Commit Graph

362 Commits

Author SHA1 Message Date
Alexandre Péré
52ad40c9cf feat(concrete-compiler): adds a tracing op in all dialects. 2023-02-17 09:04:42 +01:00
Antoniu Pop
3ab10c8d3f fix(GPU runtime): fix BS & KS GPU allocation and synchronization, avoid re-allocation of KSK. 2023-02-14 08:22:15 +00:00
youben11
7e60f87141 refactor: remove chunked_eint
the pass transforming operations on chunked_eint will operate now on
eint
2023-02-07 12:27:01 +01:00
youben11
bb87d29934 feat: support chunked integer during enc/dec/exec 2023-02-07 12:27:01 +01:00
youben11
d41d14dbb8 feat: lower FHE.add on eint64 to ops on smaller chunks
this is a first commit to support operations on U64 by decomposing them
into smaller chunks (32 chunks of 2 bits). This commit introduce the
lowering pass that will be later populated to support other operations.
2023-02-07 12:27:01 +01:00
aPere3
fb680340f9 feat(concrete-compiler): add new ciphertext multiplication operator 2023-02-06 11:30:31 +01:00
aPere3
e95c53f2ff feat: support signed execution
Author: aPere3 <alexandre.pere@zama.ai>
Co-authored-by: Umut <umutsahin@protonmail.com>
2023-02-03 16:19:43 +01:00
Andi Drebes
73fd6c5fe7 refactor(compiler): FHE to TFHE: Use OpConversionPattern for dialect conversion
Use `OpConversionPattern` instead of `OpRewritePattern` for operation
conversion during dialect conversion. This makes explicit and in-place
type conversions unnecessary, since `OpConversionPattern` already
properly converts operand types and provides them to the rewrite rule
through an operation adaptor.

The main contributions of this commit are the two class templates
`TypeConvertingReinstantiationPattern` and
`GenericOneToOneOpConversionPattern`.

The former allows for the definition of a simple replacement rule that
re-instantiates an operation after the types of its operands have been
converted. This is especially useful for type-polymorphic operations
during dialect conversion.

The latter allows for the definition of patterns, where one operation
needs to be replaced with a different operation after conversion of
its operands.

The default implementations for the class templates provide
conversions rules for operations that have a generic builder method
that takes the desired return type(s), the operands and (optionally) a
set of attributes. How attributes are discarded during a conversion
(either by omitting the builder argument or by passing an empty set of
attributes) can be defined through specialization of
`ReinstantiationAttributeDismissalStrategy`.

Custom replacement rules that deviate from the scheme above should be
implemented by specializing
`TypeConvertingReinstantiationPattern::matchAndRewrite()` and
`GenericOneToOneOpConversionPattern::matchAndRewrite()`.
2023-02-01 14:27:10 +01:00
Quentin Bourgerie
49b8bf484c fix: Do not assert fail with too large weight and fix computation of 2-Norm with negative weigth (close #892) 2023-02-01 10:50:46 +01:00
Quentin Bourgerie
d0308dda43 chore: Use parameters curves generated files and expose security level options 2023-01-30 15:14:53 +01:00
youben11
36f51ba0c2 feat: lower and exec boolean ops 2023-01-26 11:22:41 +01:00
tmontaigu
ee00672996 feat(rust): partially bind CircuitGate, EncryptionGate
This adds partial bindings of the CircuitGate, EncrytionGate
and Encoding types for the rust frontend
2023-01-24 15:49:44 +01:00
youben11
95d49f4657 feat: add boolean types/ops in FHE dialect 2023-01-18 09:13:26 +01:00
youben11
91d41a2ff8 fix: add call to init dfr from python
Co-authored-by: Antoniu Pop <antoniu.pop@zama.ai>
2023-01-11 14:45:14 +01:00
youben11
8c6a0859cd refactor(rust): add Rust wrapper for every CStruct
we want to wrap CStructs in RustStructs to own them, and free memeory
when they are no longer used. Users won't have to deal with the direct
binded CAPI, but the new wrappers
2023-01-02 12:10:14 +01:00
Antoniu Pop
aa2e0479b3 feat(compiler): add a parallel loop coalescing pass. 2023-01-02 12:04:34 +01:00
rudy
e17bbcbab5 fix: macos compiling issue with extra ; 2022-12-22 20:08:28 +01:00
rudy
de779b2f6f fix: display compilation error in benchmark
the checking of llvm expected was not correct
2022-12-14 09:21:18 +01:00
rudy
bb756a6426 fix: default is global-error-probability=1/100_000 2022-12-14 09:21:18 +01:00
rudy
4ed0c01d8e feat(optimizer): option to disable optimizer cache
--optimizer-no-cache-on-disk
2022-12-13 18:54:40 +01:00
Quentin Bourgerie
d15c9822db fix(rust-bindings): Fix rust bindings after raise-encoding 2022-12-13 17:01:52 +01:00
aPere3
2fd9b6f0e3 refactor(encodings): raise plaintext/lut encodings higher up in the pipeline 2022-12-13 17:01:52 +01:00
Andi Drebes
e2e6df322e feat(compiler): Add support for full unrolling of loops with SDFG-convertible ops
This adds a new option `--unroll-loops-with-sdfg-convertible-ops`,
which causes loops containing SDFG-convertible operations to be fully
unrolled upon the extraction of SDFG-operations using the
`--emit-sdfg-ops` switch. This avoids constant roundtrips between an
SDFG-capable accelerator and the host during execution of a loop.

