Skip to content

Navigation Menu

Sign in
Appearance settings

Search code, repositories, users, issues, pull requests...

Provide feedback

We read every piece of feedback, and take your input very seriously.

Saved searches

Use saved searches to filter your results more quickly
Sign up
Appearance settings

Commit f5534da

Browse files
committed
Auto merge of #142423 - cuviper:beta-next, r=cuviper
[beta] backports - x86 (32/64): go back to passing SIMD vectors by-ptr #141309 - use correct edition when warning for unsafe attributes #142261 r? cuviper
2 parents 645b44e + 6f3443d commit f5534da

File tree

8 files changed

+91
-52
lines changed

8 files changed

+91
-52
lines changed

‎compiler/rustc_monomorphize/src/mono_checks/abi_check.rs

Lines changed: 12 additions & 5 deletions
Original file line numberDiff line numberDiff line change
@@ -35,8 +35,6 @@ fn do_check_simd_vector_abi<'tcx>(
3535
is_call: bool,
3636
loc: impl Fn() -> (Span, HirId),
3737
) {
38-
// We check this on all functions, including those using the "Rust" ABI.
39-
// For the "Rust" ABI it would be a bug if the lint ever triggered, but better safe than sorry.
4038
let feature_def = tcx.sess.target.features_for_correct_vector_abi();
4139
let codegen_attrs = tcx.codegen_fn_attrs(def_id);
4240
let have_feature = |feat: Symbol| {
@@ -123,8 +121,9 @@ fn do_check_wasm_abi<'tcx>(
123121
is_call: bool,
124122
loc: impl Fn() -> (Span, HirId),
125123
) {
126-
// Only proceed for `extern "C" fn` on wasm32-unknown-unknown (same check as what `adjust_for_foreign_abi` uses to call `compute_wasm_abi_info`),
127-
// and only proceed if `wasm_c_abi_opt` indicates we should emit the lint.
124+
// Only proceed for `extern "C" fn` on wasm32-unknown-unknown (same check as what
125+
// `adjust_for_foreign_abi` uses to call `compute_wasm_abi_info`), and only proceed if
126+
// `wasm_c_abi_opt` indicates we should emit the lint.
128127
if !(tcx.sess.target.arch == "wasm32"
129128
&& tcx.sess.target.os == "unknown"
130129
&& tcx.wasm_c_abi_opt() == WasmCAbi::Legacy { with_lint: true }
@@ -157,8 +156,15 @@ fn check_instance_abi<'tcx>(tcx: TyCtxt<'tcx>, instance: Instance<'tcx>) {
157156
else {
158157
// An error will be reported during codegen if we cannot determine the ABI of this
159158
// function.
159+
tcx.dcx().delayed_bug("ABI computation failure should lead to compilation failure");
160160
return;
161161
};
162+
// Unlike the call-site check, we do also check "Rust" ABI functions here.
163+
// This should never trigger, *except* if we start making use of vector registers
164+
// for the "Rust" ABI and the user disables those vector registers (which should trigger a
165+
// warning as that's clearly disabling a "required" target feature for this target).
166+
// Using such a function is where disabling the vector register actually can start leading
167+
// to soundness issues, so erroring here seems good.
162168
let loc = || {
163169
let def_id = instance.def_id();
164170
(
@@ -179,7 +185,8 @@ fn check_call_site_abi<'tcx>(
179185
loc: impl Fn() -> (Span, HirId) + Copy,
180186
) {
181187
if callee.fn_sig(tcx).abi().is_rustic_abi() {
182-
// we directly handle the soundness of Rust ABIs
188+
// We directly handle the soundness of Rust ABIs -- so let's skip the majority of
189+
// call sites to avoid a perf regression.
183190
return;
184191
}
185192
let typing_env = ty::TypingEnv::fully_monomorphized();

