Merge pull request #3830 from fengb/wasm-page-allocator

WasmPageAllocator
This commit is contained in:
Andrew Kelley 2019-12-10 10:58:50 -05:00 committed by GitHub
commit cd4d638d10
No known key found for this signature in database
GPG Key ID: 4AEE18F83AFDEB23

View File

@ -254,79 +254,160 @@ const PageAllocator = struct {
extern fn @"llvm.wasm.memory.size.i32"(u32) u32;
extern fn @"llvm.wasm.memory.grow.i32"(u32, u32) i32;
/// TODO: make this re-use freed pages, and cooperate with other callers of these global intrinsics
/// by better utilizing the return value of grow()
const WasmPageAllocator = struct {
var start_ptr: [*]u8 = undefined;
var num_pages: usize = 0;
var end_index: usize = 0;
comptime {
if (builtin.arch != .wasm32) {
if (!std.Target.current.isWasm()) {
@compileError("WasmPageAllocator is only available for wasm32 arch");
}
}
fn alloc(allocator: *Allocator, size: usize, alignment: u29) ![]u8 {
const addr = @ptrToInt(start_ptr) + end_index;
const adjusted_addr = mem.alignForward(addr, alignment);
const adjusted_index = end_index + (adjusted_addr - addr);
const new_end_index = adjusted_index + size;
const PageStatus = enum(u1) {
used = 0,
free = 1,
if (new_end_index > num_pages * mem.page_size) {
const required_memory = new_end_index - (num_pages * mem.page_size);
pub const none_free: u8 = 0;
};
var inner_num_pages: usize = required_memory / mem.page_size;
if (required_memory % mem.page_size != 0) {
inner_num_pages += 1;
}
const FreeBlock = struct {
data: []u128,
const prev_page = @"llvm.wasm.memory.grow.i32"(0, @intCast(u32, inner_num_pages));
if (prev_page == -1) {
return error.OutOfMemory;
}
const Io = std.packed_int_array.PackedIntIo(u1, .Little);
num_pages += inner_num_pages;
fn totalPages(self: FreeBlock) usize {
return self.data.len * 128;
}
const result = start_ptr[adjusted_index..new_end_index];
end_index = new_end_index;
return result;
}
// Check if memory is the last "item" and is aligned correctly
fn is_last_item(memory: []u8, alignment: u29) bool {
return memory.ptr == start_ptr + end_index - memory.len and mem.alignForward(@ptrToInt(memory.ptr), alignment) == @ptrToInt(memory.ptr);
}
fn realloc(allocator: *Allocator, old_mem: []u8, old_align: u29, new_size: usize, new_align: u29) ![]u8 {
// Initialize start_ptr at the first realloc
if (num_pages == 0) {
start_ptr = @intToPtr([*]u8, @intCast(usize, @"llvm.wasm.memory.size.i32"(0)) * mem.page_size);
fn isInitialized(self: FreeBlock) bool {
return self.data.len > 0;
}
if (is_last_item(old_mem, new_align)) {
const start_index = end_index - old_mem.len;
const new_end_index = start_index + new_size;
fn getBit(self: FreeBlock, idx: usize) PageStatus {
const bit_offset = 0;
return @intToEnum(PageStatus, Io.get(@sliceToBytes(self.data), idx, bit_offset));
}
if (new_end_index > num_pages * mem.page_size) {
_ = try alloc(allocator, new_end_index - end_index, new_align);
fn setBits(self: FreeBlock, start_idx: usize, len: usize, val: PageStatus) void {
const bit_offset = 0;
var i: usize = 0;
while (i < len) : (i += 1) {
Io.set(@sliceToBytes(self.data), start_idx + i, bit_offset, @enumToInt(val));
}
const result = start_ptr[start_index..new_end_index];
}
end_index = new_end_index;
return result;
} else if (new_size <= old_mem.len and new_align <= old_align) {
// Use '0xFFFFFFFF' as a _missing_ sentinel
// This saves ~50 bytes compared to returning a nullable
// We can guarantee that conventional memory never gets this big,
// and wasm32 would not be able to address this memory (32 GB > usize).
// Revisit if this is settled: https://github.com/ziglang/zig/issues/3806
const not_found = std.math.maxInt(usize);
fn useRecycled(self: FreeBlock, num_pages: usize) usize {
@setCold(true);
for (self.data) |segment, i| {
const spills_into_next = @bitCast(i128, segment) < 0;
const has_enough_bits = @popCount(u128, segment) >= num_pages;
if (!spills_into_next and !has_enough_bits) continue;
var j: usize = i * 128;
while (j < (i + 1) * 128) : (j += 1) {
var count: usize = 0;
while (j + count < self.totalPages() and self.getBit(j + count) == .free) {
count += 1;
if (count >= num_pages) {
self.setBits(j, num_pages, .used);
return j;
}
}
j += count;
}
}
return not_found;
}
fn recycle(self: FreeBlock, start_idx: usize, len: usize) void {
self.setBits(start_idx, len, .free);
}
};
var _conventional_data = [_]u128{0} ** 16;
// Marking `conventional` as const saves ~40 bytes
const conventional = FreeBlock{ .data = &_conventional_data };
var extended = FreeBlock{ .data = &[_]u128{} };
fn extendedOffset() usize {
return conventional.totalPages();
}
fn nPages(memsize: usize) usize {
return std.mem.alignForward(memsize, std.mem.page_size) / std.mem.page_size;
}
fn alloc(allocator: *Allocator, page_count: usize, alignment: u29) error{OutOfMemory}!usize {
