[type-classes] 8380667: Prototype FP8_E5M2

[jddarcy]

First cut at a prototype of an FP8 type, this one with 5 exponent bits and 2 significand (mantissa) bits.

The base conversion functionality is not fully ported over.

---

Progress

• Change must not contain extraneous whitespace

Reviewing

Using git

Checkout this PR locally:
$ git fetch https://git.openjdk.org/valhalla.git pull/2259/head:pull/2259
$ git checkout pull/2259

Update a local copy of the PR:
$ git checkout pull/2259
$ git pull https://git.openjdk.org/valhalla.git pull/2259/head

Using Skara CLI tools

Checkout this PR locally:
$ git pr checkout 2259

View PR using the GUI difftool:
$ git pr show -t 2259

Using diff file

Download this PR as a diff file:
https://git.openjdk.org/valhalla/pull/2259.diff

Using Webrev

Link to Webrev Comment

[bridgekeeper]

👋 Welcome back darcy! A progress list of the required criteria for merging this PR into type-classes will be added to the body of your pull request. There are additional pull request commands available for use with this pull request.

[openjdk]

@jddarcy This change now passes all automated pre-integration checks.

ℹ️ This project also has non-automated pre-integration requirements. Please see the file CONTRIBUTING.md for details.

After integration, the commit message for the final commit will be:

[type-classes] 8380667: Prototype FP8_E5M2

You can use pull request commands such as /summary, /contributor and /issue to adjust it as needed.

At the time when this comment was updated there had been no new commits pushed to the type-classes branch. If another commit should be pushed before you perform the /integrate command, your PR will be automatically rebased. If you prefer to avoid any potential automatic rebasing, please check the documentation for the /integrate command for further details.

➡️ To integrate this PR with the above commit message to the type-classes branch, type /integrate in a new comment.

[mlbridge]

Webrevs

• 03: Full - Incremental (c2bbd05f)
• 02: Full - Incremental (0ea84c5a)
• 01: Full - Incremental (6efe758a)
• 00: Full (49a3ae71)

[jddarcy]

If reviewers would find it helpful, I can upload here or attach to the bug a result of diffing Float16.java compared to Float8_E5M2.java with "Float16" replaced by a "FLOAT_TYPE" token and likewise for "Float8_E5M2" in its file. This removes many inessential differences between the two files, but is still about 600 lines of changes.

Many of the remaining changes stem from different parameter names, "f16" vs "f8", and the use of "short" vs "byte".

[jddarcy]

Pushed a small refactoring: the floating-point valued constants are initialized using the valueOf factories rather than setting the bits directly. Checking the setting of the bits is moved to the regression test.