1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
|
// Copyright 2018 yuzu Emulator Project
// Licensed under GPLv2 or any later version
// Refer to the license.txt file included.
#include <cinttypes>
#include "common/assert.h"
#include "core/core.h"
#include "video_core/debug_utils/debug_utils.h"
#include "video_core/engines/maxwell_3d.h"
#include "video_core/rasterizer_interface.h"
#include "video_core/renderer_base.h"
#include "video_core/textures/decoders.h"
#include "video_core/textures/texture.h"
#include "video_core/video_core.h"
namespace Tegra {
namespace Engines {
/// First register id that is actually a Macro call.
constexpr u32 MacroRegistersStart = 0xE00;
Maxwell3D::Maxwell3D(MemoryManager& memory_manager)
: memory_manager(memory_manager), macro_interpreter(*this) {}
void Maxwell3D::CallMacroMethod(u32 method, std::vector<u32> parameters) {
auto macro_code = uploaded_macros.find(method);
// The requested macro must have been uploaded already.
ASSERT_MSG(macro_code != uploaded_macros.end(), "Macro %08X was not uploaded", method);
// Reset the current macro and execute it.
executing_macro = 0;
macro_interpreter.Execute(macro_code->second, std::move(parameters));
}
void Maxwell3D::WriteReg(u32 method, u32 value, u32 remaining_params) {
auto debug_context = Core::System::GetInstance().GetGPUDebugContext();
// It is an error to write to a register other than the current macro's ARG register before it
// has finished execution.
if (executing_macro != 0) {
ASSERT(method == executing_macro + 1);
}
// Methods after 0xE00 are special, they're actually triggers for some microcode that was
// uploaded to the GPU during initialization.
if (method >= MacroRegistersStart) {
// We're trying to execute a macro
if (executing_macro == 0) {
// A macro call must begin by writing the macro method's register, not its argument.
ASSERT_MSG((method % 2) == 0,
"Can't start macro execution by writing to the ARGS register");
executing_macro = method;
}
macro_params.push_back(value);
// Call the macro when there are no more parameters in the command buffer
if (remaining_params == 0) {
CallMacroMethod(executing_macro, std::move(macro_params));
}
return;
}
ASSERT_MSG(method < Regs::NUM_REGS,
"Invalid Maxwell3D register, increase the size of the Regs structure");
if (debug_context) {
debug_context->OnEvent(Tegra::DebugContext::Event::MaxwellCommandLoaded, nullptr);
}
regs.reg_array[method] = value;
switch (method) {
case MAXWELL3D_REG_INDEX(macros.data): {
ProcessMacroUpload(value);
break;
}
case MAXWELL3D_REG_INDEX(code_address.code_address_high):
case MAXWELL3D_REG_INDEX(code_address.code_address_low): {
// Note: For some reason games (like Puyo Puyo Tetris) seem to write 0 to the CODE_ADDRESS
// register, we do not currently know if that's intended or a bug, so we assert it lest
// stuff breaks in other places (like the shader address calculation).
ASSERT_MSG(regs.code_address.CodeAddress() == 0, "Unexpected CODE_ADDRESS register value.");
break;
}
case MAXWELL3D_REG_INDEX(const_buffer.cb_data[0]):
case MAXWELL3D_REG_INDEX(const_buffer.cb_data[1]):
case MAXWELL3D_REG_INDEX(const_buffer.cb_data[2]):
case MAXWELL3D_REG_INDEX(const_buffer.cb_data[3]):
case MAXWELL3D_REG_INDEX(const_buffer.cb_data[4]):
case MAXWELL3D_REG_INDEX(const_buffer.cb_data[5]):
case MAXWELL3D_REG_INDEX(const_buffer.cb_data[6]):
case MAXWELL3D_REG_INDEX(const_buffer.cb_data[7]):
case MAXWELL3D_REG_INDEX(const_buffer.cb_data[8]):
case MAXWELL3D_REG_INDEX(const_buffer.cb_data[9]):
case MAXWELL3D_REG_INDEX(const_buffer.cb_data[10]):
case MAXWELL3D_REG_INDEX(const_buffer.cb_data[11]):
case MAXWELL3D_REG_INDEX(const_buffer.cb_data[12]):
case MAXWELL3D_REG_INDEX(const_buffer.cb_data[13]):
case MAXWELL3D_REG_INDEX(const_buffer.cb_data[14]):
case MAXWELL3D_REG_INDEX(const_buffer.cb_data[15]): {
ProcessCBData(value);
break;
}
case MAXWELL3D_REG_INDEX(cb_bind[0].raw_config): {
ProcessCBBind(Regs::ShaderStage::Vertex);
break;
}
case MAXWELL3D_REG_INDEX(cb_bind[1].raw_config): {
ProcessCBBind(Regs::ShaderStage::TesselationControl);
break;
}
case MAXWELL3D_REG_INDEX(cb_bind[2].raw_config): {
ProcessCBBind(Regs::ShaderStage::TesselationEval);
break;
}
case MAXWELL3D_REG_INDEX(cb_bind[3].raw_config): {
ProcessCBBind(Regs::ShaderStage::Geometry);
break;
}
case MAXWELL3D_REG_INDEX(cb_bind[4].raw_config): {
ProcessCBBind(Regs::ShaderStage::Fragment);
break;
}
case MAXWELL3D_REG_INDEX(draw.vertex_end_gl): {
DrawArrays();
break;
}
case MAXWELL3D_REG_INDEX(clear_buffers): {
ProcessClearBuffers();
break;
}
case MAXWELL3D_REG_INDEX(query.query_get): {
ProcessQueryGet();
break;
}
default:
break;
}
VideoCore::g_renderer->Rasterizer()->NotifyMaxwellRegisterChanged(method);
if (debug_context) {
debug_context->OnEvent(Tegra::DebugContext::Event::MaxwellCommandProcessed, nullptr);
}
}
void Maxwell3D::ProcessMacroUpload(u32 data) {
// Store the uploaded macro code to interpret them when they're called.
auto& macro = uploaded_macros[regs.macros.entry * 2 + MacroRegistersStart];
macro.push_back(data);
}
void Maxwell3D::ProcessQueryGet() {
GPUVAddr sequence_address = regs.query.QueryAddress();
// Since the sequence address is given as a GPU VAddr, we have to convert it to an application
// VAddr before writing.
boost::optional<VAddr> address = memory_manager.GpuToCpuAddress(sequence_address);
// TODO(Subv): Support the other query units.
ASSERT_MSG(regs.query.query_get.unit == Regs::QueryUnit::Crop,
"Units other than CROP are unimplemented");
u32 value = Memory::Read32(*address);
u64 result = 0;
// TODO(Subv): Support the other query variables
switch (regs.query.query_get.select) {
case Regs::QuerySelect::Zero:
// This seems to actually write the query sequence to the query address.
result = regs.query.query_sequence;
break;
default:
UNIMPLEMENTED_MSG("Unimplemented query select type {}",
static_cast<u32>(regs.query.query_get.select.Value()));
}
// TODO(Subv): Research and implement how query sync conditions work.
struct LongQueryResult {
u64_le value;
u64_le timestamp;
};
static_assert(sizeof(LongQueryResult) == 16, "LongQueryResult has wrong size");
switch (regs.query.query_get.mode) {
case Regs::QueryMode::Write:
case Regs::QueryMode::Write2: {
u32 sequence = regs.query.query_sequence;
if (regs.query.query_get.short_query) {
// Write the current query sequence to the sequence address.
// TODO(Subv): Find out what happens if you use a long query type but mark it as a short
// query.
Memory::Write32(*address, sequence);
} else {
// Write the 128-bit result structure in long mode. Note: We emulate an infinitely fast
// GPU, this command may actually take a while to complete in real hardware due to GPU
// wait queues.
LongQueryResult query_result{};
query_result.value = result;
// TODO(Subv): Generate a real GPU timestamp and write it here instead of 0
query_result.timestamp = 0;
Memory::WriteBlock(*address, &query_result, sizeof(query_result));
}
break;
}
default:
UNIMPLEMENTED_MSG("Query mode {} not implemented",
static_cast<u32>(regs.query.query_get.mode.Value()));
}
}
void Maxwell3D::DrawArrays() {
NGLOG_DEBUG(HW_GPU, "called, topology={}, count={}",
static_cast<u32>(regs.draw.topology.Value()), regs.vertex_buffer.count);
ASSERT_MSG(!(regs.index_array.count && regs.vertex_buffer.count), "Both indexed and direct?");
auto debug_context = Core::System::GetInstance().GetGPUDebugContext();
if (debug_context) {
debug_context->OnEvent(Tegra::DebugContext::Event::IncomingPrimitiveBatch, nullptr);
}
if (debug_context) {
debug_context->OnEvent(Tegra::DebugContext::Event::FinishedPrimitiveBatch, nullptr);
}
const bool is_indexed{regs.index_array.count && !regs.vertex_buffer.count};
VideoCore::g_renderer->Rasterizer()->AccelerateDrawBatch(is_indexed);
// TODO(bunnei): Below, we reset vertex count so that we can use these registers to determine if
// the game is trying to draw indexed or direct mode. This needs to be verified on HW still -
// it's possible that it is incorrect and that there is some other register used to specify the
// drawing mode.
