Gamma pass on lsfg

2026-08-14 04:24:31 +00:00 · 2026-08-12 23:40:40 -04:00
parent 367e5084da
commit 232656df4e
3 changed files with 221 additions and 0 deletions
@@ -133,6 +133,8 @@ add_library(video_core STATIC
    renderer_vulkan/present/lsfg_beta.h
    renderer_vulkan/present/lsfg_common.cpp
    renderer_vulkan/present/lsfg_common.h
+    renderer_vulkan/present/lsfg_gamma.cpp
+    renderer_vulkan/present/lsfg_gamma.h
    renderer_vulkan/present/lsfg_mipmaps.cpp
    renderer_vulkan/present/lsfg_mipmaps.h
    renderer_vulkan/present/lsfg_shaders.cpp
@@ -0,0 +1,172 @@
+// SPDX-FileCopyrightText: Copyright 2026 Eden Emulator Project
+// SPDX-License-Identifier: GPL-3.0-or-later
+
+#include <vector>
+
+#include "video_core/frame_gen/lossless_dll.h"
+#include "video_core/renderer_vulkan/present/lsfg_gamma.h"
+#include "video_core/renderer_vulkan/present/lsfg_shaders.h"
+#include "video_core/renderer_vulkan/present/util.h"
+#include "video_core/vulkan_common/vulkan_device.h"
+
+namespace Vulkan {
+
+namespace {
+
+constexpr u32 DISPATCH_TILE_SHIFT = 3;
+
+[[nodiscard]] u32 GroupCount(u32 size) {
+    return (size + (1u << DISPATCH_TILE_SHIFT) - 1) >> DISPATCH_TILE_SHIFT;
+}
+
+} // Anonymous namespace
+
+LsfgGamma::LsfgGamma(const Device& device, MemoryAllocator& memory_allocator,
+                     const LsfgShaders& shaders, LsfgResources& resources,
+                     vk::DescriptorPool& descriptor_pool, LsfgImageHistory& inputs_,
+                     LsfgImage& flow_input_, LsfgImage* previous_)
+    : inputs{&inputs_}, flow_input{&flow_input_}, previous{previous_} {
+    using namespace VideoCore::FrameGen::PerformanceShader;
+
+    passes[0] = LsfgPass(device, shaders, GAMMA[0],
+                         {{1, VK_DESCRIPTOR_TYPE_UNIFORM_BUFFER},
+                          {2, VK_DESCRIPTOR_TYPE_SAMPLER},
+                          {5, VK_DESCRIPTOR_TYPE_SAMPLED_IMAGE},
+                          {3, VK_DESCRIPTOR_TYPE_STORAGE_IMAGE}});
+    passes[1] = LsfgPass(device, shaders, GAMMA[1],
+                         {{1, VK_DESCRIPTOR_TYPE_SAMPLER},
+                          {3, VK_DESCRIPTOR_TYPE_SAMPLED_IMAGE},
+                          {2, VK_DESCRIPTOR_TYPE_STORAGE_IMAGE}});
+    passes[2] = LsfgPass(device, shaders, GAMMA[2],
+                         {{1, VK_DESCRIPTOR_TYPE_SAMPLER},
+                          {2, VK_DESCRIPTOR_TYPE_SAMPLED_IMAGE},
+                          {2, VK_DESCRIPTOR_TYPE_STORAGE_IMAGE}});
+    passes[3] = LsfgPass(device, shaders, GAMMA[3],
+                         {{1, VK_DESCRIPTOR_TYPE_SAMPLER},
+                          {2, VK_DESCRIPTOR_TYPE_SAMPLED_IMAGE},
+                          {2, VK_DESCRIPTOR_TYPE_STORAGE_IMAGE}});
+    passes[4] = LsfgPass(device, shaders, GAMMA[4],
+                         {{1, VK_DESCRIPTOR_TYPE_UNIFORM_BUFFER},
+                          {2, VK_DESCRIPTOR_TYPE_SAMPLER},
+                          {4, VK_DESCRIPTOR_TYPE_SAMPLED_IMAGE},
