blender/blender
163
517
Fork 786
Code Issues 6.8k Pull Requests 940 Projects 19 Wiki Activity

BLI: use larger integer type in BitVector #104658

Merged
Jacques Lucke merged 4 commits from JacquesLucke/blender:bit-vector-refactor into main 2023-02-12 18:00:58 +01:00
Conversation 1 Commits 4 Files Changed 1 +85 -73
1 changed files with 85 additions and 73 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

158
source/blender/blenlib/BLI_bit_vector.hh
View File

@ -16,13 +16,13 @@
*
* The compact nature of storing bools in individual bits has some downsides that have to be kept
* in mind:
* - Writing to separate bits in the same byte is not thread-safe. Therefore, an existing vector of
* - Writing to separate bits in the same int is not thread-safe. Therefore, an existing vector of
* bool can't easily be replaced with a bit vector, if it is written to from multiple threads.
* Read-only access from multiple threads is fine though.
* - Writing individual elements is more expensive when the array is in cache already. That is
* because changing a bit is always a read-modify-write operation on the byte the bit resides in.
* because changing a bit is always a read-modify-write operation on the int the bit resides in.
* - Reading individual elements is more expensive when the array is in cache already. That is
* because additional bit-wise operations have to be applied after the corresponding byte is
* because additional bit-wise operations have to be applied after the corresponding int is
* read.
*
* Comparison to `std::vector<bool>`:
@ -42,18 +42,27 @@
#include "BLI_memory_utils.hh"
#include "BLI_span.hh"
namespace blender {
namespace blender::bits {
/**
* This is a read-only pointer to a specific bit. The value of the bit can be retrieved, but not
* changed.
* Using a large integer type is better because then it's easier to process many bits at once.
*/
using IntType = uint64_t;
static constexpr int64_t BitsPerInt = int64_t(sizeof(IntType) * 8);
static constexpr int64_t BitToIntIndexShift = 3 + (sizeof(IntType) >= 2) + (sizeof(IntType) >= 4) +
(sizeof(IntType) >= 8);
static constexpr IntType BitIndexMask = (IntType(1) << BitToIntIndexShift) - 1;
/**
* This is a read-only pointer to a specific bit. The value of the bit can be retrieved, but
* not changed.
*/
class BitRef {
private:
/** Points to the exact byte that the bit is in. */
const uint8_t *byte_ptr_;
/** Points to the integer that the bit is in. */
const IntType *ptr_;
/** All zeros except for a single one at the bit that is referenced. */
uint8_t mask_;
IntType mask_;
friend class MutableBitRef;
@ -61,13 +70,13 @@ class BitRef {
BitRef() = default;
/**
* Reference a specific bit in a byte array. Note that #byte_ptr does *not* have to point to the
* exact byte the bit is in.
* Reference a specific bit in an array. Note that #ptr does *not* have to point to the
* exact integer the bit is in.
*/
BitRef(const uint8_t *byte_ptr, const int64_t bit_index)
BitRef(const IntType *ptr, const int64_t bit_index)
{
byte_ptr_ = byte_ptr + (bit_index >> 3);
mask_ = 1 << (bit_index & 7);
ptr_ = ptr + (bit_index >> BitToIntIndexShift);
mask_ = IntType(1) << (bit_index & BitIndexMask);
}
/**
@ -75,9 +84,9 @@ class BitRef {
*/
bool test() const
{
const uint8_t byte = *byte_ptr_;
const uint8_t masked_byte = byte & mask_;
return masked_byte != 0;
const IntType value = *ptr_;
const IntType masked_value = value & mask_;
return masked_value != 0;
}
operator bool() const
@ -91,22 +100,22 @@ class BitRef {
*/
class MutableBitRef {
private:
/** Points to the exact byte that the bit is in. */
uint8_t *byte_ptr_;
/** Points to the integer that the bit is in. */
IntType *ptr_;
/** All zeros except for a single one at the bit that is referenced. */
uint8_t mask_;
IntType mask_;
public:
MutableBitRef() = default;
/**
* Reference a specific bit in a byte array. Note that #byte_ptr does *not* have to point to the
* exact byte the bit is in.
* Reference a specific bit in an array. Note that #ptr does *not* have to point to the
* exact int the bit is in.
