summaryrefslogtreecommitdiffstats
path: root/src/video_core/engines/maxwell_3d.h
diff options
context:
space:
mode:
Diffstat (limited to '')
-rw-r--r--src/video_core/engines/maxwell_3d.h186
1 files changed, 163 insertions, 23 deletions
diff --git a/src/video_core/engines/maxwell_3d.h b/src/video_core/engines/maxwell_3d.h
index f342c78e6..4c97759ed 100644
--- a/src/video_core/engines/maxwell_3d.h
+++ b/src/video_core/engines/maxwell_3d.h
@@ -15,6 +15,7 @@
#include "common/common_funcs.h"
#include "common/common_types.h"
#include "common/math_util.h"
+#include "video_core/engines/const_buffer_info.h"
#include "video_core/engines/engine_upload.h"
#include "video_core/gpu.h"
#include "video_core/macro_interpreter.h"
@@ -61,11 +62,13 @@ public:
static constexpr std::size_t NumVertexAttributes = 32;
static constexpr std::size_t NumVaryings = 31;
static constexpr std::size_t NumTextureSamplers = 32;
+ static constexpr std::size_t NumImages = 8; // TODO(Rodrigo): Investigate this number
static constexpr std::size_t NumClipDistances = 8;
static constexpr std::size_t MaxShaderProgram = 6;
static constexpr std::size_t MaxShaderStage = 5;
// Maximum number of const buffers per shader stage.
static constexpr std::size_t MaxConstBuffers = 18;
+ static constexpr std::size_t MaxConstBufferSize = 0x10000;
enum class QueryMode : u32 {
Write = 0,
@@ -88,6 +91,20 @@ public:
enum class QuerySelect : u32 {
Zero = 0,
+ TimeElapsed = 2,
+ TransformFeedbackPrimitivesGenerated = 11,
+ PrimitivesGenerated = 18,
+ SamplesPassed = 21,
+ TransformFeedbackUnknown = 26,
+ };
+
+ struct QueryCompare {
+ u32 initial_sequence;
+ u32 initial_mode;
+ u32 unknown1;
+ u32 unknown2;
+ u32 current_sequence;
+ u32 current_mode;
};
enum class QuerySyncCondition : u32 {
@@ -95,6 +112,14 @@ public:
GreaterThan = 1,
};
+ enum class ConditionMode : u32 {
+ Never = 0,
+ Always = 1,
+ ResNonZero = 2,
+ Equal = 3,
+ NotEqual = 4,
+ };
+
enum class ShaderProgram : u32 {
VertexA = 0,
VertexB = 1,
@@ -786,8 +811,9 @@ public:
INSERT_PADDING_WORDS(0x21);
u32 vb_element_base;
+ u32 vb_base_instance;
- INSERT_PADDING_WORDS(0x36);
+ INSERT_PADDING_WORDS(0x35);
union {
BitField<0, 1, u32> c0;
@@ -813,7 +839,18 @@ public:
BitField<4, 1, u32> alpha_to_one;
} multisample_control;
- INSERT_PADDING_WORDS(0x7);
+ INSERT_PADDING_WORDS(0x4);
+
+ struct {
+ u32 address_high;
+ u32 address_low;
+ ConditionMode mode;
+
+ GPUVAddr Address() const {
+ return static_cast<GPUVAddr>((static_cast<GPUVAddr>(address_high) << 32) |
+ address_low);
+ }
+ } condition;
struct {
u32 tsc_address_high;
@@ -1053,7 +1090,9 @@ public:
INSERT_PADDING_WORDS(14);
} shader_config[MaxShaderProgram];
- INSERT_PADDING_WORDS(0x80);
+ INSERT_PADDING_WORDS(0x60);
+
+ u32 firmware[0x20];
struct {
u32 cb_size;
@@ -1112,13 +1151,6 @@ public:
static_assert(std::is_trivially_copyable_v<Regs>, "Maxwell3D Regs must be trivially copyable");
struct State {
- struct ConstBufferInfo {
- GPUVAddr address;
- u32 index;
- u32 size;
- bool enabled;
- };
-
struct ShaderStageInfo {
