early-access version 2647

This commit is contained in:
pineappleEA 2022-04-06 01:05:42 +02:00
parent 016426eebf
commit 614f3e6a00
11 changed files with 61 additions and 71 deletions

View file

@ -1,7 +1,7 @@
yuzu emulator early access
=============
This is the source code for early-access 2645.
This is the source code for early-access 2647.
## Legal Notice

View file

@ -103,12 +103,6 @@ else()
-Wno-unused-parameter
)
# TODO: Remove when we update to a GCC compiler that enables this
# by default (i.e. GCC 10 or newer).
if (CMAKE_CXX_COMPILER_ID STREQUAL GNU)
add_compile_options(-fconcepts)
endif()
if (ARCHITECTURE_x86_64)
add_compile_options("-mcx16")
endif()

View file

@ -73,11 +73,13 @@ public:
}
void InterpreterFallback(u32 pc, std::size_t num_instructions) override {
parent.LogBacktrace();
UNIMPLEMENTED_MSG("This should never happen, pc = {:08X}, code = {:08X}", pc,
MemoryReadCode(pc));
}
void ExceptionRaised(u32 pc, Dynarmic::A32::Exception exception) override {
parent.LogBacktrace();
LOG_CRITICAL(Core_ARM,
"ExceptionRaised(exception = {}, pc = {:08X}, code = {:08X}, thumb = {})",
exception, pc, MemoryReadCode(pc), parent.IsInThumbMode());

View file

@ -84,6 +84,7 @@ public:
}
void InterpreterFallback(u64 pc, std::size_t num_instructions) override {
parent.LogBacktrace();
LOG_ERROR(Core_ARM,
"Unimplemented instruction @ 0x{:X} for {} instructions (instr = {:08X})", pc,
num_instructions, MemoryReadCode(pc));
@ -121,6 +122,7 @@ public:
return;
case Dynarmic::A64::Exception::Breakpoint:
default:
parent.LogBacktrace();
ASSERT_MSG(false, "ExceptionRaised(exception = {}, pc = {:08X}, code = {:08X})",
static_cast<std::size_t>(exception), pc, MemoryReadCode(pc));
}

View file

@ -200,7 +200,7 @@ void ComputePipeline::Configure(Tegra::Engines::KeplerCompute& kepler_compute,
});
}
const void* const descriptor_data{update_descriptor_queue.UpdateData()};
const bool is_rescaling = info.uses_rescaling_uniform;
const bool is_rescaling = !info.texture_descriptors.empty() || !info.image_descriptors.empty();
scheduler.Record([this, descriptor_data, is_rescaling,
rescaling_data = rescaling.Data()](vk::CommandBuffer cmdbuf) {
cmdbuf.BindPipeline(VK_PIPELINE_BIND_POINT_COMPUTE, *pipeline);

View file

@ -238,7 +238,6 @@ GraphicsPipeline::GraphicsPipeline(
enabled_uniform_buffer_masks[stage] = info->constant_buffer_mask;
std::ranges::copy(info->constant_buffer_used_sizes, uniform_buffer_sizes[stage].begin());
num_textures += Shader::NumDescriptors(info->texture_descriptors);
uses_rescale_unfiorm |= info->uses_rescaling_uniform;
}
auto func{[this, shader_notify, &render_pass_cache, &descriptor_pool, pipeline_statistics] {
DescriptorLayoutBuilder builder{MakeBuilder(device, stage_infos)};
@ -472,8 +471,7 @@ void GraphicsPipeline::ConfigureDraw(const RescalingPushConstant& rescaling) {
});
}
const bool is_rescaling{texture_cache.IsRescaling()};
const bool update_rescaling{uses_rescale_unfiorm ? scheduler.UpdateRescaling(is_rescaling)
: false};
const bool update_rescaling{scheduler.UpdateRescaling(is_rescaling)};
const bool bind_pipeline{scheduler.UpdateGraphicsPipeline(this)};
const void* const descriptor_data{update_descriptor_queue.UpdateData()};
scheduler.Record([this, descriptor_data, bind_pipeline, rescaling_data = rescaling.Data(),
@ -481,12 +479,10 @@ void GraphicsPipeline::ConfigureDraw(const RescalingPushConstant& rescaling) {
if (bind_pipeline) {
cmdbuf.BindPipeline(VK_PIPELINE_BIND_POINT_GRAPHICS, *pipeline);
}
if (uses_rescale_unfiorm) {
cmdbuf.PushConstants(*pipeline_layout, VK_SHADER_STAGE_ALL_GRAPHICS,
RESCALING_LAYOUT_WORDS_OFFSET, sizeof(rescaling_data),
rescaling_data.data());
}
if (uses_rescale_unfiorm && update_rescaling) {
if (update_rescaling) {
const f32 config_down_factor{Settings::values.resolution_info.down_factor};
const f32 scale_down_factor{is_rescaling ? config_down_factor : 1.0f};
cmdbuf.PushConstants(*pipeline_layout, VK_SHADER_STAGE_ALL_GRAPHICS,

