zig/lib/std/special/compiler_rt/atomics.zig
2021-10-04 23:48:55 -07:00

344 lines
14 KiB
Zig

const std = @import("std");
const builtin = @import("builtin");
const arch = builtin.cpu.arch;
const linkage: std.builtin.GlobalLinkage = if (builtin.is_test) .Internal else .Weak;
// This parameter is true iff the target architecture supports the bare minimum
// to implement the atomic load/store intrinsics.
// Some architectures support atomic load/stores but no CAS, but we ignore this
// detail to keep the export logic clean and because we need some kind of CAS to
// implement the spinlocks.
const supports_atomic_ops = switch (arch) {
.msp430, .avr => false,
.arm, .armeb, .thumb, .thumbeb =>
// The ARM v6m ISA has no ldrex/strex and so it's impossible to do CAS
// operations (unless we're targeting Linux, the kernel provides a way to
// perform CAS operations).
// XXX: The Linux code path is not implemented yet.
!std.Target.arm.featureSetHas(builtin.cpu.features, .has_v6m),
else => true,
};
// The size (in bytes) of the biggest object that the architecture can
// load/store atomically.
// Objects bigger than this threshold require the use of a lock.
const largest_atomic_size = switch (arch) {
// XXX: On x86/x86_64 we could check the presence of cmpxchg8b/cmpxchg16b
// and set this parameter accordingly.
else => @sizeOf(usize),
};
const cache_line_size = 64;
const SpinlockTable = struct {
// Allocate ~4096 bytes of memory for the spinlock table
const max_spinlocks = 64;
const Spinlock = struct {
// Prevent false sharing by providing enough padding between two
// consecutive spinlock elements
v: enum(usize) { Unlocked = 0, Locked } align(cache_line_size) = .Unlocked,
fn acquire(self: *@This()) void {
while (true) {
switch (@atomicRmw(@TypeOf(self.v), &self.v, .Xchg, .Locked, .Acquire)) {
.Unlocked => break,
.Locked => {},
}
}
}
fn release(self: *@This()) void {
@atomicStore(@TypeOf(self.v), &self.v, .Unlocked, .Release);
}
};
list: [max_spinlocks]Spinlock = [_]Spinlock{.{}} ** max_spinlocks,
// The spinlock table behaves as a really simple hash table, mapping
// addresses to spinlocks. The mapping is not unique but that's only a
// performance problem as the lock will be contended by more than a pair of
// threads.
fn get(self: *@This(), address: usize) *Spinlock {
var sl = &self.list[(address >> 3) % max_spinlocks];
sl.acquire();
return sl;
}
};
var spinlocks: SpinlockTable = SpinlockTable{};
// The following builtins do not respect the specified memory model and instead
// uses seq_cst, the strongest one, for simplicity sake.
// Generic version of GCC atomic builtin functions.
// Those work on any object no matter the pointer alignment nor its size.
fn __atomic_load(size: u32, src: [*]u8, dest: [*]u8, model: i32) callconv(.C) void {
_ = model;
var sl = spinlocks.get(@ptrToInt(src));
defer sl.release();
@memcpy(dest, src, size);
}
fn __atomic_store(size: u32, dest: [*]u8, src: [*]u8, model: i32) callconv(.C) void {
_ = model;
var sl = spinlocks.get(@ptrToInt(dest));
defer sl.release();
@memcpy(dest, src, size);
}
fn __atomic_exchange(size: u32, ptr: [*]u8, val: [*]u8, old: [*]u8, model: i32) callconv(.C) void {
_ = model;
var sl = spinlocks.get(@ptrToInt(ptr));
defer sl.release();
@memcpy(old, ptr, size);
@memcpy(ptr, val, size);
}
fn __atomic_compare_exchange(
size: u32,
ptr: [*]u8,
expected: [*]u8,
desired: [*]u8,
success: i32,
failure: i32,
) callconv(.C) i32 {
_ = success;
_ = failure;
var sl = spinlocks.get(@ptrToInt(ptr));
defer sl.release();
for (ptr[0..size]) |b, i| {
if (expected[i] != b) break;
} else {
// The two objects, ptr and expected, are equal
@memcpy(ptr, desired, size);
return 1;
}
@memcpy(expected, ptr, size);
return 0;
}
comptime {
if (supports_atomic_ops) {
@export(__atomic_load, .{ .name = "__atomic_load", .linkage = linkage });
@export(__atomic_store, .{ .name = "__atomic_store", .linkage = linkage });
@export(__atomic_exchange, .{ .name = "__atomic_exchange", .linkage = linkage });
@export(__atomic_compare_exchange, .{ .name = "__atomic_compare_exchange", .linkage = linkage });
}
}
// Specialized versions of the GCC atomic builtin functions.
