Skip to content

Add README.md for berkeley-softfloat-3 library - #1147

Merged
Prashanth Mundkur (pmundkur) merged 1 commit into
riscv:masterfrom
nadime15:add_berkeley_softfloat_3_readme
Jul 18, 2025
Merged

Add README.md for berkeley-softfloat-3 library#1147
Prashanth Mundkur (pmundkur) merged 1 commit into
riscv:masterfrom
nadime15:add_berkeley_softfloat_3_readme

Conversation

@nadime15

Copy link
Copy Markdown
Collaborator

GitHub did not render README.txt or README.html, so this adds a README.md.

GitHub did not render README.txt or README.html, so this adds a README.md.
@github-actions

Copy link
Copy Markdown

Test Results

2 099 tests  ±0   2 099 ✅ ±0   17m 59s ⏱️ +3s
    1 suites ±0       0 💤 ±0 
    1 files   ±0       0 ❌ ±0 

Results for commit 9dede26. ± Comparison against base commit 94a4fa0.

@pmundkur Prashanth Mundkur (pmundkur) added the will be merged Scheduled to be merged soon if nobody objects label Jul 16, 2025
@Timmmm

Copy link
Copy Markdown
Collaborator

I dunno, this seems a little unnecessary and it also means this directory isn't just a straight copy of the upstream code. I guess it doesn't matter too much.

@nadime15

Copy link
Copy Markdown
Collaborator Author

The README.md is actually part of the berkeley-softfloat-3 repository.

I recommend that you have a look at https://github.com/riscv/sail-riscv/tree/master/dependencies/softfloat/berkeley-softfloat-3 and see how its rendered now.

@nadime15

Copy link
Copy Markdown
Collaborator Author

Btw. the code that we have right now is not a straight copy. 99% of the code is the same but some files are different including (at least) one folder name that was changed from (Linux-RISCV64-GCC to Linux-RISCV-GCC) including the files in the folder.

@Timmmm

Copy link
Copy Markdown
Collaborator

Ah ok fair enough. I thought I updated it so that it was a whole ago but I must have not actually made that PR.

Copy link
Copy Markdown
Collaborator

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

This is fine for now. Might make more sense to just update the whole directory to get it back in line with upstream (or the Spike version) and bring in the bfloat16 additions.

@nadime15

Copy link
Copy Markdown
Collaborator Author

Jordan Carlin (@jordancarlin) Yes, I am working on it

@pmundkur

Prashanth Mundkur (pmundkur) commented Jul 17, 2025

Copy link
Copy Markdown
Collaborator

A Spike install creates a softfloat include directory and libsoftfloat.so library; we could use that at the expense of adding a build dependency. That's an alternative to bringing in the whole directory and then tracking upstream fixes. Either option sounds good to me. Spike's softfloat has more updates than the github ucb-bar version, so the former (Spike's) might be preferable (to ucb-bar).

@nadime15

Copy link
Copy Markdown
Collaborator Author

To be honest, I would prefer not to merge the softfloat version from Spike and stick with berkeley-softfloat-3. Its harder to maintain the library otherwise, and we would have to keep track of changes within Spikes codebase. As far as I can tell the berkeley repository has everything we need for now.

Its true that Spike has broader support (they even support direct bfloat16 arithmetic for some reason, which the upstream Berkeley version currently does not), but there is no reason Spike couldnt push those changes upstream. In fact, some of the people maintaining Spike also maintain the Berkeley library.

@nadime15

Copy link
Copy Markdown
Collaborator Author

See ucb-bar/berkeley-softfloat-3#24 (comment) and ucb-bar/berkeley-softfloat-3#26

@pmundkur
Prashanth Mundkur (pmundkur) added this pull request to the merge queue Jul 18, 2025
Merged via the queue into riscv:master with commit 86dabeb Jul 18, 2025
7 checks passed
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

will be merged Scheduled to be merged soon if nobody objects

Projects

None yet

Development

Successfully merging this pull request may close these issues.

4 participants