mips: set llvm_args -mno-check-zero-division for all mips targets by xry111 · Pull Request #157873 · rust-lang/rust

xry111 · 2026-06-13T19:21:25Z

In rust interger divide by zero is defined to panic, thus the inserted conditional trap should never trigger as the program should have panicked if the divisor is zero.

So disable the insertion of the redundant conditional trap.

rustbot · 2026-06-13T19:21:28Z

These commits modify compiler targets.
(See the Target Tier Policy.)

rustbot · 2026-06-13T19:21:30Z

r? @tiif

rustbot has assigned @tiif.
They will have a look at your PR within the next two weeks and either review your PR or reassign to another reviewer.

Use r? to explicitly pick a reviewer

Why was this reviewer chosen?

The reviewer was selected based on:

Owners of files modified in this PR: compiler
compiler expanded to 73 candidates
Random selection from 22 candidates

folkertdev · 2026-06-14T11:22:03Z

Specifically https://doc.rust-lang.org/reference/expressions/operator-expr.html#r-expr.arith-logic.behavior documents:

† For integer types, division by zero panics.

Adding the flag does in fact have an effect on codegen (I'd have assumed that LLVM would be able to spot this, but I guess not)

https://rust.godbolt.org/z/evacM46bM

It's probably a good idea to capture that in an assembly-llvm codegen test using minicore

xry111 · 2026-06-15T03:44:40Z

It's probably a good idea to capture that in an assembly-llvm codegen test using minicore

Added (but I'm unsure if I wrote it correctly, I haven't written a codegen test before).

folkertdev

r? me

View changes since this review

folkertdev · 2026-06-15T11:32:24Z

+//@ assembly-output: emit-asm
+//@ revisions: mips64el-unknown-linux-gnuabi64
+//@[mips64el-unknown-linux-gnuabi64] compile-flags: --target=mips64el-unknown-linux-gnuabi64
+//@[mips64el-unknown-linux-gnuabi64] needs-llvm-components: mips


Suggested change

//@[mips64el-unknown-linux-gnuabi64] needs-llvm-components: mips

//@ needs-llvm-components: mips

All of these tests need mips, so I think you can factor this out.

also the code's been written already anyway, but testing all targets doesn't add that much value. it would be more interesting to test the (new) default versus a configuration where you enable the div by zero trap explicitly.

hmm guess not

tidy [target-specific-tests (tests)]: /checkout/tests/assembly-llvm/mips-div-no-trap.rs: revision mipsel-unknown-netbsd should specify `needs-llvm-components: mips` as it has `--target` set tidy [target-specific-tests (tests)]: /checkout/tests/assembly-llvm/mips-div-no-trap.rs: revision mipsel-unknown-none should specify `needs-llvm-components: mips` as it has `--target` set tidy [target-specific-tests (tests)]: /checkout/tests/assembly-llvm/mips-div-no-trap.rs: revision mipsisa32r6-unknown-linux-gnu should specify `needs-llvm-components: mips` as it has `--target` set tidy [target-specific-tests (tests)]: /checkout/tests/assembly-llvm/mips-div-no-trap.rs: revision mipsisa32r6el-unknown-linux-gnu should specify `needs-llvm-components: mips` as it has `--target` set tidy [target-specific-tests (tests)]: /checkout/tests/assembly-llvm/mips-div-no-trap.rs: revision mipsisa64r6-unknown-linux-gnuabi64 should specify `needs-llvm-components: mips` as it has `--target` set tidy [target-specific-tests (tests)]: /checkout/tests/assembly-llvm/mips-div-no-trap.rs: revision mipsisa64r6el-unknown-linux-gnuabi64 should specify `needs-llvm-components: mips` as it has `--target` set tidy [target-specific-tests (tests)]: FAIL

@jieyouxu is there a reason that needs-llvm-components is not "inherited" by revisions? e.g. codegen-flags is I believe.

Not sure. Maybe when that was implemented it just wasn't "forwarded" for revisions?

folkertdev · 2026-06-15T11:33:30Z

+#![crate_type = "lib"]
+#![feature(no_core, lang_items)]
+#![no_core]
+


you can use minicore instead (there are plenty of examples of it in this directory) to remove a bunch of these manual language items.

folkertdev · 2026-06-15T11:34:56Z

+impl Div for i32 {
+    type Output = Self;
+
+    fn div(self, rhs: Self) -> Self {
+        self / rhs
+    }
+}


I think this is a recursive definition? My godbolt uses the intrinsic manually instead.

I think this is how a "language item" works and the real core uses the same pattern. But I'll use the intrinsic anyway to get rid all the panic_* language items.

rustbot · 2026-06-15T11:35:34Z

Reminder, once the PR becomes ready for a review, use @rustbot ready.

In rust interger divide by zero is defined to panic, thus the inserted conditional trap should never trigger as the program should have panicked if the divisor is zero. So disable the insertion of the redundant conditional trap.

rustbot · 2026-06-15T13:48:40Z

This PR was rebased onto a different main commit. Here's a range-diff highlighting what actually changed.

Rebasing is a normal part of keeping PRs up to date, so no action is needed—this note is just to help reviewers.

rustbot added S-waiting-on-review Status: Awaiting review from the assignee but also interested parties. T-compiler Relevant to the compiler team, which will review and decide on the PR/issue. labels Jun 13, 2026

rustbot assigned tiif Jun 13, 2026

xry111 force-pushed the mips-no-div-by-zero-trap branch from 1ae5e5c to 12b004e Compare June 15, 2026 03:40