& ci) const {
+ VkRenderPass object;
+ Check(dld->vkCreateRenderPass(handle, &ci, nullptr, &object));
+ return RenderPass(object, handle, *dld);
+}
+
+DescriptorSetLayout Device::CreateDescriptorSetLayout(
+ const VkDescriptorSetLayoutCreateInfo& ci) const {
+ VkDescriptorSetLayout object;
+ Check(dld->vkCreateDescriptorSetLayout(handle, &ci, nullptr, &object));
+ return DescriptorSetLayout(object, handle, *dld);
+}
+
+PipelineLayout Device::CreatePipelineLayout(const VkPipelineLayoutCreateInfo& ci) const {
+ VkPipelineLayout object;
+ Check(dld->vkCreatePipelineLayout(handle, &ci, nullptr, &object));
+ return PipelineLayout(object, handle, *dld);
+}
+
+Pipeline Device::CreateGraphicsPipeline(const VkGraphicsPipelineCreateInfo& ci) const {
+ VkPipeline object;
+ Check(dld->vkCreateGraphicsPipelines(handle, nullptr, 1, &ci, nullptr, &object));
+ return Pipeline(object, handle, *dld);
+}
+
+Pipeline Device::CreateComputePipeline(const VkComputePipelineCreateInfo& ci) const {
+ VkPipeline object;
+ Check(dld->vkCreateComputePipelines(handle, nullptr, 1, &ci, nullptr, &object));
+ return Pipeline(object, handle, *dld);
+}
+
+Sampler Device::CreateSampler(const VkSamplerCreateInfo& ci) const {
+ VkSampler object;
+ Check(dld->vkCreateSampler(handle, &ci, nullptr, &object));
+ return Sampler(object, handle, *dld);
+}
+
+Framebuffer Device::CreateFramebuffer(const VkFramebufferCreateInfo& ci) const {
+ VkFramebuffer object;
+ Check(dld->vkCreateFramebuffer(handle, &ci, nullptr, &object));
+ return Framebuffer(object, handle, *dld);
+}
+
+CommandPool Device::CreateCommandPool(const VkCommandPoolCreateInfo& ci) const {
+ VkCommandPool object;
+ Check(dld->vkCreateCommandPool(handle, &ci, nullptr, &object));
+ return CommandPool(object, handle, *dld);
+}
+
+DescriptorUpdateTemplateKHR Device::CreateDescriptorUpdateTemplateKHR(
+ const VkDescriptorUpdateTemplateCreateInfoKHR& ci) const {
+ VkDescriptorUpdateTemplateKHR object;
+ Check(dld->vkCreateDescriptorUpdateTemplateKHR(handle, &ci, nullptr, &object));
+ return DescriptorUpdateTemplateKHR(object, handle, *dld);
+}
+
+QueryPool Device::CreateQueryPool(const VkQueryPoolCreateInfo& ci) const {
+ VkQueryPool object;
+ Check(dld->vkCreateQueryPool(handle, &ci, nullptr, &object));
+ return QueryPool(object, handle, *dld);
+}
+
+ShaderModule Device::CreateShaderModule(const VkShaderModuleCreateInfo& ci) const {
+ VkShaderModule object;
+ Check(dld->vkCreateShaderModule(handle, &ci, nullptr, &object));
+ return ShaderModule(object, handle, *dld);
+}
+
+SwapchainKHR Device::CreateSwapchainKHR(const VkSwapchainCreateInfoKHR& ci) const {
+ VkSwapchainKHR object;
+ Check(dld->vkCreateSwapchainKHR(handle, &ci, nullptr, &object));
+ return SwapchainKHR(object, handle, *dld);
+}
+
+DeviceMemory Device::TryAllocateMemory(const VkMemoryAllocateInfo& ai) const noexcept {
+ VkDeviceMemory memory;
+ if (dld->vkAllocateMemory(handle, &ai, nullptr, &memory) != VK_SUCCESS) {
+ return {};
+ }
+ return DeviceMemory(memory, handle, *dld);
+}
+
+DeviceMemory Device::AllocateMemory(const VkMemoryAllocateInfo& ai) const {
+ VkDeviceMemory memory;
+ Check(dld->vkAllocateMemory(handle, &ai, nullptr, &memory));
+ return DeviceMemory(memory, handle, *dld);
+}
+
+VkMemoryRequirements Device::GetBufferMemoryRequirements(VkBuffer buffer) const noexcept {
+ VkMemoryRequirements requirements;
+ dld->vkGetBufferMemoryRequirements(handle, buffer, &requirements);
+ return requirements;
+}
+
+VkMemoryRequirements Device::GetImageMemoryRequirements(VkImage image) const noexcept {
+ VkMemoryRequirements requirements;
+ dld->vkGetImageMemoryRequirements(handle, image, &requirements);
+ return requirements;
+}
+
+void Device::UpdateDescriptorSets(Span<VkWriteDescriptorSet> writes,
+ Span<VkCopyDescriptorSet> copies) const noexcept {
+ dld->vkUpdateDescriptorSets(handle, writes.size(), writes.data(), copies.size(), copies.data());
+}
+
+VkPhysicalDeviceProperties PhysicalDevice::GetProperties() const noexcept {
+ VkPhysicalDeviceProperties properties;
+ dld->vkGetPhysicalDeviceProperties(physical_device, &properties);
+ return properties;
+}
+
+void PhysicalDevice::GetProperties2KHR(VkPhysicalDeviceProperties2KHR& properties) const noexcept {
+ dld->vkGetPhysicalDeviceProperties2KHR(physical_device, &properties);
+}
+
+VkPhysicalDeviceFeatures PhysicalDevice::GetFeatures() const noexcept {
+ VkPhysicalDeviceFeatures2KHR features2;
+ features2.sType = VK_STRUCTURE_TYPE_PHYSICAL_DEVICE_FEATURES_2_KHR;
+ features2.pNext = nullptr;
+ dld->vkGetPhysicalDeviceFeatures2KHR(physical_device, &features2);
+ return features2.features;
+}
+
+void PhysicalDevice::GetFeatures2KHR(VkPhysicalDeviceFeatures2KHR& features) const noexcept {
+ dld->vkGetPhysicalDeviceFeatures2KHR(physical_device, &features);
+}
+
+VkFormatProperties PhysicalDevice::GetFormatProperties(VkFormat format) const noexcept {
+ VkFormatProperties properties;
+ dld->vkGetPhysicalDeviceFormatProperties(physical_device, format, &properties);
+ return properties;
+}
+
+std::vector<VkExtensionProperties> PhysicalDevice::EnumerateDeviceExtensionProperties() const {
+ u32 num;
+ dld->vkEnumerateDeviceExtensionProperties(physical_device, nullptr, &num, nullptr);
+ std::vector<VkExtensionProperties> properties(num);
+ dld->vkEnumerateDeviceExtensionProperties(physical_device, nullptr, &num, properties.data());
+ return properties;
+}
+
+std::vector<VkQueueFamilyProperties> PhysicalDevice::GetQueueFamilyProperties() const {
+ u32 num;
+ dld->vkGetPhysicalDeviceQueueFamilyProperties(physical_device, &num, nullptr);
+ std::vector<VkQueueFamilyProperties> properties(num);
+ dld->vkGetPhysicalDeviceQueueFamilyProperties(physical_device, &num, properties.data());
+ return properties;
+}
+
+bool PhysicalDevice::GetSurfaceSupportKHR(u32 queue_family_index, VkSurfaceKHR surface) const {
+ VkBool32 supported;
+ Check(dld->vkGetPhysicalDeviceSurfaceSupportKHR(physical_device, queue_family_index, surface,
+ &supported));
+ return supported == VK_TRUE;
+}
+
+VkSurfaceCapabilitiesKHR PhysicalDevice::GetSurfaceCapabilitiesKHR(VkSurfaceKHR surface) const
+ noexcept {
+ VkSurfaceCapabilitiesKHR capabilities;
+ Check(dld->vkGetPhysicalDeviceSurfaceCapabilitiesKHR(physical_device, surface, &capabilities));
+ return capabilities;
+}
+
+std::vector<VkSurfaceFormatKHR> PhysicalDevice::GetSurfaceFormatsKHR(VkSurfaceKHR surface) const {
+ u32 num;
+ Check(dld->vkGetPhysicalDeviceSurfaceFormatsKHR(physical_device, surface, &num, nullptr));
+ std::vector<VkSurfaceFormatKHR> formats(num);
+ Check(
+ dld->vkGetPhysicalDeviceSurfaceFormatsKHR(physical_device, surface, &num, formats.data()));
+ return formats;
+}
+
+std::vector<VkPresentModeKHR> PhysicalDevice::GetSurfacePresentModesKHR(
+ VkSurfaceKHR surface) const {
+ u32 num;
+ Check(dld->vkGetPhysicalDeviceSurfacePresentModesKHR(physical_device, surface, &num, nullptr));
+ std::vector<VkPresentModeKHR> modes(num);
+ Check(dld->vkGetPhysicalDeviceSurfacePresentModesKHR(physical_device, surface, &num,
+ modes.data()));
+ return modes;
+}
+
+VkPhysicalDeviceMemoryProperties PhysicalDevice::GetMemoryProperties() const noexcept {
+ VkPhysicalDeviceMemoryProperties properties;
+ dld->vkGetPhysicalDeviceMemoryProperties(physical_device, &properties);
+ return properties;
+}
+
+std::optional<std::vector<VkExtensionProperties>> EnumerateInstanceExtensionProperties(
+ const InstanceDispatch& dld) {
+ u32 num;
+ if (dld.vkEnumerateInstanceExtensionProperties(nullptr, &num, nullptr) != VK_SUCCESS) {
+ return std::nullopt;
+ }
+ std::vector<VkExtensionProperties> properties(num);
+ if (dld.vkEnumerateInstanceExtensionProperties(nullptr, &num, properties.data()) !=
+ VK_SUCCESS) {
+ return std::nullopt;
+ }
+ return properties;
+}
+
+} // namespace Vulkan::vk
diff --git a/src/video_core/renderer_vulkan/wrapper.h b/src/video_core/renderer_vulkan/wrapper.h
new file mode 100644
index 000000000..fb3657819
--- /dev/null
+++ b/src/video_core/renderer_vulkan/wrapper.h
@@ -0,0 +1,987 @@
+// Copyright 2020 yuzu Emulator Project
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#pragma once
+
+#include <exception>
+#include <iterator>
+#include <limits>
+#include <memory>
+#include <optional>
+#include <type_traits>
+#include <utility>
+#include <vector>
+
+#define VK_NO_PROTOTYPES
+#include <vulkan/vulkan.h>
+
+#include "common/common_types.h"
+
+namespace Vulkan::vk {
+
+/**
+ * Span for Vulkan arrays.
+ * Based on std::span but optimized for array access instead of iterators.
+ * Size returns uint32_t instead of size_t to ease interaction with Vulkan functions.
+ */
+template <typename T>
+class Span {
+public:
+ using value_type = T;
+ using size_type = u32;
+ using difference_type = std::ptrdiff_t;
+ using reference = const T&;
+ using const_reference = const T&;
+ using pointer = const T*;
+ using const_pointer = const T*;
+ using iterator = const T*;
+ using const_iterator = const T*;
+
+ /// Construct an empty span.
+ constexpr Span() noexcept = default;
+
+ /// Construct a span from a single element.
+ constexpr Span(const T& value) noexcept : ptr{&value}, num{1} {}
+
+ /// Construct a span from a range.
+ template <typename Range>
+ // requires std::data(const Range&)
+ // requires std::size(const Range&)
+ constexpr Span(const Range& range) : ptr{std::data(range)}, num{std::size(range)} {}
+
+ /// Construct a span from a pointer and a size.
+ /// This is inteded for subranges.
+ constexpr Span(const T* ptr, std::size_t num) noexcept : ptr{ptr}, num{num} {}
+
+ /// Returns the data pointer by the span.
+ constexpr const T* data() const noexcept {
+ return ptr;
+ }
+
+ /// Returns the number of elements in the span.
+ /// @note Returns a 32 bits integer because most Vulkan functions expect this type.
+ constexpr u32 size() const noexcept {
+ return static_cast<u32>(num);
+ }
+
+ /// Returns true when the span is empty.
+ constexpr bool empty() const noexcept {
+ return num == 0;
+ }
+
+ /// Returns a reference to the element in the passed index.
+ /// @pre: index < size()
+ constexpr const T& operator[](std::size_t index) const noexcept {
+ return ptr[index];
+ }
+
+ /// Returns an iterator to the beginning of the span.
+ constexpr const T* begin() const noexcept {
+ return ptr;
+ }
+
+ /// Returns an iterator to the end of the span.
+ constexpr const T* end() const noexcept {
+ return ptr + num;
+ }
+
+ /// Returns an iterator to the beginning of the span.
+ constexpr const T* cbegin() const noexcept {
+ return ptr;
+ }
+
+ /// Returns an iterator to the end of the span.
+ constexpr const T* cend() const noexcept {
+ return ptr + num;
+ }
+
+private:
+ const T* ptr = nullptr;
+ std::size_t num = 0;
+};
+
+/// Vulkan exception generated from a VkResult.
+class Exception final : public std::exception {
+public:
+ /// Construct the exception with a result.
+ /// @pre result != VK_SUCCESS
+ explicit Exception(VkResult result_) : result{result_} {}
+ virtual ~Exception() = default;
+
+ const char* what() const noexcept override;
+
+private:
+ VkResult result;
+};
+
+/// Converts a VkResult enum into a rodata string
+const char* ToString(VkResult) noexcept;
+
+/// Throws a Vulkan exception if result is not success.
+inline void Check(VkResult result) {
+ if (result != VK_SUCCESS) {
+ throw Exception(result);
+ }
+}
+
+/// Throws a Vulkan exception if result is an error.
+/// @return result
+inline VkResult Filter(VkResult result) {
+ if (result < 0) {
+ throw Exception(result);
+ }
+ return result;
+}
+
+/// Table holding Vulkan instance function pointers.
+struct InstanceDispatch {
+ PFN_vkGetInstanceProcAddr vkGetInstanceProcAddr;
+
+ PFN_vkCreateInstance vkCreateInstance;
+ PFN_vkDestroyInstance vkDestroyInstance;
+ PFN_vkEnumerateInstanceExtensionProperties vkEnumerateInstanceExtensionProperties;
+
+ PFN_vkCreateDebugUtilsMessengerEXT vkCreateDebugUtilsMessengerEXT;
+ PFN_vkCreateDevice vkCreateDevice;
+ PFN_vkDestroyDebugUtilsMessengerEXT vkDestroyDebugUtilsMessengerEXT;
+ PFN_vkDestroyDevice vkDestroyDevice;
+ PFN_vkDestroySurfaceKHR vkDestroySurfaceKHR;
+ PFN_vkEnumerateDeviceExtensionProperties vkEnumerateDeviceExtensionProperties;
+ PFN_vkEnumeratePhysicalDevices vkEnumeratePhysicalDevices;
+ PFN_vkGetDeviceProcAddr vkGetDeviceProcAddr;
+ PFN_vkGetPhysicalDeviceFeatures2KHR vkGetPhysicalDeviceFeatures2KHR;
+ PFN_vkGetPhysicalDeviceFormatProperties vkGetPhysicalDeviceFormatProperties;
+ PFN_vkGetPhysicalDeviceMemoryProperties vkGetPhysicalDeviceMemoryProperties;
+ PFN_vkGetPhysicalDeviceProperties vkGetPhysicalDeviceProperties;
+ PFN_vkGetPhysicalDeviceProperties2KHR vkGetPhysicalDeviceProperties2KHR;
+ PFN_vkGetPhysicalDeviceQueueFamilyProperties vkGetPhysicalDeviceQueueFamilyProperties;
+ PFN_vkGetPhysicalDeviceSurfaceCapabilitiesKHR vkGetPhysicalDeviceSurfaceCapabilitiesKHR;
+ PFN_vkGetPhysicalDeviceSurfaceFormatsKHR vkGetPhysicalDeviceSurfaceFormatsKHR;
+ PFN_vkGetPhysicalDeviceSurfacePresentModesKHR vkGetPhysicalDeviceSurfacePresentModesKHR;
+ PFN_vkGetPhysicalDeviceSurfaceSupportKHR vkGetPhysicalDeviceSurfaceSupportKHR;
+ PFN_vkGetSwapchainImagesKHR vkGetSwapchainImagesKHR;
+ PFN_vkQueuePresentKHR vkQueuePresentKHR;
+};
+
+/// Table holding Vulkan device function pointers.
