Merge remote-tracking branch 'origin/nx2' into neo.ai/vibration-fix

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

# Conflicts:
#	src/core/hle/service/hid/hid_server.cpp
#	src/core/hle/service/hid/hid_server.h

1 files changed, 55 insertions, 6 deletions

diff --git a/src/video_core/shader_cache.cpp b/src/video_core/shader_cache.cpp
index dfe80abac3..5d41dc9697 100644
--- a/src/video_core/shader_cache.cpp
+++ b/src/video_core/shader_cache.cpp
@@ -6,6 +6,7 @@
 
 #include <algorithm>
 #include <cstring>
+#include <functional>
 #include <string>
 #include <unordered_set>
 #include <fmt/ranges.h>
@@ -14,6 +15,7 @@
 #include <boost/container/small_vector.hpp>
 
 #include "common/assert.h"
+#include "common/cityhash.h"
 #include "shader_recompiler/frontend/maxwell/control_flow.h"
 #include "shader_recompiler/exception.h"
 #include "shader_recompiler/object_pool.h"
@@ -35,6 +37,17 @@ void ShaderCache::InvalidateRegion(VAddr addr, size_t size) {
     RemovePendingShaders();
 }
 
+void ShaderCache::InvalidateRegions(std::span<const std::pair<DAddr, size_t>> sequences) {
+    if (sequences.empty()) {
+        return;
+    }
+    std::scoped_lock lock{invalidation_mutex};
+    for (const auto& [addr, size] : sequences) {
+        InvalidatePagesInRegion(addr, size);
+    }
+    RemovePendingShaders();
+}
+
 void ShaderCache::OnCacheInvalidation(VAddr addr, size_t size) {
     std::scoped_lock lock{invalidation_mutex};
     InvalidatePagesInRegion(addr, size);
@@ -48,6 +61,22 @@ void ShaderCache::SyncGuestHost() {
 ShaderCache::ShaderCache(Tegra::MaxwellDeviceMemoryManager& device_memory_)
     : device_memory{device_memory_} {}
 
+// Task and mesh programs are translated with the mesh registers and the header of the pass-through
+// geometry program (per-primitive outputs), so pipelines are keyed by them too.
+u64 ShaderCache::MeshStateHash() const {
+    using Regs = Tegra::Engines::Maxwell3D::Regs;
+    const auto& regs{maxwell3d->regs};
+    std::array<u32, 3 + 20> state{regs.reg_array[Regs::MeshShaderA], regs.reg_array[Regs::MeshShaderB],
+                                  regs.reg_array[Regs::MeshInitShader]};
+    if (regs.IsShaderConfigEnabled(Regs::ShaderType::Geometry)) {
+        const auto& geometry{regs.pipelines[static_cast<size_t>(Regs::ShaderType::Geometry)]};
+        const GPUVAddr address{(static_cast<u64>(geometry.reserved[0] & 0xff) << 32) |
+                               geometry.reserved[1]};
+        gpu_memory->ReadBlockUnsafe(address, state.data() + 3, 20 * sizeof(u32));
+    }
+    return Common::CityHash64(reinterpret_cast<const char*>(state.data()), sizeof(state)) | 1;
+}
+
 bool ShaderCache::RefreshStages(std::array<u64, 6>& unique_hashes) try {
     auto& dirty{maxwell3d->dirty.flags};
     if (!dirty[VideoCommon::Dirty::Shaders]) {
@@ -57,6 +86,7 @@ bool ShaderCache::RefreshStages(std::array<u64, 6>& unique_hashes) try {
 
