compat.h File Reference

Cross-platform aligned allocators, popcount, prefetch. More...

#include <stdint.h>
#include <stddef.h>
#include <stdlib.h>

Include dependency graph for compat.h:

This graph shows which files directly or indirectly include this file:

Go to the source code of this file.

Functions
static void	cbits_free_aligned (void *ptr)
	Free aligned memory.
static void *	cbits_malloc_aligned (size_t size, size_t align)
	Allocate aligned memory.
static void	cbits_prefetch (const void *ptr)
	Prefetch a cache line at ptr into L1.
static uint64_t	cbits_popcount64 (uint64_t x)
	Count bits set in a 64-bit word.
uint64_t	cbits_popcount_block_fallback (const uint64_t *ptr)
	Fallback popcount block implementation.
static uint64_t	cbits_popcount_block (const uint64_t *ptr)
	Inline wrapper that calls the current dispatch pointer.
void	init_cpu_dispatch (void)
	Constructor to initialize popcount dispatch pointer.

Variables
uint64_t(*	cbits_popcount_block_ptr )(const uint64_t *ptr)
	Dispatch pointer for block popcount.

Detailed Description

Cross-platform aligned allocators, popcount, prefetch.

Provides wrappers for:

• posix_memalign or _aligned_malloc/free
• cache prefetch instructions
• optimized 64-bit popcount and block-level popcount

Author

lambdaphoenix

Version

0.1.1

Copyright

Copyright (c) 2025 lambdaphoenix

Function Documentation

cbits_free_aligned()

static void cbits_free_aligned ( void *  ptr )

inlinestatic

Free aligned memory.

On MSVC uses _aligned_free, otherwise standard free.

Parameters

ptr	Pointer returned by cbits_malloc_aligned.

cbits_malloc_aligned()

static void * cbits_malloc_aligned ( size_t  size, size_t  align  )

inlinestatic

Allocate aligned memory.

Uses posix_memalign on POSIX, or _aligned_malloc on MSVC.

Parameters

size	Number of bytes to allocate.
align	Desired alignment in bytes (must be power of two).

Returns

Pointer to aligned memory, or NULL if allocation failed.

cbits_popcount64()

static uint64_t cbits_popcount64 ( uint64_t  x )

inlinestatic

Count bits set in a 64-bit word.

Parameters

ptr	Pointer to the uint64_t to count bits in.

Returns

Number of set bits in *ptr.

cbits_popcount_block()

static uint64_t cbits_popcount_block ( const uint64_t *  ptr )

inlinestatic

Inline wrapper that calls the current dispatch pointer.

Parameters

ptr	Pointer to 8 contiguous uint64_t words.

Returns

Total popcount as computed by the best available impl.

cbits_popcount_block_fallback()

uint64_t cbits_popcount_block_fallback

(

const uint64_t *

ptr

)

Fallback popcount block implementation.

Processes in 64-bit chunks, summing up cbits_popcount64 for each of 8 words. Used when no vector instructions are available.

Parameters

ptr	Pointer to at least 8 uint64_t words.

Returns

Total popcount of the 8 words.

cbits_prefetch()

static void cbits_prefetch ( const void *  ptr )

inlinestatic

Prefetch a cache line at ptr into L1.

Parameters

ptr	Address to prefetch.

init_cpu_dispatch()

void init_cpu_dispatch

(

void

)

Constructor to initialize popcount dispatch pointer.

At program start, this function checks CPU support for AVX-512VPOPCNTDQ and AVX2 via __builtin_cpu_supports, then updates cbits_popcount_block_ptr accordingly.

Parameters

void

Variable Documentation

cbits_popcount_block_ptr

uint64_t(* cbits_popcount_block_ptr) (const uint64_t *ptr) ( const uint64_t *  ptr )

extern

Dispatch pointer for block popcount.

Initially set to cbits_popcount_block_fallback, and overwritten during module initialization by init_cpu_dispatch to point to the best available implementation.