+struct DeviceDispatch : public InstanceDispatch {
+ PFN_vkAcquireNextImageKHR vkAcquireNextImageKHR;
+ PFN_vkAllocateCommandBuffers vkAllocateCommandBuffers;
+ PFN_vkAllocateDescriptorSets vkAllocateDescriptorSets;
+ PFN_vkAllocateMemory vkAllocateMemory;
+ PFN_vkBeginCommandBuffer vkBeginCommandBuffer;
+ PFN_vkBindBufferMemory vkBindBufferMemory;
+ PFN_vkBindImageMemory vkBindImageMemory;
+ PFN_vkCmdBeginQuery vkCmdBeginQuery;
+ PFN_vkCmdBeginRenderPass vkCmdBeginRenderPass;
+ PFN_vkCmdBeginTransformFeedbackEXT vkCmdBeginTransformFeedbackEXT;
+ PFN_vkCmdBindDescriptorSets vkCmdBindDescriptorSets;
+ PFN_vkCmdBindIndexBuffer vkCmdBindIndexBuffer;
+ PFN_vkCmdBindPipeline vkCmdBindPipeline;
+ PFN_vkCmdBindTransformFeedbackBuffersEXT vkCmdBindTransformFeedbackBuffersEXT;
+ PFN_vkCmdBindVertexBuffers vkCmdBindVertexBuffers;
+ PFN_vkCmdBlitImage vkCmdBlitImage;
+ PFN_vkCmdClearAttachments vkCmdClearAttachments;
+ PFN_vkCmdCopyBuffer vkCmdCopyBuffer;
+ PFN_vkCmdCopyBufferToImage vkCmdCopyBufferToImage;
+ PFN_vkCmdCopyImage vkCmdCopyImage;
+ PFN_vkCmdCopyImageToBuffer vkCmdCopyImageToBuffer;
+ PFN_vkCmdDispatch vkCmdDispatch;
+ PFN_vkCmdDraw vkCmdDraw;
+ PFN_vkCmdDrawIndexed vkCmdDrawIndexed;
+ PFN_vkCmdEndQuery vkCmdEndQuery;
+ PFN_vkCmdEndRenderPass vkCmdEndRenderPass;
+ PFN_vkCmdEndTransformFeedbackEXT vkCmdEndTransformFeedbackEXT;
+ PFN_vkCmdFillBuffer vkCmdFillBuffer;
+ PFN_vkCmdPipelineBarrier vkCmdPipelineBarrier;
+ PFN_vkCmdPushConstants vkCmdPushConstants;
+ PFN_vkCmdSetBlendConstants vkCmdSetBlendConstants;
+ PFN_vkCmdSetCheckpointNV vkCmdSetCheckpointNV;
+ PFN_vkCmdSetDepthBias vkCmdSetDepthBias;
+ PFN_vkCmdSetDepthBounds vkCmdSetDepthBounds;
+ PFN_vkCmdSetScissor vkCmdSetScissor;
+ PFN_vkCmdSetStencilCompareMask vkCmdSetStencilCompareMask;
+ PFN_vkCmdSetStencilReference vkCmdSetStencilReference;
+ PFN_vkCmdSetStencilWriteMask vkCmdSetStencilWriteMask;
+ PFN_vkCmdSetViewport vkCmdSetViewport;
+ PFN_vkCreateBuffer vkCreateBuffer;
+ PFN_vkCreateBufferView vkCreateBufferView;
+ PFN_vkCreateCommandPool vkCreateCommandPool;
+ PFN_vkCreateComputePipelines vkCreateComputePipelines;
+ PFN_vkCreateDescriptorPool vkCreateDescriptorPool;
+ PFN_vkCreateDescriptorSetLayout vkCreateDescriptorSetLayout;
+ PFN_vkCreateDescriptorUpdateTemplateKHR vkCreateDescriptorUpdateTemplateKHR;
+ PFN_vkCreateFence vkCreateFence;
+ PFN_vkCreateFramebuffer vkCreateFramebuffer;
+ PFN_vkCreateGraphicsPipelines vkCreateGraphicsPipelines;
+ PFN_vkCreateImage vkCreateImage;
+ PFN_vkCreateImageView vkCreateImageView;
+ PFN_vkCreatePipelineLayout vkCreatePipelineLayout;
+ PFN_vkCreateQueryPool vkCreateQueryPool;
+ PFN_vkCreateRenderPass vkCreateRenderPass;
+ PFN_vkCreateSampler vkCreateSampler;
+ PFN_vkCreateSemaphore vkCreateSemaphore;
+ PFN_vkCreateShaderModule vkCreateShaderModule;
+ PFN_vkCreateSwapchainKHR vkCreateSwapchainKHR;
+ PFN_vkDestroyBuffer vkDestroyBuffer;
+ PFN_vkDestroyBufferView vkDestroyBufferView;
+ PFN_vkDestroyCommandPool vkDestroyCommandPool;
+ PFN_vkDestroyDescriptorPool vkDestroyDescriptorPool;
+ PFN_vkDestroyDescriptorSetLayout vkDestroyDescriptorSetLayout;
+ PFN_vkDestroyDescriptorUpdateTemplateKHR vkDestroyDescriptorUpdateTemplateKHR;
+ PFN_vkDestroyFence vkDestroyFence;
+ PFN_vkDestroyFramebuffer vkDestroyFramebuffer;
+ PFN_vkDestroyImage vkDestroyImage;
+ PFN_vkDestroyImageView vkDestroyImageView;
+ PFN_vkDestroyPipeline vkDestroyPipeline;
+ PFN_vkDestroyPipelineLayout vkDestroyPipelineLayout;
+ PFN_vkDestroyQueryPool vkDestroyQueryPool;
+ PFN_vkDestroyRenderPass vkDestroyRenderPass;
+ PFN_vkDestroySampler vkDestroySampler;
+ PFN_vkDestroySemaphore vkDestroySemaphore;
+ PFN_vkDestroyShaderModule vkDestroyShaderModule;
+ PFN_vkDestroySwapchainKHR vkDestroySwapchainKHR;
+ PFN_vkDeviceWaitIdle vkDeviceWaitIdle;
+ PFN_vkEndCommandBuffer vkEndCommandBuffer;
+ PFN_vkFreeCommandBuffers vkFreeCommandBuffers;
+ PFN_vkFreeDescriptorSets vkFreeDescriptorSets;
+ PFN_vkFreeMemory vkFreeMemory;
+ PFN_vkGetBufferMemoryRequirements vkGetBufferMemoryRequirements;
+ PFN_vkGetDeviceQueue vkGetDeviceQueue;
+ PFN_vkGetFenceStatus vkGetFenceStatus;
+ PFN_vkGetImageMemoryRequirements vkGetImageMemoryRequirements;
+ PFN_vkGetQueryPoolResults vkGetQueryPoolResults;
+ PFN_vkGetQueueCheckpointDataNV vkGetQueueCheckpointDataNV;
+ PFN_vkMapMemory vkMapMemory;
+ PFN_vkQueueSubmit vkQueueSubmit;
+ PFN_vkResetFences vkResetFences;
+ PFN_vkResetQueryPoolEXT vkResetQueryPoolEXT;
+ PFN_vkUnmapMemory vkUnmapMemory;
+ PFN_vkUpdateDescriptorSetWithTemplateKHR vkUpdateDescriptorSetWithTemplateKHR;
+ PFN_vkUpdateDescriptorSets vkUpdateDescriptorSets;
+ PFN_vkWaitForFences vkWaitForFences;
+};
+
+/// Loads instance agnostic function pointers.
+/// @return True on success, false on error.
+bool Load(InstanceDispatch&) noexcept;
+
+/// Loads instance function pointers.
+/// @return True on success, false on error.
+bool Load(VkInstance, InstanceDispatch&) noexcept;
+
+void Destroy(VkInstance, const InstanceDispatch&) noexcept;
+void Destroy(VkDevice, const InstanceDispatch&) noexcept;
+
+void Destroy(VkDevice, VkBuffer, const DeviceDispatch&) noexcept;
+void Destroy(VkDevice, VkBufferView, const DeviceDispatch&) noexcept;
+void Destroy(VkDevice, VkCommandPool, const DeviceDispatch&) noexcept;
+void Destroy(VkDevice, VkDescriptorPool, const DeviceDispatch&) noexcept;
+void Destroy(VkDevice, VkDescriptorSetLayout, const DeviceDispatch&) noexcept;
+void Destroy(VkDevice, VkDescriptorUpdateTemplateKHR, const DeviceDispatch&) noexcept;
+void Destroy(VkDevice, VkDeviceMemory, const DeviceDispatch&) noexcept;
+void Destroy(VkDevice, VkFence, const DeviceDispatch&) noexcept;
+void Destroy(VkDevice, VkFramebuffer, const DeviceDispatch&) noexcept;
+void Destroy(VkDevice, VkImage, const DeviceDispatch&) noexcept;
+void Destroy(VkDevice, VkImageView, const DeviceDispatch&) noexcept;
+void Destroy(VkDevice, VkPipeline, const DeviceDispatch&) noexcept;
+void Destroy(VkDevice, VkPipelineLayout, const DeviceDispatch&) noexcept;
+void Destroy(VkDevice, VkQueryPool, const DeviceDispatch&) noexcept;
+void Destroy(VkDevice, VkRenderPass, const DeviceDispatch&) noexcept;
+void Destroy(VkDevice, VkSampler, const DeviceDispatch&) noexcept;
+void Destroy(VkDevice, VkSwapchainKHR, const DeviceDispatch&) noexcept;
+void Destroy(VkDevice, VkSemaphore, const DeviceDispatch&) noexcept;
+void Destroy(VkDevice, VkShaderModule, const DeviceDispatch&) noexcept;
+void Destroy(VkInstance, VkDebugUtilsMessengerEXT, const InstanceDispatch&) noexcept;
+void Destroy(VkInstance, VkSurfaceKHR, const InstanceDispatch&) noexcept;
+
+VkResult Free(VkDevice, VkDescriptorPool, Span<VkDescriptorSet>, const DeviceDispatch&) noexcept;
+VkResult Free(VkDevice, VkCommandPool, Span<VkCommandBuffer>, const DeviceDispatch&) noexcept;
+
+template <typename Type, typename OwnerType, typename Dispatch>
+class Handle;
+
+/// Handle with an owning type.
+/// Analogue to std::unique_ptr.
+template <typename Type, typename OwnerType, typename Dispatch>
+class Handle {
+public:
+ /// Construct a handle and hold it's ownership.
+ explicit Handle(Type handle_, OwnerType owner_, const Dispatch& dld_) noexcept
+ : handle{handle_}, owner{owner_}, dld{&dld_} {}
+
+ /// Construct an empty handle.
+ Handle() = default;
+
+ /// Copying Vulkan objects is not supported and will never be.
+ Handle(const Handle&) = delete;
+ Handle& operator=(const Handle&) = delete;
+
+ /// Construct a handle transfering the ownership from another handle.
+ Handle(Handle&& rhs) noexcept
+ : handle{std::exchange(rhs.handle, nullptr)}, owner{rhs.owner}, dld{rhs.dld} {}
+
+ /// Assign the current handle transfering the ownership from another handle.
+ /// Destroys any previously held object.
+ Handle& operator=(Handle&& rhs) noexcept {
+ Release();
+ handle = std::exchange(rhs.handle, nullptr);
+ owner = rhs.owner;
+ dld = rhs.dld;
+ return *this;
+ }
+
+ /// Destroys the current handle if it existed.
+ ~Handle() noexcept {
+ Release();
+ }
+
+ /// Destroys any held object.
+ void reset() noexcept {
+ Release();
+ handle = nullptr;
+ }
+
+ /// Returns the address of the held object.
+ /// Intended for Vulkan structures that expect a pointer to an array.
+ const Type* address() const noexcept {
+ return std::addressof(handle);
+ }
+
+ /// Returns the held Vulkan handle.
+ Type operator*() const noexcept {
+ return handle;
+ }
+
+ /// Returns true when there's a held object.
+ explicit operator bool() const noexcept {
+ return handle != nullptr;
+ }
+
+protected:
+ Type handle = nullptr;
+ OwnerType owner = nullptr;
+ const Dispatch* dld = nullptr;
+
+private:
+ /// Destroys the held object if it exists.
+ void Release() noexcept {
+ if (handle) {
+ Destroy(owner, handle, *dld);
+ }
+ }
+};
+
+/// Dummy type used to specify a handle has no owner.
+struct NoOwner {};
+
+/// Handle without an owning type.
+/// Analogue to std::unique_ptr
+template <typename Type, typename Dispatch>
+class Handle<Type, NoOwner, Dispatch> {
+public:
+ /// Construct a handle and hold it's ownership.
+ explicit Handle(Type handle_, const Dispatch& dld_) noexcept : handle{handle_}, dld{&dld_} {}
+
+ /// Construct an empty handle.
+ Handle() noexcept = default;
+
+ /// Copying Vulkan objects is not supported and will never be.
+ Handle(const Handle&) = delete;
+ Handle& operator=(const Handle&) = delete;
+
+ /// Construct a handle transfering ownership from another handle.
+ Handle(Handle&& rhs) noexcept : handle{std::exchange(rhs.handle, nullptr)}, dld{rhs.dld} {}
+
+ /// Assign the current handle transfering the ownership from another handle.
+ /// Destroys any previously held object.
+ Handle& operator=(Handle&& rhs) noexcept {
+ Release();
+ handle = std::exchange(rhs.handle, nullptr);
+ dld = rhs.dld;
+ return *this;
+ }
+
+ /// Destroys the current handle if it existed.
+ ~Handle() noexcept {
+ Release();
+ }
+
+ /// Destroys any held object.
+ void reset() noexcept {
+ Release();
+ handle = nullptr;
+ }
+
+ /// Returns the address of the held object.
+ /// Intended for Vulkan structures that expect a pointer to an array.
+ const Type* address() const noexcept {
+ return std::addressof(handle);
+ }
+
+ /// Returns the held Vulkan handle.
+ Type operator*() const noexcept {
+ return handle;
+ }
+
+ /// Returns true when there's a held object.
+ operator bool() const noexcept {
+ return handle != nullptr;
+ }
+
+protected:
+ Type handle = nullptr;
+ const Dispatch* dld = nullptr;
+
+private:
+ /// Destroys the held object if it exists.
+ void Release() noexcept {
+ if (handle) {
+ Destroy(handle, *dld);
+ }
+ }
+};
+
+/// Array of a pool allocation.
+/// Analogue to std::vector
+template <typename AllocationType, typename PoolType>
+class PoolAllocations {
+public:
+ /// Construct an empty allocation.
+ PoolAllocations() = default;
+
+ /// Construct an allocation. Errors are reported through IsOutOfPoolMemory().
+ explicit PoolAllocations(std::unique_ptr<AllocationType[]> allocations, std::size_t num,
+ VkDevice device, PoolType pool, const DeviceDispatch& dld) noexcept
+ : allocations{std::move(allocations)}, num{num}, device{device}, pool{pool}, dld{&dld} {}
+
+ /// Copying Vulkan allocations is not supported and will never be.
+ PoolAllocations(const PoolAllocations&) = delete;
+ PoolAllocations& operator=(const PoolAllocations&) = delete;
+
+ /// Construct an allocation transfering ownership from another allocation.
+ PoolAllocations(PoolAllocations&& rhs) noexcept
+ : allocations{std::move(rhs.allocations)}, num{rhs.num}, device{rhs.device}, pool{rhs.pool},
+ dld{rhs.dld} {}
+
+ /// Assign an allocation transfering ownership from another allocation.
+ /// Releases any previously held allocation.
+ PoolAllocations& operator=(PoolAllocations&& rhs) noexcept {
+ Release();
+ allocations = std::move(rhs.allocations);
+ num = rhs.num;
+ device = rhs.device;
+ pool = rhs.pool;
+ dld = rhs.dld;
+ return *this;
+ }
+
+ /// Destroys any held allocation.
+ ~PoolAllocations() {
+ Release();
+ }
+
+ /// Returns the number of allocations.
+ std::size_t size() const noexcept {
+ return num;
+ }
+
+ /// Returns a pointer to the array of allocations.
+ AllocationType const* data() const noexcept {
+ return allocations.get();
+ }
+
+ /// Returns the allocation in the specified index.
+ /// @pre index < size()
+ AllocationType operator[](std::size_t index) const noexcept {
+ return allocations[index];
+ }
+
+ /// True when a pool fails to construct.
+ bool IsOutOfPoolMemory() const noexcept {
+ return !device;
+ }
+
+private:
+ /// Destroys the held allocations if they exist.
+ void Release() noexcept {
+ if (!allocations) {
+ return;
+ }
+ const Span<AllocationType> span(allocations.get(), num);
+ const VkResult result = Free(device, pool, span, *dld);
+ // There's no way to report errors from a destructor.
+ if (result != VK_SUCCESS) {
+ std::terminate();
+ }
+ }
+
+ std::unique_ptr<AllocationType[]> allocations;
+ std::size_t num = 0;
+ VkDevice device = nullptr;
+ PoolType pool = nullptr;
+ const DeviceDispatch* dld = nullptr;
+};
+
+using BufferView = Handle<VkBufferView, VkDevice, DeviceDispatch>;
+using DebugCallback = Handle<VkDebugUtilsMessengerEXT, VkInstance, InstanceDispatch>;
+using DescriptorSetLayout = Handle<VkDescriptorSetLayout, VkDevice, DeviceDispatch>;
+using DescriptorUpdateTemplateKHR = Handle<VkDescriptorUpdateTemplateKHR, VkDevice, DeviceDispatch>;
+using Framebuffer = Handle<VkFramebuffer, VkDevice, DeviceDispatch>;
+using ImageView = Handle<VkImageView, VkDevice, DeviceDispatch>;
+using Pipeline = Handle<VkPipeline, VkDevice, DeviceDispatch>;
+using PipelineLayout = Handle<VkPipelineLayout, VkDevice, DeviceDispatch>;
+using QueryPool = Handle<VkQueryPool, VkDevice, DeviceDispatch>;
+using RenderPass = Handle<VkRenderPass, VkDevice, DeviceDispatch>;
+using Sampler = Handle<VkSampler, VkDevice, DeviceDispatch>;
+using Semaphore = Handle<VkSemaphore, VkDevice, DeviceDispatch>;
+using ShaderModule = Handle<VkShaderModule, VkDevice, DeviceDispatch>;
+using SurfaceKHR = Handle<VkSurfaceKHR, VkInstance, InstanceDispatch>;
+
+using DescriptorSets = PoolAllocations<VkDescriptorSet, VkDescriptorPool>;
+using CommandBuffers = PoolAllocations<VkCommandBuffer, VkCommandPool>;
+
+/// Vulkan instance owning handle.
+class Instance : public Handle<VkInstance, NoOwner, InstanceDispatch> {
+ using Handle<VkInstance, NoOwner, InstanceDispatch>::Handle;
+
+public:
+ /// Creates a Vulkan instance. Use "operator bool" for error handling.