‎compiler/rustc_parse/src/validate_attr.rs

Lines changed: 6 additions & 1 deletion
Original file line numberDiff line numberDiff line change
@@ -180,9 +180,14 @@ pub fn check_attribute_safety(
180180
let diag_span = attr_item.span();
181181

182182
// Attributes can be safe in earlier editions, and become unsafe in later ones.
183+
//
184+
// Use the span of the attribute's name to determine the edition: the span of the
185+
// attribute as a whole may be inaccurate if it was emitted by a macro.
186+
//
187+
// See https://github.com/rust-lang/rust/issues/142182.
183188
let emit_error = match unsafe_since {
184189
None => true,
185-
Some(unsafe_since) => attr.span.edition() >= unsafe_since,
190+
Some(unsafe_since) => path_span.edition() >= unsafe_since,
186191
};
187192

188193
if emit_error {

‎compiler/rustc_target/src/callconv/mod.rs

Lines changed: 5 additions & 22 deletions
Original file line numberDiff line numberDiff line change
@@ -7,7 +7,7 @@ use rustc_abi::{
77
};
88
use rustc_macros::HashStable_Generic;
99

10-
use crate::spec::{HasTargetSpec, HasWasmCAbiOpt, HasX86AbiOpt, RustcAbi,WasmCAbi};
10+
use crate::spec::{HasTargetSpec, HasWasmCAbiOpt, HasX86AbiOpt, WasmCAbi};
1111

1212
mod aarch64;
1313
mod amdgpu;
@@ -732,24 +732,6 @@ impl<'a, Ty> FnAbi<'a, Ty> {
732732
_ => {}
733733
};
734734

735-
// Decides whether we can pass the given SIMD argument via `PassMode::Direct`.
736-
// May only return `true` if the target will always pass those arguments the same way,
737-
// no matter what the user does with `-Ctarget-feature`! In other words, whatever
738-
// target features are required to pass a SIMD value in registers must be listed in
739-
// the `abi_required_features` for the current target and ABI.
740-
let can_pass_simd_directly = |arg: &ArgAbi<'_, Ty>| match &*spec.arch {
741-
// On x86, if we have SSE2 (which we have by default for x86_64), we can always pass up
742-
// to 128-bit-sized vectors.
743-
"x86" if spec.rustc_abi == Some(RustcAbi::X86Sse2) => arg.layout.size.bits() <= 128,
744-
"x86_64" if spec.rustc_abi != Some(RustcAbi::X86Softfloat) => {
745-
// FIXME once https://github.com/bytecodealliance/wasmtime/issues/10254 is fixed
746-
// accept vectors up to 128bit rather than vectors of exactly 128bit.
747-
arg.layout.size.bits() == 128
748-
}
749-
// So far, we haven't implemented this logic for any other target.
750-
_ => false,
751-
};
752-
753735
for (arg_idx, arg) in self
754736
.args
755737
.iter_mut()
@@ -849,9 +831,10 @@ impl<'a, Ty> FnAbi<'a, Ty> {
849831
// target feature sets. Some more information about this
850832
// issue can be found in #44367.
851833
//
852-
// Note that the intrinsic ABI is exempt here as those are not
853-
// real functions anyway, and the backend expects very specific types.
854-
if spec.simd_types_indirect && !can_pass_simd_directly(arg) {
834+
// We *could* do better in some cases, e.g. on x86_64 targets where SSE2 is
835+
// required. However, it turns out that that makes LLVM worse at optimizing this
836+
// code, so we pass things indirectly even there. See #139029 for more on that.
837+
if spec.simd_types_indirect {
855838
arg.make_indirect();
856839
}
857840
}

‎tests/codegen/abi-x86-sse.rs

Lines changed: 3 additions & 2 deletions
Original file line numberDiff line numberDiff line change
@@ -27,8 +27,9 @@ trait Copy {}
2727
#[repr(simd)]
2828
pub struct Sse([f32; 4]);
2929