// LLVM emits those iff the object size is known and the pointers are correctly
// aligned.
fn atomicLoadFn(comptime T: type) fn (*T, i32) callconv(.C) T {
return struct {
fn atomic_load_N(src: *T, model: i32) callconv(.C) T {
_ = model;
if (@sizeOf(T) > largest_atomic_size) {
var sl = spinlocks.get(@ptrToInt(src));
defer sl.release();
return src.*;
} else {
return @atomicLoad(T, src, .SeqCst);
}
}
}.atomic_load_N;
}
comptime {
if (supports_atomic_ops) {
const atomicLoad_u8 = atomicLoadFn(u8);
const atomicLoad_u16 = atomicLoadFn(u16);
const atomicLoad_u32 = atomicLoadFn(u32);
const atomicLoad_u64 = atomicLoadFn(u64);
@export(atomicLoad_u8, .{ .name = "__atomic_load_1", .linkage = linkage });
@export(atomicLoad_u16, .{ .name = "__atomic_load_2", .linkage = linkage });
@export(atomicLoad_u32, .{ .name = "__atomic_load_4", .linkage = linkage });
@export(atomicLoad_u64, .{ .name = "__atomic_load_8", .linkage = linkage });
}
}
fn atomicStoreFn(comptime T: type) fn (*T, T, i32) callconv(.C) void {
return struct {
fn atomic_store_N(dst: *T, value: T, model: i32) callconv(.C) void {
_ = model;
if (@sizeOf(T) > largest_atomic_size) {
var sl = spinlocks.get(@ptrToInt(dst));
defer sl.release();
dst.* = value;
} else {
@atomicStore(T, dst, value, .SeqCst);
}
}
}.atomic_store_N;
}
comptime {
if (supports_atomic_ops) {
const atomicStore_u8 = atomicStoreFn(u8);
const atomicStore_u16 = atomicStoreFn(u16);
const atomicStore_u32 = atomicStoreFn(u32);
const atomicStore_u64 = atomicStoreFn(u64);
@export(atomicStore_u8, .{ .name = "__atomic_store_1", .linkage = linkage });
@export(atomicStore_u16, .{ .name = "__atomic_store_2", .linkage = linkage });
@export(atomicStore_u32, .{ .name = "__atomic_store_4", .linkage = linkage });
@export(atomicStore_u64, .{ .name = "__atomic_store_8", .linkage = linkage });
}
}
fn atomicExchangeFn(comptime T: type) fn (*T, T, i32) callconv(.C) T {
return struct {
fn atomic_exchange_N(ptr: *T, val: T, model: i32) callconv(.C) T {
_ = model;
if (@sizeOf(T) > largest_atomic_size) {
var sl = spinlocks.get(@ptrToInt(ptr));
defer sl.release();
const value = ptr.*;
ptr.* = val;
return value;
} else {
return @atomicRmw(T, ptr, .Xchg, val, .SeqCst);
}
}
}.atomic_exchange_N;
}
comptime {
if (supports_atomic_ops) {
const atomicExchange_u8 = atomicExchangeFn(u8);
const atomicExchange_u16 = atomicExchangeFn(u16);
const atomicExchange_u32 = atomicExchangeFn(u32);
const atomicExchange_u64 = atomicExchangeFn(u64);
@export(atomicExchange_u8, .{ .name = "__atomic_exchange_1", .linkage = linkage });
@export(atomicExchange_u16, .{ .name = "__atomic_exchange_2", .linkage = linkage });
@export(atomicExchange_u32, .{ .name = "__atomic_exchange_4", .linkage = linkage });
@export(atomicExchange_u64, .{ .name = "__atomic_exchange_8", .linkage = linkage });
}
}
fn atomicCompareExchangeFn(comptime T: type) fn (*T, *T, T, i32, i32) callconv(.C) i32 {
return struct {
fn atomic_compare_exchange_N(ptr: *T, expected: *T, desired: T, success: i32, failure: i32) callconv(.C) i32 {
_ = success;
_ = failure;
if (@sizeOf(T) > largest_atomic_size) {
var sl = spinlocks.get(@ptrToInt(ptr));
defer sl.release();
const value = ptr.*;
if (value == expected.*) {
ptr.* = desired;
return 1;
}
expected.* = value;
return 0;
} else {
if (@cmpxchgStrong(T, ptr, expected.*, desired, .SeqCst, .SeqCst)) |old_value| {
expected.* = old_value;
return 0;
}
return 1;
}
}
}.atomic_compare_exchange_N;
}
comptime {
if (supports_atomic_ops) {
const atomicCompareExchange_u8 = atomicCompareExchangeFn(u8);
const atomicCompareExchange_u16 = atomicCompareExchangeFn(u16);
const atomicCompareExchange_u32 = atomicCompareExchangeFn(u32);
const atomicCompareExchange_u64 = atomicCompareExchangeFn(u64);
@export(atomicCompareExchange_u8, .{ .name = "__atomic_compare_exchange_1", .linkage = linkage });