+ static Instance Create(Span<const char*> layers, Span<const char*> extensions,
+ InstanceDispatch& dld) noexcept;
+
+ /// Enumerates physical devices.
+ /// @return Physical devices and an empty handle on failure.
+ std::optional<std::vector<VkPhysicalDevice>> EnumeratePhysicalDevices();
+
+ /// Tries to create a debug callback messenger. Returns an empty handle on failure.
+ DebugCallback TryCreateDebugCallback(PFN_vkDebugUtilsMessengerCallbackEXT callback) noexcept;
+};
+
+class Queue {
+public:
+ /// Construct an empty queue handle.
+ constexpr Queue() noexcept = default;
+
+ /// Construct a queue handle.
+ constexpr Queue(VkQueue queue, const DeviceDispatch& dld) noexcept : queue{queue}, dld{&dld} {}
+
+ /// Returns the checkpoint data.
+ /// @note Returns an empty vector when the function pointer is not present.
+ std::vector<VkCheckpointDataNV> GetCheckpointDataNV(const DeviceDispatch& dld) const;
+
+ void Submit(Span<VkSubmitInfo> submit_infos, VkFence fence) const {
+ Check(dld->vkQueueSubmit(queue, submit_infos.size(), submit_infos.data(), fence));
+ }
+
+ VkResult Present(const VkPresentInfoKHR& present_info) const noexcept {
+ return dld->vkQueuePresentKHR(queue, &present_info);
+ }
+
+private:
+ VkQueue queue = nullptr;
+ const DeviceDispatch* dld = nullptr;
+};
+
+class Buffer : public Handle<VkBuffer, VkDevice, DeviceDispatch> {
+ using Handle<VkBuffer, VkDevice, DeviceDispatch>::Handle;
+
+public:
+ /// Attaches a memory allocation.
+ void BindMemory(VkDeviceMemory memory, VkDeviceSize offset) const;
+};
+
+class Image : public Handle<VkImage, VkDevice, DeviceDispatch> {
+ using Handle<VkImage, VkDevice, DeviceDispatch>::Handle;
+
+public:
+ /// Attaches a memory allocation.
+ void BindMemory(VkDeviceMemory memory, VkDeviceSize offset) const;
+};
+
+class DeviceMemory : public Handle<VkDeviceMemory, VkDevice, DeviceDispatch> {
+ using Handle<VkDeviceMemory, VkDevice, DeviceDispatch>::Handle;
+
+public:
+ u8* Map(VkDeviceSize offset, VkDeviceSize size) const {
+ void* data;
+ Check(dld->vkMapMemory(owner, handle, offset, size, 0, &data));
+ return static_cast<u8*>(data);
+ }
+
+ void Unmap() const noexcept {
+ dld->vkUnmapMemory(owner, handle);
+ }
+};
+
+class Fence : public Handle<VkFence, VkDevice, DeviceDispatch> {
+ using Handle<VkFence, VkDevice, DeviceDispatch>::Handle;
+
+public:
+ VkResult Wait(u64 timeout = std::numeric_limits<u64>::max()) const noexcept {
+ return dld->vkWaitForFences(owner, 1, &handle, true, timeout);
+ }
+
+ VkResult GetStatus() const noexcept {
+ return dld->vkGetFenceStatus(owner, handle);
+ }
+
+ void Reset() const {
+ Check(dld->vkResetFences(owner, 1, &handle));
+ }
+};
+
+class DescriptorPool : public Handle<VkDescriptorPool, VkDevice, DeviceDispatch> {
+ using Handle<VkDescriptorPool, VkDevice, DeviceDispatch>::Handle;
+
+public:
+ DescriptorSets Allocate(const VkDescriptorSetAllocateInfo& ai) const;
+};
+
+class CommandPool : public Handle<VkCommandPool, VkDevice, DeviceDispatch> {
+ using Handle<VkCommandPool, VkDevice, DeviceDispatch>::Handle;
+
+public:
+ CommandBuffers Allocate(std::size_t num_buffers,
+ VkCommandBufferLevel level = VK_COMMAND_BUFFER_LEVEL_PRIMARY) const;
+};
+
+class SwapchainKHR : public Handle<VkSwapchainKHR, VkDevice, DeviceDispatch> {
+ using Handle<VkSwapchainKHR, VkDevice, DeviceDispatch>::Handle;
+
+public:
+ std::vector<VkImage> GetImages() const;
+};
+
+class Device : public Handle<VkDevice, NoOwner, DeviceDispatch> {
+ using Handle<VkDevice, NoOwner, DeviceDispatch>::Handle;
+
+public:
+ static Device Create(VkPhysicalDevice physical_device, Span<VkDeviceQueueCreateInfo> queues_ci,
+ Span<const char*> enabled_extensions,
+ const VkPhysicalDeviceFeatures2& enabled_features,
+ DeviceDispatch& dld) noexcept;
+
+ Queue GetQueue(u32 family_index) const noexcept;
+
+ Buffer CreateBuffer(const VkBufferCreateInfo& ci) const;
+
+ BufferView CreateBufferView(const VkBufferViewCreateInfo& ci) const;
+
+ Image CreateImage(const VkImageCreateInfo& ci) const;
+
+ ImageView CreateImageView(const VkImageViewCreateInfo& ci) const;
+
+ Semaphore CreateSemaphore() const;
+
+ Fence CreateFence(const VkFenceCreateInfo& ci) const;
+
+ DescriptorPool CreateDescriptorPool(const VkDescriptorPoolCreateInfo& ci) const;
+
+ RenderPass CreateRenderPass(const VkRenderPassCreateInfo& ci) const;
+
+ DescriptorSetLayout CreateDescriptorSetLayout(const VkDescriptorSetLayoutCreateInfo& ci) const;
+
+ PipelineLayout CreatePipelineLayout(const VkPipelineLayoutCreateInfo& ci) const;
+
+ Pipeline CreateGraphicsPipeline(const VkGraphicsPipelineCreateInfo& ci) const;
+
+ Pipeline CreateComputePipeline(const VkComputePipelineCreateInfo& ci) const;
+
+ Sampler CreateSampler(const VkSamplerCreateInfo& ci) const;
+
+ Framebuffer CreateFramebuffer(const VkFramebufferCreateInfo& ci) const;
+
+ CommandPool CreateCommandPool(const VkCommandPoolCreateInfo& ci) const;
+
+ DescriptorUpdateTemplateKHR CreateDescriptorUpdateTemplateKHR(
+ const VkDescriptorUpdateTemplateCreateInfoKHR& ci) const;
+
+ QueryPool CreateQueryPool(const VkQueryPoolCreateInfo& ci) const;
+
+ ShaderModule CreateShaderModule(const VkShaderModuleCreateInfo& ci) const;
+
+ SwapchainKHR CreateSwapchainKHR(const VkSwapchainCreateInfoKHR& ci) const;
+
+ DeviceMemory TryAllocateMemory(const VkMemoryAllocateInfo& ai) const noexcept;
+
+ DeviceMemory AllocateMemory(const VkMemoryAllocateInfo& ai) const;
+
+ VkMemoryRequirements GetBufferMemoryRequirements(VkBuffer buffer) const noexcept;
+
+ VkMemoryRequirements GetImageMemoryRequirements(VkImage image) const noexcept;
+
+ void UpdateDescriptorSets(Span<VkWriteDescriptorSet> writes,
+ Span<VkCopyDescriptorSet> copies) const noexcept;
+
+ void UpdateDescriptorSet(VkDescriptorSet set, VkDescriptorUpdateTemplateKHR update_template,
+ const void* data) const noexcept {
+ dld->vkUpdateDescriptorSetWithTemplateKHR(handle, set, update_template, data);
+ }
+
+ VkResult AcquireNextImageKHR(VkSwapchainKHR swapchain, u64 timeout, VkSemaphore semaphore,
+ VkFence fence, u32* image_index) const noexcept {
+ return dld->vkAcquireNextImageKHR(handle, swapchain, timeout, semaphore, fence,
+ image_index);
+ }
+
+ VkResult WaitIdle() const noexcept {
+ return dld->vkDeviceWaitIdle(handle);
+ }
+
+ void ResetQueryPoolEXT(VkQueryPool query_pool, u32 first, u32 count) const noexcept {
+ dld->vkResetQueryPoolEXT(handle, query_pool, first, count);
+ }
+
+ void GetQueryResults(VkQueryPool query_pool, u32 first, u32 count, std::size_t data_size,
+ void* data, VkDeviceSize stride, VkQueryResultFlags flags) const {
+ Check(dld->vkGetQueryPoolResults(handle, query_pool, first, count, data_size, data, stride,
+ flags));
+ }
+
+ template <typename T>
+ T GetQueryResult(VkQueryPool query_pool, u32 first, VkQueryResultFlags flags) const {
+ static_assert(std::is_trivially_copyable_v<T>);
+ T value;
+ GetQueryResults(query_pool, first, 1, sizeof(T), &value, sizeof(T), flags);
+ return value;
+ }
+};
+
+class PhysicalDevice {
+public:
+ constexpr PhysicalDevice() noexcept = default;
+
+ constexpr PhysicalDevice(VkPhysicalDevice physical_device, const InstanceDispatch& dld) noexcept
+ : physical_device{physical_device}, dld{&dld} {}
+
+ constexpr operator VkPhysicalDevice() const noexcept {
+ return physical_device;
+ }
+
+ VkPhysicalDeviceProperties GetProperties() const noexcept;
+
+ void GetProperties2KHR(VkPhysicalDeviceProperties2KHR&) const noexcept;
+
+ VkPhysicalDeviceFeatures GetFeatures() const noexcept;
+
+ void GetFeatures2KHR(VkPhysicalDeviceFeatures2KHR&) const noexcept;
+
+ VkFormatProperties GetFormatProperties(VkFormat) const noexcept;
+
+ std::vector<VkExtensionProperties> EnumerateDeviceExtensionProperties() const;
+
+ std::vector<VkQueueFamilyProperties> GetQueueFamilyProperties() const;
+
+ bool GetSurfaceSupportKHR(u32 queue_family_index, VkSurfaceKHR) const;
+
+ VkSurfaceCapabilitiesKHR GetSurfaceCapabilitiesKHR(VkSurfaceKHR) const noexcept;
+
+ std::vector<VkSurfaceFormatKHR> GetSurfaceFormatsKHR(VkSurfaceKHR) const;
+
+ std::vector<VkPresentModeKHR> GetSurfacePresentModesKHR(VkSurfaceKHR) const;
+
+ VkPhysicalDeviceMemoryProperties GetMemoryProperties() const noexcept;
+
+private:
+ VkPhysicalDevice physical_device = nullptr;
+ const InstanceDispatch* dld = nullptr;
+};
+
+class CommandBuffer {
+public:
+ CommandBuffer() noexcept = default;
+
+ explicit CommandBuffer(VkCommandBuffer handle, const DeviceDispatch& dld) noexcept
+ : handle{handle}, dld{&dld} {}
+
+ const VkCommandBuffer* address() const noexcept {
+ return &handle;
+ }
+
+ void Begin(const VkCommandBufferBeginInfo& begin_info) const {
+ Check(dld->vkBeginCommandBuffer(handle, &begin_info));
+ }
+
+ void End() const {
+ Check(dld->vkEndCommandBuffer(handle));
+ }
+
+ void BeginRenderPass(const VkRenderPassBeginInfo& renderpass_bi,
+ VkSubpassContents contents) const noexcept {
+ dld->vkCmdBeginRenderPass(handle, &renderpass_bi, contents);
+ }
+
+ void EndRenderPass() const noexcept {
+ dld->vkCmdEndRenderPass(handle);
+ }
+
+ void BeginQuery(VkQueryPool query_pool, u32 query, VkQueryControlFlags flags) const noexcept {
+ dld->vkCmdBeginQuery(handle, query_pool, query, flags);
+ }
+
+ void EndQuery(VkQueryPool query_pool, u32 query) const noexcept {
+ dld->vkCmdEndQuery(handle, query_pool, query);
+ }
+
+ void BindDescriptorSets(VkPipelineBindPoint bind_point, VkPipelineLayout layout, u32 first,
+ Span<VkDescriptorSet> sets, Span<u32> dynamic_offsets) const noexcept {
+ dld->vkCmdBindDescriptorSets(handle, bind_point, layout, first, sets.size(), sets.data(),
+ dynamic_offsets.size(), dynamic_offsets.data());
+ }
+
+ void BindPipeline(VkPipelineBindPoint bind_point, VkPipeline pipeline) const noexcept {
+ dld->vkCmdBindPipeline(handle, bind_point, pipeline);
+ }
+
+ void BindIndexBuffer(VkBuffer buffer, VkDeviceSize offset, VkIndexType index_type) const
+ noexcept {
+ dld->vkCmdBindIndexBuffer(handle, buffer, offset, index_type);
+ }
+
+ void BindVertexBuffers(u32 first, u32 count, const VkBuffer* buffers,
+ const VkDeviceSize* offsets) const noexcept {
+ dld->vkCmdBindVertexBuffers(handle, first, count, buffers, offsets);
+ }
+
+ void BindVertexBuffer(u32 binding, VkBuffer buffer, VkDeviceSize offset) const noexcept {
+ BindVertexBuffers(binding, 1, &buffer, &offset);
+ }
+
+ void Draw(u32 vertex_count, u32 instance_count, u32 first_vertex, u32 first_instance) const
+ noexcept {
+ dld->vkCmdDraw(handle, vertex_count, instance_count, first_vertex, first_instance);
+ }
+
+ void DrawIndexed(u32 index_count, u32 instance_count, u32 first_index, u32 vertex_offset,
+ u32 first_instance) const noexcept {
+ dld->vkCmdDrawIndexed(handle, index_count, instance_count, first_index, vertex_offset,
+ first_instance);
+ }
+
+ void ClearAttachments(Span<VkClearAttachment> attachments, Span<VkClearRect> rects) const
+ noexcept {
+ dld->vkCmdClearAttachments(handle, attachments.size(), attachments.data(), rects.size(),
+ rects.data());
+ }
+
+ void BlitImage(VkImage src_image, VkImageLayout src_layout, VkImage dst_image,
+ VkImageLayout dst_layout, Span<VkImageBlit> regions, VkFilter filter) const
+ noexcept {
+ dld->vkCmdBlitImage(handle, src_image, src_layout, dst_image, dst_layout, regions.size(),
+ regions.data(), filter);
+ }
+
+ void Dispatch(u32 x, u32 y, u32 z) const noexcept {
+ dld->vkCmdDispatch(handle, x, y, z);
+ }
+
+ void PipelineBarrier(VkPipelineStageFlags src_stage_mask, VkPipelineStageFlags dst_stage_mask,
+ VkDependencyFlags dependency_flags, Span<VkMemoryBarrier> memory_barriers,
+ Span<VkBufferMemoryBarrier> buffer_barriers,
+ Span<VkImageMemoryBarrier> image_barriers) const noexcept {
+ dld->vkCmdPipelineBarrier(handle, src_stage_mask, dst_stage_mask, dependency_flags,
+ memory_barriers.size(), memory_barriers.data(),
+ buffer_barriers.size(), buffer_barriers.data(),
+ image_barriers.size(), image_barriers.data());
+ }
+
+ void CopyBufferToImage(VkBuffer src_buffer, VkImage dst_image, VkImageLayout dst_image_layout,
+ Span<VkBufferImageCopy> regions) const noexcept {
+ dld->vkCmdCopyBufferToImage(handle, src_buffer, dst_image, dst_image_layout, regions.size(),
+ regions.data());
+ }
+
+ void CopyBuffer(VkBuffer src_buffer, VkBuffer dst_buffer, Span<VkBufferCopy> regions) const
+ noexcept {
+ dld->vkCmdCopyBuffer(handle, src_buffer, dst_buffer, regions.size(), regions.data());
+ }
+
+ void CopyImage(VkImage src_image, VkImageLayout src_layout, VkImage dst_image,
+ VkImageLayout dst_layout, Span<VkImageCopy> regions) const noexcept {
+ dld->vkCmdCopyImage(handle, src_image, src_layout, dst_image, dst_layout, regions.size(),
+ regions.data());
+ }
+
+ void CopyImageToBuffer(VkImage src_image, VkImageLayout src_layout, VkBuffer dst_buffer,
+ Span<VkBufferImageCopy> regions) const noexcept {
+ dld->vkCmdCopyImageToBuffer(handle, src_image, src_layout, dst_buffer, regions.size(),
