Add naked function attribute support.

[joseluisquiroga]

A few lines in gcc/config/epiphany.c that add naked function
attribute support to e-gcc.

• Adds the entry in 'epiphany_attribute_table'.
• Defines two functions: 'epiphany_handle_naked_attribute'
  and 'epiphany_is_naked_p'.
• Makes the respective calls.

[nickl-]

This is to add support for embedded assembly functions or is there a different reason?

[joseluisquiroga]

Exactly. I was exploring how to implement some functionality for which I thought I needed naked functions, but it turned out I did not. Just normal asm embedding (epiphany asm) was enough, and that was already provided by the Adapteva gcc hack. On Friday, July 12, 2019, 8:27:55 AM GMT-5, Nick Lombard <[email protected]> wrote: This is to add support for embedded assembly functions or is there a different reason? — You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub, or mute the thread.

[nickl-]

I would normally suggest that we mark this issue as resolved to be closed but I just noticed we don't even have issues available on this project so perhaps it is a blessing in disguise.

There were some issues that I've wanted to raise and before life happens and I forget about it again.

We might as well add new issues here:

Enable issues for this project

This has to be the first issue to raise: Please enable the issues for this project.

Issues are not only for support queries, questions, feature requests, problem reports or discussions but they also serve as reminders of outstanding tasks which contributors can pick up and resolve.

Can't have community involvement if we can't speak.

Update README.md

The README.md serves as the project introduction and at the very least should mention that this is a fork and should probably provide a reference to the origin repository.

An explanation of what was changed, why it should be used and what extra functionality is provided would be helpful. If this is documented somewhere else a reference to that would suffice.

Project behind HEAD

According to the GitHub mirror gcc-mirror/gcc project the gcc-5-branch when looking at the commit history it shows that there has been changes made till as late as Oct 10, 2017. But it would appear that we forked this project with the last commit 9d05077 made on Jun 3, 2016.

I know this is just a mirror but we should probably make an effort to rebase against the official source and catch up with any updates made.

GCC version 5

As of this writing the latest linux kernel v5.2.0 has a minimum gcc requirement of 4.6 so we are sill safe for now as far as the kernel is concerned anyway.

Is there any specific reason why these modifications were only made to version 5? Is there anything preventing us from applying these changes to a more modern release or any special considerations that should be taken?

[nickl-]

Comment on issue GCC version 5:

After further investigation the branch epiphany-gcc-8 was found which got updated to tag gcc-8_3_0-release in March 2019. With a bit of tweaking it is possible to build a gcc 8.3 version of the eSDK. Now if only we could get the updates merged then development can continue.