30-
// x86-64: <4 x float> @sse_id(<4 x float> {{[^,]*}})
31-
// x86-32: <4 x float> @sse_id(<4 x float> {{[^,]*}})
30+
// FIXME: due to #139029 we are passing them all indirectly.
31+
// x86-64: void @sse_id(ptr{{( [^,]*)?}} sret([16 x i8]){{( .*)?}}, ptr{{( [^,]*)?}})
32+
// x86-32: void @sse_id(ptr{{( [^,]*)?}} sret([16 x i8]){{( .*)?}}, ptr{{( [^,]*)?}})
3233
// x86-32-nosse: void @sse_id(ptr{{( [^,]*)?}} sret([16 x i8]){{( .*)?}}, ptr{{( [^,]*)?}})
3334
#[no_mangle]
3435
pub fn sse_id(x: Sse) -> Sse {

‎tests/codegen/simd-intrinsic/simd-intrinsic-transmute-array.rs

Lines changed: 5 additions & 15 deletions
Original file line numberDiff line numberDiff line change
@@ -1,14 +1,8 @@
11
//
22
//@ compile-flags: -C no-prepopulate-passes
3-
// LLVM IR isn't very portable and the one tested here depends on the ABI
4-
// which is different between x86 (where we use SSE registers) and others.
5-
// `x86-64` and `x86-32-sse2` are identical, but compiletest does not support
6-
// taking the union of multiple `only` annotations.
7-
//@ revisions: x86-64 x86-32-sse2 other
8-
//@[x86-64] only-x86_64
9-
//@[x86-32-sse2] only-rustc_abi-x86-sse2
10-
//@[other] ignore-rustc_abi-x86-sse2
11-
//@[other] ignore-x86_64
3+
// 32bit MSVC does not align things properly so we suppress high alignment annotations (#112480)
4+
//@ ignore-i686-pc-windows-msvc
5+
//@ ignore-i686-pc-windows-gnu
126

137
#![crate_type = "lib"]
148
#![allow(non_camel_case_types)]
@@ -47,9 +41,7 @@ pub fn build_array_s(x: [f32; 4]) -> S<4> {
4741
#[no_mangle]
4842
pub fn build_array_transmute_s(x: [f32; 4]) -> S<4> {
4943
// CHECK: %[[VAL:.+]] = load <4 x float>, ptr %x, align [[ARRAY_ALIGN]]
50-
// x86-32: ret <4 x float> %[[VAL:.+]]
51-
// x86-64: ret <4 x float> %[[VAL:.+]]
52-
// other: store <4 x float> %[[VAL:.+]], ptr %_0, align [[VECTOR_ALIGN]]
44+
// CHECK: store <4 x float> %[[VAL:.+]], ptr %_0, align [[VECTOR_ALIGN]]
5345
unsafe { std::mem::transmute(x) }
5446
}
5547

@@ -64,8 +56,6 @@ pub fn build_array_t(x: [f32; 4]) -> T {
6456
#[no_mangle]
6557
pub fn build_array_transmute_t(x: [f32; 4]) -> T {
6658
// CHECK: %[[VAL:.+]] = load <4 x float>, ptr %x, align [[ARRAY_ALIGN]]
67-
// x86-32: ret <4 x float> %[[VAL:.+]]
68-
// x86-64: ret <4 x float> %[[VAL:.+]]
69-
// other: store <4 x float> %[[VAL:.+]], ptr %_0, align [[VECTOR_ALIGN]]
59+
// CHECK: store <4 x float> %[[VAL:.+]], ptr %_0, align [[VECTOR_ALIGN]]
7060
unsafe { std::mem::transmute(x) }
7161
}

‎tests/codegen/simd/packed-simd.rs

Lines changed: 9 additions & 7 deletions
Original file line numberDiff line numberDiff line change
@@ -30,16 +30,18 @@ fn load<T, const N: usize>(v: PackedSimd<T, N>) -> FullSimd<T, N> {
3030
}
3131
}
3232