The option is limited to `scf.for` loops with static bounds and a
static step size. Since full unrolling of loops with large bounds
results in a large number of operations, the option is disabled by
default.
2022-12-13 12:03:51 +01:00
Quentin Bourgerie
07aa334d8d chore(benchmarks): Refactor the benchmark tools 2022-12-09 09:51:23 +01:00
Antoniu Pop
0dbb86bb36 feat(compiler): add lowering and bufferization for SDFG dialect, generate code to Stream Emulator API. 2022-12-08 14:54:14 +01:00
Antoniu Pop
752f0feb75 feat(runtime): add a stream emulator. 2022-12-08 14:54:14 +01:00
Andi Drebes
ee2f743a78 feat(compiler): Add new action dump-sdfg
Add a new action `dump-sdfg` to `concretecompiler` that causes the IR
to be dumpred right after the extraction of SDFG operations.
2022-12-08 14:54:14 +01:00
Andi Drebes
3da32560b7 feat(compiler): Add pass converting operations into SDFG processes
This adds a new pass `ExtractSDGOps`, which scans a function for
operations that implement `SDFGConvertibleOpInterface`, replaces them
with SDFG processes and constructs an SDFG graph around the processes.

Initialization and teardown of the SDFG graph are embedded into the
function and take place at the beginning of the function and before
the function's terminator, respectively.

The pass can be invoked using concretecompiler by specifying the new
compilation option `--emit-sdfg-ops` or programmatically on a
`CompilerEngine` using the new compilation option `extractSDFGOps`.
2022-12-08 14:54:14 +01:00
Andi Drebes
9ea6c0e8a3 enhance(compiler): Declare tensor-based BConcrete operations as side-effect-free 2022-12-08 14:54:14 +01:00
Andi Drebes
b7805f00d3 feat(compiler): Add op interface SDFGConvertibleOpInterface
This adds a new operation interface `SDFGConvertibleOpInterface` that
allows an operation to specify how it is converted to an SDFG
process. The interface consists of a single method `convert` that
receives as the arguments the DFG created using `SDFG.init`, a set of
SDFG input streams corresponding to the operands and a set of output
streams for results. The order of the input and output streams
corresponds to the order of the operands and output values,
respectively.
2022-12-08 14:54:14 +01:00
Andi Drebes
9f3615513b feat(compiler): Add new dialect SDFG for static data flow graphs
This adds a new dialect called "SDFG" for data flow graphs. An SDFG
data flow graph is composed of a set of processes, connected through
data streams. Special streams allow for data to be injected into and
to be retrieved from the data flow graph.

The dialect is intended to be lowered to API calls that allow for
offloading of the graph on hardware accelerators.
2022-12-08 14:54:14 +01:00
rudy
5516a55c1c feat: option to force cyphertext encoding 2022-12-08 14:42:25 +01:00
Quentin Bourgerie
2a89f62c1a fix(cuda): Include cuda_runtime.h in device.h to include the defininition of cudaStream_t 2022-12-07 21:32:01 +01:00
Quentin Bourgerie
3c616af622 feat(compiler): Handle batched operators for gpu codegen 2022-12-07 21:32:01 +01:00
Quentin Bourgerie
792d46fa80 enhance(runtime/gpu): Cache keys copy to gpu 2022-12-07 21:32:01 +01:00
tmontaigu
f36e1fe882 feat(CAPI): use const ptrs to create tensor lambda arguments 2022-12-07 15:55:52 +01:00
youben11
7d785eebec feat(rust): support serialization 2022-12-05 17:17:53 +01:00
youben11
fbc60097ab refactor(rust): resolve artifacts' path and better stringRef mgmnt 2022-12-05 17:17:53 +01:00
youben11
ad988de9a3 feat(CAPI): add CompilationFeedback API 2022-12-05 17:17:53 +01:00
youben11
f05b1bd1ea feat(rust): support execution with tensor args 2022-12-02 14:03:02 +01:00
youben11
16f3b0bbf6 feat(rust): manage erros coming from compiler
C struct now contains an additonal char* pointer, which can be either
NULL in case there is no error, or a buffer containing the error
message. It's the responsability of destructor function to free that
memory.
2022-12-02 14:03:02 +01:00
youben11
15b4aac0a1 feat(rust): support keygen, encryption, execution 2022-12-02 14:03:02 +01:00
youben11
7f55385ea2 feat(rust): load server lambda for later execution 2022-12-02 14:03:02 +01:00
youben11
b00115f4ae feat(rust): compile mlir into library
CAPI covering a wider API of the Support library.
Better error handling. Could also be improved by returning an error
message back from C to rust (left TODO).
2022-12-02 14:03:02 +01:00
Quentin Bourgerie
50973a39bd refactor(compiler): Remove async offloading of BS/KS 2022-11-30 10:29:19 +01:00
Quentin Bourgerie
9e16f31b87 refactor(bconcrete): Separate bufferization and CAPI call generation 2022-11-30 10:29:19 +01:00
Quentin Bourgerie
5d89ad0f84 enhance(feedback): Add p_error and global_perror to the compiler feedback 2022-11-24 09:59:19 +01:00
Umut
722e4d2eba feat: give crt decomposition feedback 2022-11-24 09:59:19 +01:00
youben11
5661b758d7 feat(CAPI): add initial API to do round-tripping with CompilerEngine 2022-11-23 14:01:25 +01:00
youben11
824aaaeff5 refactor(rust): separate generated CAPI under ffi module 2022-11-23 14:01:25 +01:00