+ regions.data());
+ }
+
+ void FillBuffer(VkBuffer dst_buffer, VkDeviceSize dst_offset, VkDeviceSize size, u32 data) const
+ noexcept {
+ dld->vkCmdFillBuffer(handle, dst_buffer, dst_offset, size, data);
+ }
+
+ void PushConstants(VkPipelineLayout layout, VkShaderStageFlags flags, u32 offset, u32 size,
+ const void* values) const noexcept {
+ dld->vkCmdPushConstants(handle, layout, flags, offset, size, values);
+ }
+
+ void SetCheckpointNV(const void* checkpoint_marker) const noexcept {
+ dld->vkCmdSetCheckpointNV(handle, checkpoint_marker);
+ }
+
+ void SetViewport(u32 first, Span<VkViewport> viewports) const noexcept {
+ dld->vkCmdSetViewport(handle, first, viewports.size(), viewports.data());
+ }
+
+ void SetScissor(u32 first, Span<VkRect2D> scissors) const noexcept {
+ dld->vkCmdSetScissor(handle, first, scissors.size(), scissors.data());
+ }
+
+ void SetBlendConstants(const float blend_constants[4]) const noexcept {
+ dld->vkCmdSetBlendConstants(handle, blend_constants);
+ }
+
+ void SetStencilCompareMask(VkStencilFaceFlags face_mask, u32 compare_mask) const noexcept {
+ dld->vkCmdSetStencilCompareMask(handle, face_mask, compare_mask);
+ }
+
+ void SetStencilReference(VkStencilFaceFlags face_mask, u32 reference) const noexcept {
+ dld->vkCmdSetStencilReference(handle, face_mask, reference);
+ }
+
+ void SetStencilWriteMask(VkStencilFaceFlags face_mask, u32 write_mask) const noexcept {
+ dld->vkCmdSetStencilWriteMask(handle, face_mask, write_mask);
+ }
+
+ void SetDepthBias(float constant_factor, float clamp, float slope_factor) const noexcept {
+ dld->vkCmdSetDepthBias(handle, constant_factor, clamp, slope_factor);
+ }
+
+ void SetDepthBounds(float min_depth_bounds, float max_depth_bounds) const noexcept {
+ dld->vkCmdSetDepthBounds(handle, min_depth_bounds, max_depth_bounds);
+ }
+
+ void BindTransformFeedbackBuffersEXT(u32 first, u32 count, const VkBuffer* buffers,
+ const VkDeviceSize* offsets,
+ const VkDeviceSize* sizes) const noexcept {
+ dld->vkCmdBindTransformFeedbackBuffersEXT(handle, first, count, buffers, offsets, sizes);
+ }
+
+ void BeginTransformFeedbackEXT(u32 first_counter_buffer, u32 counter_buffers_count,
+ const VkBuffer* counter_buffers,
+ const VkDeviceSize* counter_buffer_offsets) const noexcept {
+ dld->vkCmdBeginTransformFeedbackEXT(handle, first_counter_buffer, counter_buffers_count,
+ counter_buffers, counter_buffer_offsets);
+ }
+
+ void EndTransformFeedbackEXT(u32 first_counter_buffer, u32 counter_buffers_count,
+ const VkBuffer* counter_buffers,
+ const VkDeviceSize* counter_buffer_offsets) const noexcept {
+ dld->vkCmdEndTransformFeedbackEXT(handle, first_counter_buffer, counter_buffers_count,
+ counter_buffers, counter_buffer_offsets);
+ }
+
+private:
+ VkCommandBuffer handle;
+ const DeviceDispatch* dld;
+};
+
+std::optional<std::vector<VkExtensionProperties>> EnumerateInstanceExtensionProperties(
+ const InstanceDispatch& dld);
+
+} // namespace Vulkan::vk
diff --git a/src/video_core/shader/decode/arithmetic_integer.cpp b/src/video_core/shader/decode/arithmetic_integer.cpp
index 2fe787d6f..0f4c3103a 100644
--- a/src/video_core/shader/decode/arithmetic_integer.cpp
+++ b/src/video_core/shader/decode/arithmetic_integer.cpp
@@ -235,34 +235,30 @@ u32 ShaderIR::DecodeArithmeticInteger(NodeBlock& bb, u32 pc) {
case OpCode::Id::LEA_IMM:
case OpCode::Id::LEA_RZ:
case OpCode::Id::LEA_HI: {
- const auto [op_a, op_b, op_c] = [&]() -> std::tuple<Node, Node, Node> {
+ auto [op_a, op_b, op_c] = [&]() -> std::tuple<Node, Node, Node> {
switch (opcode->get().GetId()) {
case OpCode::Id::LEA_R2: {
return {GetRegister(instr.gpr20), GetRegister(instr.gpr39),
Immediate(static_cast<u32>(instr.lea.r2.entry_a))};
}
-
case OpCode::Id::LEA_R1: {
const bool neg = instr.lea.r1.neg != 0;
return {GetOperandAbsNegInteger(GetRegister(instr.gpr8), false, neg, true),
GetRegister(instr.gpr20),
Immediate(static_cast<u32>(instr.lea.r1.entry_a))};
}
-
case OpCode::Id::LEA_IMM: {
const bool neg = instr.lea.imm.neg != 0;
return {Immediate(static_cast<u32>(instr.lea.imm.entry_a)),
GetOperandAbsNegInteger(GetRegister(instr.gpr8), false, neg, true),
Immediate(static_cast<u32>(instr.lea.imm.entry_b))};
}
-
case OpCode::Id::LEA_RZ: {
const bool neg = instr.lea.rz.neg != 0;
return {GetConstBuffer(instr.lea.rz.cb_index, instr.lea.rz.cb_offset),
GetOperandAbsNegInteger(GetRegister(instr.gpr8), false, neg, true),
Immediate(static_cast<u32>(instr.lea.rz.entry_a))};
}
-
case OpCode::Id::LEA_HI:
default:
UNIMPLEMENTED_MSG("Unhandled LEA subinstruction: {}", opcode->get().GetName());
@@ -275,12 +271,9 @@ u32 ShaderIR::DecodeArithmeticInteger(NodeBlock& bb, u32 pc) {
UNIMPLEMENTED_IF_MSG(instr.lea.pred48 != static_cast<u64>(Pred::UnusedIndex),
"Unhandled LEA Predicate");
- const Node shifted_c =
- Operation(OperationCode::ILogicalShiftLeft, NO_PRECISE, Immediate(1), op_c);
- const Node mul_bc = Operation(OperationCode::IMul, NO_PRECISE, op_b, shifted_c);
- const Node value = Operation(OperationCode::IAdd, NO_PRECISE, op_a, mul_bc);
-
- SetRegister(bb, instr.gpr0, value);
+ Node value = Operation(OperationCode::ILogicalShiftLeft, std::move(op_a), std::move(op_c));
+ value = Operation(OperationCode::IAdd, std::move(op_b), std::move(value));
+ SetRegister(bb, instr.gpr0, std::move(value));
break;
}
diff --git a/src/video_core/shader/decode/conversion.cpp b/src/video_core/shader/decode/conversion.cpp
index 6ead42070..c72690b2b 100644
--- a/src/video_core/shader/decode/conversion.cpp
+++ b/src/video_core/shader/decode/conversion.cpp
@@ -138,18 +138,23 @@ u32 ShaderIR::DecodeConversion(NodeBlock& bb, u32 pc) {
value = GetOperandAbsNegFloat(value, instr.conversion.abs_a, instr.conversion.negate_a);
- value = [&]() {
+ value = [&] {
+ if (instr.conversion.src_size != instr.conversion.dst_size) {
+ // Rounding operations only matter when the source and destination conversion size
+ // is the same.
+ return value;
+ }
switch (instr.conversion.f2f.GetRoundingMode()) {
case Tegra::Shader::F2fRoundingOp::None:
return value;
case Tegra::Shader::F2fRoundingOp::Round:
- return Operation(OperationCode::FRoundEven, PRECISE, value);
+ return Operation(OperationCode::FRoundEven, value);
case Tegra::Shader::F2fRoundingOp::Floor:
- return Operation(OperationCode::FFloor, PRECISE, value);
+ return Operation(OperationCode::FFloor, value);
case Tegra::Shader::F2fRoundingOp::Ceil:
- return Operation(OperationCode::FCeil, PRECISE, value);
+ return Operation(OperationCode::FCeil, value);
case Tegra::Shader::F2fRoundingOp::Trunc:
- return Operation(OperationCode::FTrunc, PRECISE, value);
+ return Operation(OperationCode::FTrunc, value);
default:
UNIMPLEMENTED_MSG("Unimplemented F2F rounding mode {}",
static_cast<u32>(instr.conversion.f2f.rounding.Value()));
diff --git a/src/video_core/shader/decode/image.cpp b/src/video_core/shader/decode/image.cpp
index d2fe4ec5d..0dd7a1196 100644
--- a/src/video_core/shader/decode/image.cpp
+++ b/src/video_core/shader/decode/image.cpp
@@ -13,13 +13,247 @@
#include "video_core/engines/shader_bytecode.h"
#include "video_core/shader/node_helper.h"
#include "video_core/shader/shader_ir.h"
+#include "video_core/textures/texture.h"
namespace VideoCommon::Shader {
using Tegra::Shader::Instruction;
using Tegra::Shader::OpCode;
+using Tegra::Shader::PredCondition;
+using Tegra::Shader::StoreType;
+using Tegra::Texture::ComponentType;
+using Tegra::Texture::TextureFormat;
+using Tegra::Texture::TICEntry;
namespace {
+
+ComponentType GetComponentType(Tegra::Engines::SamplerDescriptor descriptor,
+ std::size_t component) {
+ const TextureFormat format{descriptor.format};
+ switch (format) {
+ case TextureFormat::R16_G16_B16_A16:
+ case TextureFormat::R32_G32_B32_A32:
+ case TextureFormat::R32_G32_B32:
+ case TextureFormat::R32_G32:
+ case TextureFormat::R16_G16:
+ case TextureFormat::R32:
+ case TextureFormat::R16:
+ case TextureFormat::R8:
+ case TextureFormat::R1:
+ if (component == 0) {
+ return descriptor.r_type;
+ }
+ if (component == 1) {
+ return descriptor.g_type;
+ }
+ if (component == 2) {
+ return descriptor.b_type;
+ }
+ if (component == 3) {
+ return descriptor.a_type;
+ }
+ break;
+ case TextureFormat::A8R8G8B8:
+ if (component == 0) {
+ return descriptor.a_type;
+ }
+ if (component == 1) {
+ return descriptor.r_type;
+ }
+ if (component == 2) {
+ return descriptor.g_type;
+ }
+ if (component == 3) {
+ return descriptor.b_type;
+ }
+ break;
+ case TextureFormat::A2B10G10R10:
+ case TextureFormat::A4B4G4R4:
+ case TextureFormat::A5B5G5R1:
+ case TextureFormat::A1B5G5R5:
+ if (component == 0) {
+ return descriptor.a_type;
+ }
+ if (component == 1) {
+ return descriptor.b_type;
+ }
+ if (component == 2) {
+ return descriptor.g_type;
+ }
+ if (component == 3) {
+ return descriptor.r_type;
+ }
+ break;
+ case TextureFormat::R32_B24G8:
+ if (component == 0) {
+ return descriptor.r_type;
+ }
+ if (component == 1) {
+ return descriptor.b_type;
+ }
+ if (component == 2) {
+ return descriptor.g_type;
+ }
+ break;
+ case TextureFormat::B5G6R5:
+ case TextureFormat::B6G5R5:
+ if (component == 0) {
+ return descriptor.b_type;
+ }
+ if (component == 1) {
+ return descriptor.g_type;
+ }
+ if (component == 2) {
+ return descriptor.r_type;
+ }
+ break;
+ case TextureFormat::G8R24:
+ case TextureFormat::G24R8:
+ case TextureFormat::G8R8:
+ case TextureFormat::G4R4:
+ if (component == 0) {
+ return descriptor.g_type;
+ }
+ if (component == 1) {
+ return descriptor.r_type;
+ }
+ break;
+ }
+ UNIMPLEMENTED_MSG("texture format not implement={}", format);
+ return ComponentType::FLOAT;
+}
+
+bool IsComponentEnabled(std::size_t component_mask, std::size_t component) {
+ constexpr u8 R = 0b0001;
+ constexpr u8 G = 0b0010;
+ constexpr u8 B = 0b0100;
+ constexpr u8 A = 0b1000;
+ constexpr std::array<u8, 16> mask = {
+ 0, (R), (G), (R | G), (B), (R | B), (G | B), (R | G | B),
+ (A), (R | A), (G | A), (R | G | A), (B | A), (R | B | A), (G | B | A), (R | G | B | A)};
+ return std::bitset<4>{mask.at(component_mask)}.test(component);
+}
+
+u32 GetComponentSize(TextureFormat format, std::size_t component) {
+ switch (format) {
+ case TextureFormat::R32_G32_B32_A32:
+ return 32;
+ case TextureFormat::R16_G16_B16_A16:
+ return 16;
+ case TextureFormat::R32_G32_B32:
+ return component <= 2 ? 32 : 0;
+ case TextureFormat::R32_G32:
+ return component <= 1 ? 32 : 0;
+ case TextureFormat::R16_G16:
+ return component <= 1 ? 16 : 0;
+ case TextureFormat::R32:
+ return component == 0 ? 32 : 0;
+ case TextureFormat::R16:
+ return component == 0 ? 16 : 0;
+ case TextureFormat::R8:
+ return component == 0 ? 8 : 0;
+ case TextureFormat::R1:
+ return component == 0 ? 1 : 0;
+ case TextureFormat::A8R8G8B8:
+ return 8;
+ case TextureFormat::A2B10G10R10:
+ return (component == 3 || component == 2 || component == 1) ? 10 : 2;
+ case TextureFormat::A4B4G4R4:
+ return 4;
+ case TextureFormat::A5B5G5R1:
+ return (component == 0 || component == 1 || component == 2) ? 5 : 1;
+ case TextureFormat::A1B5G5R5:
+ return (component == 1 || component == 2 || component == 3) ? 5 : 1;
+ case TextureFormat::R32_B24G8:
+ if (component == 0) {
+ return 32;
+ }
+ if (component == 1) {
+ return 24;
+ }
+ if (component == 2) {
+ return 8;
+ }
+ return 0;
+ case TextureFormat::B5G6R5:
+ if (component == 0 || component == 2) {
+ return 5;
+ }
+ if (component == 1) {
+ return 6;
+ }
+ return 0;
+ case TextureFormat::B6G5R5:
+ if (component == 1 || component == 2) {
+ return 5;
+ }
+ if (component == 0) {
+ return 6;
+ }
+ return 0;
+ case TextureFormat::G8R24:
+ if (component == 0) {
+ return 8;
+ }
+ if (component == 1) {
+ return 24;
+ }
+ return 0;
+ case TextureFormat::G24R8:
+ if (component == 0) {
+ return 8;
+ }
+ if (component == 1) {
+ return 24;
+ }
+ return 0;
+ case TextureFormat::G8R8:
+ return (component == 0 || component == 1) ? 8 : 0;
+ case TextureFormat::G4R4:
+ return (component == 0 || component == 1) ? 4 : 0;
+ default:
+ UNIMPLEMENTED_MSG("texture format not implement={}", format);
+ return 0;
+ }
+}
+
+std::size_t GetImageComponentMask(TextureFormat format) {
+ constexpr u8 R = 0b0001;
+ constexpr u8 G = 0b0010;
+ constexpr u8 B = 0b0100;
+ constexpr u8 A = 0b1000;
+ switch (format) {
+ case TextureFormat::R32_G32_B32_A32:
+ case TextureFormat::R16_G16_B16_A16:
+ case TextureFormat::A8R8G8B8:
+ case TextureFormat::A2B10G10R10:
+ case TextureFormat::A4B4G4R4:
+ case TextureFormat::A5B5G5R1:
+ case TextureFormat::A1B5G5R5:
+ return std::size_t{R | G | B | A};
+ case TextureFormat::R32_G32_B32:
+ case TextureFormat::R32_B24G8:
+ case TextureFormat::B5G6R5:
+ case TextureFormat::B6G5R5:
+ return std::size_t{R | G | B};
+ case TextureFormat::R32_G32:
+ case TextureFormat::R16_G16:
+ case TextureFormat::G8R24:
+ case TextureFormat::G24R8:
+ case TextureFormat::G8R8:
+ case TextureFormat::G4R4:
+ return std::size_t{R | G};
+ case TextureFormat::R32:
+ case TextureFormat::R16:
+ case TextureFormat::R8:
+ case TextureFormat::R1:
+ return std::size_t{R};
+ default:
+ UNIMPLEMENTED_MSG("texture format not implement={}", format);
+ return std::size_t{R | G | B | A};
+ }
+}
+
std::size_t GetImageTypeNumCoordinates(Tegra::Shader::ImageType image_type) {
switch (image_type) {
case Tegra::Shader::ImageType::Texture1D:
@@ -37,6 +271,39 @@ std::size_t GetImageTypeNumCoordinates(Tegra::Shader::ImageType image_type) {
}
} // Anonymous namespace
+std::pair<Node, bool> ShaderIR::GetComponentValue(ComponentType component_type, u32 component_size,
+ Node original_value) {
+ switch (component_type) {
+ case ComponentType::SNORM: {
+ // range [-1.0, 1.0]
+ auto cnv_value = Operation(OperationCode::FMul, original_value,
+ Immediate(static_cast<float>(1 << component_size) / 2.f - 1.f));
+ cnv_value = Operation(OperationCode::ICastFloat, std::move(cnv_value));
+ return {BitfieldExtract(std::move(cnv_value), 0, component_size), true};
+ }
+ case ComponentType::SINT:
+ case ComponentType::UNORM: {
+ bool is_signed = component_type == ComponentType::SINT;
+ // range [0.0, 1.0]
+ auto cnv_value = Operation(OperationCode::FMul, original_value,
+ Immediate(static_cast<float>(1 << component_size) - 1.f));
+ return {SignedOperation(OperationCode::ICastFloat, is_signed, std::move(cnv_value)),
+ is_signed};
+ }
+ case ComponentType::UINT: // range [0, (1 << component_size) - 1]
+ return {std::move(original_value), false};
+ case ComponentType::FLOAT:
+ if (component_size == 16) {
+ return {Operation(OperationCode::HCastFloat, original_value), true};
+ } else {
+ return {std::move(original_value), true};
+ }
+ default:
+ UNIMPLEMENTED_MSG("Unimplement component type={}", component_type);
+ return {std::move(original_value), true};
+ }
+}
+
u32 ShaderIR::DecodeImage(NodeBlock& bb, u32 pc) {
const Instruction instr = {program_code[pc]};
const auto opcode = OpCode::Decode(instr);
@@ -53,7 +320,6 @@ u32 ShaderIR::DecodeImage(NodeBlock& bb, u32 pc) {
switch (opcode->get().GetId()) {
case OpCode::Id::SULD: {
- UNIMPLEMENTED_IF(instr.suldst.mode != Tegra::Shader::SurfaceDataMode::P);
UNIMPLEMENTED_IF(instr.suldst.out_of_bounds_store !=
Tegra::Shader::OutOfBoundsStore::Ignore);
@@ -62,17 +328,89 @@ u32 ShaderIR::DecodeImage(NodeBlock& bb, u32 pc) {
: GetBindlessImage(instr.gpr39, type)};
image.MarkRead();
- u32 indexer = 0;
- for (u32 element = 0; element < 4; ++element) {
- if (!instr.suldst.IsComponentEnabled(element)) {
- continue;
+ if (instr.suldst.mode == Tegra::Shader::SurfaceDataMode::P) {
+ u32 indexer = 0;
+ for (u32 element = 0; element < 4; ++element) {
+ if (!instr.suldst.IsComponentEnabled(element)) {
+ continue;
+ }
+ MetaImage meta{image, {}, element};
+ Node value = Operation(OperationCode::ImageLoad, meta, GetCoordinates(type));
+ SetTemporary(bb, indexer++, std::move(value));
+ }
+ for (u32 i = 0; i < indexer; ++i) {
+ SetRegister(bb, instr.gpr0.Value() + i, GetTemporary(i));
+ }
+ } else if (instr.suldst.mode == Tegra::Shader::SurfaceDataMode::D_BA) {
+ UNIMPLEMENTED_IF(instr.suldst.GetStoreDataLayout() != StoreType::Bits32 &&
+ instr.suldst.GetStoreDataLayout() != StoreType::Bits64);
+
+ auto descriptor = [this, instr] {
+ std::optional<Tegra::Engines::SamplerDescriptor> descriptor;
+ if (instr.suldst.is_immediate) {
+ descriptor =
+ registry.ObtainBoundSampler(static_cast<u32>(instr.image.index.Value()));
+ } else {
+ const Node image_register = GetRegister(instr.gpr39);
+ const auto [base_image, buffer, offset] = TrackCbuf(
+ image_register, global_code, static_cast<s64>(global_code.size()));
+ descriptor = registry.ObtainBindlessSampler(buffer, offset);
+ }
+ if (!descriptor) {
+ UNREACHABLE_MSG("Failed to obtain image descriptor");
+ }
+ return *descriptor;
+ }();
+
+ const auto comp_mask = GetImageComponentMask(descriptor.format);
+
+ switch (instr.suldst.GetStoreDataLayout()) {
+ case StoreType::Bits32:
+ case StoreType::Bits64: {
+ u32 indexer = 0;
+ u32 shifted_counter = 0;
+ Node value = Immediate(0);
+ for (u32 element = 0; element < 4; ++element) {
+ if (!IsComponentEnabled(comp_mask, element)) {
+ continue;
+ }
+ const auto component_type = GetComponentType(descriptor, element);
+ const auto component_size = GetComponentSize(descriptor.format, element);
+ MetaImage meta{image, {}, element};
+
+ auto [converted_value, is_signed] = GetComponentValue(
+ component_type, component_size,
+ Operation(OperationCode::ImageLoad, meta, GetCoordinates(type)));
+
+ // shift element to correct position
+ const auto shifted = shifted_counter;
+ if (shifted > 0) {
+ converted_value =
+ SignedOperation(OperationCode::ILogicalShiftLeft, is_signed,
+ std::move(converted_value), Immediate(shifted));
+ }
+ shifted_counter += component_size;
+
+ // add value into result
+ value = Operation(OperationCode::UBitwiseOr, value, std::move(converted_value));
+
+ // if we shifted enough for 1 byte -> we save it into temp
+ if (shifted_counter >= 32) {
+ SetTemporary(bb, indexer++, std::move(value));
+ // reset counter and value to prepare pack next byte
+ value = Immediate(0);
+ shifted_counter = 0;
+ }
+ }
+ for (u32 i = 0; i < indexer; ++i) {
+ SetRegister(bb, instr.gpr0.Value() + i, GetTemporary(i));
+ }
+ break;
+ }
+ default:
+ UNREACHABLE();
+ break;
}
- MetaImage meta{image, {}, element};
- Node value = Operation(OperationCode::ImageLoad, meta, GetCoordinates(type));
- SetTemporary(bb, indexer++, std::move(value));
- }
- for (u32 i = 0; i < indexer; ++i) {
- SetRegister(bb, instr.gpr0.Value() + i, GetTemporary(i));
}
break;
}
diff --git a/src/video_core/shader/decode/memory.cpp b/src/video_core/shader/decode/memory.cpp
index b5fbc4d58..b8f63922f 100644
--- a/src/video_core/shader/decode/memory.cpp
+++ b/src/video_core/shader/decode/memory.cpp
@@ -19,7 +19,6 @@ namespace VideoCommon::Shader {
using Tegra::Shader::AtomicOp;
using Tegra::Shader::AtomicType;
using Tegra::Shader::Attribute;
-using Tegra::Shader::GlobalAtomicOp;
using Tegra::Shader::GlobalAtomicType;
using Tegra::Shader::Instruction;
using Tegra::Shader::OpCode;
@@ -28,6 +27,31 @@ using Tegra::Shader::StoreType;
namespace {
+Node GetAtomOperation(AtomicOp op, bool is_signed, Node memory, Node data) {
+ const OperationCode operation_code = [op] {
+ switch (op) {
+ case AtomicOp::Add:
+ return OperationCode::AtomicIAdd;
+ case AtomicOp::Min:
+ return OperationCode::AtomicIMin;
+ case AtomicOp::Max:
+ return OperationCode::AtomicIMax;
+ case AtomicOp::And:
+ return OperationCode::AtomicIAnd;
+ case AtomicOp::Or:
+ return OperationCode::AtomicIOr;
+ case AtomicOp::Xor:
+ return OperationCode::AtomicIXor;
+ case AtomicOp::Exch:
+ return OperationCode::AtomicIExchange;
+ default:
+ UNIMPLEMENTED_MSG("op={}", static_cast<int>(op));
+ return OperationCode::AtomicIAdd;
+ }
+ }();
+ return SignedOperation(operation_code, is_signed, std::move(memory), std::move(data));
+}
+
bool IsUnaligned(Tegra::Shader::UniformType uniform_type) {
return uniform_type == Tegra::Shader::UniformType::UnsignedByte ||
uniform_type == Tegra::Shader::UniformType::UnsignedShort;
@@ -363,10 +387,13 @@ u32 ShaderIR::DecodeMemory(NodeBlock& bb, u32 pc) {
break;
}
case OpCode::Id::ATOM: {
- UNIMPLEMENTED_IF_MSG(instr.atom.operation != GlobalAtomicOp::Add, "operation={}",
- static_cast<int>(instr.atom.operation.Value()));
- UNIMPLEMENTED_IF_MSG(instr.atom.type != GlobalAtomicType::S32, "type={}",
- static_cast<int>(instr.atom.type.Value()));
+ UNIMPLEMENTED_IF_MSG(instr.atom.operation == AtomicOp::Inc ||
+ instr.atom.operation == AtomicOp::Dec ||
+ instr.atom.operation == AtomicOp::SafeAdd,
+ "operation={}", static_cast<int>(instr.atom.operation.Value()));
+ UNIMPLEMENTED_IF_MSG(instr.atom.type == GlobalAtomicType::S64 ||
+ instr.atom.type == GlobalAtomicType::U64,
+ "type={}", static_cast<int>(instr.atom.type.Value()));
const auto [real_address, base_address, descriptor] =
TrackGlobalMemory(bb, instr, true, true);
@@ -375,25 +402,29 @@ u32 ShaderIR::DecodeMemory(NodeBlock& bb, u32 pc) {
break;
}
+ const bool is_signed =
+ instr.atoms.type == AtomicType::S32 || instr.atoms.type == AtomicType::S64;
Node gmem = MakeNode<GmemNode>(real_address, base_address, descriptor);
- Node value = Operation(OperationCode::AtomicAdd, std::move(gmem), GetRegister(instr.gpr20));
+ Node value = GetAtomOperation(static_cast<AtomicOp>(instr.atom.operation), is_signed, gmem,
+ GetRegister(instr.gpr20));
SetRegister(bb, instr.gpr0, std::move(value));
break;
}
case OpCode::Id::ATOMS: {
- UNIMPLEMENTED_IF_MSG(instr.atoms.operation != AtomicOp::Add, "operation={}",
- static_cast<int>(instr.atoms.operation.Value()));
- UNIMPLEMENTED_IF_MSG(instr.atoms.type != AtomicType::U32, "type={}",
- static_cast<int>(instr.atoms.type.Value()));
-
+ UNIMPLEMENTED_IF_MSG(instr.atoms.operation == AtomicOp::Inc ||
+ instr.atoms.operation == AtomicOp::Dec,
+ "operation={}", static_cast<int>(instr.atoms.operation.Value()));
+ UNIMPLEMENTED_IF_MSG(instr.atoms.type == AtomicType::S64 ||
+ instr.atoms.type == AtomicType::U64,
+ "type={}", static_cast<int>(instr.atoms.type.Value()));
+ const bool is_signed =
+ instr.atoms.type == AtomicType::S32 || instr.atoms.type == AtomicType::S64;
const s32 offset = instr.atoms.GetImmediateOffset();
Node address = GetRegister(instr.gpr8);
address = Operation(OperationCode::IAdd, std::move(address), Immediate(offset));
-
- Node memory = GetSharedMemory(std::move(address));
- Node data = GetRegister(instr.gpr20);
-
- Node value = Operation(OperationCode::AtomicAdd, std::move(memory), std::move(data));
+ Node value =
+ GetAtomOperation(static_cast<AtomicOp>(instr.atoms.operation), is_signed,
+ GetSharedMemory(std::move(address)), GetRegister(instr.gpr20));
SetRegister(bb, instr.gpr0, std::move(value));
break;
}
diff --git a/src/video_core/shader/decode/other.cpp b/src/video_core/shader/decode/other.cpp
index 4944e9d69..d4f95b18c 100644
--- a/src/video_core/shader/decode/other.cpp
+++ b/src/video_core/shader/decode/other.cpp
@@ -11,12 +11,17 @@
namespace VideoCommon::Shader {
+using std::move;
using Tegra::Shader::ConditionCode;
using Tegra::Shader::Instruction;
+using Tegra::Shader::IpaInterpMode;
using Tegra::Shader::OpCode;
+using Tegra::Shader::PixelImap;
using Tegra::Shader::Register;
using Tegra::Shader::SystemVariable;
+using Index = Tegra::Shader::Attribute::Index;
+
u32 ShaderIR::DecodeOther(NodeBlock& bb, u32 pc) {
const Instruction instr = {program_code[pc]};
const auto opcode = OpCode::Decode(instr);
@@ -66,18 +71,24 @@ u32 ShaderIR::DecodeOther(NodeBlock& bb, u32 pc) {
bb.push_back(Operation(OperationCode::Discard));
break;
}
- case OpCode::Id::MOV_SYS: {
+ case OpCode::Id::S2R: {
const Node value = [this, instr] {
switch (instr.sys20) {
case SystemVariable::LaneId:
- LOG_WARNING(HW_GPU, "MOV_SYS instruction with LaneId is incomplete");
+ LOG_WARNING(HW_GPU, "S2R instruction with LaneId is incomplete");
return Immediate(0U);
case SystemVariable::InvocationId:
return Operation(OperationCode::InvocationId);
case SystemVariable::Ydirection:
return Operation(OperationCode::YNegate);
case SystemVariable::InvocationInfo:
- LOG_WARNING(HW_GPU, "MOV_SYS instruction with InvocationInfo is incomplete");
+ LOG_WARNING(HW_GPU, "S2R instruction with InvocationInfo is incomplete");
+ return Immediate(0U);
+ case SystemVariable::WscaleFactorXY:
+ UNIMPLEMENTED_MSG("S2R WscaleFactorXY is not implemented");
+ return Immediate(0U);
+ case SystemVariable::WscaleFactorZ:
+ UNIMPLEMENTED_MSG("S2R WscaleFactorZ is not implemented");
return Immediate(0U);
case SystemVariable::Tid: {
Node value = Immediate(0);
@@ -213,27 +224,28 @@ u32 ShaderIR::DecodeOther(NodeBlock& bb, u32 pc) {
}
case OpCode::Id::IPA: {
const bool is_physical = instr.ipa.idx && instr.gpr8.Value() != 0xff;
-
const auto attribute = instr.attribute.fmt28;
- const Tegra::Shader::IpaMode input_mode{instr.ipa.interp_mode.Value(),
- instr.ipa.sample_mode.Value()};
+ const Index index = attribute.index;
Node value = is_physical ? GetPhysicalInputAttribute(instr.gpr8)
- : GetInputAttribute(attribute.index, attribute.element);
- const Tegra::Shader::Attribute::Index index = attribute.index.Value();
- const bool is_generic = index >= Tegra::Shader::Attribute::Index::Attribute_0 &&
- index <= Tegra::Shader::Attribute::Index::Attribute_31;
- if (is_generic || is_physical) {
- // TODO(Blinkhawk): There are cases where a perspective attribute use PASS.
- // In theory by setting them as perspective, OpenGL does the perspective correction.
- // A way must figured to reverse the last step of it.
- if (input_mode.interpolation_mode == Tegra::Shader::IpaInterpMode::Multiply) {
- value = Operation(OperationCode::FMul, PRECISE, value, GetRegister(instr.gpr20));
+ : GetInputAttribute(index, attribute.element);
+
+ // Code taken from Ryujinx.
+ if (index >= Index::Attribute_0 && index <= Index::Attribute_31) {
+ const u32 location = static_cast<u32>(index) - static_cast<u32>(Index::Attribute_0);
+ if (header.ps.GetPixelImap(location) == PixelImap::Perspective) {
+ Node position_w = GetInputAttribute(Index::Position, 3);
+ value = Operation(OperationCode::FMul, move(value), move(position_w));
}
}
- value = GetSaturatedFloat(value, instr.ipa.saturate);
- SetRegister(bb, instr.gpr0, value);
+ if (instr.ipa.interp_mode == IpaInterpMode::Multiply) {
+ value = Operation(OperationCode::FMul, move(value), GetRegister(instr.gpr20));
+ }
+
+ value = GetSaturatedFloat(move(value), instr.ipa.saturate);
+
+ SetRegister(bb, instr.gpr0, move(value));
break;
}
case OpCode::Id::OUT_R: {
diff --git a/src/video_core/shader/node.h b/src/video_core/shader/node.h
index a1828546e..5fcc9da60 100644
--- a/src/video_core/shader/node.h
+++ b/src/video_core/shader/node.h
@@ -162,7 +162,21 @@ enum class OperationCode {
AtomicImageXor, /// (MetaImage, int[N] coords) -> void
AtomicImageExchange, /// (MetaImage, int[N] coords) -> void
- AtomicAdd, /// (memory, {u}int) -> {u}int
+ AtomicUExchange, /// (memory, uint) -> uint
+ AtomicUAdd, /// (memory, uint) -> uint
+ AtomicUMin, /// (memory, uint) -> uint
+ AtomicUMax, /// (memory, uint) -> uint
+ AtomicUAnd, /// (memory, uint) -> uint
+ AtomicUOr, /// (memory, uint) -> uint
+ AtomicUXor, /// (memory, uint) -> uint
+
+ AtomicIExchange, /// (memory, int) -> int
+ AtomicIAdd, /// (memory, int) -> int
+ AtomicIMin, /// (memory, int) -> int
+ AtomicIMax, /// (memory, int) -> int
+ AtomicIAnd, /// (memory, int) -> int
+ AtomicIOr, /// (memory, int) -> int
+ AtomicIXor, /// (memory, int) -> int
Branch, /// (uint branch_target) -> void
BranchIndirect, /// (uint branch_target) -> void
diff --git a/src/video_core/shader/node_helper.cpp b/src/video_core/shader/node_helper.cpp
index 76c56abb5..7bf4ff387 100644
--- a/src/video_core/shader/node_helper.cpp
+++ b/src/video_core/shader/node_helper.cpp
@@ -86,6 +86,20 @@ OperationCode SignedToUnsignedCode(OperationCode operation_code, bool is_signed)
return OperationCode::LogicalUNotEqual;
case OperationCode::LogicalIGreaterEqual:
return OperationCode::LogicalUGreaterEqual;
+ case OperationCode::AtomicIExchange:
+ return OperationCode::AtomicUExchange;
+ case OperationCode::AtomicIAdd:
+ return OperationCode::AtomicUAdd;
+ case OperationCode::AtomicIMin:
+ return OperationCode::AtomicUMin;
+ case OperationCode::AtomicIMax:
+ return OperationCode::AtomicUMax;
+ case OperationCode::AtomicIAnd:
+ return OperationCode::AtomicUAnd;
+ case OperationCode::AtomicIOr:
+ return OperationCode::AtomicUOr;
+ case OperationCode::AtomicIXor:
+ return OperationCode::AtomicUXor;
case OperationCode::INegate:
UNREACHABLE_MSG("Can't negate an unsigned integer");
return {};
diff --git a/src/video_core/shader/shader_ir.cpp b/src/video_core/shader/shader_ir.cpp
index baf7188d2..8852c8a1b 100644
--- a/src/video_core/shader/shader_ir.cpp
+++ b/src/video_core/shader/shader_ir.cpp
@@ -359,6 +359,9 @@ Node ShaderIR::GetConditionCode(Tegra::Shader::ConditionCode cc) const {
switch (cc) {
case Tegra::Shader::ConditionCode::NEU:
return GetInternalFlag(InternalFlag::Zero, true);
+ case Tegra::Shader::ConditionCode::FCSM_TR:
+ UNIMPLEMENTED_MSG("EXIT.FCSM_TR is not implemented");
+ return MakeNode<PredicateNode>(Pred::NeverExecute, false);
default:
UNIMPLEMENTED_MSG("Unimplemented condition code: {}", static_cast<u32>(cc));
return MakeNode<PredicateNode>(Pred::NeverExecute, false);
diff --git a/src/video_core/shader/shader_ir.h b/src/video_core/shader/shader_ir.h
index 0f1ebef1b..c6e7bdf50 100644
--- a/src/video_core/shader/shader_ir.h
+++ b/src/video_core/shader/shader_ir.h
@@ -312,6 +312,10 @@ private:
/// Conditionally saturates a half float pair
Node GetSaturatedHalfFloat(Node value, bool saturate = true);
+ /// Get image component value by type and size
+ std::pair<Node, bool> GetComponentValue(Tegra::Texture::ComponentType component_type,
+ u32 component_size, Node original_value);
+
/// Returns a predicate comparing two floats
Node GetPredicateComparisonFloat(Tegra::Shader::PredCondition condition, Node op_a, Node op_b);
/// Returns a predicate comparing two integers
diff --git a/src/video_core/surface.h b/src/video_core/surface.h
index ae8817465..e0acd44d3 100644
--- a/src/video_core/surface.h
+++ b/src/video_core/surface.h
@@ -504,103 +504,6 @@ static constexpr u32 GetBytesPerPixel(PixelFormat pixel_format) {
return GetFormatBpp(pixel_format) / CHAR_BIT;
}
-enum class SurfaceCompression {
- None, // Not compressed
- Compressed, // Texture is compressed
- Converted, // Texture is converted before upload or after download
- Rearranged, // Texture is swizzled before upload or after download
-};
-
-constexpr std::array<SurfaceCompression, MaxPixelFormat> compression_type_table = {{
- SurfaceCompression::None, // ABGR8U
- SurfaceCompression::None, // ABGR8S
- SurfaceCompression::None, // ABGR8UI
- SurfaceCompression::None, // B5G6R5U
- SurfaceCompression::None, // A2B10G10R10U
- SurfaceCompression::None, // A1B5G5R5U
- SurfaceCompression::None, // R8U
- SurfaceCompression::None, // R8UI
- SurfaceCompression::None, // RGBA16F
- SurfaceCompression::None, // RGBA16U
- SurfaceCompression::None, // RGBA16S
- SurfaceCompression::None, // RGBA16UI
- SurfaceCompression::None, // R11FG11FB10F
- SurfaceCompression::None, // RGBA32UI
- SurfaceCompression::Compressed, // DXT1
- SurfaceCompression::Compressed, // DXT23
- SurfaceCompression::Compressed, // DXT45
- SurfaceCompression::Compressed, // DXN1
- SurfaceCompression::Compressed, // DXN2UNORM
- SurfaceCompression::Compressed, // DXN2SNORM
- SurfaceCompression::Compressed, // BC7U
- SurfaceCompression::Compressed, // BC6H_UF16
- SurfaceCompression::Compressed, // BC6H_SF16
- SurfaceCompression::Converted, // ASTC_2D_4X4
- SurfaceCompression::None, // BGRA8
- SurfaceCompression::None, // RGBA32F
- SurfaceCompression::None, // RG32F
- SurfaceCompression::None, // R32F
- SurfaceCompression::None, // R16F
- SurfaceCompression::None, // R16U
- SurfaceCompression::None, // R16S
- SurfaceCompression::None, // R16UI
- SurfaceCompression::None, // R16I
- SurfaceCompression::None, // RG16
- SurfaceCompression::None, // RG16F
- SurfaceCompression::None, // RG16UI
- SurfaceCompression::None, // RG16I
- SurfaceCompression::None, // RG16S
- SurfaceCompression::None, // RGB32F
- SurfaceCompression::None, // RGBA8_SRGB
- SurfaceCompression::None, // RG8U
- SurfaceCompression::None, // RG8S
- SurfaceCompression::None, // RG32UI
- SurfaceCompression::None, // RGBX16F
- SurfaceCompression::None, // R32UI
- SurfaceCompression::None, // R32I
- SurfaceCompression::Converted, // ASTC_2D_8X8
- SurfaceCompression::Converted, // ASTC_2D_8X5
- SurfaceCompression::Converted, // ASTC_2D_5X4
- SurfaceCompression::None, // BGRA8_SRGB
- SurfaceCompression::Compressed, // DXT1_SRGB
- SurfaceCompression::Compressed, // DXT23_SRGB
- SurfaceCompression::Compressed, // DXT45_SRGB
- SurfaceCompression::Compressed, // BC7U_SRGB
- SurfaceCompression::None, // R4G4B4A4U
- SurfaceCompression::Converted, // ASTC_2D_4X4_SRGB
- SurfaceCompression::Converted, // ASTC_2D_8X8_SRGB
- SurfaceCompression::Converted, // ASTC_2D_8X5_SRGB
- SurfaceCompression::Converted, // ASTC_2D_5X4_SRGB
- SurfaceCompression::Converted, // ASTC_2D_5X5
- SurfaceCompression::Converted, // ASTC_2D_5X5_SRGB
- SurfaceCompression::Converted, // ASTC_2D_10X8
- SurfaceCompression::Converted, // ASTC_2D_10X8_SRGB
- SurfaceCompression::Converted, // ASTC_2D_6X6
- SurfaceCompression::Converted, // ASTC_2D_6X6_SRGB
- SurfaceCompression::Converted, // ASTC_2D_10X10
- SurfaceCompression::Converted, // ASTC_2D_10X10_SRGB
- SurfaceCompression::Converted, // ASTC_2D_12X12
- SurfaceCompression::Converted, // ASTC_2D_12X12_SRGB
- SurfaceCompression::Converted, // ASTC_2D_8X6
- SurfaceCompression::Converted, // ASTC_2D_8X6_SRGB
- SurfaceCompression::Converted, // ASTC_2D_6X5
- SurfaceCompression::Converted, // ASTC_2D_6X5_SRGB
- SurfaceCompression::None, // E5B9G9R9F
- SurfaceCompression::None, // Z32F
- SurfaceCompression::None, // Z16
- SurfaceCompression::None, // Z24S8
- SurfaceCompression::Rearranged, // S8Z24
- SurfaceCompression::None, // Z32FS8
-}};
-
-constexpr SurfaceCompression GetFormatCompressionType(PixelFormat format) {
- if (format == PixelFormat::Invalid) {
- return SurfaceCompression::None;
- }
- DEBUG_ASSERT(static_cast<std::size_t>(format) < compression_type_table.size());
- return compression_type_table[static_cast<std::size_t>(format)];
-}
-
SurfaceTarget SurfaceTargetFromTextureType(Tegra::Texture::TextureType texture_type);
bool SurfaceTargetIsLayered(SurfaceTarget target);
diff --git a/src/video_core/texture_cache/surface_base.cpp b/src/video_core/texture_cache/surface_base.cpp
index 002df414f..7af0e792c 100644
--- a/src/video_core/texture_cache/surface_base.cpp
+++ b/src/video_core/texture_cache/surface_base.cpp
@@ -18,15 +18,20 @@ MICROPROFILE_DEFINE(GPU_Flush_Texture, "GPU", "Texture Flush", MP_RGB(128, 192,
using Tegra::Texture::ConvertFromGuestToHost;
using VideoCore::MortonSwizzleMode;
-using VideoCore::Surface::SurfaceCompression;
+using VideoCore::Surface::IsPixelFormatASTC;
+using VideoCore::Surface::PixelFormat;
StagingCache::StagingCache() = default;
StagingCache::~StagingCache() = default;
-SurfaceBaseImpl::SurfaceBaseImpl(GPUVAddr gpu_addr, const SurfaceParams& params)
- : params{params}, host_memory_size{params.GetHostSizeInBytes()}, gpu_addr{gpu_addr},
- mipmap_sizes(params.num_levels), mipmap_offsets(params.num_levels) {
+SurfaceBaseImpl::SurfaceBaseImpl(GPUVAddr gpu_addr, const SurfaceParams& params,
+ bool is_astc_supported)
+ : params{params}, gpu_addr{gpu_addr}, mipmap_sizes(params.num_levels),
+ mipmap_offsets(params.num_levels) {
+ is_converted = IsPixelFormatASTC(params.pixel_format) && !is_astc_supported;
+ host_memory_size = params.GetHostSizeInBytes(is_converted);
+
std::size_t offset = 0;
for (u32 level = 0; level < params.num_levels; ++level) {
const std::size_t mipmap_size{params.GetGuestMipmapSize(level)};
@@ -164,7 +169,7 @@ void SurfaceBaseImpl::SwizzleFunc(MortonSwizzleMode mode, u8* memory, const Surf
std::size_t guest_offset{mipmap_offsets[level]};
if (params.is_layered) {
- std::size_t host_offset{0};
+ std::size_t host_offset = 0;
const std::size_t guest_stride = layer_size;
const std::size_t host_stride = params.GetHostLayerSize(level);
for (u32 layer = 0; layer < params.depth; ++layer) {
@@ -185,28 +190,17 @@ void SurfaceBaseImpl::LoadBuffer(Tegra::MemoryManager& memory_manager,
MICROPROFILE_SCOPE(GPU_Load_Texture);
auto& staging_buffer = staging_cache.GetBuffer(0);
u8* host_ptr;
- is_continuous = memory_manager.IsBlockContinuous(gpu_addr, guest_memory_size);
-
- // Handle continuouty
- if (is_continuous) {
- // Use physical memory directly
- host_ptr = memory_manager.GetPointer(gpu_addr);
- if (!host_ptr) {
- return;
- }
- } else {
- // Use an extra temporal buffer
- auto& tmp_buffer = staging_cache.GetBuffer(1);
- tmp_buffer.resize(guest_memory_size);
- host_ptr = tmp_buffer.data();
- memory_manager.ReadBlockUnsafe(gpu_addr, host_ptr, guest_memory_size);
- }
+ // Use an extra temporal buffer
+ auto& tmp_buffer = staging_cache.GetBuffer(1);
+ tmp_buffer.resize(guest_memory_size);
+ host_ptr = tmp_buffer.data();
+ memory_manager.ReadBlockUnsafe(gpu_addr, host_ptr, guest_memory_size);
if (params.is_tiled) {
ASSERT_MSG(params.block_width == 0, "Block width is defined as {} on texture target {}",
params.block_width, static_cast<u32>(params.target));
for (u32 level = 0; level < params.num_levels; ++level) {
- const std::size_t host_offset{params.GetHostMipmapLevelOffset(level)};
+ const std::size_t host_offset{params.GetHostMipmapLevelOffset(level, false)};
SwizzleFunc(MortonSwizzleMode::MortonToLinear, host_ptr, params,
staging_buffer.data() + host_offset, level);
}
@@ -219,7 +213,7 @@ void SurfaceBaseImpl::LoadBuffer(Tegra::MemoryManager& memory_manager,
const u32 height{(params.height + block_height - 1) / block_height};
const u32 copy_size{width * bpp};
if (params.pitch == copy_size) {
- std::memcpy(staging_buffer.data(), host_ptr, params.GetHostSizeInBytes());
+ std::memcpy(staging_buffer.data(), host_ptr, params.GetHostSizeInBytes(false));
} else {
const u8* start{host_ptr};
u8* write_to{staging_buffer.data()};
@@ -231,19 +225,15 @@ void SurfaceBaseImpl::LoadBuffer(Tegra::MemoryManager& memory_manager,
}
}
- auto compression_type = params.GetCompressionType();
- if (compression_type == SurfaceCompression::None ||
- compression_type == SurfaceCompression::Compressed)
+ if (!is_converted && params.pixel_format != PixelFormat::S8Z24) {
return;
+ }
- for (u32 level_up = params.num_levels; level_up > 0; --level_up) {
- const u32 level = level_up - 1;
- const std::size_t in_host_offset{params.GetHostMipmapLevelOffset(level)};
- const std::size_t out_host_offset = compression_type == SurfaceCompression::Rearranged
- ? in_host_offset
- : params.GetConvertedMipmapOffset(level);
- u8* in_buffer = staging_buffer.data() + in_host_offset;
- u8* out_buffer = staging_buffer.data() + out_host_offset;
+ for (u32 level = params.num_levels; level--;) {
+ const std::size_t in_host_offset{params.GetHostMipmapLevelOffset(level, false)};
+ const std::size_t out_host_offset{params.GetHostMipmapLevelOffset(level, is_converted)};
+ u8* const in_buffer = staging_buffer.data() + in_host_offset;
+ u8* const out_buffer = staging_buffer.data() + out_host_offset;
ConvertFromGuestToHost(in_buffer, out_buffer, params.pixel_format,
params.GetMipWidth(level), params.GetMipHeight(level),
params.GetMipDepth(level), true, true);
@@ -256,24 +246,15 @@ void SurfaceBaseImpl::FlushBuffer(Tegra::MemoryManager& memory_manager,
auto& staging_buffer = staging_cache.GetBuffer(0);
u8* host_ptr;
- // Handle continuouty
- if (is_continuous) {
- // Use physical memory directly
- host_ptr = memory_manager.GetPointer(gpu_addr);
- if (!host_ptr) {
- return;
- }
- } else {
- // Use an extra temporal buffer
- auto& tmp_buffer = staging_cache.GetBuffer(1);
- tmp_buffer.resize(guest_memory_size);
- host_ptr = tmp_buffer.data();
- }
+ // Use an extra temporal buffer
+ auto& tmp_buffer = staging_cache.GetBuffer(1);
+ tmp_buffer.resize(guest_memory_size);
+ host_ptr = tmp_buffer.data();
if (params.is_tiled) {
ASSERT_MSG(params.block_width == 0, "Block width is defined as {}", params.block_width);
for (u32 level = 0; level < params.num_levels; ++level) {
- const std::size_t host_offset{params.GetHostMipmapLevelOffset(level)};
+ const std::size_t host_offset{params.GetHostMipmapLevelOffset(level, false)};
SwizzleFunc(MortonSwizzleMode::LinearToMorton, host_ptr, params,
staging_buffer.data() + host_offset, level);
}
@@ -299,9 +280,7 @@ void SurfaceBaseImpl::FlushBuffer(Tegra::MemoryManager& memory_manager,
}
}
}
- if (!is_continuous) {
- memory_manager.WriteBlockUnsafe(gpu_addr, host_ptr, guest_memory_size);
- }
+ memory_manager.WriteBlockUnsafe(gpu_addr, host_ptr, guest_memory_size);
}
} // namespace VideoCommon
diff --git a/src/video_core/texture_cache/surface_base.h b/src/video_core/texture_cache/surface_base.h
index 5f79bb0aa..a39a8661b 100644
--- a/src/video_core/texture_cache/surface_base.h
+++ b/src/video_core/texture_cache/surface_base.h
@@ -68,8 +68,8 @@ public:
return gpu_addr;
}
- bool Overlaps(const CacheAddr start, const CacheAddr end) const {
- return (cache_addr < end) && (cache_addr_end > start);
+ bool Overlaps(const VAddr start, const VAddr end) const {
+ return (cpu_addr < end) && (cpu_addr_end > start);
}
bool IsInside(const GPUVAddr other_start, const GPUVAddr other_end) {
@@ -86,21 +86,13 @@ public:
return cpu_addr;
}
- void SetCpuAddr(const VAddr new_addr) {
- cpu_addr = new_addr;
- }
-
- CacheAddr GetCacheAddr() const {
- return cache_addr;
+ VAddr GetCpuAddrEnd() const {
+ return cpu_addr_end;
}
- CacheAddr GetCacheAddrEnd() const {
- return cache_addr_end;
- }
-
- void SetCacheAddr(const CacheAddr new_addr) {
- cache_addr = new_addr;
- cache_addr_end = new_addr + guest_memory_size;
+ void SetCpuAddr(const VAddr new_addr) {
+ cpu_addr = new_addr;
+ cpu_addr_end = new_addr + guest_memory_size;
}
const SurfaceParams& GetSurfaceParams() const {
@@ -119,18 +111,14 @@ public:
return mipmap_sizes[level];
}
- void MarkAsContinuous(const bool is_continuous) {
- this->is_continuous = is_continuous;
- }
-
- bool IsContinuous() const {
- return is_continuous;
- }
-
bool IsLinear() const {
return !params.is_tiled;
}
+ bool IsConverted() const {
+ return is_converted;
+ }
+
bool MatchFormat(VideoCore::Surface::PixelFormat pixel_format) const {
return params.pixel_format == pixel_format;
}
@@ -160,7 +148,8 @@ public:
}
protected:
- explicit SurfaceBaseImpl(GPUVAddr gpu_addr, const SurfaceParams& params);
+ explicit SurfaceBaseImpl(GPUVAddr gpu_addr, const SurfaceParams& params,
+ bool is_astc_supported);
~SurfaceBaseImpl() = default;
virtual void DecorateSurfaceName() = 0;
@@ -168,12 +157,11 @@ protected:
const SurfaceParams params;
std::size_t layer_size;
std::size_t guest_memory_size;
- const std::size_t host_memory_size;
+ std::size_t host_memory_size;
GPUVAddr gpu_addr{};
- CacheAddr cache_addr{};
- CacheAddr cache_addr_end{};
VAddr cpu_addr{};
- bool is_continuous{};
+ VAddr cpu_addr_end{};
+ bool is_converted{};
std::vector<std::size_t> mipmap_sizes;
std::vector<std::size_t> mipmap_offsets;
@@ -288,8 +276,9 @@ public:
}
protected:
- explicit SurfaceBase(const GPUVAddr gpu_addr, const SurfaceParams& params)
- : SurfaceBaseImpl(gpu_addr, params) {}
+ explicit SurfaceBase(const GPUVAddr gpu_addr, const SurfaceParams& params,
+ bool is_astc_supported)
+ : SurfaceBaseImpl(gpu_addr, params, is_astc_supported) {}
~SurfaceBase() = default;
diff --git a/src/video_core/texture_cache/surface_params.cpp b/src/video_core/texture_cache/surface_params.cpp
index 9931c5ef7..6f3ef45be 100644
--- a/src/video_core/texture_cache/surface_params.cpp
+++ b/src/video_core/texture_cache/surface_params.cpp
@@ -113,10 +113,8 @@ SurfaceParams SurfaceParams::CreateForTexture(const FormatLookupTable& lookup_ta
params.height = tic.Height();
params.depth = tic.Depth();
params.pitch = params.is_tiled ? 0 : tic.Pitch();
- if (params.target == SurfaceTarget::Texture2D && params.depth > 1) {
- params.depth = 1;
- } else if (params.target == SurfaceTarget::TextureCubemap ||
- params.target == SurfaceTarget::TextureCubeArray) {
+ if (params.target == SurfaceTarget::TextureCubemap ||
+ params.target == SurfaceTarget::TextureCubeArray) {
params.depth *= 6;
}
params.num_levels = tic.max_mip_level + 1;
@@ -309,28 +307,26 @@ std::size_t SurfaceParams::GetGuestMipmapLevelOffset(u32 level) const {
return offset;
}
-std::size_t SurfaceParams::GetHostMipmapLevelOffset(u32 level) const {
- std::size_t offset = 0;
- for (u32 i = 0; i < level; i++) {
- offset += GetInnerMipmapMemorySize(i, true, false) * GetNumLayers();
- }
- return offset;
-}
-
-std::size_t SurfaceParams::GetConvertedMipmapOffset(u32 level) const {
+std::size_t SurfaceParams::GetHostMipmapLevelOffset(u32 level, bool is_converted) const {
std::size_t offset = 0;
- for (u32 i = 0; i < level; i++) {
- offset += GetConvertedMipmapSize(i);
+ if (is_converted) {
+ for (u32 i = 0; i < level; ++i) {
+ offset += GetConvertedMipmapSize(i) * GetNumLayers();
+ }
+ } else {
+ for (u32 i = 0; i < level; ++i) {
+ offset += GetInnerMipmapMemorySize(i, true, false) * GetNumLayers();
+ }
}
return offset;
}
std::size_t SurfaceParams::GetConvertedMipmapSize(u32 level) const {
constexpr std::size_t rgba8_bpp = 4ULL;
- const std::size_t width_t = GetMipWidth(level);
- const std::size_t height_t = GetMipHeight(level);
- const std::size_t depth_t = is_layered ? depth : GetMipDepth(level);
- return width_t * height_t * depth_t * rgba8_bpp;
+ const std::size_t mip_width = GetMipWidth(level);
+ const std::size_t mip_height = GetMipHeight(level);
+ const std::size_t mip_depth = is_layered ? 1 : GetMipDepth(level);
+ return mip_width * mip_height * mip_depth * rgba8_bpp;
}
std::size_t SurfaceParams::GetLayerSize(bool as_host_size, bool uncompressed) const {
diff --git a/src/video_core/texture_cache/surface_params.h b/src/video_core/texture_cache/surface_params.h
index 995cc3818..24957df8d 100644
--- a/src/video_core/texture_cache/surface_params.h
+++ b/src/video_core/texture_cache/surface_params.h
@@ -20,8 +20,6 @@ namespace VideoCommon {
class FormatLookupTable;
-using VideoCore::Surface::SurfaceCompression;
-
class SurfaceParams {
public:
/// Creates SurfaceCachedParams from a texture configuration.
