Skip to content

Commit 6661798

Browse files
committed
More docs
1 parent 30d5c6a commit 6661798

File tree

2 files changed

+25
-4
lines changed

2 files changed

+25
-4
lines changed

README.md

Lines changed: 2 additions & 1 deletion
Original file line numberDiff line numberDiff line change
@@ -10,7 +10,8 @@ handling various current and proposed floating point types:
1010
- [IEEE WG P3109](https://github.com/awf/P3109-Public/blob/main/Shared%20Reports/P3109%20WG%20Interim%20report.pdf): P{p} for p in 1..7
1111
- [OCP MX Formats](https://www.opencompute.org/documents/ocp-microscaling-formats-mx-v1-0-spec-final-pdf): E2M1, M2M3, E3M2, E8M0, INT8, and the MX block formats.
1212

13-
The library favours readability and extensibility over speed - for fast implementations of these datatypes see, for example, [ml_dtypes](https://github.com/jax-ml/ml_dtypes),
13+
The library favours readability and extensibility over speed (although the *_ndarray functions are reasonably fast for large arrays, see the [benchmarking notebook](docs/source/04-benchmark.ipynb)).
14+
For other implementations of these datatypes more focused on speed see, for example, [ml_dtypes](https://github.com/jax-ml/ml_dtypes),
1415
[bitstring](https://github.com/scott-griffiths/bitstring),
1516
[MX PyTorch Emulation Library](https://github.com/microsoft/microxcaling).
1617

docs/source/index.rst

Lines changed: 23 additions & 3 deletions
Original file line numberDiff line numberDiff line change
@@ -31,9 +31,29 @@ handling various current and proposed floating point types:
3131

3232
- `IEEE 754 <https://en.wikipedia.org/wiki/IEEE_754>`_: Binary16, Binary32
3333
- `Brain floating point <https://en.wikipedia.org/wiki/Bfloat16_floating-point_format>`_: BFloat16
34-
- `OCP Float8 <https://www.opencompute.org/documents/ocp-8-bit-floating-point-specification-ofp8-revision-1-0-2023-06-20-pdf>`_: E5M2, E4M3
35-
- `IEEE WG P3109 <https://github.com/awf/P3109-Public/blob/main/Shared%20Reports/P3109%20WG%20Interim%20report.pdf>`_: P{p} for p in 1..7
36-
- Types from the `OCP MX <https://www.opencompute.org/documents/ocp-microscaling-formats-mx-v1-0-spec-final-pdf>`_ spec: E8M0, INT8, and FP4, FP6 types
34+
- |ocp_link|: E5M2, E4M3
35+
- |p3109_link|: P{p} for p in 1..7
36+
- Types from the |ocp_mx_link| spec: E8M0, INT8, and FP4, FP6 types
37+
38+
As well as block formats from |ocp_mx_link|.
39+
40+
.. |ocp_mx_link| raw:: html
41+
42+
<a href="https://www.opencompute.org/documents/ocp-microscaling-formats-mx-v1-0-spec-final-pdf" target="_blank">
43+
OCP MX
44+
</a>
45+
46+
.. |ocp_link| raw:: html
47+
48+
<a href="https://www.opencompute.org/documents/ocp-8-bit-floating-point-specification-ofp8-revision-1-0-2023-06-20-pdf" target="_blank">
49+
OCP Float8
50+
</a>
51+
52+
.. |p3109_link| raw:: html
53+
54+
<a href="https://raw.githubusercontent.com/P3109/Public/main/Shared%20Reports/P3109%20WG%20Interim%20Report.pdf" target="_blank">
55+
IEEE P3109
56+
</a>
3757

3858
Example
3959
-------

0 commit comments

Comments
 (0)