llvm: Use no-builtins attribute instead of nobuiltin.

The former prevents recognizing code patterns and turning them into libcalls,
which is what we want for compiler-rt. The latter is meant to be used on call
sites to prevent them from being turned into intrinsics.

Context: https://github.com/ziglang/zig/issues/21833
This commit is contained in:
Alex Rønne Petersen 2024-11-03 14:37:42 +01:00
parent 4a3611f7d6
commit bd8ef0036d

View File

@ -3242,19 +3242,22 @@ pub const Object = struct {
if (owner_mod.unwind_tables) {
try attributes.addFnAttr(.{ .uwtable = Builder.Attribute.UwTable.default }, &o.builder);
}
if (comp.skip_linker_dependencies or comp.no_builtin) {
const target = owner_mod.resolved_target.result;
if (comp.skip_linker_dependencies or comp.no_builtin or target.cpu.arch.isBpf()) {
// The intent here is for compiler-rt and libc functions to not generate
// infinite recursion. For example, if we are compiling the memcpy function,
// and llvm detects that the body is equivalent to memcpy, it may replace the
// body of memcpy with a call to memcpy, which would then cause a stack
// overflow instead of performing memcpy.
try attributes.addFnAttr(.nobuiltin, &o.builder);
try attributes.addFnAttr(.{ .string = .{
.kind = try o.builder.string("no-builtins"),
.value = .empty,
} }, &o.builder);
}
if (owner_mod.optimize_mode == .ReleaseSmall) {
try attributes.addFnAttr(.minsize, &o.builder);
try attributes.addFnAttr(.optsize, &o.builder);
}
const target = owner_mod.resolved_target.result;
if (target.cpu.model.llvm_name) |s| {
try attributes.addFnAttr(.{ .string = .{
.kind = try o.builder.string("target-cpu"),
@ -3267,12 +3270,6 @@ pub const Object = struct {
.value = try o.builder.string(std.mem.span(s)),
} }, &o.builder);
}
if (target.cpu.arch.isBpf()) {
try attributes.addFnAttr(.{ .string = .{
.kind = try o.builder.string("no-builtins"),
.value = .empty,
} }, &o.builder);
}
if (target.floatAbi() == .soft) {
// `use-soft-float` means "use software routines for floating point computations". In
// other words, it configures how LLVM lowers basic float instructions like `fcmp`,