yuzu/src/core/hle/kernel/k_page_bitmap.h

346 lines
11 KiB
C++

// SPDX-FileCopyrightText: Copyright 2021 yuzu Emulator Project
// SPDX-License-Identifier: GPL-2.0-or-later
#pragma once
#include <array>
#include <bit>
#include "common/alignment.h"
#include "common/assert.h"
#include "common/bit_util.h"
#include "common/common_types.h"
#include "common/tiny_mt.h"
#include "core/hle/kernel/k_system_control.h"
namespace Kernel {
class KPageBitmap {
public:
class RandomBitGenerator {
public:
RandomBitGenerator() {
m_rng.Initialize(static_cast<u32>(KSystemControl::GenerateRandomU64()));
}
u64 SelectRandomBit(u64 bitmap) {
u64 selected = 0;
for (size_t cur_num_bits = Common::BitSize<decltype(bitmap)>() / 2; cur_num_bits != 0;
cur_num_bits /= 2) {
const u64 high = (bitmap >> cur_num_bits);
const u64 low = (bitmap & (~(UINT64_C(0xFFFFFFFFFFFFFFFF) << cur_num_bits)));
// Choose high if we have high and (don't have low or select high randomly).
if (high && (low == 0 || this->GenerateRandomBit())) {
bitmap = high;
selected += cur_num_bits;
} else {
bitmap = low;
selected += 0;
}
}
return selected;
}
u64 GenerateRandom(u64 max) {
// Determine the number of bits we need.
const u64 bits_needed = 1 + (Common::BitSize<decltype(max)>() - std::countl_zero(max));
// Generate a random value of the desired bitwidth.
const u64 rnd = this->GenerateRandomBits(static_cast<u32>(bits_needed));
// Adjust the value to be in range.
return rnd - ((rnd / max) * max);
}
private:
void RefreshEntropy() {
m_entropy = m_rng.GenerateRandomU32();
m_bits_available = static_cast<u32>(Common::BitSize<decltype(m_entropy)>());
}
bool GenerateRandomBit() {
if (m_bits_available == 0) {
this->RefreshEntropy();
}
const bool rnd_bit = (m_entropy & 1) != 0;
m_entropy >>= 1;
--m_bits_available;
return rnd_bit;
}
u64 GenerateRandomBits(u32 num_bits) {
u64 result = 0;
// Iteratively add random bits to our result.
while (num_bits > 0) {
// Ensure we have random bits to take from.
if (m_bits_available == 0) {
this->RefreshEntropy();
}
// Determine how many bits to take this round.
const auto cur_bits = std::min(num_bits, m_bits_available);
// Generate mask for our current bits.
const u64 mask = (static_cast<u64>(1) << cur_bits) - 1;
// Add bits to output from our entropy.
result <<= cur_bits;
result |= (m_entropy & mask);
// Remove bits from our entropy.
m_entropy >>= cur_bits;
m_bits_available -= cur_bits;
// Advance.
num_bits -= cur_bits;
}
return result;
}
private:
Common::TinyMT m_rng;
u32 m_entropy{};
u32 m_bits_available{};
};
public:
static constexpr size_t MaxDepth = 4;
public:
KPageBitmap() = default;
constexpr size_t GetNumBits() const {
return m_num_bits;
}
constexpr s32 GetHighestDepthIndex() const {
return static_cast<s32>(m_used_depths) - 1;
}
u64* Initialize(u64* storage, size_t size) {
// Initially, everything is un-set.
m_num_bits = 0;
// Calculate the needed bitmap depth.
m_used_depths = static_cast<size_t>(GetRequiredDepth(size));
ASSERT(m_used_depths <= MaxDepth);
// Set the bitmap pointers.
for (s32 depth = this->GetHighestDepthIndex(); depth >= 0; depth--) {
m_bit_storages[depth] = storage;
size = Common::AlignUp(size, Common::BitSize<u64>()) / Common::BitSize<u64>();
storage += size;
m_end_storages[depth] = storage;
}
return storage;
}
s64 FindFreeBlock(bool random) {
uintptr_t offset = 0;
s32 depth = 0;
if (random) {
do {
const u64 v = m_bit_storages[depth][offset];
if (v == 0) {
// If depth is bigger than zero, then a previous level indicated a block was
// free.
ASSERT(depth == 0);
return -1;
}
offset = offset * Common::BitSize<u64>() + m_rng.SelectRandomBit(v);
++depth;
} while (depth < static_cast<s32>(m_used_depths));
} else {
do {
const u64 v = m_bit_storages[depth][offset];
if (v == 0) {
// If depth is bigger than zero, then a previous level indicated a block was
// free.
