compiler-builtins

Porting compiler-rt intrinsics to Rust

See rust-lang/rust#35437.

When and how to use this crate?

If you are working with a target that doesn't have binary releases of std available via rustup (this probably means you are building the core crate yourself) and need compiler-rt intrinsics (i.e. you are probably getting linker errors when building an executable: undefined reference to __aeabi_memcpy), you can use this crate to get those intrinsics and solve the linker errors. To do that, add this crate somewhere in the dependency graph of the crate you are building:

# Cargo.toml
[dependencies]
compiler_builtins = { git = "https://github.com/rust-lang/compiler-builtins" }

extern crate compiler_builtins;

// ...

If you still get an "undefined reference to $INTRINSIC" error after that change, that means that we haven't ported $INTRINSIC to Rust yet! Please open an issue with the name of the intrinsic and the LLVM triple (e.g. thumbv7m-none-eabi) of the target you are using. That way we can prioritize porting that particular intrinsic.

If you've got a C compiler available for your target then while we implement this intrinsic you can temporarily enable a fallback to the actual compiler-rt implementation as well for unimplemented intrinsics:

[dependencies.compiler_builtins]
git = "https://github.com/rust-lang/compiler-builtins"
features = ["c"]

Contributing

1. Pick one or more intrinsics from the pending list.
2. Fork this repository.
3. Port the intrinsic(s) and their corresponding unit tests from their C implementation to Rust.
4. Add a test to compare the behavior of the ported intrinsic(s) with their implementation on the testing host.
5. Add the intrinsic to examples/intrinsics.rs to verify it can be linked on all targets.
6. Send a Pull Request (PR).
7. Once the PR passes our extensive testing infrastructure, we'll merge it!
8. Celebrate 🎉

Porting Reminders

1. Rust and C have slightly different operator precedence. C evaluates comparisons (== !=) before bitwise operations (& | ^), while Rust evaluates the other way.
2. C assumes wrapping operations everywhere. Rust panics on overflow when in debug mode. Consider using the Wrapping type or the explicit wrapping_* functions where applicable.
3. Note C implicit casts, especially integer promotion. Rust is much more explicit about casting, so be sure that any cast which affects the output is ported to the Rust implementation.
4. Rust has many functions for integer or floating point manipulation in the standard library. Consider using one of these functions rather than porting a new one.

Testing

The easiest way to test locally is using Docker. This can be done by running ./ci/run-docker.sh [target]. If no target is specified, all targets will be run.

In order to run the full test suite, you will also need the C compiler runtime to test against, located in a directory called compiler-rt. This can be obtained with the following:

curl -L -o rustc-llvm-19.1.tar.gz https://github.com/rust-lang/llvm-project/archive/rustc/19.1-2024-09-17.tar.gz
tar xzf rustc-llvm-19.1.tar.gz --strip-components 1 llvm-project-rustc-19.1-2024-09-17/compiler-rt

Local targets may also be tested with ./ci/run.sh [target].

Note that testing may not work on all hosts, in which cases it is acceptable to rely on CI.

Progress

• aarch64/chkstk.S
• adddf3.c
• addsf3.c
• arm/addsf3.S
• arm/aeabi_dcmp.S
• arm/aeabi_fcmp.S
• arm/aeabi_idivmod.S
• arm/aeabi_ldivmod.S
• arm/aeabi_memcpy.S
• arm/aeabi_memmove.S
• arm/aeabi_memset.S
• arm/aeabi_uidivmod.S
• arm/aeabi_uldivmod.S
• arm/chkstk.S
• arm/divmodsi4.S (generic version is done)
• arm/divsi3.S (generic version is done)
• arm/modsi3.S (generic version is done)
• arm/softfloat-alias.list
• arm/udivmodsi4.S (generic version is done)
• arm/udivsi3.S (generic version is done)
• arm/umodsi3.S (generic version is done)
• ashldi3.c
• ashrdi3.c
• avr/divmodhi4.S
• avr/divmodqi4.S
• avr/mulhi3.S
• avr/mulqi3.S
• avr/udivmodhi4.S
• avr/udivmodqi4.S
• bswapdi2.c
• bswapsi2.c
• bswapti2.c
• clzdi2.c
• clzsi2.c
• clzti2.c
• comparedf2.c
• comparesf2.c
• ctzdi2.c
• ctzsi2.c
• ctzti2.c
• divdf3.c
• divdi3.c
• divmoddi4.c
• divmodsi4.c
• divmodti4.c
• divsf3.c
• divsi3.c
• extendsfdf2.c
• fixdfdi.c
• fixdfsi.c
• fixsfdi.c
• fixsfsi.c
• fixunsdfdi.c
• fixunsdfsi.c
• fixunssfdi.c
• fixunssfsi.c
• floatdidf.c
• floatdisf.c
• floatsidf.c
• floatsisf.c
• floatundidf.c
• floatundisf.c
• floatunsidf.c
• floatunsisf.c
• i386/ashldi3.S
• i386/ashrdi3.S
• i386/chkstk.S
• i386/divdi3.S
• i386/lshrdi3.S
• i386/moddi3.S
• i386/muldi3.S
• i386/udivdi3.S
• i386/umoddi3.S
• lshrdi3.c
• moddi3.c
• modsi3.c
• muldf3.c
• muldi3.c
• mulodi4.c
• mulosi4.c
• mulsf3.c
• powidf2.c
• powisf2.c
• riscv/muldi3.S
• riscv/mulsi3.S
• subdf3.c
• subsf3.c
• truncdfsf2.c
• udivdi3.c
• udivmoddi4.c
• udivmodsi4.c
• udivsi3.c
• umoddi3.c
• umodsi3.c
• x86_64/chkstk.S

These builtins are needed to support 128-bit integers.