std::array<ConstBufferInfo, Regs::MaxConstBuffers> const_buffers;
};
@@ -1129,23 +1161,77 @@ public:
State state{};
- struct DirtyFlags {
- std::bitset<8> color_buffer{0xFF};
- std::bitset<32> vertex_array{0xFFFFFFFF};
+ struct DirtyRegs {
+ static constexpr std::size_t NUM_REGS = 256;
+ union {
+ struct {
+ bool null_dirty;
+
+ // Vertex Attributes
+ bool vertex_attrib_format;
+
+ // Vertex Arrays
+ std::array<bool, 32> vertex_array;
+
+ bool vertex_array_buffers;
+
+ // Vertex Instances
+ std::array<bool, 32> vertex_instance;
+
+ bool vertex_instances;
+
+ // Render Targets
+ std::array<bool, 8> render_target;
+ bool depth_buffer;
+
+ bool render_settings;
- bool vertex_attrib_format = true;
- bool zeta_buffer = true;
- bool shaders = true;
+ // Shaders
+ bool shaders;
+
+ // Rasterizer State
+ bool viewport;
+ bool clip_coefficient;
+ bool cull_mode;
+ bool primitive_restart;
+ bool depth_test;
+ bool stencil_test;
+ bool blend_state;
+ bool scissor_test;
+ bool transform_feedback;
+ bool color_mask;
+ bool polygon_offset;
+
+ // Complementary
+ bool viewport_transform;
+ bool screen_y_control;
+
+ bool memory_general;
+ };
+ std::array<bool, NUM_REGS> regs;
+ };
+
+ void ResetVertexArrays() {
+ vertex_array.fill(true);
+ vertex_array_buffers = true;
+ }
+
+ void ResetRenderTargets() {
+ depth_buffer = true;
+ render_target.fill(true);
+ render_settings = true;
+ }
void OnMemoryWrite() {
- zeta_buffer = true;
shaders = true;
- color_buffer.set();
- vertex_array.set();
+ memory_general = true;
+ ResetRenderTargets();
+ ResetVertexArrays();
}
- };
- DirtyFlags dirty_flags;
+ } dirty{};
+
+ std::array<u8, Regs::NUM_REGS> dirty_pointers{};
/// Reads a register value located at the input method address
u32 GetRegisterValue(u32 method) const;
@@ -1153,6 +1239,11 @@ public:
/// Write the value to the register identified by method.
void CallMethod(const GPU::MethodCall& method_call);
+ /// Write the value to the register identified by method.
+ void CallMethodFromMME(const GPU::MethodCall& method_call);
+
+ void FlushMMEInlineDraw();
+
/// Given a Texture Handle, returns the TSC and TIC entries.
Texture::FullTextureInfo GetTextureInfo(const Texture::TextureHandle tex_handle,
std::size_t offset) const;
@@ -1174,6 +1265,25 @@ public:
return macro_memory;
}
+ bool ShouldExecute() const {
+ return execute_on;
+ }
+
+ enum class MMEDrawMode : u32 {
+ Undefined,
+ Array,
+ Indexed,
+ };
+
+ struct MMEDrawState {
+ MMEDrawMode current_mode{MMEDrawMode::Undefined};
+ u32 current_count{};
+ u32 instance_count{};
+ bool instance_mode{};
+ bool gl_begin_consume{};
+ u32 gl_end_count{};
+ } mme_draw;
+
private:
void InitializeRegisterDefaults();
@@ -1184,7 +1294,9 @@ private:
MemoryManager& memory_manager;
/// Start offsets of each macro in macro_memory
- std::unordered_map<u32, u32> macro_offsets;
+ std::array<u32, 0x80> macro_positions = {};
+
+ std::array<bool, Regs::NUM_REGS> mme_inline{};
/// Memory for macro code
MacroMemory macro_memory;
@@ -1197,20 +1309,34 @@ private:
/// Interpreter for the macro codes uploaded to the GPU.
