summaryrefslogtreecommitdiffstats
path: root/src/video_core
diff options
context:
space:
mode:
Diffstat (limited to 'src/video_core')
-rw-r--r--src/video_core/renderer_opengl/maxwell_to_gl.h1
-rw-r--r--src/video_core/renderer_vulkan/blit_image.cpp4
-rw-r--r--src/video_core/renderer_vulkan/maxwell_to_vk.cpp5
-rw-r--r--src/video_core/renderer_vulkan/vk_blit_screen.cpp1
-rw-r--r--src/video_core/renderer_vulkan/vk_rasterizer.cpp3
-rw-r--r--src/video_core/renderer_vulkan/vk_staging_buffer_pool.cpp33
-rw-r--r--src/video_core/renderer_vulkan/vk_staging_buffer_pool.h5
-rw-r--r--src/video_core/renderer_vulkan/vk_texture_cache.cpp1
-rw-r--r--src/video_core/surface.cpp3
-rw-r--r--src/video_core/surface.h4
-rw-r--r--src/video_core/texture_cache/format_lookup_table.cpp4
-rw-r--r--src/video_core/texture_cache/formatter.h2
-rw-r--r--src/video_core/texture_cache/image_view_base.cpp1
-rw-r--r--src/video_core/vulkan_common/vulkan_device.cpp56
-rw-r--r--src/video_core/vulkan_common/vulkan_memory_allocator.cpp21
-rw-r--r--src/video_core/vulkan_common/vulkan_memory_allocator.h14
16 files changed, 128 insertions, 30 deletions
diff --git a/src/video_core/renderer_opengl/maxwell_to_gl.h b/src/video_core/renderer_opengl/maxwell_to_gl.h
index c7dc7e0a1..5ea9e2378 100644
--- a/src/video_core/renderer_opengl/maxwell_to_gl.h
+++ b/src/video_core/renderer_opengl/maxwell_to_gl.h
@@ -116,6 +116,7 @@ constexpr std::array<FormatTuple, VideoCore::Surface::MaxPixelFormat> FORMAT_TAB
{GL_RGB9_E5, GL_RGB, GL_UNSIGNED_INT_5_9_9_9_REV}, // E5B9G9R9_FLOAT
{GL_DEPTH_COMPONENT32F, GL_DEPTH_COMPONENT, GL_FLOAT}, // D32_FLOAT
{GL_DEPTH_COMPONENT16, GL_DEPTH_COMPONENT, GL_UNSIGNED_SHORT}, // D16_UNORM
+ {GL_DEPTH_COMPONENT24, GL_DEPTH_COMPONENT, GL_UNSIGNED_INT_24_8}, // X8_D24_UNORM
{GL_STENCIL_INDEX8, GL_STENCIL, GL_UNSIGNED_BYTE}, // S8_UINT
{GL_DEPTH24_STENCIL8, GL_DEPTH_STENCIL, GL_UNSIGNED_INT_24_8}, // D24_UNORM_S8_UINT
{GL_DEPTH24_STENCIL8, GL_DEPTH_STENCIL, GL_UNSIGNED_INT_24_8}, // S8_UINT_D24_UNORM
diff --git a/src/video_core/renderer_vulkan/blit_image.cpp b/src/video_core/renderer_vulkan/blit_image.cpp
index 4778ac4c5..f01d2394e 100644
--- a/src/video_core/renderer_vulkan/blit_image.cpp
+++ b/src/video_core/renderer_vulkan/blit_image.cpp
@@ -618,6 +618,8 @@ void BlitImageHelper::ClearDepthStencil(const Framebuffer* dst_framebuffer, bool
const VkPipelineLayout layout = *clear_color_pipeline_layout;
scheduler.RequestRenderpass(dst_framebuffer);
scheduler.Record([pipeline, layout, clear_depth, dst_region](vk::CommandBuffer cmdbuf) {
+ constexpr std::array blend_constants{0.0f, 0.0f, 0.0f, 0.0f};
+ cmdbuf.SetBlendConstants(blend_constants.data());
cmdbuf.BindPipeline(VK_PIPELINE_BIND_POINT_GRAPHICS, pipeline);
BindBlitState(cmdbuf, dst_region);
cmdbuf.PushConstants(layout, VK_SHADER_STAGE_FRAGMENT_BIT, clear_depth);
@@ -874,7 +876,7 @@ VkPipeline BlitImageHelper::FindOrEmplaceClearStencilPipeline(
.sType = VK_STRUCTURE_TYPE_PIPELINE_DEPTH_STENCIL_STATE_CREATE_INFO,
.pNext = nullptr,
.flags = 0,
- .depthTestEnable = VK_FALSE,
+ .depthTestEnable = key.depth_clear,
.depthWriteEnable = key.depth_clear,