*/
MutableBitRef(uint8_t *byte_ptr, const int64_t bit_index)
MutableBitRef(IntType *ptr, const int64_t bit_index)
{
byte_ptr_ = byte_ptr + (bit_index >> 3);
mask_ = 1 << uint8_t(bit_index & 7);
ptr_ = ptr + (bit_index >> BitToIntIndexShift);
mask_ = IntType(1) << IntType(bit_index & BitIndexMask);
}
/**
@ -115,7 +124,7 @@ class MutableBitRef {
operator BitRef() const
{
BitRef bit_ref;
bit_ref.byte_ptr_ = byte_ptr_;
bit_ref.ptr_ = ptr_;
bit_ref.mask_ = mask_;
return bit_ref;
}
@ -125,9 +134,9 @@ class MutableBitRef {
*/
bool test() const
{
const uint8_t byte = *byte_ptr_;
const uint8_t masked_byte = byte & mask_;
return masked_byte != 0;
const IntType value = *ptr_;
const IntType masked_value = value & mask_;
return masked_value != 0;
}
operator bool() const
@ -140,7 +149,7 @@ class MutableBitRef {
*/
void set()
{
*byte_ptr_ |= mask_;
*ptr_ |= mask_;
}
/**
@ -148,7 +157,7 @@ class MutableBitRef {
*/
void reset()
{
*byte_ptr_ &= ~mask_;
*ptr_ &= ~mask_;
}
/**
@ -177,21 +186,20 @@ template<
typename Allocator = GuardedAllocator>
class BitVector {
private:
static constexpr int64_t required_bytes_for_bits(const int64_t number_of_bits)
static constexpr int64_t required_ints_for_bits(const int64_t number_of_bits)
{
return (number_of_bits + BitsPerByte - 1) / BitsPerByte;
return (number_of_bits + BitsPerInt - 1) / BitsPerInt;
}
static constexpr int64_t BitsPerByte = 8;
static constexpr int64_t BytesInInlineBuffer = required_bytes_for_bits(InlineBufferCapacity);
static constexpr int64_t BitsInInlineBuffer = BytesInInlineBuffer * BitsPerByte;
static constexpr int64_t AllocationAlignment = 8;
static constexpr int64_t IntsInInlineBuffer = required_ints_for_bits(InlineBufferCapacity);
static constexpr int64_t BitsInInlineBuffer = IntsInInlineBuffer * BitsPerInt;
static constexpr int64_t AllocationAlignment = alignof(IntType);
/**
* Points to the first byte used by the vector. It might point to the memory in the inline
* Points to the first integer used by the vector. It might point to the memory in the inline
* buffer.
*/
uint8_t *data_;
IntType *data_;
/** Current size of the vector in bits. */
int64_t size_in_bits_;
@ -203,7 +211,7 @@ class BitVector {
BLI_NO_UNIQUE_ADDRESS Allocator allocator_;
/** Contains the bits as long as the vector is small enough. */
BLI_NO_UNIQUE_ADDRESS TypedBuffer<uint8_t, BytesInInlineBuffer> inline_buffer_;
BLI_NO_UNIQUE_ADDRESS TypedBuffer<IntType, IntsInInlineBuffer> inline_buffer_;
public:
BitVector(Allocator allocator = {}) noexcept : allocator_(allocator)
@ -211,7 +219,7 @@ class BitVector {
data_ = inline_buffer_;
size_in_bits_ = 0;
capacity_in_bits_ = BitsInInlineBuffer;
uninitialized_fill_n(data_, BytesInInlineBuffer, uint8_t(0));
uninitialized_fill_n(data_, IntsInInlineBuffer, IntType(0));
}
BitVector(NoExceptConstructor, Allocator allocator = {}) noexcept : BitVector(allocator)
@ -220,29 +228,29 @@ class BitVector {
BitVector(const BitVector &other) : BitVector(NoExceptConstructor(), other.allocator_)
{
const int64_t bytes_to_copy = other.used_bytes_amount();
const int64_t ints_to_copy = other.used_ints_amount();
if (other.size_in_bits_ <= BitsInInlineBuffer) {
/* The data is copied into the owned inline buffer. */
data_ = inline_buffer_;
capacity_in_bits_ = BitsInInlineBuffer;
}
else {
/* Allocate a new byte array because the inline buffer is too small. */
data_ = static_cast<uint8_t *>(
allocator_.allocate(bytes_to_copy, AllocationAlignment, __func__));
capacity_in_bits_ = bytes_to_copy * BitsPerByte;
/* Allocate a new array because the inline buffer is too small. */
data_ = static_cast<IntType *>(
allocator_.allocate(ints_to_copy * sizeof(IntType), AllocationAlignment, __func__));
capacity_in_bits_ = ints_to_copy * BitsPerInt;
}
size_in_bits_ = other.size_in_bits_;
uninitialized_copy_n(other.data_, bytes_to_copy, data_);
uninitialized_copy_n(other.data_, ints_to_copy, data_);
}
BitVector(BitVector &&other) noexcept : BitVector(NoExceptConstructor(), other.allocator_)
{
if (other.is_inline()) {
/* Copy the data into the inline buffer. */
const int64_t bytes_to_copy = other.used_bytes_amount();
const int64_t ints_to_copy = other.used_ints_amount();
data_ = inline_buffer_;
uninitialized_copy_n(other.data_, bytes_to_copy, data_);
uninitialized_copy_n(other.data_, ints_to_copy, data_);
}
else {
/* Steal the pointer. */
@ -305,7 +313,7 @@ class BitVector {
bool is_empty() const
{
return this->size() == 0;
return size_in_bits_ == 0;
}
/**
@ -442,20 +450,20 @@ class BitVector {
*/
void fill_range(const IndexRange range, const bool value)
{
const AlignedIndexRanges aligned_ranges = split_index_range_by_alignment(range, BitsPerByte);
const AlignedIndexRanges aligned_ranges = split_index_range_by_alignment(range, BitsPerInt);
/* Fill first few bits. */
for (const int64_t i : aligned_ranges.prefix) {
(*this)[i].set(value);
}
/* Fill entire bytes at once. */
const int64_t start_fill_byte_index = aligned_ranges.aligned.start() / BitsPerByte;
const int64_t bytes_to_fill = aligned_ranges.aligned.size() / BitsPerByte;
const uint8_t fill_value = value ? uint8_t(0xff) : uint8_t(0x00);
initialized_fill_n(data_ + start_fill_byte_index, bytes_to_fill, fill_value);
/* Fill entire ints at once. */
const int64_t start_fill_int_index = aligned_ranges.aligned.start() / BitsPerInt;
const int64_t ints_to_fill = aligned_ranges.aligned.size() / BitsPerInt;
const IntType fill_value = value ? IntType(-1) : IntType(0);
initialized_fill_n(data_ + start_fill_int_index, ints_to_fill, fill_value);
/* Fill bits in the end that don't cover a full byte. */
/* Fill bits in the end that don't cover a full int. */
for (const int64_t i : aligned_ranges.suffix) {
(*this)[i].set(value);
}
@ -509,37 +517,35 @@ class BitVector {
}
BLI_NOINLINE void realloc_to_at_least(const int64_t min_capacity_in_bits,
const uint8_t initial_value_for_new_bytes = 0x00)
const IntType initial_value_for_new_ints = 0x00)
{
if (capacity_in_bits_ >= min_capacity_in_bits) {
return;
}
const int64_t min_capacity_in_bytes = this->required_bytes_for_bits(min_capacity_in_bits);
const int64_t min_capacity_in_ints = this->required_ints_for_bits(min_capacity_in_bits);
/* At least double the size of the previous allocation. */
const int64_t min_new_capacity_in_bytes = capacity_in_bits_ * 2;
const int64_t min_new_capacity_in_ints = 2 * this->required_ints_for_bits(capacity_in_bits_);
const int64_t new_capacity_in_bytes = std::max(min_capacity_in_bytes,
min_new_capacity_in_bytes);
const int64_t bytes_to_copy = this->used_bytes_amount();
const int64_t new_capacity_in_ints = std::max(min_capacity_in_ints, min_new_capacity_in_ints);
const int64_t ints_to_copy = this->used_ints_amount();
uint8_t *new_data = static_cast<uint8_t *>(
allocator_.allocate(new_capacity_in_bytes, AllocationAlignment, __func__));
uninitialized_copy_n(data_, bytes_to_copy, new_data);
IntType *new_data = static_cast<IntType *>(allocator_.allocate(
new_capacity_in_ints * sizeof(IntType), AllocationAlignment, __func__));
uninitialized_copy_n(data_, ints_to_copy, new_data);
/* Always initialize new capacity even if it isn't used yet. That's necessary to avoid warnings
* caused by using uninitialized memory. This happens when e.g. setting a clearing a bit in an
* uninitialized byte. */
uninitialized_fill_n(new_data + bytes_to_copy,
new_capacity_in_bytes - bytes_to_copy,
uint8_t(initial_value_for_new_bytes));
* uninitialized int. */
uninitialized_fill_n(
new_data + ints_to_copy, new_capacity_in_ints - ints_to_copy, initial_value_for_new_ints);
if (!this->is_inline()) {
allocator_.deallocate(data_);
}
data_ = new_data;
capacity_in_bits_ = new_capacity_in_bytes * BitsPerByte;
capacity_in_bits_ = new_capacity_in_ints * BitsPerInt;
}
bool is_inline() const
@ -547,10 +553,16 @@ class BitVector {
return data_ == inline_buffer_;
}
int64_t used_bytes_amount() const
int64_t used_ints_amount() const
{
return this->required_bytes_for_bits(size_in_bits_);
return this->required_ints_for_bits(size_in_bits_);
}
};
} // namespace blender::bits
namespace blender {
using bits::BitRef;
using bits::BitVector;
using bits::MutableBitRef;
} // namespace blender