MacroInterpreter macro_interpreter;
+ static constexpr u32 null_cb_data = 0xFFFFFFFF;
+ struct {
+ std::array<std::array<u32, 0x4000>, 16> buffer;
+ u32 current{null_cb_data};
+ u32 id{null_cb_data};
+ u32 start_pos{};
+ u32 counter{};
+ } cb_data_state;
+
Upload::State upload_state;
+ bool execute_on{true};
+
/// Retrieves information about a specific TIC entry from the TIC buffer.
Texture::TICEntry GetTICEntry(u32 tic_index) const;
/// Retrieves information about a specific TSC entry from the TSC buffer.
Texture::TSCEntry GetTSCEntry(u32 tsc_index) const;
+ void InitDirtySettings();
+
/**
* Call a macro on this engine.
* @param method Method to call
+ * @param num_parameters Number of arguments
* @param parameters Arguments to the method call
*/
- void CallMacroMethod(u32 method, std::vector<u32> parameters);
+ void CallMacroMethod(u32 method, std::size_t num_parameters, const u32* parameters);
/// Handles writes to the macro uploading register.
void ProcessMacroUpload(u32 data);
@@ -1218,23 +1344,34 @@ private:
/// Handles writes to the macro bind register.
void ProcessMacroBind(u32 data);
+ /// Handles firmware blob 4
+ void ProcessFirmwareCall4();
+
/// Handles a write to the CLEAR_BUFFERS register.
void ProcessClearBuffers();
/// Handles a write to the QUERY_GET register.
void ProcessQueryGet();
+ // Handles Conditional Rendering
+ void ProcessQueryCondition();
+
/// Handles writes to syncing register.
void ProcessSyncPoint();
/// Handles a write to the CB_DATA[i] register.
+ void StartCBData(u32 method);
void ProcessCBData(u32 value);
+ void FinishCBData();
/// Handles a write to the CB_BIND register.
void ProcessCBBind(Regs::ShaderStage stage);
/// Handles a write to the VERTEX_END_GL register, triggering a draw.
void DrawArrays();
+
+ // Handles a instance drawcall from MME
+ void StepInstance(MMEDrawMode expected_mode, u32 count);
};
#define ASSERT_REG_POSITION(field_name, position) \
@@ -1291,10 +1428,12 @@ ASSERT_REG_POSITION(stencil_front_mask, 0x4E7);
ASSERT_REG_POSITION(frag_color_clamp, 0x4EA);
ASSERT_REG_POSITION(screen_y_control, 0x4EB);
ASSERT_REG_POSITION(vb_element_base, 0x50D);
+ASSERT_REG_POSITION(vb_base_instance, 0x50E);
ASSERT_REG_POSITION(clip_distance_enabled, 0x544);
ASSERT_REG_POSITION(point_size, 0x546);
ASSERT_REG_POSITION(zeta_enable, 0x54E);
ASSERT_REG_POSITION(multisample_control, 0x54F);
+ASSERT_REG_POSITION(condition, 0x554);
ASSERT_REG_POSITION(tsc, 0x557);
ASSERT_REG_POSITION(polygon_offset_factor, 0x55b);
ASSERT_REG_POSITION(tic, 0x55D);
@@ -1324,6 +1463,7 @@ ASSERT_REG_POSITION(vertex_array[0], 0x700);
ASSERT_REG_POSITION(independent_blend, 0x780);
ASSERT_REG_POSITION(vertex_array_limit[0], 0x7C0);
ASSERT_REG_POSITION(shader_config[0], 0x800);
+ASSERT_REG_POSITION(firmware, 0x8C0);
ASSERT_REG_POSITION(const_buffer, 0x8E0);
ASSERT_REG_POSITION(cb_bind[0], 0x904);
ASSERT_REG_POSITION(tex_cb_index, 0x982);
generated by cgit v1.2.3 (git 2.25.1) at 2024-08-17 01:41:06 +0200