Commit Graph

9 Commits

Author SHA1 Message Date
Andi Drebes
55a6041943 feat(compiler): Add support for tiled TFHE IR in TFHE parametrization pass
Add the required patterns for `tensor.empty` and `scf.forall` to the
TFHE circuit parametrization pass.
2024-03-14 06:32:19 +01:00
Benoit Chevallier-Mames
e37a840e7e docs(compiler): adding API doc 2024-03-13 17:10:00 +01:00
Andi Drebes
9b6878316f fix(compiler): Preserve explicit optimizer partition boundaries through the pipeline
The Concrete Optimizer is invoked on a representation of the program
in the high-level FHELinalg / FHE Dialects and yields a solution with
a one-to-one mapping of operations to keys. However, the abstractions
used by these dialects do not allow for references to keys and the
application of the solution is delayed until the pipeline reaches a
representation of the program in the lower-level TFHE dialect. Various
transformations applied by the pipeline along the way may break the
one-to-one mapping and add indirections into producer-consumer
relationships, resulting in ambiguous or partial mappings of TFHE
operations to the keys. In particular, explicit frontiers between
optimizer partitions may not be recovered.

This commit preserves explicit frontiers between optimizer partitions
as `optimizer.partition_frontier` operations and lowers these to
keyswitch operations before parametrization of TFHE operations.
2024-03-07 15:42:26 +01:00
Andi Drebes
a701b3a742 feat(compiler): Add support for tensor.empty in the pipeline from FHE to std 2024-03-06 14:50:27 +01:00
Andi Drebes
21a7eead6c refactor(compiler): Re-implement TFHE multi-parameter parametrization with type inference
The current pass applying the parameters determined by the optimizer
to the IR propagates the parametrized TFHE types to operations not
directly tagged with an optimizer ID only under certain conditions. In
particular, it does not always properly propagate types into nested
regions (e.g., of `scf.for` loops).

This burdens preceding transformations that are applied in between the
invocation of the optimizer and the parametrization pass with
data-flow analysis and book-keeping in order to tag newly inserted
operations with the right optimizer IDs that ensure proper
parametrization.

This commit replaces the current parametrization pass with a new pass
that propagates parametrized TFHE types up and down def-use chains
using type inference and a proper rewriter. The pass is limited to the
operations supported by `TFHEParametrizationTypeResolver::resolve`.
2024-02-01 16:18:45 +01:00
Bourgerie Quentin
b6da228dd4 fix(compiler): Add extra conversion keyswitch just after bootstrap
Rely on strong asumptions of the optimization (see comment)

close https://github.com/zama-ai/concrete-internal/issues/352
2023-06-21 08:53:36 +02:00
Quentin Bourgerie
d973f068f6 fix(compiler): Fixing extra conversion fixup when producer is not directly a TFHE operator 2023-05-12 09:47:35 +02:00
Quentin Bourgerie
0fa1b4ba1d refactor(compiler): Cleanup TFHECircuitSolutionParametrization and make it less verbose in verbose mode 2023-05-12 09:47:35 +02:00
Quentin Bourgerie
7d1c43bc47 feat(compiler/multi-parameters): Create a pass to apply the multi-parameter circuit solution of the optimize 2023-04-14 15:01:18 +02:00