summaryrefslogtreecommitdiffstats
path: root/src/video_core/memory_manager.cpp
blob: 1e090279f0f680349550783ff2ec1c3db00efebb (plain) (blame)
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
// SPDX-FileCopyrightText: Copyright 2018 yuzu Emulator Project
// SPDX-License-Identifier: GPL-2.0-or-later

#include <algorithm>

#include "common/alignment.h"
#include "common/assert.h"
#include "common/logging/log.h"
#include "core/core.h"
#include "core/hle/kernel/k_page_table.h"
#include "core/hle/kernel/k_process.h"
#include "core/memory.h"
#include "video_core/memory_manager.h"
#include "video_core/rasterizer_interface.h"
#include "video_core/renderer_base.h"

namespace Tegra {

MemoryManager::MemoryManager(Core::System& system_, u64 address_space_bits_, u64 page_bits_)
    : system{system_}, address_space_bits{address_space_bits_}, page_bits{page_bits_}, entries{},
      page_table{address_space_bits, address_space_bits + page_bits - 38, page_bits} {
    address_space_size = 1ULL << address_space_bits;
    allocate_start = address_space_bits > 32 ? 1ULL << 32 : 0;
    page_size = 1ULL << page_bits;
    page_mask = page_size - 1ULL;
    const u64 page_table_bits = address_space_bits - cpu_page_bits;
    const u64 page_table_size = 1ULL << page_table_bits;
    page_table_mask = page_table_size - 1;

    entries.resize(page_table_size / 32, 0);
}

MemoryManager::~MemoryManager() = default;

MemoryManager::EntryType MemoryManager::GetEntry(size_t position) const {
    position = position >> page_bits;
    const u64 entry_mask = entries[position / 32];
    const size_t sub_index = position % 32;
    return static_cast<EntryType>((entry_mask >> (2 * sub_index)) & 0x03ULL);
}

void MemoryManager::SetEntry(size_t position, MemoryManager::EntryType entry) {
    position = position >> page_bits;
    const u64 entry_mask = entries[position / 32];
    const size_t sub_index = position % 32;
    entries[position / 32] =
        (~(3ULL << sub_index * 2) & entry_mask) | (static_cast<u64>(entry) << sub_index * 2);
}

template <MemoryManager::EntryType entry_type>
GPUVAddr MemoryManager::PageTableOp(GPUVAddr gpu_addr, [[maybe_unused]] VAddr cpu_addr,
                                    size_t size) {
    u64 remaining_size{size};
    if constexpr (entry_type == EntryType::Mapped) {
        page_table.ReserveRange(gpu_addr, size);
    }
    for (u64 offset{}; offset < size; offset += page_size) {
        const GPUVAddr current_gpu_addr = gpu_addr + offset;
        SetEntry(current_gpu_addr, entry_type);
        if constexpr (entry_type == EntryType::Mapped) {
            const VAddr current_cpu_addr = cpu_addr + offset;
            const auto index = PageEntryIndex(current_gpu_addr);
            page_table[index] = static_cast<u32>(current_cpu_addr >> 12ULL);
        }
        remaining_size -= page_size;
    }
    return gpu_addr;
}

void MemoryManager::BindRasterizer(VideoCore::RasterizerInterface* rasterizer_) {
    rasterizer = rasterizer_;
}

GPUVAddr MemoryManager::Map(VAddr cpu_addr, GPUVAddr gpu_addr, std::size_t size) {
    return PageTableOp<EntryType::Mapped>(gpu_addr, cpu_addr, size);
}

GPUVAddr MemoryManager::MapAllocate(VAddr cpu_addr, std::size_t size, std::size_t align) {
    return Map(cpu_addr, *FindFreeRange(size, align), size);
}

