output size of keyswiting wasn't set properly. As this information must
come from the selected parameters, it should goes down from the MidLFHE
to the appropriate call to ciphertext allocation
The polynomialSize is currently holding its log2 instead of the actual
value. This should be fixed later, but in the meantime, we need to
compute it from log2
This changes the semantics of `HLFHE.dot_eint_int` from memref-based
reference semantics to tensor-based value semantics. The former:
"HLFHE.dot_eint_int"(%arg0, %arg1, %arg2) :
(memref<Nx!HLFHE.eint<0>>, memref<Nxi32>, memref<!HLFHE.eint<0>>) -> ()
becomes:
"HLFHE.dot_eint_int"(%arg0, %arg1) :
(tensor<Nx!HLFHE.eint<0>>, tensor<Nxi32>) -> !HLFHE.eint<0>
As a side effect, data-flow analyses become much easier. With the
previous memref type of the plaintext argument it is difficult to
check whether the plaintext values are statically defined constants or
originate from a memory region changed at execution time (e.g., for
analyses evaluating the impact on noise). Changing the plaintext type
from `memref` to `vector` makes such analyses significantly easier.
-Encoding Cleartext to Plaintext Op
-Encoding Int to Plaintext Op
-Cleartext/Plaintext should add a bits field as `p`
-Op to create a constant Cleartext
-IntToCleartextOp: kind of casting an int to be later encoded, I'm not
sure if there is a better mechanism for this (e.g. auto casting in Ops),
but we currently need a way to encode int into plaintext, and we need to
go through cleartexts
also fixes an issue regarding populateWithGenerated, which can be
duplicated across different pattern files. So I redefined a different
function that is more unique to the pass that should be ran, and hide
the populateWithGenerated from the global namespace
* feat(compiler): low level fhe dialect
* feat(compiler): using generated printer/parser in LowLFHE
* feat(compiler): new types and ops for LowLFHE
* tests(compiler): LowLFHE types and ops
* feat(compiler): fill ops
* cleanup
* summary + description
* tests(compiler): use new CLI args
* formatting
- feat(compiler): python bindings
- build: update docker image for python bindings
- pin pybind11 to 2.6.2, 2.7 is not having correct include_dirs set (still
a question why?)
- using generated parser/printer
Add a simple dot product between a vector of encrypted integers and a
vector of plaintext integers to the HLFHE dialect.
The operation takes two input operands and one output operand, all
modeled as memrefs. Example:
"HLFHE.dot_eint_int"(%lhs, %rhs, %out) :
(memref<?x!HLFHE.eint<0>>, memref<?xi32>, memref<!HLFHE.eint<0>>) -> ()
* fix(compiler): Generate and include autogenerated dialect definitions
Since commit 485cc55edfb875628e19bb6d9de4706af2865d3e in llvm-project,
dialects are now required to include a file with autogenerated
definitions. This commit generates the definition files and includes
them in appropriate dialect source files for MidLFHE and HLFHE.
* CI: update docker image ref
Co-authored-by: Andi Drebes <andi.drebes@zama.ai>
Co-authored-by: Ayoub Benaissa <ayoub.benaissa@zama.ai>
Make `HLFHE.EncryptedInteger` implement the
`MemRefElementTypeInterface` trait to allow the type to be used for
elements of memrefs.
The invocation of the `add_mlir_dialect` macro in
`include/zamalang/Dialect/HLFHE/IR/CMakeLists.txt` had to be broken
into its parts in order to pass `-typedefs-dialect=HLFHE` to
`mlir-tablegen`. This is required due to the inclusion of
`mlir/IR/BuiltinTypes.td` in
`include/zamalang/Dialect/HLFHE/IR/HLFHETypes.td`, resulting in
definitions from multiple dialects.
Add a small test actually using `HLFE.EncryptedInteger` in a `memref`.
