diff options
Diffstat (limited to 'src/video_core')
-rw-r--r-- | src/video_core/command_classes/codecs/codec.h | 1 | ||||
-rw-r--r-- | src/video_core/engines/maxwell_3d.h | 7 | ||||
-rw-r--r-- | src/video_core/gpu.cpp | 29 | ||||
-rw-r--r-- | src/video_core/renderer_vulkan/fixed_pipeline_state.cpp | 7 | ||||
-rw-r--r-- | src/video_core/renderer_vulkan/fixed_pipeline_state.h | 8 | ||||
-rw-r--r-- | src/video_core/renderer_vulkan/vk_pipeline_cache.cpp | 1 | ||||
-rw-r--r-- | src/video_core/renderer_vulkan/vk_shader_decompiler.cpp | 6 | ||||
-rw-r--r-- | src/video_core/renderer_vulkan/vk_shader_decompiler.h | 1 |
8 files changed, 38 insertions, 22 deletions
diff --git a/src/video_core/command_classes/codecs/codec.h b/src/video_core/command_classes/codecs/codec.h index 5bbe6a332..ee5d62540 100644 --- a/src/video_core/command_classes/codecs/codec.h +++ b/src/video_core/command_classes/codecs/codec.h @@ -10,6 +10,7 @@ extern "C" { #if defined(__GNUC__) || defined(__clang__) +#pragma GCC diagnostic push #pragma GCC diagnostic ignored "-Wconversion" #endif #include <libavcodec/avcodec.h> diff --git a/src/video_core/engines/maxwell_3d.h b/src/video_core/engines/maxwell_3d.h index 1cbe8fe67..b0d9559d0 100644 --- a/src/video_core/engines/maxwell_3d.h +++ b/src/video_core/engines/maxwell_3d.h @@ -755,7 +755,11 @@ public: u32 data_upload; - INSERT_UNION_PADDING_WORDS(0x44); + INSERT_UNION_PADDING_WORDS(0x16); + + u32 force_early_fragment_tests; + + INSERT_UNION_PADDING_WORDS(0x2D); struct { union { @@ -1572,6 +1576,7 @@ ASSERT_REG_POSITION(shadow_ram_control, 0x49); ASSERT_REG_POSITION(upload, 0x60); ASSERT_REG_POSITION(exec_upload, 0x6C); ASSERT_REG_POSITION(data_upload, 0x6D); +ASSERT_REG_POSITION(force_early_fragment_tests, 0x84); ASSERT_REG_POSITION(sync_info, 0xB2); ASSERT_REG_POSITION(tess_mode, 0xC8); ASSERT_REG_POSITION(tess_level_outer, 0xC9); diff --git a/src/video_core/gpu.cpp b/src/video_core/gpu.cpp index ebd149c3a..e91f52938 100644 --- a/src/video_core/gpu.cpp +++ b/src/video_core/gpu.cpp @@ -95,22 +95,29 @@ void GPU::WaitFence(u32 syncpoint_id, u32 value) { if (!is_async) { return; } + if (syncpoint_id == UINT32_MAX) { + // TODO: Research what this does. + LOG_ERROR(HW_GPU, "Waiting for syncpoint -1 not implemented"); + return; + } MICROPROFILE_SCOPE(GPU_wait); std::unique_lock lock{sync_mutex}; - sync_cv.wait(lock, [=, this] { return syncpoints[syncpoint_id].load() >= value; }); + sync_cv.wait(lock, [=, this] { return syncpoints.at(syncpoint_id).load() >= value; }); } void GPU::IncrementSyncPoint(const u32 syncpoint_id) { - syncpoints[syncpoint_id]++; + auto& syncpoint = syncpoints.at(syncpoint_id); + syncpoint++; std::lock_guard lock{sync_mutex}; sync_cv.notify_all(); - if (!syncpt_interrupts[syncpoint_id].empty()) { - u32 value = syncpoints[syncpoint_id].load(); - auto it = syncpt_interrupts[syncpoint_id].begin(); - while (it != syncpt_interrupts[syncpoint_id].end()) { + auto& interrupt = syncpt_interrupts.at(syncpoint_id); + if (!interrupt.empty()) { + u32 