Concrete v2.3.0
What's Changed
-
feat(frontend-python): Introduce fhe.hint to allows the end user to specify the operation bit width by @umut-sahin in #542
-
feat(frontend-python): Improve performance on comparison operators by using multi-precision by @umut-sahin in #549
-
docs(frontend-python): Improve comparisons tutorial by @umut-sahin in #556
-
feat(compiler): Add to the compiler feedback the estimated memory footprint per location by @youben11 in #538
-
feat(compiler): Expose more compilation options to the python bindings by @youben11 in #552
-
feat(optimizer): Improve optimization of levelled only circuit by removing the power of two constraint on lwe size by @rudy-6-4 in #541
-
fix(compiler): Fixing the lowering of fhe.zero to tfhe with crt encoding by @youben11 in #548
-
fix(optimizer): Fixing woppbs cost model by @rudy-6-4 in #540
-
fix(compiler): Disable high error warning when an explicit global_p_error is set by @rudy-6-4 in #511
-
fix(compiler): do not optimize llvm twice by @rudy-6-4 in #502
-
feat(compiler): Allow arbitrary integer bit width of the apply lookup table operator by @aPere3 in #554
Internal
- feat(concrete-cuda): Add multiplication of radix encoded integers by @bbarbakadze in #547
- refactor(compiler): clean statistic passes by @youben11 in #551
- chore(compiler/benchmark): Activate levelled + cifar-16 benchmark by @rudy-6-4 in #553
- fix(compiler): ci, format_cpp.sh, bug + better message by @rudy-6-4 in #557
- fix(compiler): Use FHE.zero_tensor instead of bufferization.alloc_ten… by @BourgerieQuentin in #561
- refactor(compiler): Interfaces to generalize noise calculation in FHE/FHELinalg ops by @mkmks in #416
New Contributors
Full Changelog: v2.2.0...v2.3.0