• ashlti3.c
• ashrti3.c
• divti3.c
• fixdfti.c
• fixsfti.c
• fixunsdfti.c
• fixunssfti.c
• floattidf.c
• floattisf.c
• floatuntidf.c
• floatuntisf.c
• lshrti3.c
• modti3.c
• muloti4.c
• multi3.c
• udivmodti4.c
• udivti3.c
• umodti3.c

These builtins are needed to support f16 and f128, which are in the process of being added to Rust.

• addtf3.c
• comparetf2.c
• divtf3.c
• extenddftf2.c
• extendhfsf2.c
• extendhftf2.c
• extendsftf2.c
• fixtfdi.c
• fixtfsi.c
• fixtfti.c
• fixunstfdi.c
• fixunstfsi.c
• fixunstfti.c
• floatditf.c
• floatsitf.c
• floattitf.c
• floatunditf.c
• floatunsitf.c
• floatuntitf.c
• multf3.c
• powitf2.c
• subtf3.c
• truncdfhf2.c
• truncsfhf2.c
• trunctfdf2.c
• trunctfhf2.c
• trunctfsf2.c

These builtins are used by the Hexagon DSP

• hexagon/common_entry_exit_abi1.S
• hexagon/common_entry_exit_abi2.S
• hexagon/common_entry_exit_legacy.S
• hexagon/dfaddsub.S~~
• hexagon/dfdiv.S~~
• hexagon/dffma.S~~
• hexagon/dfminmax.S~~
• hexagon/dfmul.S~~
• hexagon/dfsqrt.S~~
• hexagon/divdi3.S~~
• hexagon/divsi3.S~~
• hexagon/fastmath2_dlib_asm.S~~
• hexagon/fastmath2_ldlib_asm.S~~
• hexagon/fastmath_dlib_asm.S~~
• hexagon/memcpy_forward_vp4cp4n2.S~~
• hexagon/memcpy_likely_aligned.S~~
• hexagon/moddi3.S~~
• hexagon/modsi3.S~~
• hexagon/sfdiv_opt.S~~
• hexagon/sfsqrt_opt.S~~
• hexagon/udivdi3.S~~
• hexagon/udivmoddi4.S~~
• hexagon/udivmodsi4.S~~
• hexagon/udivsi3.S~~
• hexagon/umoddi3.S~~
• hexagon/umodsi3.S~~

Unimplemented functions

These builtins are for x87 f80 floating-point numbers that are not supported by Rust.

• extendxftf2.c
• fixunsxfdi.c
• fixunsxfsi.c
• fixunsxfti.c
• fixxfdi.c
• fixxfti.c
• floatdixf.c
• floattixf.c
• floatundixf.c
• floatuntixf.c
• i386/floatdixf.S
• i386/floatundixf.S
• x86_64/floatdixf.c
• x86_64/floatundixf.S

These builtins are for IBM "extended double" non-IEEE 128-bit floating-point numbers.

• ppc/divtc3.c
• ppc/fixtfdi.c
• ppc/fixtfti.c
• ppc/fixunstfdi.c
• ppc/fixunstfti.c
• ppc/floatditf.c
• ppc/floattitf.c
• ppc/floatunditf.c
• ppc/gcc_qadd.c
• ppc/gcc_qdiv.c
• ppc/gcc_qmul.c
• ppc/gcc_qsub.c
• ppc/multc3.c

These builtins are for 16-bit brain floating-point numbers that are not supported by Rust.

• truncdfbf2.c
• truncsfbf2.c
• trunctfxf2.c

These builtins involve complex floating-point types that are not supported by Rust.

• divdc3.c
• divsc3.c
• divtc3.c
• divxc3.c
• muldc3.c
• mulsc3.c
• multc3.c
• mulxc3.c
• powixf2.c

These builtins are never called by LLVM.