+                          {1, VK_DESCRIPTOR_TYPE_STORAGE_IMAGE}});
+
+    const VkExtent2D extent = (*inputs)[0][0].Extent();
+    for (auto& image : temp1) {
+        image = LsfgImage(device, memory_allocator, extent);
+    }
+    for (auto& image : temp2) {
+        image = LsfgImage(device, memory_allocator, extent);
+    }
+    out_image = LsfgImage(device, memory_allocator, extent, LSFG_MOTION_FORMAT);
+
+    std::vector<VkDescriptorSetLayout> layouts;
+    for (size_t i = 0; i < LSFG_HISTORY_SLOTS; ++i) {
+        layouts.push_back(passes[0].SetLayout());
+    }
+    for (size_t i = 1; i < LSFG_GAMMA_STAGES; ++i) {
+        layouts.push_back(passes[i].SetLayout());
+    }
+    owned_sets = CreateWrappedDescriptorSets(descriptor_pool, layouts);
+
+    for (size_t i = 0; i < LSFG_HISTORY_SLOTS; ++i) {
+        first_descriptor_sets[i] = owned_sets[i];
+    }
+    for (size_t i = 0; i < LSFG_GAMMA_STAGES - 1; ++i) {
+        descriptor_sets[i] = owned_sets[LSFG_HISTORY_SLOTS + i];
+    }
+
+    const VkSampler sampler = resources.GetSampler();
+    const VkSampler border_sampler = resources.GetSampler(
+        VK_SAMPLER_ADDRESS_MODE_CLAMP_TO_BORDER, VK_COMPARE_OP_NEVER, true);
+    const VkSampler edge_sampler =
+        resources.GetSampler(VK_SAMPLER_ADDRESS_MODE_CLAMP_TO_EDGE, VK_COMPARE_OP_ALWAYS, false);
+    const VkBuffer buffer = resources.GetBuffer(LSFG_TIMESTAMP, previous == nullptr);
+
+    for (size_t i = 0; i < LSFG_HISTORY_SLOTS; ++i) {
+        LsfgDescriptorWriter(first_descriptor_sets[i])
+            .AddUniformBuffer(buffer, LsfgResources::BufferSize())
+            .AddSampler(border_sampler)
+            .AddSampler(edge_sampler)
+            .AddSampledImages((*inputs)[(i + 2) % LSFG_HISTORY_SLOTS])
+            .AddSampledImages((*inputs)[i % LSFG_HISTORY_SLOTS])
+            .AddSampledImage(previous)
+            .AddStorageImages(temp1)
+            .Build(device);
+    }
+    LsfgDescriptorWriter(descriptor_sets[0])
+        .AddSampler(sampler)
+        .AddSampledImages(temp1)
+        .AddStorageImages(temp2)
+        .Build(device);
+    LsfgDescriptorWriter(descriptor_sets[1])
+        .AddSampler(sampler)
+        .AddSampledImages(temp2)
+        .AddStorageImage(temp1[0])
+        .AddStorageImage(temp1[1])
+        .Build(device);
+    LsfgDescriptorWriter(descriptor_sets[2])
+        .AddSampler(sampler)
+        .AddSampledImage(temp1[0])
+        .AddSampledImage(temp1[1])
+        .AddStorageImages(temp2)
+        .Build(device);
+    LsfgDescriptorWriter(descriptor_sets[3])
+        .AddUniformBuffer(buffer, LsfgResources::BufferSize())
+        .AddSampler(sampler)
+        .AddSampler(edge_sampler)
+        .AddSampledImages(temp2)
+        .AddSampledImage(previous)
+        .AddSampledImage(*flow_input)
+        .AddStorageImage(out_image)
+        .Build(device);
+}
+