GPUVAddr MemoryManager::MapAllocate32(VAddr cpu_addr, std::size_t size) {
    const std::optional<GPUVAddr> gpu_addr = FindFreeRange(size, 1, true);
    ASSERT(gpu_addr);
    return Map(cpu_addr, *gpu_addr, size);
}

void MemoryManager::Unmap(GPUVAddr gpu_addr, std::size_t size) {
    if (size == 0) {
        return;
    }
    const auto submapped_ranges = GetSubmappedRange(gpu_addr, size);

    for (const auto& [map_addr, map_size] : submapped_ranges) {
        // Flush and invalidate through the GPU interface, to be asynchronous if possible.
        const std::optional<VAddr> cpu_addr = GpuToCpuAddress(map_addr);
        ASSERT(cpu_addr);

        rasterizer->UnmapMemory(*cpu_addr, map_size);
    }

    PageTableOp<EntryType::Free>(gpu_addr, 0, size);
}

std::optional<GPUVAddr> MemoryManager::AllocateFixed(GPUVAddr gpu_addr, std::size_t size) {
    for (u64 offset{}; offset < size; offset += page_size) {
        if (GetEntry(gpu_addr + offset) != EntryType::Free) {
            return std::nullopt;
        }
    }

    return PageTableOp<EntryType::Reserved>(gpu_addr, 0, size);
}

GPUVAddr MemoryManager::Allocate(std::size_t size, std::size_t align) {
    return *AllocateFixed(*FindFreeRange(size, align), size);
}

std::optional<GPUVAddr> MemoryManager::FindFreeRange(std::size_t size, std::size_t align,
                                                     bool start_32bit_address) const {
    if (!align) {
        align = page_size;
    } else {
        align = Common::AlignUp(align, page_size);
    }

    u64 available_size{};
    GPUVAddr gpu_addr{allocate_start};
    while (gpu_addr + available_size < address_space_size) {
        if (GetEntry(gpu_addr + available_size) == EntryType::Free) {
            available_size += page_size;

            if (available_size >= size) {
                return gpu_addr;
            }
        } else {
            gpu_addr += available_size + page_size;
            available_size = 0;

            const auto remainder{gpu_addr % align};
            if (remainder) {
                gpu_addr = (gpu_addr - remainder) + align;
            }
        }
    }

    return std::nullopt;
}

std::optional<VAddr> MemoryManager::GpuToCpuAddress(GPUVAddr gpu_addr) const {
    if (GetEntry(gpu_addr) != EntryType::Mapped) {
        return std::nullopt;
    }

    const VAddr cpu_addr_base = static_cast<VAddr>(page_table[PageEntryIndex(gpu_addr)]) << 12ULL;
    return cpu_addr_base + (gpu_addr & page_mask);
}

std::optional<VAddr> MemoryManager::GpuToCpuAddress(GPUVAddr addr, std::size_t size) const {
    size_t page_index{addr >> page_bits};
    const size_t page_last{(addr + size + page_size - 1) >> page_bits};
    while (page_index < page_last) {
        const auto page_addr{GpuToCpuAddress(page_index << page_bits)};
        if (page_addr && *page_addr != 0) {
            return page_addr;
        }
        ++page_index;
    }
    return std::nullopt;
}

template <typename T>
T MemoryManager::Read(GPUVAddr addr) const {
    if (auto page_pointer{GetPointer(addr)}; page_pointer) {
        // NOTE: Avoid adding any extra logic to this fast-path block
        T value;
        std::memcpy(&value, page_pointer, sizeof(T));
        return value;
    }

    ASSERT(false);

    return {};
}

template <typename T>
void MemoryManager::Write(GPUVAddr addr, T data) {
    if (auto page_pointer{GetPointer(addr)}; page_pointer) {
        // NOTE: Avoid adding any extra logic to this fast-path block
        std::memcpy(page_pointer, &data, sizeof(T));
        return;
    }

