-
Notifications
You must be signed in to change notification settings - Fork 64
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
gold / lld support for parallel linking (GCC-368) #57
Comments
@igrr how feasible is this? i took a quick stab at
after that i ran into "unsupported ELF machine" error (i was using an amd64 build of gold, so not surpirsing) |
@rojer , gold has no xtensa support yet. It could not process required xtensa-specific relocations until not implemented. It will likely not get it due to very low priority. |
ok, what about |
@rojer , you can refer to LLVM project (but it is not stable yet). Seems LLD is supported for versions greater than esp-15.0.0-20221201. (See espressif/llvm-project#53) |
ok, does this mean that it can be integrated into IDF as an option, or even the default? i'm looking at the toolchain distribution archive and not sure if the xtensa or risc-v variant is included, i only see amd64 and wasm variants (
|
@gerekon PTAL |
@rojer LLD supplied with Espressif Clang toolchain supports Xtensa and RISCV archs, but it has some issues.
Currently we are working on fixing IDF builds with LLD for RISCV chips. Hopefully it will get into the next release. |
@rojer But note, that Xtensa support in Clang itself is not optimal as of now (code size and perf), e.g. direct calls are not implemented yet. espressif/llvm-project#90. So it seems to be early to switch to Clang toolchain completely. May be we can find a way to use LLD with GCC as an option. |
oh yes, clang is hopeless in ok, thank you for the update, please know that this is eagerly awaited :) |
Linking phase is taking substantial amount of time on larger projects, it would be great to get support for
-fuse-ld=gold
and/or-fuse-ld=lld
, both of which support parallel linking and are much faster than the default BFD linker.The text was updated successfully, but these errors were encountered: