aboutsummaryrefslogtreecommitdiff
path: root/src/video_core
diff options
context:
space:
mode:
Diffstat (limited to 'src/video_core')
-rw-r--r--src/video_core/engines/maxwell_3d.h4
-rw-r--r--src/video_core/host1x/syncpoint_manager.cpp12
-rw-r--r--src/video_core/renderer_base.cpp8
-rw-r--r--src/video_core/renderer_opengl/gl_shader_cache.cpp4
-rw-r--r--src/video_core/renderer_vulkan/fixed_pipeline_state.cpp4
-rw-r--r--src/video_core/renderer_vulkan/vk_staging_buffer_pool.cpp84
-rw-r--r--src/video_core/renderer_vulkan/vk_staging_buffer_pool.h3
7 files changed, 51 insertions, 68 deletions
diff --git a/src/video_core/engines/maxwell_3d.h b/src/video_core/engines/maxwell_3d.h
index 12dbd9cc4..75e3b868d 100644
--- a/src/video_core/engines/maxwell_3d.h
+++ b/src/video_core/engines/maxwell_3d.h
@@ -390,7 +390,7 @@ public:
FractionalEven = 2,
};
- enum class OutputPrimitves : u32 {
+ enum class OutputPrimitives : u32 {
Points = 0,
Lines = 1,
Triangles_CW = 2,
@@ -401,7 +401,7 @@ public:
union {
BitField<0, 2, DomainType> domain_type;
BitField<4, 2, Spacing> spacing;
- BitField<8, 2, OutputPrimitves> output_primitives;
+ BitField<8, 2, OutputPrimitives> output_primitives;
};
} params;
diff --git a/src/video_core/host1x/syncpoint_manager.cpp b/src/video_core/host1x/syncpoint_manager.cpp
index 326e8355a..a44fc83d3 100644
--- a/src/video_core/host1x/syncpoint_manager.cpp
+++ b/src/video_core/host1x/syncpoint_manager.cpp
@@ -36,7 +36,17 @@ SyncpointManager::ActionHandle SyncpointManager::RegisterAction(
void SyncpointManager::DeregisterAction(std::list<RegisteredAction>& action_storage,
ActionHandle& handle) {
std::unique_lock lk(guard);
- action_storage.erase(handle);
+
+ // We want to ensure the iterator still exists prior to erasing it
+ // Otherwise, if an invalid iterator was passed in then it could lead to UB
+ // It is important to avoid UB in that case since the deregister isn't called from a locked
+ // context
+ for (auto it = action_storage.begin(); it != action_storage.end(); it++) {
+ if (it == handle) {
+ action_storage.erase(it);
+ return;
+ }
+ }
}
void SyncpointManager::DeregisterGuestAction(u32 syncpoint_id, ActionHandle& handle) {
diff --git a/src/video_core/renderer_base.cpp b/src/video_core/renderer_base.cpp
index 45791aa75..e8761a747 100644
--- a/src/video_core/renderer_base.cpp
+++ b/src/video_core/renderer_base.cpp
@@ -1,6 +1,8 @@
// SPDX-FileCopyrightText: 2015 Citra Emulator Project
// SPDX-License-Identifier: GPL-2.0-or-later
+#include <thread>
+
#include "common/logging/log.h"
#include "core/frontend/emu_window.h"
#include "video_core/renderer_base.h"
@@ -35,8 +37,12 @@ void RendererBase::RequestScreenshot(void* data, std::function<void(bool)> callb
LOG_ERROR(Render, "A screenshot is already requested or in progress, ignoring the request");
return;
}
+ auto async_callback{[callback = std::move(callback)](bool invert_y) {
+ std::thread t{callback, invert_y};
+ t.detach();
+ }};
renderer_settings.screenshot_bits = data;
- renderer_settings.screenshot_complete_callback = std::move(callback);
+ renderer_settings.screenshot_complete_callback = async_callback;
renderer_settings.screenshot_framebuffer_layout = layout;
renderer_settings.screenshot_requested = true;
}
diff --git a/src/video_core/renderer_opengl/gl_shader_cache.cpp b/src/video_core/renderer_opengl/gl_shader_cache.cpp
index 6bdb0b645..609f0a772 100644