• absvdi2.c
• absvsi2.c
• absvti2.c
• addvdi3.c
• addvsi3.c
• addvti3.c
• arm/aeabi_cdcmp.S
• arm/aeabi_cdcmpeq_check_nan.c
• arm/aeabi_cfcmp.S
• arm/aeabi_cfcmpeq_check_nan.c
• arm/aeabi_div0.c
• arm/aeabi_drsub.c
• arm/aeabi_frsub.c
• arm/aeabi_memcmp.S
• arm/bswapdi2.S
• arm/bswapsi2.S
• arm/clzdi2.S
• arm/clzsi2.S
• arm/comparesf2.S
• arm/restore_vfp_d8_d15_regs.S
• arm/save_vfp_d8_d15_regs.S
• arm/switch16.S
• arm/switch32.S
• arm/switch8.S
• arm/switchu8.S
• cmpdi2.c
• cmpti2.c
• ffssi2.c
• ffsdi2.c - this is called by gcc though!
• ffsti2.c
• mulvdi3.c
• mulvsi3.c
• mulvti3.c
• negdf2.c
• negdi2.c
• negsf2.c
• negti2.c
• negvdi2.c
• negvsi2.c
• negvti2.c
• paritydi2.c
• paritysi2.c
• parityti2.c
• popcountdi2.c
• popcountsi2.c
• popcountti2.c
• ppc/restFP.S
• ppc/saveFP.S
• subvdi3.c
• subvsi3.c
• subvti3.c
• ucmpdi2.c
• ucmpti2.c
• udivmodti4.c

Rust only exposes atomic types on platforms that support them, and therefore does not need to fall back to software implementations.

• arm/sync_fetch_and_add_4.S
• arm/sync_fetch_and_add_8.S
• arm/sync_fetch_and_and_4.S
• arm/sync_fetch_and_and_8.S
• arm/sync_fetch_and_max_4.S
• arm/sync_fetch_and_max_8.S
• arm/sync_fetch_and_min_4.S
• arm/sync_fetch_and_min_8.S
• arm/sync_fetch_and_nand_4.S
• arm/sync_fetch_and_nand_8.S
• arm/sync_fetch_and_or_4.S
• arm/sync_fetch_and_or_8.S
• arm/sync_fetch_and_sub_4.S
• arm/sync_fetch_and_sub_8.S
• arm/sync_fetch_and_umax_4.S
• arm/sync_fetch_and_umax_8.S
• arm/sync_fetch_and_umin_4.S
• arm/sync_fetch_and_umin_8.S
• arm/sync_fetch_and_xor_4.S
• arm/sync_fetch_and_xor_8.S
• arm/sync_synchronize.S
• atomic.c
• atomic_flag_clear.c
• atomic_flag_clear_explicit.c
• atomic_flag_test_and_set.c
• atomic_flag_test_and_set_explicit.c
• atomic_signal_fence.c
• atomic_thread_fence.c

Miscellaneous functionality that is not used by Rust.

• aarch64/fp_mode.c
• aarch64/lse.S (LSE atomics)
• aarch64/sme-abi-init.c (matrix extension)
• aarch64/sme-abi.S (matrix extension)
• aarch64/sme-libc-routines.c (matrix extension)
• apple_versioning.c
• arm/fp_mode.c
• avr/exit.S
• clear_cache.c
• cpu_model/aarch64.c
• cpu_model/x86.c
• crtbegin.c
• crtend.c
• emutls.c
• enable_execute_stack.c
• eprintf.c
• fp_mode.c (float exception handling)
• gcc_personality_v0.c
• i386/fp_mode.c
• int_util.c
• loongarch/fp_mode.c
• os_version_check.c
• riscv/fp_mode.c
• riscv/restore.S (callee-saved registers)
• riscv/save.S (callee-saved registers)
• trampoline_setup.c
• ve/grow_stack.S
• ve/grow_stack_align.S

Floating-point implementations of builtins that are only called from soft-float code. It would be better to simply use the generic soft-float versions in this case.

• i386/floatdidf.S
• i386/floatdisf.S
• i386/floatundidf.S
• i386/floatundisf.S
• x86_64/floatundidf.S
• x86_64/floatundisf.S
• x86_64/floatdidf.c
• x86_64/floatdisf.c

Unsupported in any current target: used on old versions of 32-bit iOS with ARMv5.

• arm/adddf3vfp.S
• arm/addsf3vfp.S
• arm/divdf3vfp.S
• arm/divsf3vfp.S
• arm/eqdf2vfp.S
• arm/eqsf2vfp.S
• arm/extendsfdf2vfp.S
• arm/fixdfsivfp.S
• arm/fixsfsivfp.S
• arm/fixunsdfsivfp.S
• arm/fixunssfsivfp.S
• arm/floatsidfvfp.S
• arm/floatsisfvfp.S
• arm/floatunssidfvfp.S
• arm/floatunssisfvfp.S
• arm/gedf2vfp.S
• arm/gesf2vfp.S
• arm/gtdf2vfp.S
• arm/gtsf2vfp.S
• arm/ledf2vfp.S
• arm/lesf2vfp.S
• arm/ltdf2vfp.S
• arm/ltsf2vfp.S
• arm/muldf3vfp.S
• arm/mulsf3vfp.S
• arm/nedf2vfp.S
• arm/negdf2vfp.S
• arm/negsf2vfp.S
• arm/nesf2vfp.S
• arm/subdf3vfp.S
• arm/subsf3vfp.S
• arm/truncdfsf2vfp.S
• arm/unorddf2vfp.S
• arm/unordsf2vfp.S

License

The compiler-builtins crate is dual licensed under both the University of Illinois "BSD-Like" license and the MIT license. As a user of this code you may choose to use it under either license. As a contributor, you agree to allow your code to be used under both.

Full text of the relevant licenses is in LICENSE.TXT.