@@ -67,16 +65,14 @@ public:
return GetInnerMemorySize(false, false, false);
}
- std::size_t GetHostSizeInBytes() const {
- std::size_t host_size_in_bytes;
- if (GetCompressionType() == SurfaceCompression::Converted) {
- // ASTC is uncompressed in software, in emulated as RGBA8
- host_size_in_bytes = 0;
- for (u32 level = 0; level < num_levels; ++level) {
- host_size_in_bytes += GetConvertedMipmapSize(level);
- }
- } else {
- host_size_in_bytes = GetInnerMemorySize(true, false, false);
+ std::size_t GetHostSizeInBytes(bool is_converted) const {
+ if (!is_converted) {
+ return GetInnerMemorySize(true, false, false);
+ }
+ // ASTC is uncompressed in software, in emulated as RGBA8
+ std::size_t host_size_in_bytes = 0;
+ for (u32 level = 0; level < num_levels; ++level) {
+ host_size_in_bytes += GetConvertedMipmapSize(level) * GetNumLayers();
}
return host_size_in_bytes;
}
@@ -107,9 +103,8 @@ public:
u32 GetMipBlockDepth(u32 level) const;
/// Returns the best possible row/pitch alignment for the surface.
- u32 GetRowAlignment(u32 level) const {
- const u32 bpp =
- GetCompressionType() == SurfaceCompression::Converted ? 4 : GetBytesPerPixel();
+ u32 GetRowAlignment(u32 level, bool is_converted) const {
+ const u32 bpp = is_converted ? 4 : GetBytesPerPixel();
return 1U << Common::CountTrailingZeroes32(GetMipWidth(level) * bpp);
}
@@ -117,11 +112,7 @@ public:
std::size_t GetGuestMipmapLevelOffset(u32 level) const;
/// Returns the offset in bytes in host memory (linear) of a given mipmap level.
- std::size_t GetHostMipmapLevelOffset(u32 level) const;
-
- /// Returns the offset in bytes in host memory (linear) of a given mipmap level
- /// for a texture that is converted in host gpu.
- std::size_t GetConvertedMipmapOffset(u32 level) const;
+ std::size_t GetHostMipmapLevelOffset(u32 level, bool is_converted) const;
/// Returns the size in bytes in guest memory of a given mipmap level.
std::size_t GetGuestMipmapSize(u32 level) const {
@@ -196,11 +187,6 @@ public:
pixel_format < VideoCore::Surface::PixelFormat::MaxDepthStencilFormat;
}
- /// Returns how the compression should be handled for this texture.
- SurfaceCompression GetCompressionType() const {
- return VideoCore::Surface::GetFormatCompressionType(pixel_format);
- }
-
/// Returns is the surface is a TextureBuffer type of surface.
bool IsBuffer() const {
return target == VideoCore::Surface::SurfaceTarget::TextureBuffer;
diff --git a/src/video_core/texture_cache/texture_cache.h b/src/video_core/texture_cache/texture_cache.h
index 6cdbe63d0..88fe3e25f 100644
--- a/src/video_core/texture_cache/texture_cache.h
+++ b/src/video_core/texture_cache/texture_cache.h
@@ -52,11 +52,9 @@ using RenderTargetConfig = Tegra::Engines::Maxwell3D::Regs::RenderTargetConfig;
template <typename TSurface, typename TView>
class TextureCache {
- using IntervalMap = boost::icl::interval_map<CacheAddr, std::set<TSurface>>;
- using IntervalType = typename IntervalMap::interval_type;
public:
- void InvalidateRegion(CacheAddr addr, std::size_t size) {
+ void InvalidateRegion(VAddr addr, std::size_t size) {
std::lock_guard lock{mutex};
for (const auto& surface : GetSurfacesInRegion(addr, size)) {
@@ -76,7 +74,7 @@ public:
guard_samplers = new_guard;
}
- void FlushRegion(CacheAddr addr, std::size_t size) {
+ void FlushRegion(VAddr addr, std::size_t size) {
std::lock_guard lock{mutex};
auto surfaces = GetSurfacesInRegion(addr, size);
@@ -99,9 +97,9 @@ public:
return GetNullSurface(SurfaceParams::ExpectedTarget(entry));
}
- const auto host_ptr{system.GPU().MemoryManager().GetPointer(gpu_addr)};
- const auto cache_addr{ToCacheAddr(host_ptr)};
- if (!cache_addr) {
+ const std::optional<VAddr> cpu_addr =
+ system.GPU().MemoryManager().GpuToCpuAddress(gpu_addr);
+ if (!cpu_addr) {
return GetNullSurface(SurfaceParams::ExpectedTarget(entry));
}
@@ -110,7 +108,7 @@ public:
}
const auto params{SurfaceParams::CreateForTexture(format_lookup_table, tic, entry)};
- const auto [surface, view] = GetSurface(gpu_addr, cache_addr, params, true, false);
+ const auto [surface, view] = GetSurface(gpu_addr, *cpu_addr, params, true, false);
if (guard_samplers) {
sampled_textures.push_back(surface);
}
@@ -124,13 +122,13 @@ public:
if (!gpu_addr) {
return GetNullSurface(SurfaceParams::ExpectedTarget(entry));
}
- const auto host_ptr{system.GPU().MemoryManager().GetPointer(gpu_addr)};
- const auto cache_addr{ToCacheAddr(host_ptr)};
- if (!cache_addr) {
+ const std::optional<VAddr> cpu_addr =
+ system.GPU().MemoryManager().GpuToCpuAddress(gpu_addr);
+ if (!cpu_addr) {
return GetNullSurface(SurfaceParams::ExpectedTarget(entry));
}
const auto params{SurfaceParams::CreateForImage(format_lookup_table, tic, entry)};
- const auto [surface, view] = GetSurface(gpu_addr, cache_addr, params, true, false);
+ const auto [surface, view] = GetSurface(gpu_addr, *cpu_addr, params, true, false);
if (guard_samplers) {
sampled_textures.push_back(surface);
}
@@ -159,14 +157,14 @@ public:
SetEmptyDepthBuffer();
return {};
}
- const auto host_ptr{system.GPU().MemoryManager().GetPointer(gpu_addr)};
- const auto cache_addr{ToCacheAddr(host_ptr)};
- if (!cache_addr) {
+ const std::optional<VAddr> cpu_addr =
+ system.GPU().MemoryManager().GpuToCpuAddress(gpu_addr);
+ if (!cpu_addr) {
SetEmptyDepthBuffer();
return {};
}
const auto depth_params{SurfaceParams::CreateForDepthBuffer(system)};
- auto surface_view = GetSurface(gpu_addr, cache_addr, depth_params, preserve_contents, true);
+ auto surface_view = GetSurface(gpu_addr, *cpu_addr, depth_params, preserve_contents, true);
if (depth_buffer.target)
depth_buffer.target->MarkAsRenderTarget(false, NO_RT);
depth_buffer.target = surface_view.first;
@@ -199,15 +197,15 @@ public:
return {};
}
- const auto host_ptr{system.GPU().MemoryManager().GetPointer(gpu_addr)};
- const auto cache_addr{ToCacheAddr(host_ptr)};
- if (!cache_addr) {
+ const std::optional<VAddr> cpu_addr =
+ system.GPU().MemoryManager().GpuToCpuAddress(gpu_addr);
+ if (!cpu_addr) {
SetEmptyColorBuffer(index);
return {};
}
auto surface_view =
- GetSurface(gpu_addr, cache_addr, SurfaceParams::CreateForFramebuffer(system, index),
+ GetSurface(gpu_addr, *cpu_addr, SurfaceParams::CreateForFramebuffer(system, index),
preserve_contents, true);
if (render_targets[index].target)
render_targets[index].target->MarkAsRenderTarget(false, NO_RT);
@@ -257,27 +255,26 @@ public:
const GPUVAddr src_gpu_addr = src_config.Address();
const GPUVAddr dst_gpu_addr = dst_config.Address();
DeduceBestBlit(src_params, dst_params, src_gpu_addr, dst_gpu_addr);
- const auto dst_host_ptr{system.GPU().MemoryManager().GetPointer(dst_gpu_addr)};
- const auto dst_cache_addr{ToCacheAddr(dst_host_ptr)};
- const auto src_host_ptr{system.GPU().MemoryManager().GetPointer(src_gpu_addr)};
- const auto src_cache_addr{ToCacheAddr(src_host_ptr)};
+ const std::optional<VAddr> dst_cpu_addr =
+ system.GPU().MemoryManager().GpuToCpuAddress(dst_gpu_addr);
+ const std::optional<VAddr> src_cpu_addr =
+ system.GPU().MemoryManager().GpuToCpuAddress(src_gpu_addr);
std::pair<TSurface, TView> dst_surface =
- GetSurface(dst_gpu_addr, dst_cache_addr, dst_params, true, false);
+ GetSurface(dst_gpu_addr, *dst_cpu_addr, dst_params, true, false);
std::pair<TSurface, TView> src_surface =
- GetSurface(src_gpu_addr, src_cache_addr, src_params, true, false);
+ GetSurface(src_gpu_addr, *src_cpu_addr, src_params, true, false);
ImageBlit(src_surface.second, dst_surface.second, copy_config);
dst_surface.first->MarkAsModified(true, Tick());
}
- TSurface TryFindFramebufferSurface(const u8* host_ptr) {
- const CacheAddr cache_addr = ToCacheAddr(host_ptr);
- if (!cache_addr) {
+ TSurface TryFindFramebufferSurface(VAddr addr) {
+ if (!addr) {
return nullptr;
}
- const CacheAddr page = cache_addr >> registry_page_bits;
+ const VAddr page = addr >> registry_page_bits;
std::vector<TSurface>& list = registry[page];
for (auto& surface : list) {
- if (surface->GetCacheAddr() == cache_addr) {
+ if (surface->GetCpuAddr() == addr) {
return surface;
}
}
@@ -289,8 +286,9 @@ public:
}
protected:
- TextureCache(Core::System& system, VideoCore::RasterizerInterface& rasterizer)
- : system{system}, rasterizer{rasterizer} {
+ explicit TextureCache(Core::System& system, VideoCore::RasterizerInterface& rasterizer,
+ bool is_astc_supported)
+ : system{system}, is_astc_supported{is_astc_supported}, rasterizer{rasterizer} {
for (std::size_t i = 0; i < Tegra::Engines::Maxwell3D::Regs::NumRenderTargets; i++) {
SetEmptyColorBuffer(i);
}
@@ -337,18 +335,14 @@ protected:
void Register(TSurface surface) {
const GPUVAddr gpu_addr = surface->GetGpuAddr();
- const CacheAddr cache_ptr = ToCacheAddr(system.GPU().MemoryManager().GetPointer(gpu_addr));
const std::size_t size = surface->GetSizeInBytes();
const std::optional<VAddr> cpu_addr =
system.GPU().MemoryManager().GpuToCpuAddress(gpu_addr);
- if (!cache_ptr || !cpu_addr) {
+ if (!cpu_addr) {
LOG_CRITICAL(HW_GPU, "Failed to register surface with unmapped gpu_address 0x{:016x}",
gpu_addr);
return;
}
- const bool continuous = system.GPU().MemoryManager().IsBlockContinuous(gpu_addr, size);
- surface->MarkAsContinuous(continuous);
- surface->SetCacheAddr(cache_ptr);
surface->SetCpuAddr(*cpu_addr);
RegisterInnerCache(surface);
surface->MarkAsRegistered(true);
@@ -381,6 +375,7 @@ protected:
}
Core::System& system;
+ const bool is_astc_supported;
private:
enum class RecycleStrategy : u32 {
@@ -632,7 +627,7 @@ private:
std::optional<std::pair<TSurface, TView>> Manage3DSurfaces(std::vector<TSurface>& overlaps,
const SurfaceParams& params,
const GPUVAddr gpu_addr,
- const CacheAddr cache_addr,
+ const VAddr cpu_addr,
bool preserve_contents) {
if (params.target == SurfaceTarget::Texture3D) {
bool failed = false;
@@ -657,7 +652,7 @@ private:
failed = true;
break;
}
- const u32 offset = static_cast<u32>(surface->GetCacheAddr() - cache_addr);
+ const u32 offset = static_cast<u32>(surface->GetCpuAddr() - cpu_addr);
const auto [x, y, z] = params.GetBlockOffsetXYZ(offset);
modified |= surface->IsModified();
const CopyParams copy_params(0, 0, 0, 0, 0, z, 0, 0, params.width, params.height,
@@ -677,7 +672,7 @@ private:
} else {
for (const auto& surface : overlaps) {
if (!surface->MatchTarget(params.target)) {
- if (overlaps.size() == 1 && surface->GetCacheAddr() == cache_addr) {
+ if (overlaps.size() == 1 && surface->GetCpuAddr() == cpu_addr) {
if (Settings::values.use_accurate_gpu_emulation) {
return std::nullopt;
}
@@ -686,7 +681,7 @@ private:
}
return std::nullopt;
}
- if (surface->GetCacheAddr() != cache_addr) {
+ if (surface->GetCpuAddr() != cpu_addr) {
continue;
}
if (surface->MatchesStructure(params) == MatchStructureResult::FullMatch) {
@@ -720,13 +715,13 @@ private:
* left blank.
* @param is_render Whether or not the surface is a render target.
**/
- std::pair<TSurface, TView> GetSurface(const GPUVAddr gpu_addr, const CacheAddr cache_addr,
+ std::pair<TSurface, TView> GetSurface(const GPUVAddr gpu_addr, const VAddr cpu_addr,
const SurfaceParams& params, bool preserve_contents,
bool is_render) {
// Step 1
// Check Level 1 Cache for a fast structural match. If candidate surface
// matches at certain level we are pretty much done.
- if (const auto iter = l1_cache.find(cache_addr); iter != l1_cache.end()) {
+ if (const auto iter = l1_cache.find(cpu_addr); iter != l1_cache.end()) {
TSurface& current_surface = iter->second;
const auto topological_result = current_surface->MatchesTopology(params);
if (topological_result != MatchTopologyResult::FullMatch) {
@@ -753,7 +748,7 @@ private:
// Step 2
// Obtain all possible overlaps in the memory region
const std::size_t candidate_size = params.GetGuestSizeInBytes();
- auto overlaps{GetSurfacesInRegion(cache_addr, candidate_size)};
+ auto overlaps{GetSurfacesInRegion(cpu_addr, candidate_size)};
// If none are found, we are done. we just load the surface and create it.
if (overlaps.empty()) {
@@ -775,7 +770,7 @@ private:
// Check if it's a 3D texture
if (params.block_depth > 0) {
auto surface =
- Manage3DSurfaces(overlaps, params, gpu_addr, cache_addr, preserve_contents);
+ Manage3DSurfaces(overlaps, params, gpu_addr, cpu_addr, preserve_contents);
if (surface) {
return *surface;
}
@@ -850,16 +845,16 @@ private:
* @param params The parameters on the candidate surface.