+void LsfgGamma::Dispatch(vk::CommandBuffer cmdbuf, u64 frame_count) {
+    const VkExtent2D extent = temp1[0].Extent();
+    const u32 groups_x = GroupCount(extent.width);
+    const u32 groups_y = GroupCount(extent.height);
+
+    const size_t slot = frame_count % LSFG_HISTORY_SLOTS;
+    const size_t previous_slot = (frame_count + 2) % LSFG_HISTORY_SLOTS;
+
+    LsfgBarriers(cmdbuf)
+        .WriteToReadAll((*inputs)[previous_slot])
+        .WriteToReadAll((*inputs)[slot])
+        .WriteToRead(previous)
+        .ReadToWriteAll(temp1)
+        .Build();
+    passes[0].Bind(cmdbuf, first_descriptor_sets[slot]);
+    cmdbuf.Dispatch(groups_x, groups_y, 1);
+
+    LsfgBarriers(cmdbuf).WriteToReadAll(temp1).ReadToWriteAll(temp2).Build();
+    passes[1].Bind(cmdbuf, descriptor_sets[0]);
+    cmdbuf.Dispatch(groups_x, groups_y, 1);
+
+    LsfgBarriers(cmdbuf)
+        .WriteToReadAll(temp2)
+        .ReadToWrite(temp1[0])
+        .ReadToWrite(temp1[1])
+        .Build();
+    passes[2].Bind(cmdbuf, descriptor_sets[1]);
+    cmdbuf.Dispatch(groups_x, groups_y, 1);
+
+    LsfgBarriers(cmdbuf)
+        .WriteToRead(temp1[0])
+        .WriteToRead(temp1[1])
+        .ReadToWriteAll(temp2)
+        .Build();
+    passes[3].Bind(cmdbuf, descriptor_sets[2]);
+    cmdbuf.Dispatch(groups_x, groups_y, 1);
+
+    LsfgBarriers(cmdbuf)
+        .WriteToReadAll(temp2)
+        .WriteToRead(previous)
+        .WriteToRead(*flow_input)
+        .ReadToWrite(out_image)
+        .Build();
+    passes[4].Bind(cmdbuf, descriptor_sets[3]);
+    cmdbuf.Dispatch(groups_x, groups_y, 1);
+}
+
+} // namespace Vulkan
@@ -0,0 +1,47 @@
+// SPDX-FileCopyrightText: Copyright 2026 Eden Emulator Project
+// SPDX-License-Identifier: GPL-3.0-or-later
+
+#pragma once
+
+#include <array>
+
+#include "common/common_types.h"
+#include "video_core/renderer_vulkan/present/lsfg_common.h"
+
+namespace Vulkan {
+
+class Device;
+class LsfgShaders;
+
+constexpr size_t LSFG_GAMMA_STAGES = 5;
+constexpr size_t LSFG_GAMMA_TEMPS = 3;
+
+class LsfgGamma {
+public:
+    LsfgGamma() = default;
+    LsfgGamma(const Device& device, MemoryAllocator& memory_allocator, const LsfgShaders& shaders,
+              LsfgResources& resources, vk::DescriptorPool& descriptor_pool,
+              LsfgImageHistory& inputs, LsfgImage& flow_input, LsfgImage* previous);
+
+    void Dispatch(vk::CommandBuffer cmdbuf, u64 frame_count);
+
+    [[nodiscard]] LsfgImage& Output() {
+        return out_image;
+    }
+
+private:
+    LsfgImageHistory* inputs{};
+    LsfgImage* flow_input{};
+    LsfgImage* previous{};
+
+    std::array<LsfgPass, LSFG_GAMMA_STAGES> passes;
+    std::array<VkDescriptorSet, LSFG_HISTORY_SLOTS> first_descriptor_sets{};
+    std::array<VkDescriptorSet, LSFG_GAMMA_STAGES - 1> descriptor_sets{};
+    vk::DescriptorSets owned_sets;
+
+    std::array<LsfgImage, LSFG_GAMMA_TEMPS> temp1;
+    LsfgImagePair temp2;
+    LsfgImage out_image;
+};
+
+} // namespace Vulkan