Android #166

README.md

@@ -1,3 +1,15 @@
+| Pull Request | Commit | Title | Author | Merged? |
+|----|----|----|----|----|
+| [12390](https://github.com/yuzu-emu/yuzu//pull/12390) | [`fba3fa705`](https://github.com/yuzu-emu/yuzu//pull/12390/files) | renderer_vulkan: work around turnip binding bug in a610 | [liamwhite](https://github.com/liamwhite/) | Yes |
+| [12400](https://github.com/yuzu-emu/yuzu//pull/12400) | [`a2b567dfd`](https://github.com/yuzu-emu/yuzu//pull/12400/files) | vk_query_cache: Fix prefix sum max_accumulation_limit logic | [ameerj](https://github.com/ameerj/) | Yes |
+| [12403](https://github.com/yuzu-emu/yuzu//pull/12403) | [`fcfa8b680`](https://github.com/yuzu-emu/yuzu//pull/12403/files) |  shader_recompiler: use minimal clip distance array  | [liamwhite](https://github.com/liamwhite/) | Yes |
+| [12409](https://github.com/yuzu-emu/yuzu//pull/12409) | [`93c19a40b`](https://github.com/yuzu-emu/yuzu//pull/12409/files) | nce: fix read size in simd immediate emulation | [liamwhite](https://github.com/liamwhite/) | Yes |
+
+
+End of merge log. You can find the original README.md below the break.
+
+-----
+
 <!--
 SPDX-FileCopyrightText: 2018 yuzu Emulator Project
 SPDX-License-Identifier: GPL-2.0-or-later

src/core/arm/nce/arm_nce.cpp

@@ -39,7 +39,7 @@ fpsimd_context* GetFloatingPointState(mcontext_t& host_ctx) {
 }
 
 using namespace Common::Literals;
-constexpr u32 StackSize = 32_KiB;
+constexpr u32 StackSize = 128_KiB;
 
 } // namespace
 

src/core/arm/nce/interpreter_visitor.cpp

@@ -249,6 +249,7 @@ bool InterpreterVisitor::LDR_lit_fpsimd(Imm<2> opc, Imm<19> imm19, Vec Vt) {
         return false;
     }
 
+    // Size in bytes
     const u64 size = 4 << opc.ZeroExtend();
     const u64 offset = imm19.SignExtend<u64>() << 2;
     const u64 address = this->GetPc() + offset;
@@ -530,7 +531,7 @@ bool InterpreterVisitor::SIMDImmediate(bool wback, bool postindex, size_t scale,
     }
     case MemOp::Load: {
         u128 data{};
-        m_memory.ReadBlock(address, &data, datasize);
+        m_memory.ReadBlock(address, &data, datasize / 8);
         this->SetVec(Vt, data);
         break;
     }

src/shader_recompiler/backend/spirv/emit_spirv_context_get_set.cpp

@@ -74,6 +74,11 @@ std::optional<OutAttr> OutputAttrPointer(EmitContext& ctx, IR::Attribute attr) {
     case IR::Attribute::ClipDistance7: {
         const u32 base{static_cast<u32>(IR::Attribute::ClipDistance0)};
         const u32 index{static_cast<u32>(attr) - base};
+        if (index >= ctx.profile.max_user_clip_distances) {
+            LOG_WARNING(Shader, "Ignoring clip distance store {} >= {} supported", index,
+                        ctx.profile.max_user_clip_distances);
+            return std::nullopt;
+        }
         const Id clip_num{ctx.Const(index)};
         return OutputAccessChain(ctx, ctx.output_f32, ctx.clip_distances, clip_num);
     }

src/shader_recompiler/backend/spirv/spirv_emit_context.cpp

@@ -1528,7 +1528,8 @@ void EmitContext::DefineOutputs(const IR::Program& program) {
         if (stage == Stage::Fragment) {
             throw NotImplementedException("Storing ClipDistance in fragment stage");
         }
-        const Id type{TypeArray(F32[1], Const(8U))};
+        const Id type{TypeArray(
+            F32[1], Const(std::min(info.used_clip_distances, profile.max_user_clip_distances)))};
         clip_distances = DefineOutput(*this, type, invocations, spv::BuiltIn::ClipDistance);
     }
     if (info.stores[IR::Attribute::Layer] &&

src/shader_recompiler/ir_opt/collect_shader_info_pass.cpp

@@ -913,7 +913,11 @@ void GatherInfoFromHeader(Environment& env, Info& info) {
         }
         for (size_t index = 0; index < 8; ++index) {
             const u16 mask{header.vtg.omap_systemc.clip_distances};
-            info.stores.Set(IR::Attribute::ClipDistance0 + index, ((mask >> index) & 1) != 0);
+            const bool used{((mask >> index) & 1) != 0};
+            info.stores.Set(IR::Attribute::ClipDistance0 + index, used);
+            if (used) {
+                info.used_clip_distances = static_cast<u32>(index) + 1;
+            }
         }
         info.stores.Set(IR::Attribute::PrimitiveId,
                         header.vtg.omap_systemb.primitive_array_id != 0);

src/shader_recompiler/profile.h

@@ -87,6 +87,8 @@ struct Profile {
     bool has_broken_robust{};
 
     u64 min_ssbo_alignment{};
+
+    u32 max_user_clip_distances{};
 };
 
 } // namespace Shader

src/shader_recompiler/shader_info.h

@@ -227,6 +227,8 @@ struct Info {
     bool requires_layer_emulation{};
     IR::Attribute emulated_layer{};
 
