Deduplicating atomics written in terms of CAS

[Rich Felker <dalias@libc.org>]

[-- Attachment #1: Type: text/plain, Size: 1293 bytes --]

Lots of archs define most or all of their atomics except a_cas in
terms of a_cas. The attached atomic.h is a proposed replacement for
arch-specific atomic.h that would live in src/internal. The arch
atomic.h files would be replaced with atomic_arch.h, which could opt
to define nothing but a_cas, or could define more primitives itself if
it can do so more efficiently.

The second attachment, atomic_generic.h, is an implementation of the
atomics (and other non-atomic ops we've traditionally had in atomic.h)
using GNU C builtins. This file can be used as-is for any new archs
that satisfy the following conditions:

- They're not supported by compilers too old to have the __sync_*
  builtins.

- They don't need runtime switching/detection of atomic
  implementations.

- GCC doesn't generate pathologically bad code for the builtins.

Even if some of these conditions aren't met, atomic_generic.h may be
useful when porting to help determine the right way to implement
atomics by watching what the compiler does, and it may be suitable for
early porting stages before a proper atomic.h is written.

I'm not sure if I'll try to integrate this stuff right away or as part
of the bits deduplication and build system overhaul that's been
coming-soon for a long time.

Comments welcome.

Rich

[-- Attachment #2: atomic.h --]
[-- Type: text/plain, Size: 3009 bytes --]

#ifndef _ATOMIC_H
#define _ATOMIC_H

#include "atomic_arch.h"

#ifndef a_swap
static inline int a_swap(volatile int *p, int v)
{
	int old;
	do old = *p;
	while (a_cas(p, old, v) != old);
	return old;
}
#endif

#ifndef a_fetch_add
static inline int a_fetch_add(volatile int *p, int v)
{
	int old;
	do old = *p;
	while (a_cas(p, old, old+v) != old);
	return old;
}
#endif

#ifndef a_and
static inline void a_and(volatile int *p, int v)
{
	int old;
	do old = *p;
	while (a_cas(p, old, old&v) != old);
}
#endif

#ifndef a_or
static inline void a_or(volatile int *p, int v)
{
	int old;
	do old = *p;
	while (a_cas(p, old, old|v) != old);
}
#endif

#ifndef a_inc
static inline void a_inc(volatile int *p)
{
	a_fetch_add(p, 1);
}
#endif

#ifndef a_dec
static inline void a_dec(volatile int *p)
{
	a_fetch_add(p, -1);
}
#endif

#ifndef a_store
static inline void a_store(volatile int *p, int v)
{
#ifdef a_barrier
	a_barrier();
	*p = v;
	a_barrier();
#else
	a_swap(p, v);
#endif
}
#endif

#ifndef a_barrier
static void a_barrier()
{
	volatile int tmp = 0;
	a_cas(&tmp, 0, 0);
}
#endif

#ifndef a_spin
#define a_spin a_barrier
#endif

#ifndef a_and_64
static inline void a_and_64(volatile uint64_t *p, uint64_t v)
{
	union { uint64_t v; uint32_t r[2]; } u = { v };
	if (u.r[0]+1) a_and((int *)p, u.r[0]);
	if (u.r[1]+1) a_and((int *)p+1, u.r[1]);
}
#endif

#ifndef a_or_64
static inline void a_or_64(volatile uint64_t *p, uint64_t v)
{
	union { uint64_t v; uint32_t r[2]; } u = { v };
	if (u.r[0]) a_or((int *)p, u.r[0]);
	if (u.r[1]) a_or((int *)p+1, u.r[1]);
}
#endif

#ifndef a_cas_p
static inline void *a_cas_p(volatile void *p, void *t, void *s)
{
	return (void *)a_cas(p, (int)t, (int)s);
}
#endif

#ifndef a_or_l
static inline void a_or_l(volatile void *p, long v)
{
	if (sizeof(long) == sizeof(int)) a_or(p, v);
	else a_or_64(p, v);
}
#endif

#ifndef a_crash
static inline void a_crash()
{
	*(volatile char *)0=0;
}
#endif

#ifndef a_ctz_64
static inline int a_ctz_64(uint64_t x)
{
	static const char debruijn64[64] = {
		0, 1, 2, 53, 3, 7, 54, 27, 4, 38, 41, 8, 34, 55, 48, 28,
		62, 5, 39, 46, 44, 42, 22, 9, 24, 35, 59, 56, 49, 18, 29, 11,
		63, 52, 6, 26, 37, 40, 33, 47, 61, 45, 43, 21, 23, 58, 17, 10,
		51, 25, 36, 32, 60, 20, 57, 16, 50, 31, 19, 15, 30, 14, 13, 12
	};
	static const char debruijn32[32] = {
		0, 1, 23, 2, 29, 24, 19, 3, 30, 27, 25, 11, 20, 8, 4, 13,
		31, 22, 28, 18, 26, 10, 7, 12, 21, 17, 9, 6, 16, 5, 15, 14
	};
	if (sizeof(long) < 8) {
		uint32_t y = x;
		if (!y) {
			y = x>>32;
			return 32 + debruijn32[(y&-y)*0x076be629 >> 27];
		}
		return debruijn32[(y&-y)*0x076be629 >> 27];
	}
	return debruijn64[(x&-x)*0x022fdd63cc95386dull >> 58];
}
#endif

#ifndef a_ctz_l
static inline int a_ctz_l(unsigned long x)
{
	static const char debruijn32[32] = {
		0, 1, 23, 2, 29, 24, 19, 3, 30, 27, 25, 11, 20, 8, 4, 13,
		31, 22, 28, 18, 26, 10, 7, 12, 21, 17, 9, 6, 16, 5, 15, 14
	};
	if (sizeof(long) == 8) return a_ctz_64(x);
	return debruijn32[(x&-x)*0x076be629 >> 27];
}
#endif

#endif

[-- Attachment #3: atomic_generic.h --]
[-- Type: text/plain, Size: 1063 bytes --]

static inline int a_cas(volatile int *p, int t, int s)
{
	return __sync_val_compare_and_swap(p, t, s);
}

static inline void *a_cas_p(volatile void **p, void *t, void *s)
{
	return __sync_val_compare_and_swap(p, t, s);
}

static inline int a_swap(volatile int *p, int v)
{
	int old;
	do old = *p;
	while (!__sync_bool_compare_and_swap(p, old, v));
	return old;
}

static inline int a_fetch_add(volatile int *p, int v)
{
	return __sync_fetch_and_add(p, v);
}

static inline void a_and(volatile int *p, int v)
{
	__sync_fetch_and_and(p, v);
}

static inline void a_or(volatile int *p, int v)
{
	__sync_fetch_and_or(p, v);
}

static void a_barrier()
{
	__sync_synchronize();
}

static inline void a_crash()
{
	__builtin_trap();
}

static inline int a_ctz_l(unsigned long x)
{
	return __builtin_ctzl(x);
}

static inline int a_ctz_64(uint64_t x)
{
	return __builtin_ctzll(x);
}

#define a_cas a_cas
#define a_cas_p a_cas_p
#define a_swap a_swap
#define a_fetch_add
#define a_and
#define a_or
#define a_barrier
#define a_crash
#define a_ctz_l
#define a_ctz_64

#endif