introduce std.heap.SmpAllocator

bootstrap.c

@@ -139,7 +139,7 @@ int main(int argc, char **argv) {
             "pub const enable_tracy = false;\n"
             "pub const value_tracing = false;\n"
             "pub const skip_non_native = false;\n"
-            "pub const force_gpa = false;\n"
+            "pub const debug_gpa = false;\n"
             "pub const dev = .core;\n"
             "pub const value_interpret_mode = .direct;\n"
         , zig_version);

build.zig

@@ -171,7 +171,7 @@ pub fn build(b: *std.Build) !void {
     const tracy_callstack = b.option(bool, "tracy-callstack", "Include callstack information with Tracy data. Does nothing if -Dtracy is not provided") orelse (tracy != null);
     const tracy_allocation = b.option(bool, "tracy-allocation", "Include allocation information with Tracy data. Does nothing if -Dtracy is not provided") orelse (tracy != null);
     const tracy_callstack_depth: u32 = b.option(u32, "tracy-callstack-depth", "Declare callstack depth for Tracy data. Does nothing if -Dtracy_callstack is not provided") orelse 10;
-    const force_gpa = b.option(bool, "force-gpa", "Force the compiler to use GeneralPurposeAllocator") orelse false;
+    const debug_gpa = b.option(bool, "debug-allocator", "Force the compiler to use DebugAllocator") orelse false;
     const link_libc = b.option(bool, "force-link-libc", "Force self-hosted compiler to link libc") orelse (enable_llvm or only_c);
     const sanitize_thread = b.option(bool, "sanitize-thread", "Enable thread-sanitization") orelse false;
     const strip = b.option(bool, "strip", "Omit debug information");
@@ -233,7 +233,7 @@ pub fn build(b: *std.Build) !void {
     exe_options.addOption(bool, "llvm_has_csky", llvm_has_csky);
     exe_options.addOption(bool, "llvm_has_arc", llvm_has_arc);
     exe_options.addOption(bool, "llvm_has_xtensa", llvm_has_xtensa);
-    exe_options.addOption(bool, "force_gpa", force_gpa);
+    exe_options.addOption(bool, "debug_gpa", debug_gpa);
     exe_options.addOption(DevEnv, "dev", b.option(DevEnv, "dev", "Build a compiler with a reduced feature set for development of specific features") orelse if (only_c) .bootstrap else .full);
     exe_options.addOption(ValueInterpretMode, "value_interpret_mode", value_interpret_mode);
 
@@ -608,7 +608,7 @@ fn addWasiUpdateStep(b: *std.Build, version: [:0]const u8) !void {
 
     exe_options.addOption(u32, "mem_leak_frames", 0);
     exe_options.addOption(bool, "have_llvm", false);
-    exe_options.addOption(bool, "force_gpa", false);
+    exe_options.addOption(bool, "debug_gpa", false);
     exe_options.addOption([:0]const u8, "version", version);
     exe_options.addOption(std.SemanticVersion, "semver", semver);
     exe_options.addOption(bool, "enable_debug_extensions", false);

lib/libc/musl/src/thread/riscv32/clone.s

@@ -7,6 +7,8 @@
 .global __clone
 .type  __clone, %function
 __clone:
+	andi a1, a1, -16
+
 	# Save func and arg to stack
 	addi a1, a1, -16
 	sw a0, 0(a1)

lib/libc/musl/src/thread/riscv64/clone.s

@@ -7,6 +7,8 @@
 .global __clone
 .type  __clone, %function
 __clone:
+	andi a1, a1, -16
+
 	# Save func and arg to stack
 	addi a1, a1, -16
 	sd a0, 0(a1)

lib/std/heap.zig

@@ -9,11 +9,12 @@ const Allocator = std.mem.Allocator;
 const windows = std.os.windows;
 
 pub const ArenaAllocator = @import("heap/arena_allocator.zig").ArenaAllocator;
-pub const WasmAllocator = @import("heap/WasmAllocator.zig");
+pub const SmpAllocator = @import("heap/SmpAllocator.zig");
+pub const FixedBufferAllocator = @import("heap/FixedBufferAllocator.zig");
 pub const PageAllocator = @import("heap/PageAllocator.zig");
-pub const ThreadSafeAllocator = @import("heap/ThreadSafeAllocator.zig");
 pub const SbrkAllocator = @import("heap/sbrk_allocator.zig").SbrkAllocator;
-pub const FixedBufferAllocator = @import("heap/FixedBufferAllocator.zig");
+pub const ThreadSafeAllocator = @import("heap/ThreadSafeAllocator.zig");
+pub const WasmAllocator = @import("heap/WasmAllocator.zig");
 
 pub const DebugAllocatorConfig = @import("heap/debug_allocator.zig").Config;
 pub const DebugAllocator = @import("heap/debug_allocator.zig").DebugAllocator;
@@ -358,6 +359,11 @@ else if (builtin.target.isWasm()) .{
     .vtable = &PageAllocator.vtable,
 };
 
