apache / apache/arrow

[Format] Serialization of fp16 literals via Substrait is under-defined

Open
#47,846 0 comments 0 reactions 1 assignee Claimed by @westonpace View on GitHub
Component: Format Type: enhancement
Dominant language
C++
Stars
17.1k
Forks
4.3k
Avg merge
3d 13h
Merged PRs (30d)
88

Description

### Describe the enhancement requested

Currently we state:

```
# fp16 literals are encoded as user defined literals with
# a google.protobuf.UInt32Value message where the lower 16 bits are
# the fp16 value.
```

I think "lower 16 bits of UInt32Value" is well defined. However, it is not clear how the fp16 value should be encoded. Should the upper byte of the fp16 value go in the lowest 8 bits of the UInt32Value or the highest 8 bits of the UInt32Value?

Hopefully this is not controversial and just under-specified. I will add a PR to adopt little-endian encoding.

### Component(s)

Format

Contributor guide

Open the contributing guide

Assessment

This issue has not been assessed yet.

Get new issues in your inbox

A short digest of beginner-friendly GitHub issues.