nightly

fix: fix lwe-to-lattigo for preprocessing funcs w/ no args

When there are no crypto arguments, the preprocessing function won't have params/encoder added, so this fixes that. It updates the legality condition on the call op to match with the func op signature, which makes it a lot cleaner to deduce legality than checking for which evaluators are needed in the callsite.

PiperOrigin-RevId: 874222580

v2026.02.01

What's Changed

• Upgrade GitHub Actions for Node 24 compatibility by @salmanmkc in #2467
• migrate pocketfft and cereal to MODULE.bazel by @j2kun in #2470
• Add a faster heuristic for IntegerRelation::isEqual by @j2kun in #2505
• fix: loosen tolerance on polynomial eval fuzz test by @copybara-service[bot] in #2503
• rework community page, add bookable slots by @copybara-service[bot] in #2508
• fix: fix bootstrap insertion pattern for CKKS by @copybara-service[bot] in #2511
• Bump the pip group across 2 directories with 4 updates by @dependabot[bot] in #2512
• Issue #202 : Add tensor shape to polynomial mod function name by @TomiAkeju in #2502
• Use opaque ciphertext/plaintext type for openfhe by @copybara-service[bot] in #2471
• Replace unicode escaped characters in ipynb files by @copybara-service[bot] in #2518
• add openfhe benchmarks for fast rotations by @asraa in #2524
• bump rust dependencies by @j2kun in #2528
• Add in-place OpenFHE ops and in-place transform by @copybara-service[bot] in #2520
• migrate llvm dependency to bazelmod by @j2kun in #2472
• Peel first iteration of loops with ciphertext iter-args (Halo Solution A-1) by @j2kun in #2534
• Bootstrap loop iter args (Halo Solution A-2) by @j2kun in #2536
• Rewrite level analysis for loop support by @j2kun in #2551
• add AI tool use policy by @copybara-service[bot] in #2530
• pre-commit: use python3 instead of python by @AlexanderViand in #2547
• make README example also work with pypi version by @AlexanderViand in #2562
• frontend: use signed extension when changing bitwidth by @AlexanderViand in #2546
• support parallelization in openfhe targets by @copybara-service[bot] in #2535
• split-preprocessing: handle plaintext function arguments that have layouts assigned by @copybara-service[bot] in #2563
• Upgrade GitHub Actions for Node 24 compatibility by @salmanmkc in #2567
• make ai tool use policy page visible by @j2kun in #2566
• Frontend literal upcasting by @AlexanderViand in #2552
• Partially unrolling loops for better level consumption (Halo B-2) by @j2kun in #2538
• Migrate OpenFHE to bazelmod by @j2kun in #2548
• Optimisation: Extend the Lut generation to LUT4 elements by @WoutLegiest in #2458
• fix build by @copybara-service[bot] in #2573
• bazelrc: update openfhe flags to bzlmod version by @AlexanderViand in #2576
• openfhe: use openfhe API GetBootstrapDepth to get actual bootstrap depth by @copybara-service[bot] in #2578
• layout: generate packing functions for tensor_ext.assign_layout lowerings by @copybara-service[bot] in #2569
• remove linalgDetensorize and some other lingering tosa stuff by @copybara-service[bot] in #2582
• Add primitive root calculator based on sympy by @copybara-service[bot] in #2583
• range analysis: propagate state on plaintexts operations by @copybara-service[bot] in #2584
• openfhe: support multiple function return types in emitter by @copybara-service[bot] in #2585

New Contributors

• @TomiAkeju made their first contribution in #2502

Full Changelog: v2026.01.01...v2026.02.01

v2026.01.01

What's Changed

• support dense resources in openfhe pke emitter by @j2kun in #2465
• layout: ensure code-genned cmpi has agreeing arg types by @j2kun in #2469
• Upgrade GitHub Actions to latest versions by @salmanmkc in #2468
• Fix pypa/gh-action-pypi-publish version to use SHA pinning by @salmanmkc in #2473
• test: add LeNet MNIST CNN model e2e test by @copybara-service[bot] in #2475
• Migrates builder.create<Op>() => Op::create() in tablegen files by @copybara-service[bot] in #2490

New Contributors

• @salmanmkc made their first contribution in #2468

Full Changelog: v2025.12.11...v2026.01.01

v2025.12.11

What's Changed

• fix: fix openfhe e2e 2-d conv test by @copybara-service[bot] in #2433
• add openfhe ckks rotom chained_matmul test by @edwjchen in #2432
• OpenFHE interpreter overhead reduction (lenet example) by @j2kun in #2445
• Convert orion to lattigo by @j2kun in #2431
• remove deprecated rewriter.create calls by @copybara-service[bot] in #2452
• linalg: ensure linalg.generic attrs are preserved in GenericToElementwise transform by @copybara-service[bot] in #2459
• performance: skip over any zero diagonal rows in a 2-D conv by @copybara-service[bot] in #2460
• add binaries (heir-opt etc) to monthly releases by @copybara-service[bot] in #2461

New Contributors

• @edwjchen made their first contribution in #2432

Full Changelog: v2025.12.01...v2025.12.11

v2025.12.01

What's Changed

• add pass to forward values to tensor.extract_slice by @copybara-service[bot] in #2367
• update lockfile by @j2kun in #2373
• enable layout propagation for tensor.extract_slice by @copybara-service[bot] in #2353
• implement-shift-network: de-dupe ctslot mappings to targets by @copybara-service[bot] in #2354
• ckks: support tensor.insert_slice and tensor.extract_slice in secret-to-ckks by @copybara-service[bot] in #2374
• Docs add test example by @isabelfaulds in #2377
• Add fuzz testing framework by @j2kun in #2196
• Add BSGS support for bicyclic matmul by @j2kun in #2366
• secret: allow plaintext linalg ops to be hoisted out of secret.generic by @copybara-service[bot] in #2390
• secret-to-ckks: support arith.divf by a plaintext by @copybara-service[bot] in #2393
• Replace some uses of return failure() with notifyMatchFailure by @j2kun in #2382
• Add bicyclic matmul to layout pipeline by @j2kun in #2360
• Refactor other lower-polynomial-eval methods to use ArithmeticDag, drop small coeffs by @j2kun in #2167
• Modified BSGS implementation to address #2162 for suboptimal implementation of BSGS choices by @j2kun in #2392
• migrate some rules_hdl deps to BCR equivalents with better macos support by @j2kun in #2406
• Address performance of secret-insert-mgmt passes by @j2kun in #2391
• #2388 - added tricyclic batch ctct matmul by @rishabhjain1712 in #2399
• Add type inference for LWE/CKKS ops by @j2kun in #2412
• Follow up Remove Tensor from CGGI: Update the Emitters and tests by @WoutLegiest in #2278
• CKKS: Use prime selection logic from OpenFHE by @ZenithalHourlyRate in #2402
• Improve Chebyshev Paterson-Stockmeyer evaluation by @j2kun in #2411
• Add timing table to interpreter mode by @j2kun in #2423
• internal change by @copybara-service[bot] in #2426
• migrate some bazel rules to MODULE.bazel, bump some versions by @j2kun in #2419
• Bump golang.org/x/crypto from 0.35.0 to 0.45.0 in the go_modules group across 1 directory by @dependabot[bot] in #2427
• add openfhe benchmark binary by @j2kun in #2425
• fix: openfhe interpreter fails to update SSA values in loops by @copybara-service[bot] in #2428
• add canonicalization pattern for sigmoid operation by @copybara-service[bot] in #2396
• fix benchmark copybara transform by @copybara-service[bot] in #2430

New Contributors

• @isabelfaulds made their first contribution in #2377
• @rishabhjain1712 made their first contribution in #2399

Full Changelog: v2025.11.01...v2025.12.01

v2025.11.01

What's Changed

• Make BGV/CKKS ops elementwise-mappable, and add ElementwiseByOperandOpInterface by @j2kun in #2277
• Update plaintext tests by @Time0o in #2245
• Add insert/extract slice to noise analysis by @j2kun in #2273
• Remove AutoHog from HEIR by @copybara-service[bot] in #2287
• blog post for 2025-10-02 HEIR meeting by @copybara-service[bot] in #2286
• Run Vos-Vos-Erkin with multiple random shift orderings by @Time0o in #2271
• ensure debug helper fn can be linked properly by @copybara-service[bot] in #2288
• add extract_slice_of_splat pattern by @j2kun in #2290
• Generalize tensor_ext_to_tensor to support dynamic and multidimensional rotate ops by @j2kun in #2291
• rewrite linalg.pooling_nchw_sum as linalg.conv_2d_nchw_fchw. by @copybara-service[bot] in #2279
• Remove old layout code by @j2kun in #2199
• Remove ckks.extract by @j2kun in #2302
• lockfile update by @j2kun in #2300
• [NFC] rename 'NewLayout' to Layout in its various forms by @j2kun in #2299
• [NFC]: Rename tensor_ext.permute to tensor_ext.remap by @j2kun in #2301
• fix bug in conv2d layout by @j2kun in #2307
• remove redundant pattern by @copybara-service[bot] in #2308
• linalg canonicalizations and drop unit dims enhancements by @copybara-service[bot] in #2283
• relu-canonicalizations: add pass to rewrite select(a >= c, a, c) to max(a, c) by @copybara-service[bot] in #2285
• layout: handle non-rank reducing collapse_shape by @copybara-service[bot] in #2284
• Put convert-layout folder into layout-propagation as cleanup by @j2kun in #2305
• Add doctests to all passes in layout design doc (#2296) by @j2kun in #2306
• Fix random network comparison in shift network tests by @Time0o in #2315
• layout: relax layout conditions to lower halevi-shoup kernel by @copybara-service[bot] in #2313
• Fix #2047 by cleaning up last todos by @j2kun in #2322
• Add packing design doc to the HEIR website by @j2kun in #2296
• Disable mdformat for good by @j2kun in #2324
• fix frontmatter for real this time by @copybara-service[bot] in #2327
• fix formatting issues in layout design doc by @copybara-service[bot] in #2333
• Add a test for lower poly eval. Also add --lower-polynomial-eval pass to heir-polynomial-to-llvm pipeline. by @copybara-service[bot] in #2329
• BGV: NoiseAnalysis KPZ21: more conservative on ct-pt mul by @ZenithalHourlyRate in #2325
• Remove manual cyclic repetition in LattigoEmitter by @j2kun in #2319
• Add an OpenFHE dialect interpreter by @j2kun in #2314
• add canonicalizer pass before implement-rotate-and-reduce by @j2kun in #2303
• fix: tensor-linalg-to-affine-loops fix for 0-D tensors by @copybara-service[bot] in #2341
• [docs] update docs to match current pipelines by @copybara-service[bot] in #2340
• Add fold-plaintext-masks for simple cleanup after implement-shift-network by @j2kun in #2343
• Remove todo about emitting Yosys's internal representation by @copybara-service[bot] in #2346
• add support for tensor.insert_slice in layout optimization by @copybara-service[bot] in #2335
• Add orion entry dialect by @j2kun in #2330
• Separate TODO into isolated issue by @j2kun in #2321
• test: add e2e test for a 2-D convolution in isolation by @copybara-service[bot] in #2357
• Add DAG-based kernel cost model for layout optimization by @bon-cdp in #2347
• Add bicyclic layout and matmul kernel by @j2kun in #2359