if (is_indexed) {
regs.index_array.count = 0;
} else {
regs.vertex_buffer.count = 0;
}
}
void Maxwell3D::ProcessCBBind(Regs::ShaderStage stage) {
// Bind the buffer currently in CB_ADDRESS to the specified index in the desired shader stage.
auto& shader = state.shader_stages[static_cast<size_t>(stage)];
auto& bind_data = regs.cb_bind[static_cast<size_t>(stage)];
auto& buffer = shader.const_buffers[bind_data.index];
buffer.enabled = bind_data.valid.Value() != 0;
buffer.index = bind_data.index;
buffer.address = regs.const_buffer.BufferAddress();
buffer.size = regs.const_buffer.cb_size;
}
void Maxwell3D::ProcessCBData(u32 value) {
// Write the input value to the current const buffer at the current position.
GPUVAddr buffer_address = regs.const_buffer.BufferAddress();
ASSERT(buffer_address != 0);
// Don't allow writing past the end of the buffer.
ASSERT(regs.const_buffer.cb_pos + sizeof(u32) <= regs.const_buffer.cb_size);
boost::optional<VAddr> address =
memory_manager.GpuToCpuAddress(buffer_address + regs.const_buffer.cb_pos);
Memory::Write32(*address, value);
// Increment the current buffer position.
regs.const_buffer.cb_pos = regs.const_buffer.cb_pos + 4;
}
Texture::TICEntry Maxwell3D::GetTICEntry(u32 tic_index) const {
GPUVAddr tic_base_address = regs.tic.TICAddress();
GPUVAddr tic_address_gpu = tic_base_address + tic_index * sizeof(Texture::TICEntry);
boost::optional<VAddr> tic_address_cpu = memory_manager.GpuToCpuAddress(tic_address_gpu);
Texture::TICEntry tic_entry;
Memory::ReadBlock(*tic_address_cpu, &tic_entry, sizeof(Texture::TICEntry));
ASSERT_MSG(tic_entry.header_version == Texture::TICHeaderVersion::BlockLinear ||
tic_entry.header_version == Texture::TICHeaderVersion::Pitch,
"TIC versions other than BlockLinear or Pitch are unimplemented");
ASSERT_MSG((tic_entry.texture_type == Texture::TextureType::Texture2D) ||
(tic_entry.texture_type == Texture::TextureType::Texture2DNoMipmap),
"Texture types other than Texture2D are unimplemented");
auto r_type = tic_entry.r_type.Value();
auto g_type = tic_entry.g_type.Value();
auto b_type = tic_entry.b_type.Value();
auto a_type = tic_entry.a_type.Value();
// TODO(Subv): Different data types for separate components are not supported
ASSERT(r_type == g_type && r_type == b_type && r_type == a_type);
// TODO(Subv): Only UNORM formats are supported for now.
ASSERT(r_type == Texture::ComponentType::UNORM);
return tic_entry;
}
Texture::TSCEntry Maxwell3D::GetTSCEntry(u32 tsc_index) const {
GPUVAddr tsc_base_address = regs.tsc.TSCAddress();
GPUVAddr tsc_address_gpu = tsc_base_address + tsc_index * sizeof(Texture::TSCEntry);
boost::optional<VAddr> tsc_address_cpu = memory_manager.GpuToCpuAddress(tsc_address_gpu);
Texture::TSCEntry tsc_entry;
Memory::ReadBlock(*tsc_address_cpu, &tsc_entry, sizeof(Texture::TSCEntry));
return tsc_entry;
}
std::vector<Texture::FullTextureInfo> Maxwell3D::GetStageTextures(Regs::ShaderStage stage) const {
std::vector<Texture::FullTextureInfo> textures;
auto& fragment_shader = state.shader_stages[static_cast<size_t>(stage)];
auto& tex_info_buffer = fragment_shader.const_buffers[regs.tex_cb_index];
ASSERT(tex_info_buffer.enabled && tex_info_buffer.address != 0);
GPUVAddr tic_base_address = regs.tic.TICAddress();
GPUVAddr tex_info_buffer_end = tex_info_buffer.address + tex_info_buffer.size;
// Offset into the texture constbuffer where the texture info begins.