@export(atomicCompareExchange_u16, .{ .name = "__atomic_compare_exchange_2", .linkage = linkage });
@export(atomicCompareExchange_u32, .{ .name = "__atomic_compare_exchange_4", .linkage = linkage });
@export(atomicCompareExchange_u64, .{ .name = "__atomic_compare_exchange_8", .linkage = linkage });
}
}
fn fetchFn(comptime T: type, comptime op: std.builtin.AtomicRmwOp) fn (*T, T, i32) callconv(.C) T {
return struct {
pub fn fetch_op_N(ptr: *T, val: T, model: i32) callconv(.C) T {
_ = model;
if (@sizeOf(T) > largest_atomic_size) {
var sl = spinlocks.get(@ptrToInt(ptr));
defer sl.release();
const value = ptr.*;
ptr.* = switch (op) {
.Add => value +% val,
.Sub => value -% val,
.And => value & val,
.Nand => ~(value & val),
.Or => value | val,
.Xor => value ^ val,
else => @compileError("unsupported atomic op"),
};
return value;
}
return @atomicRmw(T, ptr, op, val, .SeqCst);
}
}.fetch_op_N;
}
comptime {
if (supports_atomic_ops) {
const fetch_add_u8 = fetchFn(u8, .Add);
const fetch_add_u16 = fetchFn(u16, .Add);
const fetch_add_u32 = fetchFn(u32, .Add);
const fetch_add_u64 = fetchFn(u64, .Add);
@export(fetch_add_u8, .{ .name = "__atomic_fetch_add_1", .linkage = linkage });
@export(fetch_add_u16, .{ .name = "__atomic_fetch_add_2", .linkage = linkage });
@export(fetch_add_u32, .{ .name = "__atomic_fetch_add_4", .linkage = linkage });
@export(fetch_add_u64, .{ .name = "__atomic_fetch_add_8", .linkage = linkage });
const fetch_sub_u8 = fetchFn(u8, .Sub);
const fetch_sub_u16 = fetchFn(u16, .Sub);
const fetch_sub_u32 = fetchFn(u32, .Sub);
const fetch_sub_u64 = fetchFn(u64, .Sub);
@export(fetch_sub_u8, .{ .name = "__atomic_fetch_sub_1", .linkage = linkage });
@export(fetch_sub_u16, .{ .name = "__atomic_fetch_sub_2", .linkage = linkage });
@export(fetch_sub_u32, .{ .name = "__atomic_fetch_sub_4", .linkage = linkage });
@export(fetch_sub_u64, .{ .name = "__atomic_fetch_sub_8", .linkage = linkage });
const fetch_and_u8 = fetchFn(u8, .And);
const fetch_and_u16 = fetchFn(u16, .And);
const fetch_and_u32 = fetchFn(u32, .And);
const fetch_and_u64 = fetchFn(u64, .And);
@export(fetch_and_u8, .{ .name = "__atomic_fetch_and_1", .linkage = linkage });
@export(fetch_and_u16, .{ .name = "__atomic_fetch_and_2", .linkage = linkage });
@export(fetch_and_u32, .{ .name = "__atomic_fetch_and_4", .linkage = linkage });
@export(fetch_and_u64, .{ .name = "__atomic_fetch_and_8", .linkage = linkage });
const fetch_or_u8 = fetchFn(u8, .Or);
const fetch_or_u16 = fetchFn(u16, .Or);
const fetch_or_u32 = fetchFn(u32, .Or);
const fetch_or_u64 = fetchFn(u64, .Or);
@export(fetch_or_u8, .{ .name = "__atomic_fetch_or_1", .linkage = linkage });
@export(fetch_or_u16, .{ .name = "__atomic_fetch_or_2", .linkage = linkage });
@export(fetch_or_u32, .{ .name = "__atomic_fetch_or_4", .linkage = linkage });
@export(fetch_or_u64, .{ .name = "__atomic_fetch_or_8", .linkage = linkage });
const fetch_xor_u8 = fetchFn(u8, .Xor);
const fetch_xor_u16 = fetchFn(u16, .Xor);
const fetch_xor_u32 = fetchFn(u32, .Xor);
const fetch_xor_u64 = fetchFn(u64, .Xor);
@export(fetch_xor_u8, .{ .name = "__atomic_fetch_xor_1", .linkage = linkage });
@export(fetch_xor_u16, .{ .name = "__atomic_fetch_xor_2", .linkage = linkage });
@export(fetch_xor_u32, .{ .name = "__atomic_fetch_xor_4", .linkage = linkage });
@export(fetch_xor_u64, .{ .name = "__atomic_fetch_xor_8", .linkage = linkage });
const fetch_nand_u8 = fetchFn(u8, .Nand);
const fetch_nand_u16 = fetchFn(u16, .Nand);
const fetch_nand_u32 = fetchFn(u32, .Nand);
const fetch_nand_u64 = fetchFn(u64, .Nand);
@export(fetch_nand_u8, .{ .name = "__atomic_fetch_nand_1", .linkage = linkage });
@export(fetch_nand_u16, .{ .name = "__atomic_fetch_nand_2", .linkage = linkage });
@export(fetch_nand_u32, .{ .name = "__atomic_fetch_nand_4", .linkage = linkage });
@export(fetch_nand_u64, .{ .name = "__atomic_fetch_nand_8", .linkage = linkage });
}
}