    ASSERT(false);
}

template u8 MemoryManager::Read<u8>(GPUVAddr addr) const;
template u16 MemoryManager::Read<u16>(GPUVAddr addr) const;
template u32 MemoryManager::Read<u32>(GPUVAddr addr) const;
template u64 MemoryManager::Read<u64>(GPUVAddr addr) const;
template void MemoryManager::Write<u8>(GPUVAddr addr, u8 data);
template void MemoryManager::Write<u16>(GPUVAddr addr, u16 data);
template void MemoryManager::Write<u32>(GPUVAddr addr, u32 data);
template void MemoryManager::Write<u64>(GPUVAddr addr, u64 data);

u8* MemoryManager::GetPointer(GPUVAddr gpu_addr) {
    const auto address{GpuToCpuAddress(gpu_addr)};
    if (!address) {
        return {};
    }

    return system.Memory().GetPointer(*address);
}

const u8* MemoryManager::GetPointer(GPUVAddr gpu_addr) const {
    const auto address{GpuToCpuAddress(gpu_addr)};
    if (!address) {
        return {};
    }

    return system.Memory().GetPointer(*address);
}

void MemoryManager::ReadBlockImpl(GPUVAddr gpu_src_addr, void* dest_buffer, std::size_t size,
                                  bool is_safe) const {
    std::size_t remaining_size{size};
    std::size_t page_index{gpu_src_addr >> page_bits};
    std::size_t page_offset{gpu_src_addr & page_mask};

    while (remaining_size > 0) {
        const std::size_t copy_amount{
            std::min(static_cast<std::size_t>(page_size) - page_offset, remaining_size)};
        const auto page_addr{GpuToCpuAddress(page_index << page_bits)};
        if (page_addr) {
            const auto src_addr{*page_addr + page_offset};
            if (is_safe) {
                // Flush must happen on the rasterizer interface, such that memory is always
                // synchronous when it is read (even when in asynchronous GPU mode).
                // Fixes Dead Cells title menu.
                rasterizer->FlushRegion(src_addr, copy_amount);
            }
            system.Memory().ReadBlockUnsafe(src_addr, dest_buffer, copy_amount);
        } else {
            std::memset(dest_buffer, 0, copy_amount);
        }

        page_index++;
        page_offset = 0;
        dest_buffer = static_cast<u8*>(dest_buffer) + copy_amount;
        remaining_size -= copy_amount;
    }
}

void MemoryManager::ReadBlock(GPUVAddr gpu_src_addr, void* dest_buffer, std::size_t size) const {
    ReadBlockImpl(gpu_src_addr, dest_buffer, size, true);
}

void MemoryManager::ReadBlockUnsafe(GPUVAddr gpu_src_addr, void* dest_buffer,
                                    const std::size_t size) const {
    ReadBlockImpl(gpu_src_addr, dest_buffer, size, false);
}

void MemoryManager::WriteBlockImpl(GPUVAddr gpu_dest_addr, const void* src_buffer, std::size_t size,
                                   bool is_safe) {
    std::size_t remaining_size{size};
    std::size_t page_index{gpu_dest_addr >> page_bits};
    std::size_t page_offset{gpu_dest_addr & page_mask};

    while (remaining_size > 0) {
        const std::size_t copy_amount{
            std::min(static_cast<std::size_t>(page_size) - page_offset, remaining_size)};
        const auto page_addr{GpuToCpuAddress(page_index << page_bits)};
        if (page_addr) {
            const auto dest_addr{*page_addr + page_offset};

            if (is_safe) {
                // Invalidate must happen on the rasterizer interface, such that memory is always
                // synchronous when it is written (even when in asynchronous GPU mode).
                rasterizer->InvalidateRegion(dest_addr, copy_amount);
            }
            system.Memory().WriteBlockUnsafe(dest_addr, src_buffer, copy_amount);
        }

        page_index++;
        page_offset = 0;
        src_buffer = static_cast<const u8*>(src_buffer) + copy_amount;
        remaining_size -= copy_amount;
    }
}

void MemoryManager::WriteBlock(GPUVAddr gpu_dest_addr, const void* src_buffer, std::size_t size) {
    WriteBlockImpl(gpu_dest_addr, src_buffer, size, true);
}

void MemoryManager::WriteBlockUnsafe(GPUVAddr gpu_dest_addr, const void* src_buffer,
                                     std::size_t size) {
    WriteBlockImpl(gpu_dest_addr, src_buffer, size, false);
}