--- a/src/video_core/renderer_opengl/gl_shader_cache.cpp
+++ b/src/video_core/renderer_opengl/gl_shader_cache.cpp
@@ -317,8 +317,8 @@ GraphicsPipeline* ShaderCache::CurrentGraphicsPipeline() {
graphics_key.tessellation_primitive.Assign(regs.tessellation.params.domain_type.Value());
graphics_key.tessellation_spacing.Assign(regs.tessellation.params.spacing.Value());
graphics_key.tessellation_clockwise.Assign(
- regs.tessellation.params.output_primitives.Value() !=
- Maxwell::Tessellation::OutputPrimitves::Triangles_CCW);
+ regs.tessellation.params.output_primitives.Value() ==
+ Maxwell::Tessellation::OutputPrimitives::Triangles_CW);
graphics_key.xfb_enabled.Assign(regs.transform_feedback_enabled != 0 ? 1 : 0);
if (graphics_key.xfb_enabled) {
SetXfbState(graphics_key.xfb_state, regs);
diff --git a/src/video_core/renderer_vulkan/fixed_pipeline_state.cpp b/src/video_core/renderer_vulkan/fixed_pipeline_state.cpp
index eab1b8f93..f85ed8e5b 100644
--- a/src/video_core/renderer_vulkan/fixed_pipeline_state.cpp
+++ b/src/video_core/renderer_vulkan/fixed_pipeline_state.cpp
@@ -73,8 +73,8 @@ void FixedPipelineState::Refresh(Tegra::Engines::Maxwell3D& maxwell3d,
patch_control_points_minus_one.Assign(regs.patch_vertices - 1);
tessellation_primitive.Assign(static_cast<u32>(regs.tessellation.params.domain_type.Value()));
tessellation_spacing.Assign(static_cast<u32>(regs.tessellation.params.spacing.Value()));
- tessellation_clockwise.Assign(regs.tessellation.params.output_primitives.Value() !=
- Maxwell::Tessellation::OutputPrimitves::Triangles_CCW);
+ tessellation_clockwise.Assign(regs.tessellation.params.output_primitives.Value() ==
+ Maxwell::Tessellation::OutputPrimitives::Triangles_CW);
logic_op_enable.Assign(regs.logic_op.enable != 0 ? 1 : 0);
logic_op.Assign(PackLogicOp(regs.logic_op.op));
topology.Assign(regs.draw.topology);
diff --git a/src/video_core/renderer_vulkan/vk_staging_buffer_pool.cpp b/src/video_core/renderer_vulkan/vk_staging_buffer_pool.cpp
index 7fb256953..06f68d09a 100644
--- a/src/video_core/renderer_vulkan/vk_staging_buffer_pool.cpp
+++ b/src/video_core/renderer_vulkan/vk_staging_buffer_pool.cpp
@@ -26,39 +26,20 @@ using namespace Common::Literals;
constexpr VkDeviceSize MAX_ALIGNMENT = 256;
// Maximum size to put elements in the stream buffer
constexpr VkDeviceSize MAX_STREAM_BUFFER_REQUEST_SIZE = 8_MiB;
+// Stream buffer size in bytes
+constexpr VkDeviceSize STREAM_BUFFER_SIZE = 128_MiB;
+constexpr VkDeviceSize REGION_SIZE = STREAM_BUFFER_SIZE / StagingBufferPool::NUM_SYNCS;
constexpr VkMemoryPropertyFlags HOST_FLAGS =
VK_MEMORY_PROPERTY_HOST_VISIBLE_BIT | VK_MEMORY_PROPERTY_HOST_COHERENT_BIT;
constexpr VkMemoryPropertyFlags STREAM_FLAGS = VK_MEMORY_PROPERTY_DEVICE_LOCAL_BIT | HOST_FLAGS;
-static bool IsStreamHeap(VkMemoryHeap heap, size_t staging_buffer_size) noexcept {
- return staging_buffer_size < (heap.size * 2) / 3;
-}
-
-static bool HasLargeDeviceLocalHostVisibleMemory(const VkPhysicalDeviceMemoryProperties& props) {
- const auto flags{VK_MEMORY_PROPERTY_DEVICE_LOCAL_BIT | VK_MEMORY_PROPERTY_HOST_VISIBLE_BIT};
-
- for (u32 type_index = 0; type_index < props.memoryTypeCount; ++type_index) {
- const auto& memory_type{props.memoryTypes[type_index]};
-
- if ((memory_type.propertyFlags & flags) != flags) {
- // Memory must be device local and host visible
- continue;
- }
-