33-
// CHECK-LABEL: define <3 x float> @square_packed_full(ptr{{[a-z_ ]*}} align 4 {{[^,]*}})
33+
// CHECK-LABEL: square_packed_full
34+
// CHECK-SAME: ptr{{[a-z_ ]*}} sret([[RET_TYPE:[^)]+]]) [[RET_ALIGN:align (8|16)]]{{[^%]*}} [[RET_VREG:%[_0-9]*]]
35+
// CHECK-SAME: ptr{{[a-z_ ]*}} align 4
3436
#[no_mangle]
3537
pub fn square_packed_full(x: PackedSimd<f32, 3>) -> FullSimd<f32, 3> {
36-
// The unoptimized version of this is not very interesting to check
37-
// since `load` does not get inlined.
38-
// opt3-NEXT: start:
39-
// opt3-NEXT: load <3 x float>
38+
// CHECK-NEXT: start
39+
// noopt: alloca [[RET_TYPE]], [[RET_ALIGN]]
40+
// CHECK: load <3 x float>
4041
let x = load(x);
41-
// opt3-NEXT: [[VREG:%[a-z0-9_]+]] = fmul <3 x float>
42-
// opt3-NEXT: ret <3 x float> [[VREG:%[a-z0-9_]+]]
42+
// CHECK: [[VREG:%[a-z0-9_]+]] = fmul <3 x float>
43+
// CHECK-NEXT: store <3 x float> [[VREG]], ptr [[RET_VREG]], [[RET_ALIGN]]
44+
// CHECK-NEXT: ret void
4345
unsafe { intrinsics::simd_mul(x, x) }
4446
}
4547

Lines changed: 24 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -0,0 +1,24 @@
1+
error: unsafe attribute used without unsafe
2+
--> $DIR/unsafe-attr-edition-span.rs:21:3
3+
|
4+
LL | #[no_mangle]
5+
| ^^^^^^^^^ usage of unsafe attribute
6+
|
7+
help: wrap the attribute in `unsafe(...)`
8+
|
9+
LL | #[unsafe(no_mangle)]
10+
| +++++++ +
11+
12+
error: unsafe attribute used without unsafe
13+
--> $DIR/unsafe-attr-edition-span.rs:25:7
14+
|
15+
LL | #[no_mangle]
16+
| ^^^^^^^^^ usage of unsafe attribute
17+
|
18+
help: wrap the attribute in `unsafe(...)`
19+
|
20+
LL | #[unsafe(no_mangle)]
21+
| +++++++ +
22+
23+
error: aborting due to 2 previous errors
24+
Lines changed: 27 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -0,0 +1,27 @@
1+
// Tests that the correct span is used to determine the edition of an attribute that was safe to use
2+
// in earlier editions, but has become `unsafe` in later editions.
3+
//
4+
// Determining the correct edition is non-trivial because of macro expansion. For instance,
5+
// the `thread_local!` macro (defined in std and hence using the most recent edition) parses the
6+
// attribute, and then re-emits it. Therefore, the span of the `#` token starting the
7+
// `#[no_mangle]` attribute has std's edition, while the attribute name has the edition of this
8+
// file, which may be different.
9+
10+
//@ revisions: e2015 e2018 e2021 e2024
11+
12+
//@[e2018] edition:2018
13+
//@[e2021] edition:2021
14+
//@[e2024] edition:2024
15+
//
16+
//@[e2015] check-pass
17+
//@[e2018] check-pass
18+
//@[e2021] check-pass
19+
#![crate_type = "lib"]
20+
21+
#[no_mangle] //[e2024]~ ERROR unsafe attribute used without unsafe
22+
static TEST_OUTSIDE: usize = 0;
23+
24+
thread_local! {
25+
#[no_mangle]//[e2024]~ ERROR unsafe attribute used without unsafe
26+
static TEST: usize = 0;
27+
}

0 commit comments

Comments
(0)

AltStyle によって変換されたページ (->オリジナル) /