.depthCompareOp = VK_COMPARE_OP_ALWAYS,
.depthBoundsTestEnable = VK_FALSE,
diff --git a/src/video_core/renderer_vulkan/maxwell_to_vk.cpp b/src/video_core/renderer_vulkan/maxwell_to_vk.cpp
index 208e88533..a08f2f67f 100644
--- a/src/video_core/renderer_vulkan/maxwell_to_vk.cpp
+++ b/src/video_core/renderer_vulkan/maxwell_to_vk.cpp
@@ -214,8 +214,9 @@ struct FormatTuple {
{VK_FORMAT_E5B9G9R9_UFLOAT_PACK32}, // E5B9G9R9_FLOAT
// Depth formats
- {VK_FORMAT_D32_SFLOAT, Attachable}, // D32_FLOAT
- {VK_FORMAT_D16_UNORM, Attachable}, // D16_UNORM
+ {VK_FORMAT_D32_SFLOAT, Attachable}, // D32_FLOAT
+ {VK_FORMAT_D16_UNORM, Attachable}, // D16_UNORM
+ {VK_FORMAT_X8_D24_UNORM_PACK32, Attachable}, // X8_D24_UNORM
// Stencil formats
{VK_FORMAT_S8_UINT, Attachable}, // S8_UINT
diff --git a/src/video_core/renderer_vulkan/vk_blit_screen.cpp b/src/video_core/renderer_vulkan/vk_blit_screen.cpp
index 31928bb94..52fc142d1 100644
--- a/src/video_core/renderer_vulkan/vk_blit_screen.cpp
+++ b/src/video_core/renderer_vulkan/vk_blit_screen.cpp
@@ -96,6 +96,7 @@ std::size_t GetSizeInBytes(const Tegra::FramebufferConfig& framebuffer) {
VkFormat GetFormat(const Tegra::FramebufferConfig& framebuffer) {
switch (framebuffer.pixel_format) {
case Service::android::PixelFormat::Rgba8888:
+ case Service::android::PixelFormat::Rgbx8888:
return VK_FORMAT_A8B8G8R8_UNORM_PACK32;
case Service::android::PixelFormat::Rgb565:
return VK_FORMAT_R5G6B5_UNORM_PACK16;
diff --git a/src/video_core/renderer_vulkan/vk_rasterizer.cpp b/src/video_core/renderer_vulkan/vk_rasterizer.cpp
index 1628d76d6..83f2b6045 100644
--- a/src/video_core/renderer_vulkan/vk_rasterizer.cpp
+++ b/src/video_core/renderer_vulkan/vk_rasterizer.cpp
@@ -422,7 +422,8 @@ void RasterizerVulkan::Clear(u32 layer_count) {
return;
}
- if (use_stencil && regs.stencil_front_mask != 0xFF && regs.stencil_front_mask != 0) {
+ if (use_stencil && framebuffer->HasAspectStencilBit() && regs.stencil_front_mask != 0xFF &&
+ regs.stencil_front_mask != 0) {
Region2D dst_region = {
Offset2D{.x = clear_rect.rect.offset.x, .y = clear_rect.rect.offset.y},
Offset2D{.x = clear_rect.rect.offset.x + static_cast<s32>(clear_rect.rect.extent.width),
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 ce92f66ab..b278614e6 100644
--- a/src/video_core/renderer_vulkan/vk_staging_buffer_pool.cpp
+++ b/src/video_core/renderer_vulkan/vk_staging_buffer_pool.cpp
@@ -24,25 +24,38 @@ using namespace Common::Literals;
// Maximum potential alignment of a Vulkan buffer
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 VkDeviceSize MAX_STREAM_BUFFER_SIZE = 128_MiB;
-size_t Region(size_t iterator) noexcept {
- return iterator / REGION_SIZE;
+size_t GetStreamBufferSize(const Device& device) {
+ VkDeviceSize size{0};
+ if (device.HasDebuggingToolAttached()) {
+ ForEachDeviceLocalHostVisibleHeap(device, [&size](size_t index, VkMemoryHeap& heap) {
+ size = std::max(size, heap.size);
+ });
+ // If rebar is not supported, cut the max heap size to 40%. This will allow 2 captures to be
+ // loaded at the same time in RenderDoc. If rebar is supported, this shouldn't be an issue
+ // as the heap will be much larger.
+ if (size <= 256_MiB) {
+ size = size * 40 / 100;
+ }
+ } else {
+ size = MAX_STREAM_BUFFER_SIZE;
+ }
+ return std::min(Common::AlignUp(size, MAX_ALIGNMENT), MAX_STREAM_BUFFER_SIZE);
}
} // Anonymous namespace
StagingBufferPool::StagingBufferPool(const Device& device_, MemoryAllocator& memory_allocator_,
Scheduler& scheduler_)
- : device{device_}, memory_allocator{memory_allocator_}, scheduler{scheduler_} {
+ : device{device_}, memory_allocator{memory_allocator_}, scheduler{scheduler_},
+ stream_buffer_size{GetStreamBufferSize(device)}, region_size{stream_buffer_size /
+ StagingBufferPool::NUM_SYNCS} {
VkBufferCreateInfo stream_ci = {
.sType = VK_STRUCTURE_TYPE_BUFFER_CREATE_INFO,
.pNext = nullptr,
.flags = 0,
- .size = STREAM_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,
@@ -63,7 +76,7 @@ StagingBufferPool::StagingBufferPool(const Device& device_, MemoryAllocator& mem
StagingBufferPool::~StagingBufferPool() = default;
StagingBufferRef StagingBufferPool::Request(size_t size, MemoryUsage usage, bool deferred) {
- if (!deferred && usage == MemoryUsage::Upload && size <= MAX_STREAM_BUFFER_REQUEST_SIZE) {
+ if (!deferred && usage == MemoryUsage::Upload && size <= region_size) {
return GetStreamBuffer(size);
}
return GetStagingBuffer(size, usage, deferred);
@@ -101,7 +114,7 @@ StagingBufferRef StagingBufferPool::GetStreamBuffer(size_t size) {
used_iterator = iterator;
free_iterator = std::max(free_iterator, iterator + size);
- if (iterator + size >= STREAM_BUFFER_SIZE) {
+ if (iterator + size >= stream_buffer_size) {
std::fill(sync_ticks.begin() + Region(used_iterator), sync_ticks.begin() + NUM_SYNCS,
current_tick);
used_iterator = 0;
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 5f69f08b1..d3deb9072 100644
--- a/src/video_core/renderer_vulkan/vk_staging_buffer_pool.h
+++ b/src/video_core/renderer_vulkan/vk_staging_buffer_pool.h
@@ -90,6 +90,9 @@ private:
void ReleaseCache(MemoryUsage usage);
void ReleaseLevel(StagingBuffersCache& cache, size_t log2);
+ size_t Region(size_t iter) const noexcept {
+ return iter / region_size;
+ }
const Device& device;
MemoryAllocator& memory_allocator;
@@ -97,6 +100,8 @@ private:
vk::Buffer stream_buffer;
std::span<u8> stream_pointer;
+ VkDeviceSize stream_buffer_size;
+ VkDeviceSize region_size;
size_t iterator = 0;
size_t used_iterator = 0;
diff --git a/src/video_core/renderer_vulkan/vk_texture_cache.cpp b/src/video_core/renderer_vulkan/vk_texture_cache.cpp
index f02d3e8b8..00ab47268 100644
--- a/src/video_core/renderer_vulkan/vk_texture_cache.cpp
+++ b/src/video_core/renderer_vulkan/vk_texture_cache.cpp
@@ -238,6 +238,7 @@ constexpr VkBorderColor ConvertBorderColor(const std::array<float, 4>& color) {
return any_r ? VK_IMAGE_ASPECT_STENCIL_BIT : VK_IMAGE_ASPECT_DEPTH_BIT;
case PixelFormat::D16_UNORM:
case PixelFormat::D32_FLOAT:
+ case PixelFormat::X8_D24_UNORM:
return VK_IMAGE_ASPECT_DEPTH_BIT;
case PixelFormat::S8_UINT:
return VK_IMAGE_ASPECT_STENCIL_BIT;
diff --git a/src/video_core/surface.cpp b/src/video_core/surface.cpp
index e16cd5e73..5b3c7aa5a 100644
--- a/src/video_core/surface.cpp
+++ b/src/video_core/surface.cpp