- const auto& heap{props.memoryHeaps[memory_type.heapIndex]};
- if (heap.size >= 7168_MiB) {
- // This is the right type of memory
- return true;
- }
- }
-
- return false;
+bool IsStreamHeap(VkMemoryHeap heap) noexcept {
+ return STREAM_BUFFER_SIZE < (heap.size * 2) / 3;
}
std::optional<u32> FindMemoryTypeIndex(const VkPhysicalDeviceMemoryProperties& props, u32 type_mask,
- VkMemoryPropertyFlags flags,
- size_t staging_buffer_size) noexcept {
+ VkMemoryPropertyFlags flags) noexcept {
for (u32 type_index = 0; type_index < props.memoryTypeCount; ++type_index) {
if (((type_mask >> type_index) & 1) == 0) {
// Memory type is incompatible
@@ -69,7 +50,7 @@ std::optional<u32> FindMemoryTypeIndex(const VkPhysicalDeviceMemoryProperties& p
// Memory type doesn't have the flags we want
continue;
}
- if (!IsStreamHeap(props.memoryHeaps[memory_type.heapIndex], staging_buffer_size)) {
+ if (!IsStreamHeap(props.memoryHeaps[memory_type.heapIndex])) {
// Memory heap is not suitable for streaming
continue;
}
@@ -80,17 +61,17 @@ std::optional<u32> FindMemoryTypeIndex(const VkPhysicalDeviceMemoryProperties& p
}
u32 FindMemoryTypeIndex(const VkPhysicalDeviceMemoryProperties& props, u32 type_mask,
- bool try_device_local, size_t staging_buffer_size) {
+ bool try_device_local) {
std::optional<u32> type;
if (try_device_local) {
// Try to find a DEVICE_LOCAL_BIT type, Nvidia and AMD have a dedicated heap for this
- type = FindMemoryTypeIndex(props, type_mask, STREAM_FLAGS, staging_buffer_size);
+ type = FindMemoryTypeIndex(props, type_mask, STREAM_FLAGS);
if (type) {
return *type;
}
}
// Otherwise try without the DEVICE_LOCAL_BIT
- type = FindMemoryTypeIndex(props, type_mask, HOST_FLAGS, staging_buffer_size);
+ type = FindMemoryTypeIndex(props, type_mask, HOST_FLAGS);
if (type) {
return *type;
}
@@ -98,32 +79,20 @@ u32 FindMemoryTypeIndex(const VkPhysicalDeviceMemoryProperties& props, u32 type_
throw vk::Exception(VK_ERROR_OUT_OF_DEVICE_MEMORY);
}
-size_t Region(size_t iterator, size_t region_size) noexcept {
- return iterator / region_size;
+size_t Region(size_t iterator) noexcept {
+ return iterator / REGION_SIZE;
}
} // Anonymous namespace
StagingBufferPool::StagingBufferPool(const Device& device_, MemoryAllocator& memory_allocator_,
Scheduler& scheduler_)
: device{device_}, memory_allocator{memory_allocator_}, scheduler{scheduler_} {
-
- const auto memory_properties{device.GetPhysical().GetMemoryProperties().memoryProperties};
- if (HasLargeDeviceLocalHostVisibleMemory(memory_properties)) {
- // Possible on many integrated and newer discrete cards
- staging_buffer_size = 1_GiB;
- } else {
- // Well-supported default size used by most Vulkan PC games
- staging_buffer_size = 256_MiB;
- }
-
- region_size = staging_buffer_size / StagingBufferPool::NUM_SYNCS;
-
const vk::Device& dev = device.GetLogical();
stream_buffer = dev.CreateBuffer(VkBufferCreateInfo{
.sType = VK_STRUCTURE_TYPE_BUFFER_CREATE_INFO,
.pNext = nullptr,
.flags = 0,
- .size = staging_buffer_size,
+ .size = STREAM_BUFFER_SIZE,
.usage = VK_BUFFER_USAGE_TRANSFER_SRC_BIT | VK_BUFFER_USAGE_UNIFORM_BUFFER_BIT |
VK_BUFFER_USAGE_INDEX_BUFFER_BIT | VK_BUFFER_USAGE_STORAGE_BUFFER_BIT,
.sharingMode = VK_SHARING_MODE_EXCLUSIVE,
@@ -148,18 +117,19 @@ StagingBufferPool::StagingBufferPool(const Device& device_, MemoryAllocator& mem
.image = nullptr,
.buffer = *stream_buffer,
};
+ const auto memory_properties = device.GetPhysical().GetMemoryProperties().memoryProperties;
VkMemoryAllocateInfo stream_memory_info{
.sType = VK_STRUCTURE_TYPE_MEMORY_ALLOCATE_INFO,
.pNext = make_dedicated ? &dedicated_info : nullptr,
.allocationSize = requirements.size,
- .memoryTypeIndex = FindMemoryTypeIndex(memory_properties, requirements.memoryTypeBits, true,
- staging_buffer_size),
+ .memoryTypeIndex =
+ FindMemoryTypeIndex(memory_properties, requirements.memoryTypeBits, true),
};
stream_memory = dev.TryAllocateMemory(stream_memory_info);
if (!stream_memory) {
LOG_INFO(Render_Vulkan, "Dynamic memory allocation failed, trying with system memory");
- stream_memory_info.memoryTypeIndex = FindMemoryTypeIndex(
- memory_properties, requirements.memoryTypeBits, false, staging_buffer_size);
+ stream_memory_info.memoryTypeIndex =
+ FindMemoryTypeIndex(memory_properties, requirements.memoryTypeBits, false);
stream_memory = dev.AllocateMemory(stream_memory_info);
}
@@ -167,7 +137,7 @@ StagingBufferPool::StagingBufferPool(const Device& device_, MemoryAllocator& mem
stream_memory.SetObjectNameEXT("Stream Buffer Memory");
}
stream_buffer.BindMemory(*stream_memory, 0);
- stream_pointer = stream_memory.Map(0, staging_buffer_size);
+ stream_pointer = stream_memory.Map(0, STREAM_BUFFER_SIZE);
}
StagingBufferPool::~StagingBufferPool() = default;
@@ -188,25 +158,25 @@ void StagingBufferPool::TickFrame() {
}
StagingBufferRef StagingBufferPool::GetStreamBuffer(size_t size) {
- if (AreRegionsActive(Region(free_iterator, region_size) + 1,
- std::min(Region(iterator + size, region_size) + 1, NUM_SYNCS))) {
+ if (AreRegionsActive(Region(free_iterator) + 1,
+ std::min(Region(iterator + size) + 1, NUM_SYNCS))) {
// Avoid waiting for the previous usages to be free
return GetStagingBuffer(size, MemoryUsage::Upload);
}
const u64 current_tick = scheduler.CurrentTick();
- std::fill(sync_ticks.begin() + Region(used_iterator, region_size),
- sync_ticks.begin() + Region(iterator, region_size), current_tick);
+ std::fill(sync_ticks.begin() + Region(used_iterator), sync_ticks.begin() + Region(iterator),
+ current_tick);
used_iterator = iterator;
free_iterator = std::max(free_iterator, iterator + size);
- if (iterator + size >= staging_buffer_size) {
- std::fill(sync_ticks.begin() + Region(used_iterator, region_size),
- sync_ticks.begin() + NUM_SYNCS, current_tick);
+ if (iterator + size >= STREAM_BUFFER_SIZE) {
+ std::fill(sync_ticks.begin() + Region(used_iterator), sync_ticks.begin() + NUM_SYNCS,
+ current_tick);
used_iterator = 0;
iterator = 0;
free_iterator = size;
- if (AreRegionsActive(0, Region(size, region_size) + 1)) {
+ if (AreRegionsActive(0, Region(size) + 1)) {
// Avoid waiting for the previous usages to be free
return GetStagingBuffer(size, MemoryUsage::Upload);
}
diff --git a/src/video_core/renderer_vulkan/vk_staging_buffer_pool.h b/src/video_core/renderer_vulkan/vk_staging_buffer_pool.h
index 90c67177f..91dc84da8 100644
--- a/src/video_core/renderer_vulkan/vk_staging_buffer_pool.h
+++ b/src/video_core/renderer_vulkan/vk_staging_buffer_pool.h
@@ -93,9 +93,6 @@ private:
size_t free_iterator = 0;
std::array<u64, NUM_SYNCS> sync_ticks{};
- size_t staging_buffer_size = 0;
- size_t region_size = 0;
-
StagingBuffersCache device_local_cache;
StagingBuffersCache upload_cache;
StagingBuffersCache download_cache;
generated by cgit v1.2.3 (git 2.25.1) at 2025-12-27 11:07:31 +0000