value = syncpoint.load(); + auto it = interrupt.begin(); + while (it != interrupt.end()) { if (value >= *it) { TriggerCpuInterrupt(syncpoint_id, *it); - it = syncpt_interrupts[syncpoint_id].erase(it); + it = interrupt.erase(it); continue; } it++; @@ -119,22 +126,22 @@ void GPU::IncrementSyncPoint(const u32 syncpoint_id) { } u32 GPU::GetSyncpointValue(const u32 syncpoint_id) const { - return syncpoints[syncpoint_id].load(); + return syncpoints.at(syncpoint_id).load(); } void GPU::RegisterSyncptInterrupt(const u32 syncpoint_id, const u32 value) { - auto& interrupt = syncpt_interrupts[syncpoint_id]; + auto& interrupt = syncpt_interrupts.at(syncpoint_id); bool contains = std::any_of(interrupt.begin(), interrupt.end(), [value](u32 in_value) { return in_value == value; }); if (contains) { return; } - syncpt_interrupts[syncpoint_id].emplace_back(value); + interrupt.emplace_back(value); } bool GPU::CancelSyncptInterrupt(const u32 syncpoint_id, const u32 value) { std::lock_guard lock{sync_mutex}; - auto& interrupt = syncpt_interrupts[syncpoint_id]; + auto& interrupt = syncpt_interrupts.at(syncpoint_id); const auto iter = std::find_if(interrupt.begin(), interrupt.end(), [value](u32 interrupt_value) { return value == interrupt_value; }); diff --git a/src/video_core/renderer_vulkan/fixed_pipeline_state.cpp b/src/video_core/renderer_vulkan/fixed_pipeline_state.cpp index fffae528e..5ec43db11 100644 --- a/src/video_core/renderer_vulkan/fixed_pipeline_state.cpp +++ b/src/video_core/renderer_vulkan/fixed_pipeline_state.cpp @@ -46,7 +46,7 @@ void FixedPipelineState::Fill(const Maxwell& regs, bool has_extended_dynamic_sta regs.polygon_offset_fill_enable}; const u32 topology_index = static_cast<u32>(regs.draw.topology.Value()); - raw = 0; + raw1 = 0; primitive_restart_enable.Assign(regs.primitive_restart.enabled != 0 ? 1 : 0); depth_bias_enable.Assign(enabled_lut[POLYGON_OFFSET_ENABLE_LUT[topology_index]] != 0 ? 1 : 0); depth_clamp_disabled.Assign(regs.view_volume_clip_control.depth_clamp_disabled.Value()); @@ -61,12 +61,13 @@ void FixedPipelineState::Fill(const Maxwell& regs, bool has_extended_dynamic_sta rasterize_enable.Assign(regs.rasterize_enable != 0 ? 1 : 0); topology.Assign(regs.draw.topology); - alpha_raw = 0; + raw2 = 0; const auto test_func = regs.alpha_test_enabled == 1 ? regs.alpha_test_func : Maxwell::ComparisonOp::Always; alpha_test_func.Assign(PackComparisonOp(test_func)); - alpha_test_ref = Common::BitCast<u32>(regs.alpha_test_ref); + early_z.Assign(regs.force_early_fragment_tests != 0 ? 1 : 0); + alpha_test_ref = Common::BitCast<u32>(regs.alpha_test_ref); point_size = Common::BitCast<u32>(regs.point_size); for (std::size_t index = 0; index < Maxwell::NumVertexArrays; ++index) { diff --git a/src/video_core/renderer_vulkan/fixed_pipeline_state.h b/src/video_core/renderer_vulkan/fixed_pipeline_state.h index 42480e8d0..c26b77790 100644 --- a/src/video_core/renderer_vulkan/fixed_pipeline_state.h +++ b/src/video_core/renderer_vulkan/fixed_pipeline_state.h @@ -171,7 +171,7 @@ struct FixedPipelineState { }; union { - u32 raw; + u32 raw1; BitField<0, 1, u32> no_extended_dynamic_state; BitField<2, 1, u32> primitive_restart_enable; BitField<3, 1, u32> depth_bias_enable; @@ -187,13 +187,13 @@ struct FixedPipelineState { BitField<23, 1, u32> rasterize_enable; BitField<24, 4, Maxwell::PrimitiveTopology> topology; }; - - u32 alpha_test_ref; ///< Alpha test reference value union { - u32 alpha_raw; + u32 raw2; BitField<0, 3, u32> alpha_test_func; + BitField<3, 1, u32> early_z; }; + u32 alpha_test_ref; u32 point_size; std::array<u32, Maxwell::NumVertexArrays> binding_divisors; std::array<VertexAttribute, Maxwell::NumVertexAttributes> attributes; diff --git a/src/video_core/renderer_vulkan/vk_pipeline_cache.cpp b/src/video_core/renderer_vulkan/vk_pipeline_cache.cpp index f9efe526d..df7e8c864 100644 --- a/src/video_core/renderer_vulkan/vk_pipeline_cache.cpp +++ b/src/video_core/renderer_vulkan/vk_pipeline_cache.cpp @@ -344,6 +344,7 @@ VKPipelineCache::DecompileShaders(const FixedPipelineState& fixed_state) { specialization.attribute_types[i] = attribute.Type(); } specialization.ndc_minus_one_to_one = fixed_state.ndc_minus_one_to_one; + specialization.early_fragment_tests = fixed_state.early_z; // Alpha test specialization.alpha_test_func = diff --git a/src/video_core/renderer_vulkan/vk_shader_decompiler.cpp b/src/video_core/renderer_vulkan/vk_shader_decompiler.cpp index 1c52f40bb..fed9ebecd 100644 --- a/src/video_core/renderer_vulkan/vk_shader_decompiler.cpp +++ b/src/video_core/renderer_vulkan/vk_shader_decompiler.cpp @@ -315,7 +315,6 @@ public: "supported on this device"); } } - if (ir.UsesLayer() || ir.UsesViewportIndex()) { if (ir.UsesViewportIndex()) { AddCapability(spv::Capability::MultiViewport); @@ -325,11 +324,9 @@ public: AddCapability(spv::Capability::ShaderViewportIndexLayerEXT); } } - if (device.IsFormatlessImageLoadSupported()) { AddCapability(spv::Capability::StorageImageReadWithoutFormat); } - if (device.IsFloat16Supported()) { AddCapability(spv::Capability::Float16); } @@ -377,6 +374,9 @@ public: if (header.ps.omap.depth) { AddExecutionMode(main, spv::ExecutionMode::DepthReplacing); } + if (specialization.early_fragment_tests) { + AddExecutionMode(main, spv::ExecutionMode::EarlyFragmentTests); + } break; case ShaderType::Compute: const auto workgroup_size = specialization.workgroup_size; diff --git a/src/video_core/renderer_vulkan/vk_shader_decompiler.h b/src/video_core/renderer_vulkan/vk_shader_decompiler.h index cd3d0a415..110848922 100644 --- a/src/video_core/renderer_vulkan/vk_shader_decompiler.h +++ b/src/video_core/renderer_vulkan/vk_shader_decompiler.h @@ -95,6 +95,7 @@ struct Specialization final { std::bitset<Maxwell::NumVertexAttributes> enabled_attributes; std::array<Maxwell::VertexAttribute::Type, Maxwell::NumVertexAttributes> attribute_types{}; bool ndc_minus_one_to_one{}; + bool early_fragment_tests{}; float alpha_test_ref{}; Maxwell::ComparisonOp alpha_test_func{}; }; |