Skip to content

Releases: google/xls

v0.0.0-8793-g23d50957d

10 Oct 07:52

Choose a tag to compare

Remove capped `next_value_opt(3)` pass from optimization pipeline.

The `next_value_opt` pass is now run without a cap on its optimization level, as the default level is sufficient. The compound pass definition for `next_value_opt(3)` is also removed.

This has no affect on optimization decisions as nothing changes when opt level is higher than 3 (which is considered max level). It is mostly just a cleanup.

PiperOrigin-RevId: 817373790

v0.0.0-8788-geb308f0da

09 Oct 07:49

Choose a tag to compare

Changes to avoid extra memory reads with operator [] = form

PiperOrigin-RevId: 816927874

v0.0.0-8778-gfb15b8bf0

08 Oct 07:49

Choose a tag to compare

Merge pull request #2652 from antmicro:80072-code-location

PiperOrigin-RevId: 816381164

v0.0.0-8769-ga60e7a690

07 Oct 07:48

Choose a tag to compare

[Proc-scoped channels] Fix the toggling of converting test procs by s…

v0.0.0-8764-g31f1e4993

05 Oct 07:49

Choose a tag to compare

Automated Code Change

PiperOrigin-RevId: 815109595

v0.0.0-8763-g54c929f49

04 Oct 07:48

Choose a tag to compare

Model OHS 2-layer nand reduce for logical-effort delay.

This is not actually used in any delay model but this is a better model for how much logical effort is actually needed to implement a OHS.

PiperOrigin-RevId: 814811564

v0.0.0-8761-g634753551

03 Oct 07:51

Choose a tag to compare

Merge pull request #3133 from xlsynth:cdleary/2025-09-29-package-name…

v0.0.0-8758-g6904e585f

02 Oct 07:51

Choose a tag to compare

[xls][mlir] Move conversions between int/floats and `xls::Value`s to …

v0.0.0-8751-g772e7891a

01 Oct 07:51

Choose a tag to compare

Integrate LLVM at llvm/llvm-project@d28c07b7550a

Updates LLVM usage to match
[d28c07b7550a](https://github.com/llvm/llvm-project/commit/d28c07b7550a)

PiperOrigin-RevId: 813490970

v0.0.0-8744-g7677774fc

30 Sep 07:52

Choose a tag to compare

Make bdd_query_engine calculate AtMostOneTrue out to 300 bits

We previously wouldn't even try for more than 64 bits. By allowing more depth we get better results.

This also rewrites the checker slightly to avoid creating as large of a BDD improving performance.

PiperOrigin-RevId: 812934517