static constexpr size_t TextureInfoOffset = 0x20;
for (GPUVAddr current_texture = tex_info_buffer.address + TextureInfoOffset;
current_texture < tex_info_buffer_end; current_texture += sizeof(Texture::TextureHandle)) {
Texture::TextureHandle tex_handle{
Memory::Read32(*memory_manager.GpuToCpuAddress(current_texture))};
Texture::FullTextureInfo tex_info{};
// TODO(Subv): Use the shader to determine which textures are actually accessed.
tex_info.index =
static_cast<u32>(current_texture - tex_info_buffer.address - TextureInfoOffset) /
sizeof(Texture::TextureHandle);
// Load the TIC data.
if (tex_handle.tic_id != 0) {
tex_info.enabled = true;
auto tic_entry = GetTICEntry(tex_handle.tic_id);
// TODO(Subv): Workaround for BitField's move constructor being deleted.
std::memcpy(&tex_info.tic, &tic_entry, sizeof(tic_entry));
}
// Load the TSC data
if (tex_handle.tsc_id != 0) {
auto tsc_entry = GetTSCEntry(tex_handle.tsc_id);
// TODO(Subv): Workaround for BitField's move constructor being deleted.
std::memcpy(&tex_info.tsc, &tsc_entry, sizeof(tsc_entry));
}
if (tex_info.enabled)
textures.push_back(tex_info);
}
return textures;
}
Texture::FullTextureInfo Maxwell3D::GetStageTexture(Regs::ShaderStage stage, size_t offset) const {
auto& shader = state.shader_stages[static_cast<size_t>(stage)];
auto& tex_info_buffer = shader.const_buffers[regs.tex_cb_index];
ASSERT(tex_info_buffer.enabled && tex_info_buffer.address != 0);
GPUVAddr tex_info_address = tex_info_buffer.address + offset * sizeof(Texture::TextureHandle);
ASSERT(tex_info_address < tex_info_buffer.address + tex_info_buffer.size);
boost::optional<VAddr> tex_address_cpu = memory_manager.GpuToCpuAddress(tex_info_address);
Texture::TextureHandle tex_handle{Memory::Read32(*tex_address_cpu)};
Texture::FullTextureInfo tex_info{};
tex_info.index = static_cast<u32>(offset);
// Load the TIC data.
if (tex_handle.tic_id != 0) {
tex_info.enabled = true;
auto tic_entry = GetTICEntry(tex_handle.tic_id);
// TODO(Subv): Workaround for BitField's move constructor being deleted.
std::memcpy(&tex_info.tic, &tic_entry, sizeof(tic_entry));
}
// Load the TSC data
if (tex_handle.tsc_id != 0) {
auto tsc_entry = GetTSCEntry(tex_handle.tsc_id);
// TODO(Subv): Workaround for BitField's move constructor being deleted.
std::memcpy(&tex_info.tsc, &tsc_entry, sizeof(tsc_entry));
}
return tex_info;
}
u32 Maxwell3D::GetRegisterValue(u32 method) const {
ASSERT_MSG(method < Regs::NUM_REGS, "Invalid Maxwell3D register");
return regs.reg_array[method];
}
bool Maxwell3D::IsShaderStageEnabled(Regs::ShaderStage stage) const {
// The Vertex stage is always enabled.
if (stage == Regs::ShaderStage::Vertex)
return true;
switch (stage) {
case Regs::ShaderStage::TesselationControl:
return regs.shader_config[static_cast<size_t>(Regs::ShaderProgram::TesselationControl)]
.enable != 0;
case Regs::ShaderStage::TesselationEval:
return regs.shader_config[static_cast<size_t>(Regs::ShaderProgram::TesselationEval)]
.enable != 0;
case Regs::ShaderStage::Geometry:
return regs.shader_config[static_cast<size_t>(Regs::ShaderProgram::Geometry)].enable != 0;
case Regs::ShaderStage::Fragment:
return regs.shader_config[static_cast<size_t>(Regs::ShaderProgram::Fragment)].enable != 0;
}
UNREACHABLE();
}
void Maxwell3D::ProcessClearBuffers() {
ASSERT(regs.clear_buffers.R == regs.clear_buffers.G &&
regs.clear_buffers.R == regs.clear_buffers.B &&
regs.clear_buffers.R == regs.clear_buffers.A);
VideoCore::g_renderer->Rasterizer()->Clear();
}
} // namespace Engines
} // namespace Tegra
|