New Contributors

• @bon-cdp made their first contribution in #2347

Full Changelog: v2025.10.01...v2025.11.01

v2025.10.01

What's Changed

• migrate halevi-shoup matvec rewrite to kernel implementation using arithmetic DAG by @copybara-service[bot] in #2183
• add blog post for 2025-09-04 by @copybara-service[bot] in #2194
• Add a split-preprocessing pass by @copybara-service[bot] in #2185
• Automated Code Change by @copybara-service[bot] in #2197
• update lockfile from #2195 by @asraa in #2198
• workflows: install llvm tools from apt.llvm.org by @j2kun in #2179
• Add more doctests by @j2kun in #2186
• Add support for assigning new layouts for scalars by @copybara-service[bot] in #2200
• Add a tensor.insert kernel for new layouts by @j2kun in #2204
• fix mod_arith_to_arith runner tests by @j2kun in #2212
• diable allowPatternRollback in non-context-aware dialect conversion drivers by @j2kun in #2211
• Add layout propagation for vecmat and matmuls by @copybara-service[bot] in #2207
• tensor.extract kernel with new layout by @j2kun in #2205
• poly approximation: emit a chebyshev poly from the C-F approx to enable chebyshev basis lowerings by @copybara-service[bot] in #2191
• update lockfile by @j2kun in #2220
• add insert to new-layout-propagation by @j2kun in #2219
• Fix one missed API update by @copybara-service[bot] in #2221
• split out utils for making constants out of values by @copybara-service[bot] in #2222
• remove old TODOs we won't do by @j2kun in #2230
• Partition RNS limbs (not poly coefficients) in key_switch_inner lowering by @j2kun in #2208
• fix ubsan error: use OpConversionPattern with conversion driver by @copybara-service[bot] in #2232
• Remove TOSA from HEIR by @j2kun in #2229
• Minor cleanup of emitters by @j2kun in #2228
• fix build by @j2kun in #2233
• Fix workflows by doing checkout first by @j2kun in #2224
• Fix issues with Docker setup by @mabhay3420 in #2235
• Layout folding analysis by @j2kun in #2141
• Add integer relations for a 2-D convolution by @copybara-service[bot] in #2218
• try experimenting with macos ci by @j2kun in #2225
• Initial Sbox by @WoutLegiest in #2012
• update lockfile by @j2kun in #2243
• frontend: update to numba 0.62, simplify dependencies by @AlexanderViand in #2244
• add 2d convolution support in new-layout-propagation by @copybara-service[bot] in #2242
• Refactor ISL codegen using visitor pattern. by @copybara-service[bot] in #2247
• frontend: Emit ordered comparison predicates for floats by @Time0o in #2246
• MNIST via py_test and python torch dep by @j2kun in #2226
• fix requirements.txt breakage by @j2kun in #2250
• Add utils for vos-vos-erkin generalization by @j2kun in #2252
• remove stale code parsing integerrelations from integersets by @copybara-service[bot] in #2261
• Lower linalg.conv2d to tensor_ext in convert-to-ciphertext-semantics by @copybara-service[bot] in #2260
• Generalize Vos-Vos-Erkin algorithm to support arbitrary mappings by @j2kun in #2240
• test: add kernel testing with layout assignment by @copybara-service[bot] in #2266
• remove remaining uses of mlir_runner by @j2kun in #2274
• bump rules_hdl version for fixing fetch failures by @j2kun in #2276
• GenerateParam: update doc by @ZenithalHourlyRate in #2272

New Contributors

• @mabhay3420 made their first contribution in #2235
• @Time0o made their first contribution in #2246

Full Changelog: v2025.09.05...v2025.10.01

v0.0.2

This is the first release of the HEIR compiler.

For now we have no special launch or feature set. We're just tying GitHub releases to the publishing of heir_py on PyPI. v0.0.1 was a manual push of the wheels to PyPI. This release is to test the automated GitHub release workflow.