void MemoryManager::FlushRegion(GPUVAddr gpu_addr, size_t size) const {
    size_t remaining_size{size};
    size_t page_index{gpu_addr >> page_bits};
    size_t page_offset{gpu_addr & page_mask};
    while (remaining_size > 0) {
        const size_t num_bytes{std::min(page_size - page_offset, remaining_size)};
        if (const auto page_addr{GpuToCpuAddress(page_index << page_bits)}; page_addr) {
            rasterizer->FlushRegion(*page_addr + page_offset, num_bytes);
        }
        ++page_index;
        page_offset = 0;
        remaining_size -= num_bytes;
    }
}

void MemoryManager::CopyBlock(GPUVAddr gpu_dest_addr, GPUVAddr gpu_src_addr, std::size_t size) {
    std::vector<u8> tmp_buffer(size);
    ReadBlock(gpu_src_addr, tmp_buffer.data(), size);

    // The output block must be flushed in case it has data modified from the GPU.
    // Fixes NPC geometry in Zombie Panic in Wonderland DX
    FlushRegion(gpu_dest_addr, size);
    WriteBlock(gpu_dest_addr, tmp_buffer.data(), size);
}

bool MemoryManager::IsGranularRange(GPUVAddr gpu_addr, std::size_t size) const {
    const auto cpu_addr{GpuToCpuAddress(gpu_addr)};
    if (!cpu_addr) {
        return false;
    }
    const std::size_t page{(*cpu_addr & Core::Memory::YUZU_PAGEMASK) + size};
    return page <= Core::Memory::YUZU_PAGESIZE;
}

bool MemoryManager::IsContinousRange(GPUVAddr gpu_addr, std::size_t size) const {
    size_t page_index{gpu_addr >> page_bits};
    const size_t page_last{(gpu_addr + size + page_size - 1) >> page_bits};
    std::optional<VAddr> old_page_addr{};
    while (page_index != page_last) {
        const auto page_addr{GpuToCpuAddress(page_index << page_bits)};
        if (!page_addr || *page_addr == 0) {
            return false;
        }
        if (old_page_addr) {
            if (*old_page_addr + page_size != *page_addr) {
                return false;
            }
        }
        old_page_addr = page_addr;
        ++page_index;
    }
    return true;
}

bool MemoryManager::IsFullyMappedRange(GPUVAddr gpu_addr, std::size_t size) const {
    size_t page_index{gpu_addr >> page_bits};
    const size_t page_last{(gpu_addr + size + page_size - 1) >> page_bits};
    while (page_index < page_last) {
        if (GetEntry(page_index << page_bits) == EntryType::Free) {
            return false;
        }
        ++page_index;
    }
    return true;
}

std::vector<std::pair<GPUVAddr, std::size_t>> MemoryManager::GetSubmappedRange(
    GPUVAddr gpu_addr, std::size_t size) const {
    std::vector<std::pair<GPUVAddr, std::size_t>> result{};
    size_t page_index{gpu_addr >> page_bits};
    size_t remaining_size{size};
    size_t page_offset{gpu_addr & page_mask};
    std::optional<std::pair<GPUVAddr, std::size_t>> last_segment{};
    std::optional<VAddr> old_page_addr{};
    const auto extend_size = [this, &last_segment, &page_index, &page_offset](std::size_t bytes) {
        if (!last_segment) {
            const GPUVAddr new_base_addr = (page_index << page_bits) + page_offset;
            last_segment = {new_base_addr, bytes};
        } else {
            last_segment->second += bytes;
        }
    };
    const auto split = [&last_segment, &result] {
        if (last_segment) {
            result.push_back(*last_segment);
            last_segment = std::nullopt;
        }
    };
    while (remaining_size > 0) {
        const size_t num_bytes{std::min(page_size - page_offset, remaining_size)};
        const auto page_addr{GpuToCpuAddress(page_index << page_bits)};
        if (!page_addr || *page_addr == 0) {
            split();
        } else if (old_page_addr) {
            if (*old_page_addr + page_size != *page_addr) {
                split();
            }
            extend_size(num_bytes);
        } else {
            extend_size(num_bytes);
        }
        ++page_index;
        page_offset = 0;
        remaining_size -= num_bytes;
        old_page_addr = page_addr;
    }
    split();
    return result;
}

} // namespace Tegra