View file

@ -151,7 +151,6 @@ private:
std::mutex build_mutex;
std::atomic_bool is_built{false};
bool uses_push_descriptor{false};
bool uses_rescale_unfiorm{false};
};
} // namespace Vulkan

View file

@ -234,9 +234,12 @@ void RasterizerVulkan::Clear() {
const VkExtent2D render_area = framebuffer->RenderArea();
scheduler.RequestRenderpass(framebuffer);
const bool is_rescaling = texture_cache.IsRescaling();
const u32 up_scale = is_rescaling ? Settings::values.resolution_info.up_scale : 1U;
const u32 down_shift = is_rescaling ? Settings::values.resolution_info.down_shift : 0U;
u32 up_scale = 1;
u32 down_shift = 0;
if (texture_cache.IsRescaling()) {
up_scale = Settings::values.resolution_info.up_scale;
down_shift = Settings::values.resolution_info.down_shift;
}
UpdateViewportsState(regs);
VkClearRect clear_rect{
@ -692,9 +695,12 @@ void RasterizerVulkan::UpdateScissorsState(Tegra::Engines::Maxwell3D::Regs& regs
if (!state_tracker.TouchScissors()) {
return;
}
const bool is_rescaling = texture_cache.IsRescaling();
const u32 up_scale = is_rescaling ? Settings::values.resolution_info.up_scale : 1U;
const u32 down_shift = is_rescaling ? Settings::values.resolution_info.down_shift : 0U;
u32 up_scale = 1;
u32 down_shift = 0;
if (texture_cache.IsRescaling()) {
up_scale = Settings::values.resolution_info.up_scale;
down_shift = Settings::values.resolution_info.down_shift;
}
const std::array scissors{
GetScissorState(regs, 0, up_scale, down_shift),
GetScissorState(regs, 1, up_scale, down_shift),

View file

@ -29,10 +29,7 @@ constexpr VkDeviceSize MAX_ALIGNMENT = 256;
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_STREAM_REGIONS;
static_assert(Common::IsAligned(REGION_SIZE, MAX_ALIGNMENT),
"Stream buffer region size must be VK buffer aligned");
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;
@ -86,17 +83,6 @@ u32 FindMemoryTypeIndex(const VkPhysicalDeviceMemoryProperties& props, u32 type_
size_t Region(size_t iterator) noexcept {
return iterator / REGION_SIZE;
}
constexpr std::array<VkDeviceSize, StagingBufferPool::NUM_STREAM_REGIONS> MakeStreamBufferOffset() {
std::array<VkDeviceSize, StagingBufferPool::NUM_STREAM_REGIONS> offsets{};
for (size_t i = 0; i < StagingBufferPool::NUM_STREAM_REGIONS; ++i) {
offsets[i] = static_cast<VkDeviceSize>(i * REGION_SIZE);
}
return offsets;
}
constexpr auto STREAM_BUFFER_OFFSETS_LUT = MakeStreamBufferOffset();
} // Anonymous namespace
StagingBufferPool::StagingBufferPool(const Device& device_, MemoryAllocator& memory_allocator_,
@ -173,19 +159,31 @@ void StagingBufferPool::TickFrame() {
}
StagingBufferRef StagingBufferPool::GetStreamBuffer(size_t size) {
const auto num_requested_regions = Region(size) + 1;
const auto available_index = NextAvailableStreamIndex(num_requested_regions);
if (!available_index) {
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();
const auto begin_itr = sync_ticks.begin() + *available_index;
std::fill(begin_itr, begin_itr + num_requested_regions, 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);
const VkDeviceSize offset = STREAM_BUFFER_OFFSETS_LUT[*available_index];
ASSERT(offset + size <= STREAM_BUFFER_SIZE);
next_index = (*available_index + num_requested_regions) % NUM_STREAM_REGIONS;
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) + 1)) {
// Avoid waiting for the previous usages to be free
return GetStagingBuffer(size, MemoryUsage::Upload);
}
}
const size_t offset = iterator;
iterator = Common::AlignUp(iterator + size, MAX_ALIGNMENT);
return StagingBufferRef{
.buffer = *stream_buffer,
.offset = static_cast<VkDeviceSize>(offset),
@ -193,22 +191,10 @@ StagingBufferRef StagingBufferPool::GetStreamBuffer(size_t size) {
};
}
std::optional<size_t> StagingBufferPool::NextAvailableStreamIndex(size_t num_regions) const {
const auto is_index_available = [this, num_regions](size_t begin_offset) {
bool StagingBufferPool::AreRegionsActive(size_t region_begin, size_t region_end) const {
const u64 gpu_tick = scheduler.GetMasterSemaphore().KnownGpuTick();
const auto tick_check = [gpu_tick](u64 sync_tick) { return gpu_tick >= sync_tick; };
const auto begin_itr = sync_ticks.begin() + begin_offset;
const bool is_available = std::all_of(begin_itr, begin_itr + num_regions, tick_check);
return is_available ? std::optional(begin_offset) : std::nullopt;
};
// Avoid overflow
if (next_index + num_regions <= NUM_STREAM_REGIONS) {
return is_index_available(next_index);
}
// Not enough contiguous regions at the next_index,
// Check if the contiguous range in the front is available
return is_index_available(0);
return std::any_of(sync_ticks.begin() + region_begin, sync_ticks.begin() + region_end,
[gpu_tick](u64 sync_tick) { return gpu_tick < sync_tick; });
};
StagingBufferRef StagingBufferPool::GetStagingBuffer(size_t size, MemoryUsage usage) {