+    u32 used_clip_distances{};
+
     boost::container::static_vector<ConstantBufferDescriptor, MAX_CBUFS>
         constant_buffer_descriptors;
     boost::container::static_vector<StorageBufferDescriptor, MAX_SSBOS> storage_buffers_descriptors;

src/video_core/renderer_opengl/gl_shader_cache.cpp

@@ -233,6 +233,7 @@ ShaderCache::ShaderCache(RasterizerOpenGL& rasterizer_, Core::Frontend::EmuWindo
           .ignore_nan_fp_comparisons = true,
           .gl_max_compute_smem_size = device.GetMaxComputeSharedMemorySize(),
           .min_ssbo_alignment = device.GetShaderStorageBufferAlignment(),
+          .max_user_clip_distances = 8,
       },
       host_info{
           .support_float64 = true,

src/video_core/renderer_vulkan/vk_buffer_cache.cpp

@@ -563,22 +563,27 @@ void BufferCacheRuntime::BindVertexBuffers(VideoCommon::HostBindings<Buffer>& bi
         }
         buffer_handles.push_back(handle);
     }
+    const u32 device_max = device.GetMaxVertexInputBindings();
+    const u32 min_binding = std::min(bindings.min_index, device_max);
+    const u32 max_binding = std::min(bindings.max_index, device_max);
+    const u32 binding_count = max_binding - min_binding;
+    if (binding_count == 0) {
+        return;
+    }
     if (device.IsExtExtendedDynamicStateSupported()) {
-        scheduler.Record([this, bindings_ = std::move(bindings),
+        scheduler.Record([bindings_ = std::move(bindings),
-                          buffer_handles_ = std::move(buffer_handles)](vk::CommandBuffer cmdbuf) {
+                          buffer_handles_ = std::move(buffer_handles),
-            cmdbuf.BindVertexBuffers2EXT(bindings_.min_index,
+                          binding_count](vk::CommandBuffer cmdbuf) {
-                                         std::min(bindings_.max_index - bindings_.min_index,
+            cmdbuf.BindVertexBuffers2EXT(bindings_.min_index, binding_count, buffer_handles_.data(),
-                                                  device.GetMaxVertexInputBindings()),
+                                         bindings_.offsets.data(), bindings_.sizes.data(),
-                                         buffer_handles_.data(), bindings_.offsets.data(),
+                                         bindings_.strides.data());
-                                         bindings_.sizes.data(), bindings_.strides.data());
         });
     } else {
-        scheduler.Record([this, bindings_ = std::move(bindings),
+        scheduler.Record([bindings_ = std::move(bindings),
-                          buffer_handles_ = std::move(buffer_handles)](vk::CommandBuffer cmdbuf) {
+                          buffer_handles_ = std::move(buffer_handles),
-            cmdbuf.BindVertexBuffers(bindings_.min_index,
+                          binding_count](vk::CommandBuffer cmdbuf) {
-                                     std::min(bindings_.max_index - bindings_.min_index,
+            cmdbuf.BindVertexBuffers(bindings_.min_index, binding_count, buffer_handles_.data(),
-                                              device.GetMaxVertexInputBindings()),
+                                     bindings_.offsets.data());
-                                     buffer_handles_.data(), bindings_.offsets.data());
         });
     }
 }

src/video_core/renderer_vulkan/vk_pipeline_cache.cpp

@@ -374,6 +374,7 @@ PipelineCache::PipelineCache(RasterizerVulkan& rasterizer_, const Device& device
         .has_broken_robust =
             device.IsNvidia() && device.GetNvidiaArch() <= NvidiaArchitecture::Arch_Pascal,
         .min_ssbo_alignment = device.GetStorageBufferAlignment(),
+        .max_user_clip_distances = device.GetMaxUserClipDistances(),
     };
 
     host_info = Shader::HostTranslateInfo{

src/video_core/renderer_vulkan/vk_query_cache.cpp

@@ -289,12 +289,15 @@ public:
         }
 
         if (has_multi_queries) {
-            size_t intermediary_buffer_index = ObtainBuffer<false>(num_slots_used);
+            const size_t min_accumulation_limit =
+                std::min(first_accumulation_checkpoint, num_slots_used);
+            const size_t max_accumulation_limit =
+                std::max(last_accumulation_checkpoint, num_slots_used);
+            const size_t intermediary_buffer_index = ObtainBuffer<false>(num_slots_used);
             resolve_buffers.push_back(intermediary_buffer_index);
             queries_prefix_scan_pass->Run(*accumulation_buffer, *buffers[intermediary_buffer_index],
                                           *buffers[resolve_buffer_index], num_slots_used,
-                                          std::min(first_accumulation_checkpoint, num_slots_used),
+                                          min_accumulation_limit, max_accumulation_limit);
-                                          last_accumulation_checkpoint);
 
         } else {
             scheduler.RequestOutsideRenderPassOperationContext();

src/video_core/vulkan_common/vulkan_device.cpp

@@ -695,6 +695,11 @@ Device::Device(VkInstance instance_, vk::PhysicalDevice physical_, VkSurfaceKHR
             std::min(properties.properties.limits.maxVertexInputBindings, 16U);
     }
 
+    if (is_turnip) {
+        LOG_WARNING(Render_Vulkan, "Turnip requires higher-than-reported binding limits");
+        properties.properties.limits.maxVertexInputBindings = 32;
+    }
+
     if (!extensions.extended_dynamic_state && extensions.extended_dynamic_state2) {
         LOG_INFO(Render_Vulkan,
                  "Removing extendedDynamicState2 due to missing extendedDynamicState");

src/video_core/vulkan_common/vulkan_device.h

@@ -665,6 +665,10 @@ public:
         return properties.properties.limits.maxViewports;
     }
 
+    u32 GetMaxUserClipDistances() const {
+        return properties.properties.limits.maxClipDistances;
+    }
+
     bool SupportsConditionalBarriers() const {
         return supports_conditional_barriers;
     }