**/
Deduction DeduceSurface(const GPUVAddr gpu_addr, const SurfaceParams& params) {
- const auto host_ptr{system.GPU().MemoryManager().GetPointer(gpu_addr)};
- const auto cache_addr{ToCacheAddr(host_ptr)};
+ const std::optional<VAddr> cpu_addr =
+ system.GPU().MemoryManager().GpuToCpuAddress(gpu_addr);
- if (!cache_addr) {
+ if (!cpu_addr) {
Deduction result{};
result.type = DeductionType::DeductionFailed;
return result;
}
- if (const auto iter = l1_cache.find(cache_addr); iter != l1_cache.end()) {
+ if (const auto iter = l1_cache.find(*cpu_addr); iter != l1_cache.end()) {
TSurface& current_surface = iter->second;
const auto topological_result = current_surface->MatchesTopology(params);
if (topological_result != MatchTopologyResult::FullMatch) {
@@ -878,7 +873,7 @@ private:
}
const std::size_t candidate_size = params.GetGuestSizeInBytes();
- auto overlaps{GetSurfacesInRegion(cache_addr, candidate_size)};
+ auto overlaps{GetSurfacesInRegion(*cpu_addr, candidate_size)};
if (overlaps.empty()) {
Deduction result{};
@@ -1022,10 +1017,10 @@ private:
}
void RegisterInnerCache(TSurface& surface) {
- const CacheAddr cache_addr = surface->GetCacheAddr();
- CacheAddr start = cache_addr >> registry_page_bits;
- const CacheAddr end = (surface->GetCacheAddrEnd() - 1) >> registry_page_bits;
- l1_cache[cache_addr] = surface;
+ const VAddr cpu_addr = surface->GetCpuAddr();
+ VAddr start = cpu_addr >> registry_page_bits;
+ const VAddr end = (surface->GetCpuAddrEnd() - 1) >> registry_page_bits;
+ l1_cache[cpu_addr] = surface;
while (start <= end) {
registry[start].push_back(surface);
start++;
@@ -1033,10 +1028,10 @@ private:
}
void UnregisterInnerCache(TSurface& surface) {
- const CacheAddr cache_addr = surface->GetCacheAddr();
- CacheAddr start = cache_addr >> registry_page_bits;
- const CacheAddr end = (surface->GetCacheAddrEnd() - 1) >> registry_page_bits;
- l1_cache.erase(cache_addr);
+ const VAddr cpu_addr = surface->GetCpuAddr();
+ VAddr start = cpu_addr >> registry_page_bits;
+ const VAddr end = (surface->GetCpuAddrEnd() - 1) >> registry_page_bits;
+ l1_cache.erase(cpu_addr);
while (start <= end) {
auto& reg{registry[start]};
reg.erase(std::find(reg.begin(), reg.end(), surface));
@@ -1044,18 +1039,18 @@ private:
}
}
- std::vector<TSurface> GetSurfacesInRegion(const CacheAddr cache_addr, const std::size_t size) {
+ std::vector<TSurface> GetSurfacesInRegion(const VAddr cpu_addr, const std::size_t size) {
if (size == 0) {
return {};
}
- const CacheAddr cache_addr_end = cache_addr + size;
- CacheAddr start = cache_addr >> registry_page_bits;
- const CacheAddr end = (cache_addr_end - 1) >> registry_page_bits;
+ const VAddr cpu_addr_end = cpu_addr + size;
+ VAddr start = cpu_addr >> registry_page_bits;
+ const VAddr end = (cpu_addr_end - 1) >> registry_page_bits;
std::vector<TSurface> surfaces;
while (start <= end) {
std::vector<TSurface>& list = registry[start];
for (auto& surface : list) {
- if (!surface->IsPicked() && surface->Overlaps(cache_addr, cache_addr_end)) {
+ if (!surface->IsPicked() && surface->Overlaps(cpu_addr, cpu_addr_end)) {
surface->MarkAsPicked(true);
surfaces.push_back(surface);
}
@@ -1144,14 +1139,14 @@ private:
// large in size.
static constexpr u64 registry_page_bits{20};
static constexpr u64 registry_page_size{1 << registry_page_bits};
- std::unordered_map<CacheAddr, std::vector<TSurface>> registry;
+ std::unordered_map<VAddr, std::vector<TSurface>> registry;
static constexpr u32 DEPTH_RT = 8;
static constexpr u32 NO_RT = 0xFFFFFFFF;
// The L1 Cache is used for fast texture lookup before checking the overlaps
// This avoids calculating size and other stuffs.
- std::unordered_map<CacheAddr, TSurface> l1_cache;
+ std::unordered_map<VAddr, TSurface> l1_cache;
/// The surface reserve is a "backup" cache, this is where we put unique surfaces that have
/// previously been used. This is to prevent surfaces from being constantly created and
diff --git a/src/video_core/textures/texture.cpp b/src/video_core/textures/texture.cpp
new file mode 100644
index 000000000..d1939d744
--- /dev/null
+++ b/src/video_core/textures/texture.cpp
@@ -0,0 +1,80 @@
+// Copyright 2020 yuzu Emulator Project
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#include <algorithm>
+#include <array>
+
+#include "core/settings.h"
+#include "video_core/textures/texture.h"
+
+namespace Tegra::Texture {
+
+namespace {
+
+constexpr std::array<float, 256> SRGB_CONVERSION_LUT = {
+ 0.000000f, 0.000000f, 0.000000f, 0.000012f, 0.000021f, 0.000033f, 0.000046f, 0.000062f,
+ 0.000081f, 0.000102f, 0.000125f, 0.000151f, 0.000181f, 0.000214f, 0.000251f, 0.000293f,
+ 0.000338f, 0.000388f, 0.000443f, 0.000503f, 0.000568f, 0.000639f, 0.000715f, 0.000798f,
+ 0.000887f, 0.000983f, 0.001085f, 0.001195f, 0.001312f, 0.001437f, 0.001569f, 0.001710f,
+ 0.001860f, 0.002019f, 0.002186f, 0.002364f, 0.002551f, 0.002748f, 0.002955f, 0.003174f,
+ 0.003403f, 0.003643f, 0.003896f, 0.004160f, 0.004436f, 0.004725f, 0.005028f, 0.005343f,
+ 0.005672f, 0.006015f, 0.006372f, 0.006744f, 0.007130f, 0.007533f, 0.007950f, 0.008384f,
+ 0.008834f, 0.009301f, 0.009785f, 0.010286f, 0.010805f, 0.011342f, 0.011898f, 0.012472f,
+ 0.013066f, 0.013680f, 0.014313f, 0.014967f, 0.015641f, 0.016337f, 0.017054f, 0.017793f,
+ 0.018554f, 0.019337f, 0.020144f, 0.020974f, 0.021828f, 0.022706f, 0.023609f, 0.024536f,
+ 0.025489f, 0.026468f, 0.027473f, 0.028504f, 0.029563f, 0.030649f, 0.031762f, 0.032904f,
+ 0.034074f, 0.035274f, 0.036503f, 0.037762f, 0.039050f, 0.040370f, 0.041721f, 0.043103f,
+ 0.044518f, 0.045964f, 0.047444f, 0.048956f, 0.050503f, 0.052083f, 0.053699f, 0.055349f,
+ 0.057034f, 0.058755f, 0.060513f, 0.062307f, 0.064139f, 0.066008f, 0.067915f, 0.069861f,
+ 0.071845f, 0.073869f, 0.075933f, 0.078037f, 0.080182f, 0.082369f, 0.084597f, 0.086867f,
+ 0.089180f, 0.091535f, 0.093935f, 0.096378f, 0.098866f, 0.101398f, 0.103977f, 0.106601f,
+ 0.109271f, 0.111988f, 0.114753f, 0.117565f, 0.120426f, 0.123335f, 0.126293f, 0.129301f,
+ 0.132360f, 0.135469f, 0.138629f, 0.141841f, 0.145105f, 0.148421f, 0.151791f, 0.155214f,
+ 0.158691f, 0.162224f, 0.165810f, 0.169453f, 0.173152f, 0.176907f, 0.180720f, 0.184589f,
+ 0.188517f, 0.192504f, 0.196549f, 0.200655f, 0.204820f, 0.209046f, 0.213334f, 0.217682f,
+ 0.222093f, 0.226567f, 0.231104f, 0.235704f, 0.240369f, 0.245099f, 0.249894f, 0.254754f,
+ 0.259681f, 0.264674f, 0.269736f, 0.274864f, 0.280062f, 0.285328f, 0.290664f, 0.296070f,
+ 0.301546f, 0.307094f, 0.312713f, 0.318404f, 0.324168f, 0.330006f, 0.335916f, 0.341902f,
+ 0.347962f, 0.354097f, 0.360309f, 0.366597f, 0.372961f, 0.379403f, 0.385924f, 0.392524f,
+ 0.399202f, 0.405960f, 0.412798f, 0.419718f, 0.426719f, 0.433802f, 0.440967f, 0.448216f,
+ 0.455548f, 0.462965f, 0.470465f, 0.478052f, 0.485725f, 0.493484f, 0.501329f, 0.509263f,
+ 0.517285f, 0.525396f, 0.533595f, 0.541885f, 0.550265f, 0.558736f, 0.567299f, 0.575954f,
+ 0.584702f, 0.593542f, 0.602477f, 0.611507f, 0.620632f, 0.629852f, 0.639168f, 0.648581f,
+ 0.658092f, 0.667700f, 0.677408f, 0.687214f, 0.697120f, 0.707127f, 0.717234f, 0.727443f,
+ 0.737753f, 0.748167f, 0.758685f, 0.769305f, 0.780031f, 0.790861f, 0.801798f, 0.812839f,
+ 0.823989f, 0.835246f, 0.846611f, 0.858085f, 0.869668f, 0.881360f, 0.893164f, 0.905078f,
+ 0.917104f, 0.929242f, 0.941493f, 0.953859f, 0.966338f, 1.000000f, 1.000000f, 1.000000f,
+};
+
+unsigned SettingsMinimumAnisotropy() noexcept {
+ switch (static_cast<Anisotropy>(Settings::values.max_anisotropy)) {
+ default:
+ case Anisotropy::Default:
+ return 1U;
+ case Anisotropy::Filter2x:
+ return 2U;
+ case Anisotropy::Filter4x:
+ return 4U;
+ case Anisotropy::Filter8x:
+ return 8U;
+ case Anisotropy::Filter16x:
+ return 16U;
+ }
+}
+
+} // Anonymous namespace
+
+std::array<float, 4> TSCEntry::GetBorderColor() const noexcept {
+ if (!srgb_conversion) {
+ return border_color;
+ }
+ return {SRGB_CONVERSION_LUT[srgb_border_color_r], SRGB_CONVERSION_LUT[srgb_border_color_g],
+ SRGB_CONVERSION_LUT[srgb_border_color_b], border_color[3]};
+}
+
+float TSCEntry::GetMaxAnisotropy() const noexcept {
+ return static_cast<float>(std::max(1U << max_anisotropy, SettingsMinimumAnisotropy()));
+}
+
+} // namespace Tegra::Texture
diff --git a/src/video_core/textures/texture.h b/src/video_core/textures/texture.h
index 7edc4abe1..eba05aced 100644
--- a/src/video_core/textures/texture.h
+++ b/src/video_core/textures/texture.h
@@ -8,7 +8,6 @@
#include "common/assert.h"
#include "common/bit_field.h"
#include "common/common_types.h"
-#include "core/settings.h"
namespace Tegra::Texture {
@@ -132,6 +131,20 @@ enum class SwizzleSource : u32 {
OneFloat = 7,
};
+enum class MsaaMode : u32 {
+ Msaa1x1 = 0,
+ Msaa2x1 = 1,
+ Msaa2x2 = 2,
+ Msaa4x2 = 3,
+ Msaa4x2_D3D = 4,
+ Msaa2x1_D3D = 5,
+ Msaa4x4 = 6,
+ Msaa2x2_VC4 = 8,
+ Msaa2x2_VC12 = 9,
+ Msaa4x2_VC8 = 10,
+ Msaa4x2_VC24 = 11,
+};
+
union TextureHandle {
TextureHandle(u32 raw) : raw{raw} {}
@@ -198,6 +211,7 @@ struct TICEntry {
union {
BitField<0, 4, u32> res_min_mip_level;
BitField<4, 4, u32> res_max_mip_level;
+ BitField<8, 4, MsaaMode> msaa_mode;
BitField<12, 12, u32> min_lod_clamp;
};
@@ -336,24 +350,9 @@ struct TSCEntry {
std::array<u8, 0x20> raw;
};
- float GetMaxAnisotropy() const {
- const u32 min_value = [] {
- switch (static_cast<Anisotropy>(Settings::values.max_anisotropy)) {
- default:
- case Anisotropy::Default:
- return 1U;
- case Anisotropy::Filter2x:
- return 2U;
- case Anisotropy::Filter4x:
- return 4U;
- case Anisotropy::Filter8x:
- return 8U;
- case Anisotropy::Filter16x:
- return 16U;
- }
- }();
- return static_cast<float>(std::max(1U << max_anisotropy, min_value));
- }
+ std::array<float, 4> GetBorderColor() const noexcept;
+
+ float GetMaxAnisotropy() const noexcept;
float GetMinLod() const {
return static_cast<float>(min_lod_clamp) / 256.0f;
@@ -368,15 +367,6 @@ struct TSCEntry {
constexpr u32 mask = 1U << (13 - 1);
return static_cast<float>(static_cast<s32>((mip_lod_bias ^ mask) - mask)) / 256.0f;
}
-
- std::array<float, 4> GetBorderColor() const {
- if (srgb_conversion) {
- return {static_cast<float>(srgb_border_color_r) / 255.0f,
- static_cast<float>(srgb_border_color_g) / 255.0f,
- static_cast<float>(srgb_border_color_b) / 255.0f, border_color[3]};
- }
- return border_color;
- }
};
static_assert(sizeof(TSCEntry) == 0x20, "TSCEntry has wrong size");
diff --git a/src/video_core/video_core.cpp b/src/video_core/video_core.cpp
index a5f81a8a0..f60bdc60a 100644
--- a/src/video_core/video_core.cpp
+++ b/src/video_core/video_core.cpp
@@ -15,13 +15,13 @@
#endif
#include "video_core/video_core.h"
-namespace VideoCore {
-
-std::unique_ptr<RendererBase> CreateRenderer(Core::Frontend::EmuWindow& emu_window,
- Core::System& system) {
+namespace {
+std::unique_ptr<VideoCore::RendererBase> CreateRenderer(Core::Frontend::EmuWindow& emu_window,
+ Core::System& system,
+ Core::Frontend::GraphicsContext& context) {
switch (Settings::values.renderer_backend) {
case Settings::RendererBackend::OpenGL:
- return std::make_unique<OpenGL::RendererOpenGL>(emu_window, system);
+ return std::make_unique<OpenGL::RendererOpenGL>(emu_window, system, context);
#ifdef HAS_VULKAN
case Settings::RendererBackend::Vulkan:
return std::make_unique<Vulkan::RendererVulkan>(emu_window, system);
@@ -30,13 +30,23 @@ std::unique_ptr<RendererBase> CreateRenderer(Core::Frontend::EmuWindow& emu_wind
return nullptr;
}
}
+} // Anonymous namespace
-std::unique_ptr<Tegra::GPU> CreateGPU(Core::System& system) {
- if (Settings::values.use_asynchronous_gpu_emulation) {
- return std::make_unique<VideoCommon::GPUAsynch>(system, system.Renderer());
+namespace VideoCore {
+
+std::unique_ptr<Tegra::GPU> CreateGPU(Core::Frontend::EmuWindow& emu_window, Core::System& system) {
+ auto context = emu_window.CreateSharedContext();
+ const auto scope = context->Acquire();
+ auto renderer = CreateRenderer(emu_window, system, *context);
+ if (!renderer->Init()) {
+ return nullptr;
}
- return std::make_unique<VideoCommon::GPUSynch>(system, system.Renderer());
+ if (Settings::values.use_asynchronous_gpu_emulation) {
+ return std::make_unique<VideoCommon::GPUAsynch>(system, std::move(renderer),
+ std::move(context));
+ }
+ return std::make_unique<VideoCommon::GPUSynch>(system, std::move(renderer), std::move(context));
}
u16 GetResolutionScaleFactor(const RendererBase& renderer) {
diff --git a/src/video_core/video_core.h b/src/video_core/video_core.h
index b8e0ac372..f5c27125d 100644
--- a/src/video_core/video_core.h
+++ b/src/video_core/video_core.h
@@ -22,17 +22,8 @@ namespace VideoCore {
class RendererBase;
-/**
- * Creates a renderer instance.
- *
- * @note The returned renderer instance is simply allocated. Its Init()
- * function still needs to be called to fully complete its setup.
- */
-std::unique_ptr<RendererBase> CreateRenderer(Core::Frontend::EmuWindow& emu_window,
- Core::System& system);
-
/// Creates an emulated GPU instance using the given system context.
-std::unique_ptr<Tegra::GPU> CreateGPU(Core::System& system);
+std::unique_ptr<Tegra::GPU> CreateGPU(Core::Frontend::EmuWindow& emu_window, Core::System& system);
u16 GetResolutionScaleFactor(const RendererBase& renderer);