ASSERT(depth == 0);
return -1;
}
offset = offset * Common::BitSize<u64>() + std::countr_zero(v);
++depth;
} while (depth < static_cast<s32>(m_used_depths));
}
return static_cast<s64>(offset);
}
s64 FindFreeRange(size_t count) {
// Check that it is possible to find a range.
const u64* const storage_start = m_bit_storages[m_used_depths - 1];
const u64* const storage_end = m_end_storages[m_used_depths - 1];
// If we don't have a storage to iterate (or want more blocks than fit in a single storage),
// we can't find a free range.
if (!(storage_start < storage_end && count <= Common::BitSize<u64>())) {
return -1;
}
// Walk the storages to select a random free range.
const size_t options_per_storage = std::max<size_t>(Common::BitSize<u64>() / count, 1);
const size_t num_entries = std::max<size_t>(storage_end - storage_start, 1);
const u64 free_mask = (static_cast<u64>(1) << count) - 1;
size_t num_valid_options = 0;
s64 chosen_offset = -1;
for (size_t storage_index = 0; storage_index < num_entries; ++storage_index) {
u64 storage = storage_start[storage_index];
for (size_t option = 0; option < options_per_storage; ++option) {
if ((storage & free_mask) == free_mask) {
// We've found a new valid option.
++num_valid_options;
// Select the Kth valid option with probability 1/K. This leads to an overall
// uniform distribution.
if (num_valid_options == 1 || m_rng.GenerateRandom(num_valid_options) == 0) {
// This is our first option, so select it.
chosen_offset = storage_index * Common::BitSize<u64>() + option * count;
}
}
storage >>= count;
}
}
// Return the random offset we chose.*/
return chosen_offset;
}
void SetBit(size_t offset) {
this->SetBit(this->GetHighestDepthIndex(), offset);
m_num_bits++;
}
void ClearBit(size_t offset) {
this->ClearBit(this->GetHighestDepthIndex(), offset);
m_num_bits--;
}
bool ClearRange(size_t offset, size_t count) {
s32 depth = this->GetHighestDepthIndex();
u64* bits = m_bit_storages[depth];
size_t bit_ind = offset / Common::BitSize<u64>();
if (count < Common::BitSize<u64>()) [[likely]] {
const size_t shift = offset % Common::BitSize<u64>();
ASSERT(shift + count <= Common::BitSize<u64>());
// Check that all the bits are set.
const u64 mask = ((u64(1) << count) - 1) << shift;
u64 v = bits[bit_ind];
if ((v & mask) != mask) {
return false;
}
// Clear the bits.
v &= ~mask;
bits[bit_ind] = v;
if (v == 0) {
this->ClearBit(depth - 1, bit_ind);
}
} else {
ASSERT(offset % Common::BitSize<u64>() == 0);
ASSERT(count % Common::BitSize<u64>() == 0);
// Check that all the bits are set.
size_t remaining = count;
size_t i = 0;
do {
if (bits[bit_ind + i++] != ~u64(0)) {
return false;
}
remaining -= Common::BitSize<u64>();
} while (remaining > 0);
// Clear the bits.
remaining = count;
i = 0;
do {
bits[bit_ind + i] = 0;
this->ClearBit(depth - 1, bit_ind + i);
i++;
remaining -= Common::BitSize<u64>();
} while (remaining > 0);
}
m_num_bits -= count;
return true;
}
private:
void SetBit(s32 depth, size_t offset) {
while (depth >= 0) {
size_t ind = offset / Common::BitSize<u64>();
size_t which = offset % Common::BitSize<u64>();
const u64 mask = u64(1) << which;
u64* bit = std::addressof(m_bit_storages[depth][ind]);
u64 v = *bit;
ASSERT((v & mask) == 0);
*bit = v | mask;
if (v) {
break;
}
offset = ind;
depth--;
}
}
void ClearBit(s32 depth, size_t offset) {
while (depth >= 0) {
size_t ind = offset / Common::BitSize<u64>();
size_t which = offset % Common::BitSize<u64>();
const u64 mask = u64(1) << which;
u64* bit = std::addressof(m_bit_storages[depth][ind]);
u64 v = *bit;
ASSERT((v & mask) != 0);
v &= ~mask;
*bit = v;
if (v) {
break;
}
offset = ind;
depth--;
}
}
private:
static constexpr s32 GetRequiredDepth(size_t region_size) {
s32 depth = 0;
while (true) {
region_size /= Common::BitSize<u64>();
depth++;
if (region_size == 0) {
return depth;
}
}
}
public:
static constexpr size_t CalculateManagementOverheadSize(size_t region_size) {
size_t overhead_bits = 0;
for (s32 depth = GetRequiredDepth(region_size) - 1; depth >= 0; depth--) {
region_size =
Common::AlignUp(region_size, Common::BitSize<u64>()) / Common::BitSize<u64>();
overhead_bits += region_size;
}
return overhead_bits * sizeof(u64);
}
private:
std::array<u64*, MaxDepth> m_bit_storages{};
std::array<u64*, MaxDepth> m_end_storages{};
RandomBitGenerator m_rng;
size_t m_num_bits{};
size_t m_used_depths{};
};
} // namespace Kernel