+pub const smp_allocator: Allocator = .{
+    .ptr = undefined,
+    .vtable = &SmpAllocator.vtable,
+};
+
 /// This allocator is fast, small, and specific to WebAssembly. In the future,
 /// this will be the implementation automatically selected by
 /// `GeneralPurposeAllocator` when compiling in `ReleaseSmall` mode for wasm32
@@ -475,7 +481,7 @@ pub fn StackFallbackAllocator(comptime size: usize) type {
     };
 }
 
-test "c_allocator" {
+test c_allocator {
     if (builtin.link_libc) {
         try testAllocator(c_allocator);
         try testAllocatorAligned(c_allocator);
@@ -484,12 +490,20 @@ test "c_allocator" {
     }
 }
 
-test "raw_c_allocator" {
+test raw_c_allocator {
     if (builtin.link_libc) {
         try testAllocator(raw_c_allocator);
     }
 }
 
+test smp_allocator {
+    if (builtin.single_threaded) return;
+    try testAllocator(smp_allocator);
+    try testAllocatorAligned(smp_allocator);
+    try testAllocatorLargeAlignment(smp_allocator);
+    try testAllocatorAlignedShrink(smp_allocator);
+}
+
 test PageAllocator {
     const allocator = page_allocator;
     try testAllocator(allocator);
@@ -978,4 +992,5 @@ test {
     if (builtin.target.isWasm()) {
         _ = WasmAllocator;
     }
+    if (!builtin.single_threaded) _ = smp_allocator;
 }

lib/std/heap/PageAllocator.zig

@@ -16,11 +16,7 @@ pub const vtable: Allocator.VTable = .{
     .free = free,
 };
 
-fn alloc(context: *anyopaque, n: usize, alignment: mem.Alignment, ra: usize) ?[*]u8 {
-    _ = context;
-    _ = ra;
-    assert(n > 0);
-
+pub fn map(n: usize, alignment: mem.Alignment) ?[*]u8 {
     const page_size = std.heap.pageSize();
     if (n >= maxInt(usize) - page_size) return null;
     const alignment_bytes = alignment.toByteUnits();
@@ -101,6 +97,13 @@ fn alloc(context: *anyopaque, n: usize, alignment: mem.Alignment, ra: usize) ?[*
     return result_ptr;
 }
 
+fn alloc(context: *anyopaque, n: usize, alignment: mem.Alignment, ra: usize) ?[*]u8 {
+    _ = context;
+    _ = ra;
+    assert(n > 0);
+    return map(n, alignment);
+}
+
 fn resize(
     context: *anyopaque,
     memory: []u8,
@@ -114,7 +117,7 @@ fn resize(
     return realloc(memory, new_len, false) != null;
 }
 
-pub fn remap(
+fn remap(
     context: *anyopaque,
     memory: []u8,
     alignment: mem.Alignment,
@@ -127,21 +130,24 @@ pub fn remap(
     return realloc(memory, new_len, true);
 }
 
-fn free(context: *anyopaque, slice: []u8, alignment: mem.Alignment, return_address: usize) void {
+fn free(context: *anyopaque, memory: []u8, alignment: mem.Alignment, return_address: usize) void {
     _ = context;
     _ = alignment;
     _ = return_address;
+    return unmap(@alignCast(memory));
+}
 
+pub fn unmap(memory: []align(page_size_min) u8) void {
     if (native_os == .windows) {
-        windows.VirtualFree(slice.ptr, 0, windows.MEM_RELEASE);
+        windows.VirtualFree(memory.ptr, 0, windows.MEM_RELEASE);
     } else {
-        const buf_aligned_len = mem.alignForward(usize, slice.len, std.heap.pageSize());
-        posix.munmap(@alignCast(slice.ptr[0..buf_aligned_len]));
+        const page_aligned_len = mem.alignForward(usize, memory.len, std.heap.pageSize());
+        posix.munmap(memory.ptr[0..page_aligned_len]);
     }
 }
 
-fn realloc(uncasted_memory: []u8, new_len: usize, may_move: bool) ?[*]u8 {
-    const memory: []align(std.heap.page_size_min) u8 = @alignCast(uncasted_memory);
+pub fn realloc(uncasted_memory: []u8, new_len: usize, may_move: bool) ?[*]u8 {
+    const memory: []align(page_size_min) u8 = @alignCast(uncasted_memory);
     const page_size = std.heap.pageSize();
     const new_size_aligned = mem.alignForward(usize, new_len, page_size);