@@ -85,6 +85,8 @@ PixelFormat PixelFormatFromDepthFormat(Tegra::DepthFormat format) {
return PixelFormat::S8_UINT;
case Tegra::DepthFormat::Z32_FLOAT_X24S8_UINT:
return PixelFormat::D32_FLOAT_S8_UINT;
+ case Tegra::DepthFormat::X8Z24_UNORM:
+ return PixelFormat::X8_D24_UNORM;
default:
UNIMPLEMENTED_MSG("Unimplemented format={}", format);
return PixelFormat::S8_UINT_D24_UNORM;
@@ -202,6 +204,7 @@ PixelFormat PixelFormatFromRenderTargetFormat(Tegra::RenderTargetFormat format)
PixelFormat PixelFormatFromGPUPixelFormat(Service::android::PixelFormat format) {
switch (format) {
case Service::android::PixelFormat::Rgba8888:
+ case Service::android::PixelFormat::Rgbx8888:
return PixelFormat::A8B8G8R8_UNORM;
case Service::android::PixelFormat::Rgb565:
return PixelFormat::R5G6B5_UNORM;
diff --git a/src/video_core/surface.h b/src/video_core/surface.h
index 9b9c4d9bc..a5e8e2f62 100644
--- a/src/video_core/surface.h
+++ b/src/video_core/surface.h
@@ -115,6 +115,7 @@ enum class PixelFormat {
// Depth formats
D32_FLOAT = MaxColorFormat,
D16_UNORM,
+ X8_D24_UNORM,
MaxDepthFormat,
@@ -251,6 +252,7 @@ constexpr std::array<u8, MaxPixelFormat> BLOCK_WIDTH_TABLE = {{
1, // E5B9G9R9_FLOAT
1, // D32_FLOAT
1, // D16_UNORM
+ 1, // X8_D24_UNORM
1, // S8_UINT
1, // D24_UNORM_S8_UINT
1, // S8_UINT_D24_UNORM
@@ -360,6 +362,7 @@ constexpr std::array<u8, MaxPixelFormat> BLOCK_HEIGHT_TABLE = {{
1, // E5B9G9R9_FLOAT
1, // D32_FLOAT
1, // D16_UNORM
+ 1, // X8_D24_UNORM
1, // S8_UINT
1, // D24_UNORM_S8_UINT
1, // S8_UINT_D24_UNORM
@@ -469,6 +472,7 @@ constexpr std::array<u8, MaxPixelFormat> BITS_PER_BLOCK_TABLE = {{
32, // E5B9G9R9_FLOAT
32, // D32_FLOAT
16, // D16_UNORM
+ 32, // X8_D24_UNORM
8, // S8_UINT
32, // D24_UNORM_S8_UINT
32, // S8_UINT_D24_UNORM
diff --git a/src/video_core/texture_cache/format_lookup_table.cpp b/src/video_core/texture_cache/format_lookup_table.cpp
index 56307d030..3162c8f5e 100644
--- a/src/video_core/texture_cache/format_lookup_table.cpp
+++ b/src/video_core/texture_cache/format_lookup_table.cpp
@@ -142,6 +142,10 @@ PixelFormat PixelFormatFromTextureInfo(TextureFormat format, ComponentType red,
return PixelFormat::D16_UNORM;
case Hash(TextureFormat::Z16, UNORM, UINT, UINT, UINT, LINEAR):
return PixelFormat::D16_UNORM;
+ case Hash(TextureFormat::X8Z24, UNORM):
+ return PixelFormat::X8_D24_UNORM;
+ case Hash(TextureFormat::X8Z24, UNORM, UINT, UINT, UINT, LINEAR):
+ return PixelFormat::X8_D24_UNORM;
case Hash(TextureFormat::Z24S8, UINT, UNORM, UNORM, UNORM, LINEAR):
return PixelFormat::S8_UINT_D24_UNORM;
case Hash(TextureFormat::Z24S8, UINT, UNORM, UINT, UINT, LINEAR):
diff --git a/src/video_core/texture_cache/formatter.h b/src/video_core/texture_cache/formatter.h
index 9ee57a076..cabbfcb2d 100644
--- a/src/video_core/texture_cache/formatter.h
+++ b/src/video_core/texture_cache/formatter.h
@@ -211,6 +211,8 @@ struct fmt::formatter<VideoCore::Surface::PixelFormat> : fmt::formatter<fmt::str
return "D32_FLOAT";
case PixelFormat::D16_UNORM:
return "D16_UNORM";
+ case PixelFormat::X8_D24_UNORM:
+ return "X8_D24_UNORM";
case PixelFormat::S8_UINT:
return "S8_UINT";
case PixelFormat::D24_UNORM_S8_UINT:
diff --git a/src/video_core/texture_cache/image_view_base.cpp b/src/video_core/texture_cache/image_view_base.cpp
index 0c5f4450d..18b9250f9 100644
--- a/src/video_core/texture_cache/image_view_base.cpp
+++ b/src/video_core/texture_cache/image_view_base.cpp
@@ -85,6 +85,7 @@ bool ImageViewBase::SupportsAnisotropy() const noexcept {
// Depth formats
case PixelFormat::D32_FLOAT:
case PixelFormat::D16_UNORM:
+ case PixelFormat::X8_D24_UNORM:
// Stencil formats
case PixelFormat::S8_UINT:
// DepthStencil formats
diff --git a/src/video_core/vulkan_common/vulkan_device.cpp b/src/video_core/vulkan_common/vulkan_device.cpp
index 3960b135a..876cec2e8 100644
--- a/src/video_core/vulkan_common/vulkan_device.cpp
+++ b/src/video_core/vulkan_common/vulkan_device.cpp
@@ -84,9 +84,12 @@ constexpr std::array VK_FORMAT_A4B4G4R4_UNORM_PACK16{
} // namespace Alternatives
enum class NvidiaArchitecture {
- AmpereOrNewer,
+ KeplerOrOlder,
+ Maxwell,
+ Pascal,
+ Volta,
Turing,
- VoltaOrOlder,
+ AmpereOrNewer,
};
template <typename T>
@@ -200,6 +203,7 @@ std::unordered_map<VkFormat, VkFormatProperties> GetFormatProperties(vk::Physica
VK_FORMAT_BC7_UNORM_BLOCK,
VK_FORMAT_D16_UNORM,
VK_FORMAT_D16_UNORM_S8_UINT,
+ VK_FORMAT_X8_D24_UNORM_PACK32,
VK_FORMAT_D24_UNORM_S8_UINT,
VK_FORMAT_D32_SFLOAT,
VK_FORMAT_D32_SFLOAT_S8_UINT,
@@ -321,13 +325,38 @@ NvidiaArchitecture GetNvidiaArchitecture(vk::PhysicalDevice physical,
physical.GetProperties2(physical_properties);
if (shading_rate_props.primitiveFragmentShadingRateWithMultipleViewports) {
// Only Ampere and newer support this feature
+ // TODO: Find a way to differentiate Ampere and Ada
return NvidiaArchitecture::AmpereOrNewer;
}
- }
- if (exts.contains(VK_NV_SHADING_RATE_IMAGE_EXTENSION_NAME)) {
return NvidiaArchitecture::Turing;
}
- return NvidiaArchitecture::VoltaOrOlder;
+
+ if (exts.contains(VK_EXT_BLEND_OPERATION_ADVANCED_EXTENSION_NAME)) {
+ VkPhysicalDeviceBlendOperationAdvancedPropertiesEXT advanced_blending_props{};
+ advanced_blending_props.sType =
+ VK_STRUCTURE_TYPE_PHYSICAL_DEVICE_BLEND_OPERATION_ADVANCED_PROPERTIES_EXT;
+ VkPhysicalDeviceProperties2 physical_properties{};
+ physical_properties.sType = VK_STRUCTURE_TYPE_PHYSICAL_DEVICE_PROPERTIES_2;
+ physical_properties.pNext = &advanced_blending_props;
+ physical.GetProperties2(physical_properties);
+ if (advanced_blending_props.advancedBlendMaxColorAttachments == 1) {
+ return NvidiaArchitecture::Maxwell;
+ }
+
+ if (exts.contains(VK_EXT_CONSERVATIVE_RASTERIZATION_EXTENSION_NAME)) {
+ VkPhysicalDeviceConservativeRasterizationPropertiesEXT conservative_raster_props{};
+ conservative_raster_props.sType =
+ VK_STRUCTURE_TYPE_PHYSICAL_DEVICE_CONSERVATIVE_RASTERIZATION_PROPERTIES_EXT;
+ physical_properties.pNext = &conservative_raster_props;
+ physical.GetProperties2(physical_properties);
+ if (conservative_raster_props.degenerateLinesRasterized) {
+ return NvidiaArchitecture::Volta;
+ }
+ return NvidiaArchitecture::Pascal;
+ }
+ }
+
+ return NvidiaArchitecture::KeplerOrOlder;
}
std::vector<const char*> ExtensionListForVulkan(
@@ -504,19 +533,14 @@ Device::Device(VkInstance instance_, vk::PhysicalDevice physical_, VkSurfaceKHR
if (is_nvidia) {
const u32 nv_major_version = (properties.properties.driverVersion >> 22) & 0x3ff;
const auto arch = GetNvidiaArchitecture(physical, supported_extensions);
- switch (arch) {
- case NvidiaArchitecture::AmpereOrNewer:
+ if (arch >= NvidiaArchitecture::AmpereOrNewer) {
LOG_WARNING(Render_Vulkan, "Ampere and newer have broken float16 math");
features.shader_float16_int8.shaderFloat16 = false;
- break;
- case NvidiaArchitecture::Turing:
- break;
- case NvidiaArchitecture::VoltaOrOlder:
+ } else if (arch <= NvidiaArchitecture::Volta) {
if (nv_major_version < 527) {
LOG_WARNING(Render_Vulkan, "Volta and older have broken VK_KHR_push_descriptor");
RemoveExtension(extensions.push_descriptor, VK_KHR_PUSH_DESCRIPTOR_EXTENSION_NAME);
}
- break;
}
if (nv_major_version >= 510) {
LOG_WARNING(Render_Vulkan, "NVIDIA Drivers >= 510 do not support MSAA image blits");
@@ -661,7 +685,15 @@ Device::Device(VkInstance instance_, vk::PhysicalDevice physical_, VkSurfaceKHR
"ANV drivers 22.3.0 to 23.1.0 have broken VK_KHR_push_descriptor");
RemoveExtension(extensions.push_descriptor, VK_KHR_PUSH_DESCRIPTOR_EXTENSION_NAME);
}
+ } else if (extensions.push_descriptor && is_nvidia) {
+ const auto arch = GetNvidiaArchitecture(physical, supported_extensions);
+ if (arch <= NvidiaArchitecture::Pascal) {
+ LOG_WARNING(Render_Vulkan,
+ "Pascal and older architectures have broken VK_KHR_push_descriptor");
+ RemoveExtension(extensions.push_descriptor, VK_KHR_PUSH_DESCRIPTOR_EXTENSION_NAME);
+ }
}
+
if (is_mvk) {
LOG_WARNING(Render_Vulkan,
"MVK driver breaks when using more than 16 vertex attributes/bindings");
diff --git a/src/video_core/vulkan_common/vulkan_memory_allocator.cpp b/src/video_core/vulkan_common/vulkan_memory_allocator.cpp
index 3ef381a38..82767fdf0 100644
--- a/src/video_core/vulkan_common/vulkan_memory_allocator.cpp
+++ b/src/video_core/vulkan_common/vulkan_memory_allocator.cpp
@@ -9,6 +9,7 @@
#include "common/alignment.h"
#include "common/assert.h"
#include "common/common_types.h"
+#include "common/literals.h"
#include "common/logging/log.h"
#include "common/polyfill_ranges.h"
#include "video_core/vulkan_common/vma.h"
@@ -69,8 +70,7 @@ struct Range {
case MemoryUsage::Download:
return VMA_ALLOCATION_CREATE_HOST_ACCESS_RANDOM_BIT;
case MemoryUsage::DeviceLocal:
- return VMA_ALLOCATION_CREATE_HOST_ACCESS_SEQUENTIAL_WRITE_BIT |
- VMA_ALLOCATION_CREATE_HOST_ACCESS_ALLOW_TRANSFER_INSTEAD_BIT;
+ return {};
}
return {};
}
@@ -212,7 +212,20 @@ MemoryAllocator::MemoryAllocator(const Device& device_)
: device{device_}, allocator{device.GetAllocator()},
properties{device_.GetPhysical().GetMemoryProperties().memoryProperties},
buffer_image_granularity{
- device_.GetPhysical().GetProperties().limits.bufferImageGranularity} {}
+ device_.GetPhysical().GetProperties().limits.bufferImageGranularity} {
+ // GPUs not supporting rebar may only have a region with less than 256MB host visible/device
+ // local memory. In that case, opening 2 RenderDoc captures side-by-side is not possible due to
+ // the heap running out of memory. With RenderDoc attached and only a small host/device region,
+ // only allow the stream buffer in this memory heap.