     const GPUVAddr base_addr{maxwell3d->regs.program_region.Address()};
     const bool native = maxwell3d->GetNativeMacroEngine().native_turing;
+    const bool mesh{native && maxwell3d->regs.IsMeshEnabled()};
     for (size_t index = 0; index < Tegra::Engines::Maxwell3D::Regs::MaxShaderProgram; ++index) {
         if (!maxwell3d->regs.IsShaderConfigEnabled(index)) {
             unique_hashes[index] = 0;
@@ -69,6 +99,11 @@ bool ShaderCache::RefreshStages(std::array<u64, 6>& unique_hashes) try {
             unique_hashes[index] = 0;
             continue;
         }
+        if (mesh && program == Tegra::Engines::Maxwell3D::Regs::ShaderType::Geometry) {
+            // A header without code: the mesh program's environment reads its output map.
+            unique_hashes[index] = 0;
+            continue;
+        }
         const GPUVAddr shader_addr{native
             ? (static_cast<u64>(shader_config.reserved[0] & 0xff) << 32) | shader_config.reserved[1]
             : base_addr + shader_config.offset};
@@ -89,7 +124,8 @@ bool ShaderCache::RefreshStages(std::array<u64, 6>& unique_hashes) try {
         if (!shader_info) {
             const u32 start_address{native ? 0 : shader_config.offset};
             try {
-                // Native programs are lowered while the environment is built.
+                // Native programs are lowered while the environment is built (once: the
+                // ShaderInfo keeps the result for the pipeline's environments).
                 RethrowIfRejected(*cpu_shader_addr);
                 GraphicsEnvironment env{*maxwell3d, *gpu_memory, program, native ? shader_addr : base_addr, start_address};
                 shader_info = MakeShaderInfo(env, *cpu_shader_addr);
@@ -111,6 +147,10 @@ bool ShaderCache::RefreshStages(std::array<u64, 6>& unique_hashes) try {
         }
         shader_infos[index] = shader_info;
         unique_hashes[index] = shader_info->unique_hash;
+        if (mesh && (program == Tegra::Engines::Maxwell3D::Regs::ShaderType::VertexB ||
+                     program == Tegra::Engines::Maxwell3D::Regs::ShaderType::Tessellation)) {
+            unique_hashes[index] ^= MeshStateHash();
+        }
     }
     last_shaders_valid = true;
     return true;
@@ -202,7 +242,8 @@ void ShaderCache::GetGraphicsEnvironments(GraphicsEnvironments& result,
         const GPUVAddr address = native
             ? (static_cast<u64>(config.reserved[0] & 0xff) << 32) | config.reserved[1]
             : base_addr;
-        env = GraphicsEnvironment{*maxwell3d, *gpu_memory, program, address, start_address};
+        env = GraphicsEnvironment{*maxwell3d, *gpu_memory, program, address, start_address,
+                                  shader_infos[index]->lowered_native};
         env.SetCachedSize(shader_infos[index]->size_bytes);
         result.env_ptrs[env_index++] = &env;
     }
@@ -275,10 +316,17 @@ void ShaderCache::RemovePendingShaders() {
     marked_for_removal.clear();
 
     if (!removed_shaders.empty()) {
-        // Remove the given shaders from the cache
-        std::erase_if(storage, [&removed_shaders](const std::unique_ptr<ShaderInfo>& shader) {
-            return std::ranges::find(removed_shaders, shader.get()) != removed_shaders.end();
-        });
+        if (removed_shaders.size() > 16) {
+            std::ranges::sort(removed_shaders, std::less<ShaderInfo*>{});
+            std::erase_if(storage, [&removed_shaders](const std::unique_ptr<ShaderInfo>& shader) {
+                return std::binary_search(removed_shaders.begin(), removed_shaders.end(),
+                                          shader.get(), std::less<ShaderInfo*>{});
+            });
+        } else {
+            std::erase_if(storage, [&removed_shaders](const std::unique_ptr<ShaderInfo>& shader) {
+                return std::ranges::find(removed_shaders, shader.get()) != removed_shaders.end();
+            });
+        }
     }
 }
 
@@ -342,6 +390,7 @@ const ShaderInfo* ShaderCache::MakeShaderInfo(GenericEnvironment& env, VAddr cpu
         info->unique_hash = env.CalculateHash();
         info->size_bytes = env.ReadSizeBytes();
     }
+    info->lowered_native = env.LoweredNative();
     const size_t size_bytes{env.SourceSizeBytes() ? env.SourceSizeBytes() : info->size_bytes};
     const ShaderInfo* const result{info.get()};
     Register(std::move(info), cpu_addr, size_bytes);