In both backends they did not observe the Liveness information for try
instructions. Now they do. For the C backend this is necessary for
correctness; for the LLVM backend, it improves code generation.
`@llvm.dbg.value` is absolutely useless, adding a temporary alloca
to store the constant in will make it actually show up in debuggers.
The effect on performance should be minimal since there is only one
store and it the change is not applied to ReleaseSafe builds.
```zig
fn foo(a: u32, b: []const u8, c: bool, d: enum { yes, no }) void {
_ = a; _ = b; _ = c; _ = d;
}
```
before:
```
Breakpoint 1, a.foo (a=<optimized out>, b=..., c=<optimized out>, d=<optimized out>) at a.zig:18
18 _ = d;
```
after:
```
Breakpoint 1, a.foo (a=1, b=..., c=false, d=yes) at a.zig:15
15 _ = a; _ = b; _ = c; _ = d;
(gdb) p b
$1 = {ptr = 0x20854f <a.main.anon_3888> "bar", len = 3}
```
This also modifies the inline assembly to be more optimizable - instead of
doing explicit movs, we instead communicate to LLVM which registers we
would like to, somehow, have the correct values. This is how the x86_64
code already worked and thus allows the code to be unified across the
two architectures.
As a bonus, I threw in x86 support.
Adds optimizations for by-ref types to:
- .struct_field_val
- .slice_elem_val
- .ptr_elem_val
I would have expected LLVM to be able to optimize away these
temporaries since we don't leak pointers to them and they are fed
straight from def to use, but empirically it does not.
Resolves https://github.com/ziglang/zig/issues/12713
Resolves https://github.com/ziglang/zig/issues/12638
This change makes any of the `*_val` instructions check whether it's
safe to elide copies for by-ref types rather than performing this
elision blindly.
AIR instructions fixed:
- .array_elem_val
- .struct_field_val
- .unwrap_errunion_payload
- .try
- .optional_payload
These now all respect value semantics, as expected.
P.S. Thanks to Andrew for the new way to approach this. Many of the
lines here are from his recommended change, which comes with the
significant advantage that loads are now as small as the intervening
memory access allows.
Co-authored by: Andrew Kelley <andrew@ziglang.org>
When Wasm extern functions contain the same name, but have a
different module name such as `extern "a"` vs `extern "b"` LLVM will
currently resolve the two functions to the same symbol. By mangling
the name of the symbol, we ensure the functions are resolved
seperately. We mangle the name by applying <name>|<module> where
module is also known as the library name.
This function is redundant with CType.sizeInBits(), and until the
previous commit they disagreed about the correct long double type
for several targets. Although they're all synced up now, it's much
simpler just to have a single source of truth.
This is encoded as a primitive AIR instruction to resolve one corner
case: A function may include a `catch { ... }` or `else |err| { ... }`
block but not call any errorable fn. In that case, there is no error
return trace to save the index of and codegen needs to avoid
interacting with the non-existing error trace.
By using a primitive AIR op, we can depend on Liveness to mark this
unused in this corner case.
