Skip to content

Tags: google/heir

Tags

nightly-2026.06.22

Toggle nightly-2026.06.22's commit message
bump BCR deps for hermetic LLVM (extracted from #3048)

PiperOrigin-RevId: 934564902

nightly-2026.06.21

Toggle nightly-2026.06.21's commit message
bump BCR deps for hermetic LLVM (extracted from #3048)

PiperOrigin-RevId: 934564902

nightly-2026.06.20

Toggle nightly-2026.06.20's commit message
bump BCR deps for hermetic LLVM (extracted from #3048)

PiperOrigin-RevId: 934564902

nightly-2026.06.19

Toggle nightly-2026.06.19's commit message
bump BCR deps for hermetic LLVM (extracted from #3048)

PiperOrigin-RevId: 934564902

nightly-2026.06.18

Toggle nightly-2026.06.18's commit message
Add an NTT/INTT and equivalence tests with OpenFHE

Nb., I am not an expert in NTT/INTT implementations, so I relied heavily on Gemini for the implementation. The equivalence tests should provide some additional certainty about the correctness of the implementation. The benchmark shows that for N=8192, Q=60 bits, this implementation is ~713.0 us, while OpenFHE's implementation is ~188.0 us.

One might ask: why not just use OpenFHE's implementation directly? We could, but at least having a fallback free of any dependencies is probably a good idea.

PiperOrigin-RevId: 934004133

nightly

Toggle nightly's commit message
bump BCR deps for hermetic LLVM (extracted from #3048)

PiperOrigin-RevId: 934564902

nightly-2026.06.17

Toggle nightly-2026.06.17's commit message
Optimize APInt factorization for 64-bit ints

This speeds up parameter generation specifically in the `-c dbg` path.

## Results (`-c opt`)

| Benchmark | Path | Time (ns) | Time/Iter (ns) | Speedup |
| :--- | :--- | :--- | :--- | :--- |
| **Small Prime** | Fast Path | 263,002 | 5.67 | **10.4x** |
| | Slow Path | 2,750,183 | 59.3 | |
| **Medium Composite** | Fast Path | 6,284,008 | 6.28 | **9.37x** |
| | Slow Path | 58,868,774 | 58.86 | |

---

## Results (`-c dbg`)

| Benchmark | Path | Time (ns) | Time/Iter (ns) | Speedup |
| :--- | :--- | :--- | :--- | :--- |
| **Small Prime** | Fast Path | 259,877 | 5.61 | **883x** |
| | Slow Path | 229,695,511 | 4956 | |
| **Medium Composite** | Fast Path | 6,184,147 | 6.18 | **784x** |
| | Slow Path | 4,848,338,349 | 4848 | |

PiperOrigin-RevId: 933376498

nightly-2026.06.16

Toggle nightly-2026.06.16's commit message
Decrease CKKS scaling factor in cc_mm test

PiperOrigin-RevId: 932721995

nightly-2026.06.15

Toggle nightly-2026.06.15's commit message
misc changes to support optimal bufferization

PiperOrigin-RevId: 931403381

nightly-2026.06.14

Toggle nightly-2026.06.14's commit message
misc changes to support optimal bufferization

PiperOrigin-RevId: 931403381