var idx = conventional.useRecycled(page_count);
if (idx != FreeBlock.not_found) {
return idx;
}
idx = extended.useRecycled(page_count);
if (idx != FreeBlock.not_found) {
return idx + extendedOffset();
}
const prev_page_count = @"llvm.wasm.memory.grow.i32"(0, @intCast(u32, page_count));
if (prev_page_count <= 0) {
return error.OutOfMemory;
}
return @intCast(usize, prev_page_count);
}
pub fn realloc(allocator: *Allocator, old_mem: []u8, old_align: u29, new_size: usize, new_align: u29) Allocator.Error![]u8 {
if (new_align > std.mem.page_size) {
return error.OutOfMemory;
}
if (nPages(new_size) == nPages(old_mem.len)) {
return old_mem.ptr[0..new_size];
} else if (new_size < old_mem.len) {
return shrink(allocator, old_mem, old_align, new_size, new_align);
} else {
const result = try alloc(allocator, new_size, new_align);
mem.copy(u8, result, old_mem);
return result;
const page_idx = try alloc(allocator, nPages(new_size), new_align);
const new_mem = @intToPtr([*]u8, page_idx * std.mem.page_size)[0..new_size];
std.mem.copy(u8, new_mem, old_mem);
_ = shrink(allocator, old_mem, old_align, 0, 0);
return new_mem;
}
}
fn shrink(allocator: *Allocator, old_mem: []u8, old_align: u29, new_size: usize, new_align: u29) []u8 {
pub fn shrink(allocator: *Allocator, old_mem: []u8, old_align: u29, new_size: usize, new_align: u29) []u8 {
@setCold(true);
const free_start = nPages(@ptrToInt(old_mem.ptr) + new_size);
var free_end = nPages(@ptrToInt(old_mem.ptr) + old_mem.len);
if (free_end > free_start) {
if (free_start < extendedOffset()) {
const clamped_end = std.math.min(extendedOffset(), free_end);
conventional.recycle(free_start, clamped_end - free_start);
}
if (free_end > extendedOffset()) {
if (!extended.isInitialized()) {
// Steal the last page from the memory currently being recycled
// TODO: would it be better if we use the first page instead?
free_end -= 1;
extended.data = @intToPtr([*]u128, free_end * std.mem.page_size)[0 .. std.mem.page_size / @sizeOf(u128)];
// Since this is the first page being freed and we consume it, assume *nothing* is free.
std.mem.set(u128, extended.data, PageStatus.none_free);
}
const clamped_start = std.math.max(extendedOffset(), free_start);
extended.recycle(clamped_start - extendedOffset(), free_end - clamped_start);
}
}
return old_mem[0..new_size];
}
};
@ -724,12 +805,51 @@ test "c_allocator" {
}
}
test "WasmPageAllocator internals" {
if (comptime std.Target.current.isWasm()) {
const conventional_memsize = WasmPageAllocator.conventional.totalPages() * std.mem.page_size;
const initial = try page_allocator.alloc(u8, std.mem.page_size);
std.debug.assert(@ptrToInt(initial.ptr) < conventional_memsize); // If this isn't conventional, the rest of these tests don't make sense. Also we have a serious memory leak in the test suite.
var inplace = try page_allocator.realloc(initial, 1);
testing.expectEqual(initial.ptr, inplace.ptr);
inplace = try page_allocator.realloc(inplace, 4);
testing.expectEqual(initial.ptr, inplace.ptr);
page_allocator.free(inplace);
const reuse = try page_allocator.alloc(u8, 1);
testing.expectEqual(initial.ptr, reuse.ptr);
page_allocator.free(reuse);
// This segment may span conventional and extended which has really complex rules so we're just ignoring it for now.
const padding = try page_allocator.alloc(u8, conventional_memsize);
page_allocator.free(padding);
const extended = try page_allocator.alloc(u8, conventional_memsize);
testing.expect(@ptrToInt(extended.ptr) >= conventional_memsize);
const use_small = try page_allocator.alloc(u8, 1);
testing.expectEqual(initial.ptr, use_small.ptr);
page_allocator.free(use_small);
inplace = try page_allocator.realloc(extended, 1);
testing.expectEqual(extended.ptr, inplace.ptr);
page_allocator.free(inplace);
const reuse_extended = try page_allocator.alloc(u8, conventional_memsize);
testing.expectEqual(extended.ptr, reuse_extended.ptr);
page_allocator.free(reuse_extended);
}
}
test "PageAllocator" {
const allocator = page_allocator;
try testAllocator(allocator);
try testAllocatorAligned(allocator, 16);
try testAllocatorLargeAlignment(allocator);
try testAllocatorAlignedShrink(allocator);
if (!std.Target.current.isWasm()) {
try testAllocatorLargeAlignment(allocator);
try testAllocatorAlignedShrink(allocator);
}
if (builtin.os == .windows) {
// Trying really large alignment. As mentionned in the implementation,
@ -766,7 +886,7 @@ test "ArenaAllocator" {
try testAllocatorAlignedShrink(&arena_allocator.allocator);
}
var test_fixed_buffer_allocator_memory: [80000 * @sizeOf(u64)]u8 = undefined;
var test_fixed_buffer_allocator_memory: [800000 * @sizeOf(u64)]u8 = undefined;
test "FixedBufferAllocator" {
var fixed_buffer_allocator = FixedBufferAllocator.init(test_fixed_buffer_allocator_memory[0..]);