Clarification on x86_64 support in ELD

[sunildora]

Hi ELD team,

I noticed a small mismatch in the documentation and wanted to confirm the current status:

The Supported features / targets section of the README lists ARM, AArch64, RISCV, and Hexagon.

The build instructions and CI show x86_64 being built — for example, the Nightly builder for testing eld against musl-test-suite for x86_64 workflow configures -DELD_TARGETS_TO_BUILD='x86_64' and runs the musl test-suite.

Could you please confirm:

Is x86_64 officially supported by ELD?

If yes, does x86_64 have the same level of feature coverage (static linking, dynamic linking, LTO, linker scripts, etc.) as the other backends?

Are there any known limitations or experimental aspects for x86_64 that users should be aware of?

If x86_64 is supported, would you consider updating the README’s “Supported targets” section for consistency?

Thanks for your work on ELD — it’s great to see CI coverage for x86_64.

[quic-seaswara]

Thanks for your question and your interest in eld!

x86_64 is currently experimental; eld supports only static linking currently on x86_64.

We are working on the handful of tests that are currently XFAILED in the CI builder.

We are still continuing to add x86_64 support, and would be able to see full support in few months.

Could you please let me know more about what functionality that you are interested in testing with eld ?

To speed up the process, we are always welcome to review any patches / merge patches from the eld community.

@sunildora I converted this query to a discussion and hope it is fine with you.

[sunildora1988]

Hi @quic-seaswara,

Thank you, Shankar, for the clarification and for converting this to a discussion — that’s perfectly fine with me.
It’s encouraging to see that x86_64 support is progressing and currently available for static linking.

I had a few follow-up questions:

Are there any public reports or benchmarks comparing ELD’s performance (link time or output size)
with traditional linkers like ld.lld or bfd?

Will the long-term goal be to maintain feature parity across all supported targets (ARM, AArch64, RISC-V, Hexagon, x86_64),
or will some targets remain primarily focused on embedded use cases?

Does ELD have a public summary or dashboard showing CI test results per target (e.g., pass/fail rates), or
are these currently only available through the workflow logs?

Thanks again for your time and for your efforts in maintaining this project — it’s much appreciated!

[quic-seaswara]

We’re actively enhancing x86_64 support in our linker, and expect to have libc tests passing soon—including both static and dynamic linking.

Currently, eld does not fully support x86_64, which makes it challenging to measure link-time performance accurately. For embedded workflows, link time and image debuggability are critical. Eld is designed with both in mind, and we’re exploring ways to improve link time by benchmarking against other linkers like lld and mold.

Our primary goal is to achieve feature parity across all supported targets, especially for embedded use cases. However, we’re also interested in enabling host linking, where glibc support becomes essential for broader adoption.

You may have seen some of the workflows already. We’re working on improving the dashboard to better showcase our progress and milestones.

We’re currently focused on importing features from community linkers and welcome contributions—whether in testing, development, or feature parity efforts.

If you're interested in contributing to any part of the linker, please feel free to submit patches. Every contribution is valuable and appreciated!