+ if (device.HasDebuggingToolAttached()) {
+ using namespace Common::Literals;
+ ForEachDeviceLocalHostVisibleHeap(device, [this](size_t index, VkMemoryHeap& heap) {
+ if (heap.size <= 256_MiB) {
+ valid_memory_types &= ~(1u << index);
+ }
+ });
+ }
+}
MemoryAllocator::~MemoryAllocator() = default;
@@ -244,7 +257,7 @@ vk::Buffer MemoryAllocator::CreateBuffer(const VkBufferCreateInfo& ci, MemoryUsa
.usage = MemoryUsageVma(usage),
.requiredFlags = 0,
.preferredFlags = MemoryUsagePreferedVmaFlags(usage),
- .memoryTypeBits = 0,
+ .memoryTypeBits = usage == MemoryUsage::Stream ? 0u : valid_memory_types,
.pool = VK_NULL_HANDLE,
.pUserData = nullptr,
.priority = 0.f,
diff --git a/src/video_core/vulkan_common/vulkan_memory_allocator.h b/src/video_core/vulkan_common/vulkan_memory_allocator.h
index f449bc8d0..38a182bcb 100644
--- a/src/video_core/vulkan_common/vulkan_memory_allocator.h
+++ b/src/video_core/vulkan_common/vulkan_memory_allocator.h
@@ -7,6 +7,7 @@
#include <span>
#include <vector>
#include "common/common_types.h"
+#include "video_core/vulkan_common/vulkan_device.h"
#include "video_core/vulkan_common/vulkan_wrapper.h"
VK_DEFINE_HANDLE(VmaAllocator)
@@ -26,6 +27,18 @@ enum class MemoryUsage {
Stream, ///< Requests device local host visible buffer, falling back host memory.
};
+template <typename F>
+void ForEachDeviceLocalHostVisibleHeap(const Device& device, F&& f) {
+ auto memory_props = device.GetPhysical().GetMemoryProperties().memoryProperties;
+ for (size_t i = 0; i < memory_props.memoryTypeCount; i++) {
+ auto& memory_type = memory_props.memoryTypes[i];
+ if ((memory_type.propertyFlags & VK_MEMORY_PROPERTY_DEVICE_LOCAL_BIT) &&
+ (memory_type.propertyFlags & VK_MEMORY_PROPERTY_HOST_VISIBLE_BIT)) {
+ f(memory_type.heapIndex, memory_props.memoryHeaps[memory_type.heapIndex]);
+ }
+ }
+}
+
/// Ownership handle of a memory commitment.
/// Points to a subregion of a memory allocation.
class MemoryCommit {
@@ -124,6 +137,7 @@ private:
std::vector<std::unique_ptr<MemoryAllocation>> allocations; ///< Current allocations.
VkDeviceSize buffer_image_granularity; // The granularity for adjacent offsets between buffers
// and optimal images
+ u32 valid_memory_types{~0u};
};
} // namespace Vulkan
generated by cgit v1.2.3 (git 2.25.1) at 2024-11-21 23:46:43 +0100