Skip to content

Commit 1ad7689

Browse files
committed
Update docs for 1.1.0 release
Signed-off-by: Jay Gu <jagu@nvidia.com>
1 parent 5ade61c commit 1ad7689

21 files changed

+182
-171
lines changed

CHANGELOG.md

Lines changed: 32 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -3,6 +3,38 @@
33

44
Release Notes
55
=============
6+
1.1.0 (2025-01-30)
7+
------------------
8+
### Features
9+
- Add support for nested functions and lambdas.
10+
- Add support for custom reduction via `ct.reduce()`.
11+
- Add `Array.slice(axis, start, stop)` to create a view of an array sliced along a single axis.
12+
The result shares memory with the original array (no data copy).
13+
14+
### Bug Fixes
15+
- Fix reductions with multiple axes specified in non-increasing order.
16+
- Fix a bug when pattern matching (FusedMultiplyAdd) attempts to remove a value that is used by the new operation.
17+
18+
### Enhancements
19+
- Allow assignments with type annotations. Type annotations are ignored.
20+
- Support constructors of built-in numeric types (bool, int, float), e.g., `float('inf')`.
21+
- Lift the ban on recursive helper function calls. Instead, add a limit on recursion depth.
22+
Add a new exception class `TileRecursionError`, thrown at compile time when the recursion limit
23+
is reached during function call inlining.
24+
- Improve error messages for type mismatches in control flow statements.
25+
- Relax type checking rules for variables that are assigned a different type
26+
depending on the branch taken: it is now only an error if the variable is used
27+
afterwards.
28+
- Stricter rules for potentially-undefined variable detection: if a variable
29+
is first assigned inside a `for` loop, and then used after the loop,
30+
it is now an error because the loop may take zero iterations, resulting
31+
in a use of an undefined variable.
32+
- Include a full cuTile traceback in error messages. Improve formatting of code locations;
33+
include function names, remove unnecessary characters to reduce line lengths.
34+
- Delay the loading of CUDA driver until kernel launch.
35+
- Expose the `TileError` base class in the public API.
36+
- Add `ct.abs()` for completeness.
37+
638

739
1.0.1 (2025-12-18)
840
------------------

changelog.d/allow-recursion.md

Lines changed: 0 additions & 10 deletions
This file was deleted.

changelog.d/ann-assign.md

Lines changed: 0 additions & 4 deletions
This file was deleted.

changelog.d/array-slice.md

Lines changed: 0 additions & 5 deletions
This file was deleted.

changelog.d/closures.md

Lines changed: 0 additions & 4 deletions
This file was deleted.

changelog.d/ctabs.md

Lines changed: 0 additions & 4 deletions
This file was deleted.

changelog.d/custom-reduction.md

Lines changed: 0 additions & 4 deletions
This file was deleted.

changelog.d/fix-reduce-multiaxis.md

Lines changed: 0 additions & 4 deletions
This file was deleted.

changelog.d/fix-rewrite-pattern.md

Lines changed: 0 additions & 4 deletions
Original file line numberDiff line numberDiff line change
@@ -1,4 +0,0 @@
1-
<!--- SPDX-FileCopyrightText: Copyright (c) <2025> NVIDIA CORPORATION & AFFILIATES. All rights reserved. -->
2-
<!--- SPDX-License-Identifier: Apache-2.0 -->
3-
4-
- Fixed a bug when pattern match attempted to remove a value that is used by the new operation

changelog.d/lazy-load-cuda-driver.md

Lines changed: 0 additions & 4 deletions
This file was deleted.

0 commit comments

Comments
 (0)