View file

@ -25,7 +25,7 @@ struct StagingBufferRef {
class StagingBufferPool {
public:
static constexpr size_t NUM_STREAM_REGIONS = 0x8000;
static constexpr size_t NUM_SYNCS = 16;
explicit StagingBufferPool(const Device& device, MemoryAllocator& memory_allocator,
VKScheduler& scheduler);
@ -67,7 +67,7 @@ private:
StagingBufferRef GetStreamBuffer(size_t size);
std::optional<size_t> NextAvailableStreamIndex(size_t num_regions) const;
bool AreRegionsActive(size_t region_begin, size_t region_end) const;
StagingBufferRef GetStagingBuffer(size_t size, MemoryUsage usage);
@ -89,8 +89,10 @@ private:
vk::DeviceMemory stream_memory;
u8* stream_pointer = nullptr;
size_t next_index = 0;
std::array<u64, NUM_STREAM_REGIONS> sync_ticks{};
size_t iterator = 0;
size_t used_iterator = 0;
size_t free_iterator = 0;
std::array<u64, NUM_SYNCS> sync_ticks{};
StagingBuffersCache device_local_cache;
StagingBuffersCache upload_cache;

View file

@ -328,8 +328,7 @@ void TextureCache<P>::UpdateRenderTargets(bool is_clear) {
}
const bool rescaled = RescaleRenderTargets(is_clear);
const auto& resolution_info = Settings::values.resolution_info;
if (resolution_info.active && is_rescaling != rescaled) {
if (is_rescaling != rescaled) {
flags[Dirty::RescaleViewports] = true;
flags[Dirty::RescaleScissors] = true;
is_rescaling = rescaled;
@ -346,8 +345,12 @@ void TextureCache<P>::UpdateRenderTargets(bool is_clear) {
for (size_t index = 0; index < NUM_RT; ++index) {
render_targets.draw_buffers[index] = static_cast<u8>(maxwell3d.regs.rt_control.Map(index));
}
const u32 up_scale = is_rescaling ? resolution_info.up_scale : 1U;
const u32 down_shift = is_rescaling ? resolution_info.down_shift : 0U;
u32 up_scale = 1;
u32 down_shift = 0;
if (is_rescaling) {
up_scale = Settings::values.resolution_info.up_scale;
down_shift = Settings::values.resolution_info.down_shift;
}
render_targets.size = Extent2D{
(maxwell3d.regs.render_area.width * up_scale) >> down_shift,
(maxwell3d.regs.render_area.height * up_scale) >> down_shift,